diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,119226 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 19866,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3557046979865774e-08,
+      "loss": 1.5199,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.711409395973155e-08,
+      "loss": 1.6068,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0067114093959732e-07,
+      "loss": 1.5329,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.342281879194631e-07,
+      "loss": 1.8072,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6778523489932889e-07,
+      "loss": 1.4519,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0134228187919465e-07,
+      "loss": 1.4996,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3489932885906043e-07,
+      "loss": 1.6043,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.684563758389262e-07,
+      "loss": 0.3937,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.02013422818792e-07,
+      "loss": 1.5009,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3557046979865777e-07,
+      "loss": 1.5145,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6912751677852356e-07,
+      "loss": 1.7132,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.026845637583893e-07,
+      "loss": 1.4423,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.362416107382551e-07,
+      "loss": 1.5034,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6979865771812087e-07,
+      "loss": 1.596,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.033557046979866e-07,
+      "loss": 1.4662,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.369127516778524e-07,
+      "loss": 1.4837,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.704697986577181e-07,
+      "loss": 1.5547,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.04026845637584e-07,
+      "loss": 1.496,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.375838926174498e-07,
+      "loss": 1.4477,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.711409395973155e-07,
+      "loss": 1.5493,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.046979865771812e-07,
+      "loss": 1.5507,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.382550335570471e-07,
+      "loss": 1.5263,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.718120805369129e-07,
+      "loss": 1.653,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.053691275167786e-07,
+      "loss": 0.3416,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.389261744966444e-07,
+      "loss": 1.5985,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.724832214765102e-07,
+      "loss": 1.4179,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.060402684563759e-07,
+      "loss": 1.64,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.395973154362417e-07,
+      "loss": 1.4729,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.731543624161074e-07,
+      "loss": 1.4494,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.006711409395973e-06,
+      "loss": 1.421,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.040268456375839e-06,
+      "loss": 1.5142,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0738255033557048e-06,
+      "loss": 1.3601,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1073825503355707e-06,
+      "loss": 1.4267,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1409395973154363e-06,
+      "loss": 1.4159,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.174496644295302e-06,
+      "loss": 1.4551,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.208053691275168e-06,
+      "loss": 1.3831,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2416107382550337e-06,
+      "loss": 1.2966,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2751677852348996e-06,
+      "loss": 1.3796,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3087248322147653e-06,
+      "loss": 1.3548,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.342281879194631e-06,
+      "loss": 1.3182,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3758389261744968e-06,
+      "loss": 1.2788,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4093959731543623e-06,
+      "loss": 1.3689,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4429530201342285e-06,
+      "loss": 1.295,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4765100671140942e-06,
+      "loss": 1.3579,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.51006711409396e-06,
+      "loss": 0.3872,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5436241610738257e-06,
+      "loss": 1.3534,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5771812080536914e-06,
+      "loss": 1.4165,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6107382550335572e-06,
+      "loss": 1.5217,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6442953020134231e-06,
+      "loss": 1.2716,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6778523489932889e-06,
+      "loss": 1.2806,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7114093959731546e-06,
+      "loss": 1.2958,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7449664429530203e-06,
+      "loss": 1.2912,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.778523489932886e-06,
+      "loss": 1.3429,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8120805369127518e-06,
+      "loss": 1.3006,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8456375838926177e-06,
+      "loss": 1.1472,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8791946308724835e-06,
+      "loss": 1.2363,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.912751677852349e-06,
+      "loss": 1.2719,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9463087248322147e-06,
+      "loss": 1.2353,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9798657718120805e-06,
+      "loss": 1.1443,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.013422818791946e-06,
+      "loss": 1.2637,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0469798657718124e-06,
+      "loss": 1.0585,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.080536912751678e-06,
+      "loss": 1.2558,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.114093959731544e-06,
+      "loss": 1.1395,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1476510067114096e-06,
+      "loss": 1.2486,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1812080536912753e-06,
+      "loss": 1.3411,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2147651006711415e-06,
+      "loss": 1.2538,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2483221476510068e-06,
+      "loss": 1.2352,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2818791946308725e-06,
+      "loss": 1.2931,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3154362416107382e-06,
+      "loss": 1.2271,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.348993288590604e-06,
+      "loss": 1.2058,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3825503355704697e-06,
+      "loss": 1.0626,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.416107382550336e-06,
+      "loss": 1.1576,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4496644295302016e-06,
+      "loss": 1.3223,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4832214765100673e-06,
+      "loss": 1.201,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5167785234899335e-06,
+      "loss": 1.1589,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5503355704697992e-06,
+      "loss": 1.1961,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.583892617449665e-06,
+      "loss": 0.4277,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6174496644295307e-06,
+      "loss": 1.2036,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6510067114093964e-06,
+      "loss": 1.218,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.684563758389262e-06,
+      "loss": 1.1816,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.718120805369128e-06,
+      "loss": 1.2788,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7516778523489936e-06,
+      "loss": 1.24,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.785234899328859e-06,
+      "loss": 1.0352,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8187919463087247e-06,
+      "loss": 1.2423,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8523489932885904e-06,
+      "loss": 0.3853,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.885906040268457e-06,
+      "loss": 1.2158,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9194630872483227e-06,
+      "loss": 1.1332,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9530201342281885e-06,
+      "loss": 1.1946,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.986577181208054e-06,
+      "loss": 1.1448,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.02013422818792e-06,
+      "loss": 1.1569,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0536912751677857e-06,
+      "loss": 1.1632,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0872483221476514e-06,
+      "loss": 1.1483,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.120805369127517e-06,
+      "loss": 1.1341,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.154362416107383e-06,
+      "loss": 1.1582,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1879194630872486e-06,
+      "loss": 1.1548,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2214765100671143e-06,
+      "loss": 1.0326,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.25503355704698e-06,
+      "loss": 1.1497,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2885906040268462e-06,
+      "loss": 1.1591,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.322147651006712e-06,
+      "loss": 1.2194,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3557046979865777e-06,
+      "loss": 1.1127,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3892617449664434e-06,
+      "loss": 1.1905,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.422818791946309e-06,
+      "loss": 1.2272,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.456375838926175e-06,
+      "loss": 1.1438,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4899328859060407e-06,
+      "loss": 1.1316,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5234899328859064e-06,
+      "loss": 1.0854,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.557046979865772e-06,
+      "loss": 1.2703,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.590604026845638e-06,
+      "loss": 1.1565,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6241610738255036e-06,
+      "loss": 1.1533,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6577181208053697e-06,
+      "loss": 1.1033,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6912751677852355e-06,
+      "loss": 1.1031,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7248322147651012e-06,
+      "loss": 1.0808,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.758389261744967e-06,
+      "loss": 1.1443,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7919463087248327e-06,
+      "loss": 1.2217,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.825503355704698e-06,
+      "loss": 1.2133,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.859060402684564e-06,
+      "loss": 1.155,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8926174496644295e-06,
+      "loss": 1.0743,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.926174496644295e-06,
+      "loss": 1.1259,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.959731543624161e-06,
+      "loss": 1.1599,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.993288590604027e-06,
+      "loss": 1.1108,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.026845637583892e-06,
+      "loss": 1.1725,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.060402684563759e-06,
+      "loss": 1.1439,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.093959731543625e-06,
+      "loss": 1.1605,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1275167785234905e-06,
+      "loss": 1.1155,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.161073825503356e-06,
+      "loss": 1.1405,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.194630872483222e-06,
+      "loss": 1.0718,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.228187919463088e-06,
+      "loss": 1.0854,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.261744966442953e-06,
+      "loss": 1.2319,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.295302013422819e-06,
+      "loss": 1.1169,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.328859060402685e-06,
+      "loss": 1.2139,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.362416107382551e-06,
+      "loss": 1.1047,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.395973154362416e-06,
+      "loss": 1.1092,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.429530201342283e-06,
+      "loss": 1.0579,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.463087248322149e-06,
+      "loss": 1.1346,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4966442953020135e-06,
+      "loss": 1.2008,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.530201342281879e-06,
+      "loss": 1.0988,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.563758389261745e-06,
+      "loss": 1.0405,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.597315436241611e-06,
+      "loss": 1.0965,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6308724832214765e-06,
+      "loss": 1.0654,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.664429530201342e-06,
+      "loss": 1.2358,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.697986577181208e-06,
+      "loss": 1.0952,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.731543624161074e-06,
+      "loss": 1.0039,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.765100671140939e-06,
+      "loss": 1.1413,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.798657718120805e-06,
+      "loss": 1.0757,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.832214765100672e-06,
+      "loss": 1.1255,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8657718120805375e-06,
+      "loss": 1.1319,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.899328859060403e-06,
+      "loss": 1.1039,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.932885906040269e-06,
+      "loss": 1.0424,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.966442953020135e-06,
+      "loss": 1.0344,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 1.1502,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.033557046979867e-06,
+      "loss": 1.1308,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.067114093959732e-06,
+      "loss": 1.1005,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1006711409395985e-06,
+      "loss": 1.1485,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.134228187919463e-06,
+      "loss": 1.0858,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.16778523489933e-06,
+      "loss": 1.144,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.201342281879195e-06,
+      "loss": 1.0484,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.234899328859061e-06,
+      "loss": 1.079,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.268456375838926e-06,
+      "loss": 1.0624,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.302013422818793e-06,
+      "loss": 1.1422,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.335570469798658e-06,
+      "loss": 1.0736,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.369127516778524e-06,
+      "loss": 1.201,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.40268456375839e-06,
+      "loss": 1.2214,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.436241610738256e-06,
+      "loss": 1.1599,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4697986577181215e-06,
+      "loss": 1.1565,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.503355704697987e-06,
+      "loss": 1.0865,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.536912751677853e-06,
+      "loss": 1.1181,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.570469798657718e-06,
+      "loss": 1.0703,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.6040268456375845e-06,
+      "loss": 0.998,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.637583892617449e-06,
+      "loss": 1.1467,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.671140939597316e-06,
+      "loss": 1.1109,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.704697986577181e-06,
+      "loss": 1.1556,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.738255033557047e-06,
+      "loss": 0.9959,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.771812080536914e-06,
+      "loss": 1.1045,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.805369127516779e-06,
+      "loss": 1.0826,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8389261744966455e-06,
+      "loss": 1.0626,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.87248322147651e-06,
+      "loss": 1.0375,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.906040268456377e-06,
+      "loss": 1.1457,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.939597315436242e-06,
+      "loss": 1.1166,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.973154362416108e-06,
+      "loss": 1.1196,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.006711409395973e-06,
+      "loss": 1.162,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.04026845637584e-06,
+      "loss": 1.1013,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.073825503355705e-06,
+      "loss": 1.0646,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.107382550335571e-06,
+      "loss": 1.1094,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.140939597315436e-06,
+      "loss": 1.0516,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.174496644295303e-06,
+      "loss": 1.0772,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2080536912751686e-06,
+      "loss": 1.065,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.241610738255034e-06,
+      "loss": 0.3481,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2751677852349e-06,
+      "loss": 0.9981,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.308724832214766e-06,
+      "loss": 1.0402,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3422818791946315e-06,
+      "loss": 1.0337,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.375838926174497e-06,
+      "loss": 0.968,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.409395973154363e-06,
+      "loss": 1.0704,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.442953020134229e-06,
+      "loss": 1.0553,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.4765100671140944e-06,
+      "loss": 1.0443,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.51006711409396e-06,
+      "loss": 1.0654,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.543624161073826e-06,
+      "loss": 1.128,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5771812080536925e-06,
+      "loss": 1.1109,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.610738255033557e-06,
+      "loss": 1.1867,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.644295302013424e-06,
+      "loss": 1.0875,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.677852348993289e-06,
+      "loss": 0.9852,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.711409395973155e-06,
+      "loss": 1.0096,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.74496644295302e-06,
+      "loss": 1.0733,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.778523489932887e-06,
+      "loss": 1.0896,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.812080536912752e-06,
+      "loss": 1.1265,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.845637583892618e-06,
+      "loss": 1.0887,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.879194630872483e-06,
+      "loss": 1.2117,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.91275167785235e-06,
+      "loss": 1.083,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.9463087248322156e-06,
+      "loss": 1.1484,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.979865771812081e-06,
+      "loss": 1.0317,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.013422818791947e-06,
+      "loss": 1.1371,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.046979865771813e-06,
+      "loss": 1.0437,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.0805369127516785e-06,
+      "loss": 1.0629,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.114093959731544e-06,
+      "loss": 1.0804,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.14765100671141e-06,
+      "loss": 0.9986,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.181208053691276e-06,
+      "loss": 1.106,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.2147651006711414e-06,
+      "loss": 1.0453,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.248322147651007e-06,
+      "loss": 1.022,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.281879194630873e-06,
+      "loss": 1.1152,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.3154362416107395e-06,
+      "loss": 1.0304,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.348993288590604e-06,
+      "loss": 1.1202,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.382550335570471e-06,
+      "loss": 1.0278,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.416107382550336e-06,
+      "loss": 1.058,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.4496644295302024e-06,
+      "loss": 1.0008,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.483221476510067e-06,
+      "loss": 0.411,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.516778523489934e-06,
+      "loss": 1.077,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.550335570469799e-06,
+      "loss": 1.0765,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.583892617449665e-06,
+      "loss": 1.0413,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.61744966442953e-06,
+      "loss": 1.0419,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.651006711409396e-06,
+      "loss": 1.0911,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.684563758389262e-06,
+      "loss": 1.1135,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.718120805369127e-06,
+      "loss": 1.0743,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.751677852348995e-06,
+      "loss": 1.0509,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.785234899328859e-06,
+      "loss": 0.9843,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.818791946308726e-06,
+      "loss": 0.9511,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.85234899328859e-06,
+      "loss": 0.9681,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.885906040268458e-06,
+      "loss": 1.1326,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.919463087248322e-06,
+      "loss": 1.1006,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.95302013422819e-06,
+      "loss": 1.1611,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.986577181208053e-06,
+      "loss": 1.0927,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.02013422818792e-06,
+      "loss": 1.0908,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.053691275167785e-06,
+      "loss": 1.1639,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.087248322147652e-06,
+      "loss": 1.121,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.120805369127518e-06,
+      "loss": 1.1105,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.154362416107384e-06,
+      "loss": 0.9988,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.18791946308725e-06,
+      "loss": 1.0728,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.221476510067115e-06,
+      "loss": 1.1173,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.255033557046981e-06,
+      "loss": 1.0893,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.288590604026847e-06,
+      "loss": 1.0237,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.322147651006712e-06,
+      "loss": 1.0105,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.355704697986578e-06,
+      "loss": 1.1247,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.389261744966444e-06,
+      "loss": 1.0418,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.42281879194631e-06,
+      "loss": 1.1164,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.456375838926175e-06,
+      "loss": 1.0627,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.489932885906041e-06,
+      "loss": 1.1649,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.523489932885907e-06,
+      "loss": 1.0296,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.557046979865773e-06,
+      "loss": 1.1027,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.590604026845638e-06,
+      "loss": 1.0925,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.624161073825504e-06,
+      "loss": 1.0998,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.65771812080537e-06,
+      "loss": 1.0377,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.691275167785235e-06,
+      "loss": 1.0825,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.724832214765101e-06,
+      "loss": 1.1021,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.758389261744967e-06,
+      "loss": 1.1139,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.791946308724833e-06,
+      "loss": 1.0652,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.825503355704698e-06,
+      "loss": 1.0825,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.859060402684566e-06,
+      "loss": 1.1082,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.89261744966443e-06,
+      "loss": 1.0344,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.926174496644297e-06,
+      "loss": 1.1046,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.959731543624161e-06,
+      "loss": 1.016,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.993288590604027e-06,
+      "loss": 1.0334,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.026845637583893e-06,
+      "loss": 1.1113,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.060402684563759e-06,
+      "loss": 1.0334,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.093959731543624e-06,
+      "loss": 1.0577,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.12751677852349e-06,
+      "loss": 1.0621,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.161073825503356e-06,
+      "loss": 1.1468,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.194630872483221e-06,
+      "loss": 1.1211,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.228187919463087e-06,
+      "loss": 1.0743,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.261744966442953e-06,
+      "loss": 1.0429,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.29530201342282e-06,
+      "loss": 0.9983,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.328859060402684e-06,
+      "loss": 1.1,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.362416107382552e-06,
+      "loss": 1.0454,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.395973154362416e-06,
+      "loss": 1.0297,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.429530201342283e-06,
+      "loss": 0.9662,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.463087248322147e-06,
+      "loss": 1.0512,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.496644295302015e-06,
+      "loss": 1.0204,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.530201342281879e-06,
+      "loss": 1.1009,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.563758389261746e-06,
+      "loss": 1.0481,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.59731543624161e-06,
+      "loss": 1.0904,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.630872483221478e-06,
+      "loss": 1.0281,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.664429530201343e-06,
+      "loss": 0.4094,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.69798657718121e-06,
+      "loss": 1.0977,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.731543624161075e-06,
+      "loss": 1.0101,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.76510067114094e-06,
+      "loss": 0.4353,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.798657718120806e-06,
+      "loss": 1.0734,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.832214765100672e-06,
+      "loss": 1.0884,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.865771812080538e-06,
+      "loss": 1.0074,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.899328859060404e-06,
+      "loss": 1.092,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.93288590604027e-06,
+      "loss": 1.1103,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.966442953020135e-06,
+      "loss": 1.0022,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1e-05,
+      "loss": 0.3732,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0033557046979867e-05,
+      "loss": 1.0269,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0067114093959734e-05,
+      "loss": 1.1365,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0100671140939598e-05,
+      "loss": 0.4671,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0134228187919464e-05,
+      "loss": 1.0883,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.016778523489933e-05,
+      "loss": 1.0823,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0201342281879197e-05,
+      "loss": 1.0725,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0234899328859061e-05,
+      "loss": 1.1031,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0268456375838927e-05,
+      "loss": 1.0252,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0302013422818792e-05,
+      "loss": 1.1018,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.033557046979866e-05,
+      "loss": 1.0953,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0369127516778524e-05,
+      "loss": 1.0496,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.040268456375839e-05,
+      "loss": 1.0615,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0436241610738255e-05,
+      "loss": 1.0574,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0469798657718123e-05,
+      "loss": 1.1123,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0503355704697989e-05,
+      "loss": 1.0605,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0536912751677853e-05,
+      "loss": 1.0323,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0570469798657718e-05,
+      "loss": 1.1464,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0604026845637586e-05,
+      "loss": 1.1688,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0637583892617451e-05,
+      "loss": 1.1365,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0671140939597316e-05,
+      "loss": 1.0179,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0704697986577181e-05,
+      "loss": 1.111,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0738255033557049e-05,
+      "loss": 1.0871,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0771812080536914e-05,
+      "loss": 1.0511,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.080536912751678e-05,
+      "loss": 1.074,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0838926174496644e-05,
+      "loss": 0.3361,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0872483221476512e-05,
+      "loss": 1.08,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0906040268456377e-05,
+      "loss": 1.0615,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0939597315436243e-05,
+      "loss": 1.1252,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0973154362416107e-05,
+      "loss": 1.1132,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1006711409395975e-05,
+      "loss": 1.1195,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.104026845637584e-05,
+      "loss": 0.9608,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1073825503355706e-05,
+      "loss": 1.0077,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.110738255033557e-05,
+      "loss": 1.116,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1140939597315436e-05,
+      "loss": 1.021,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1174496644295303e-05,
+      "loss": 1.0413,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1208053691275169e-05,
+      "loss": 0.4165,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1241610738255035e-05,
+      "loss": 1.0521,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1275167785234899e-05,
+      "loss": 1.1688,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1308724832214766e-05,
+      "loss": 0.432,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1342281879194632e-05,
+      "loss": 1.0723,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1375838926174498e-05,
+      "loss": 1.0669,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1409395973154362e-05,
+      "loss": 1.0942,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1442953020134229e-05,
+      "loss": 1.0377,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1476510067114095e-05,
+      "loss": 0.9859,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.151006711409396e-05,
+      "loss": 0.3576,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1543624161073828e-05,
+      "loss": 1.0197,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1577181208053692e-05,
+      "loss": 1.1231,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1610738255033558e-05,
+      "loss": 1.072,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1644295302013424e-05,
+      "loss": 1.0378,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1677852348993291e-05,
+      "loss": 1.0619,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1711409395973155e-05,
+      "loss": 1.1158,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.174496644295302e-05,
+      "loss": 1.0239,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1778523489932886e-05,
+      "loss": 1.1775,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1812080536912754e-05,
+      "loss": 0.9655,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1845637583892618e-05,
+      "loss": 1.1413,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1879194630872484e-05,
+      "loss": 1.1034,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.191275167785235e-05,
+      "loss": 1.087,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1946308724832217e-05,
+      "loss": 1.05,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1979865771812083e-05,
+      "loss": 0.9975,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2013422818791947e-05,
+      "loss": 1.0158,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2046979865771812e-05,
+      "loss": 1.0716,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.208053691275168e-05,
+      "loss": 1.0209,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2114093959731545e-05,
+      "loss": 1.0844,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.214765100671141e-05,
+      "loss": 1.0163,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2181208053691275e-05,
+      "loss": 1.0245,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2214765100671143e-05,
+      "loss": 1.1059,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2248322147651008e-05,
+      "loss": 1.0525,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2281879194630872e-05,
+      "loss": 1.0062,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2315436241610738e-05,
+      "loss": 1.1189,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2348993288590606e-05,
+      "loss": 1.0542,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2382550335570471e-05,
+      "loss": 1.0819,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2416107382550337e-05,
+      "loss": 1.1233,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2449664429530201e-05,
+      "loss": 1.1152,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2483221476510069e-05,
+      "loss": 1.0083,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2516778523489934e-05,
+      "loss": 1.0994,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.25503355704698e-05,
+      "loss": 1.0235,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2583892617449664e-05,
+      "loss": 0.9616,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2617449664429532e-05,
+      "loss": 1.0316,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2651006711409397e-05,
+      "loss": 1.0879,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2684563758389263e-05,
+      "loss": 0.3388,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.271812080536913e-05,
+      "loss": 1.02,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2751677852348994e-05,
+      "loss": 1.0227,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.278523489932886e-05,
+      "loss": 1.1413,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2818791946308726e-05,
+      "loss": 1.093,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2852348993288593e-05,
+      "loss": 1.0504,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2885906040268457e-05,
+      "loss": 1.0504,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2919463087248323e-05,
+      "loss": 0.9951,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2953020134228189e-05,
+      "loss": 1.0346,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2986577181208055e-05,
+      "loss": 1.0732,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.302013422818792e-05,
+      "loss": 1.0746,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3053691275167786e-05,
+      "loss": 1.0647,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3087248322147652e-05,
+      "loss": 1.0092,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3120805369127518e-05,
+      "loss": 1.1107,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3154362416107385e-05,
+      "loss": 0.4044,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3187919463087249e-05,
+      "loss": 1.0149,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3221476510067115e-05,
+      "loss": 1.0769,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.325503355704698e-05,
+      "loss": 1.0708,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3288590604026848e-05,
+      "loss": 1.0296,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3322147651006712e-05,
+      "loss": 0.9653,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3355704697986578e-05,
+      "loss": 1.1191,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3389261744966443e-05,
+      "loss": 1.0989,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.342281879194631e-05,
+      "loss": 1.1157,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3456375838926175e-05,
+      "loss": 1.0498,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.348993288590604e-05,
+      "loss": 0.9963,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3523489932885906e-05,
+      "loss": 1.0774,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3557046979865774e-05,
+      "loss": 1.0936,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.359060402684564e-05,
+      "loss": 1.1184,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3624161073825504e-05,
+      "loss": 1.0699,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.365771812080537e-05,
+      "loss": 1.0398,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3691275167785237e-05,
+      "loss": 1.1092,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3724832214765102e-05,
+      "loss": 1.1246,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3758389261744966e-05,
+      "loss": 1.0413,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3791946308724832e-05,
+      "loss": 0.9533,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.38255033557047e-05,
+      "loss": 1.0791,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3859060402684565e-05,
+      "loss": 1.0619,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3892617449664431e-05,
+      "loss": 1.0119,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3926174496644295e-05,
+      "loss": 0.993,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3959731543624163e-05,
+      "loss": 0.9676,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3993288590604028e-05,
+      "loss": 1.0515,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4026845637583894e-05,
+      "loss": 1.091,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4060402684563758e-05,
+      "loss": 1.0473,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4093959731543626e-05,
+      "loss": 1.0754,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4127516778523491e-05,
+      "loss": 1.0706,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4161073825503357e-05,
+      "loss": 1.1156,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4194630872483221e-05,
+      "loss": 1.0398,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4228187919463088e-05,
+      "loss": 1.1077,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4261744966442954e-05,
+      "loss": 1.0332,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.429530201342282e-05,
+      "loss": 0.3643,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4328859060402687e-05,
+      "loss": 1.0487,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4362416107382551e-05,
+      "loss": 1.0018,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4395973154362417e-05,
+      "loss": 0.341,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4429530201342283e-05,
+      "loss": 1.0637,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.446308724832215e-05,
+      "loss": 1.0356,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4496644295302014e-05,
+      "loss": 0.3279,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.453020134228188e-05,
+      "loss": 1.0939,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4563758389261746e-05,
+      "loss": 1.1327,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4597315436241613e-05,
+      "loss": 1.062,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4630872483221479e-05,
+      "loss": 0.9375,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4664429530201343e-05,
+      "loss": 1.0861,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4697986577181209e-05,
+      "loss": 0.9442,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4731543624161076e-05,
+      "loss": 1.1202,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4765100671140942e-05,
+      "loss": 1.1445,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4798657718120806e-05,
+      "loss": 1.0339,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4832214765100672e-05,
+      "loss": 1.0812,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4865771812080539e-05,
+      "loss": 1.0394,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4899328859060405e-05,
+      "loss": 0.936,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4932885906040269e-05,
+      "loss": 0.9464,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4966442953020135e-05,
+      "loss": 0.8927,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 1.0891,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5033557046979868e-05,
+      "loss": 1.042,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5067114093959734e-05,
+      "loss": 1.0362,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5100671140939598e-05,
+      "loss": 1.0353,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5134228187919463e-05,
+      "loss": 1.0332,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.516778523489933e-05,
+      "loss": 1.0246,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5201342281879196e-05,
+      "loss": 0.9751,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.523489932885906e-05,
+      "loss": 1.0239,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5268456375838928e-05,
+      "loss": 1.1249,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5302013422818792e-05,
+      "loss": 1.0658,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.533557046979866e-05,
+      "loss": 1.0384,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5369127516778523e-05,
+      "loss": 1.025,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.540268456375839e-05,
+      "loss": 1.0485,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5436241610738255e-05,
+      "loss": 1.0539,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5469798657718122e-05,
+      "loss": 1.0105,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.550335570469799e-05,
+      "loss": 1.1211,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5536912751677854e-05,
+      "loss": 1.0245,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5570469798657718e-05,
+      "loss": 1.048,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5604026845637585e-05,
+      "loss": 1.0274,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5637583892617453e-05,
+      "loss": 1.0111,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5671140939597317e-05,
+      "loss": 1.0524,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.570469798657718e-05,
+      "loss": 1.0924,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5738255033557048e-05,
+      "loss": 1.0153,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5771812080536916e-05,
+      "loss": 1.1116,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.580536912751678e-05,
+      "loss": 1.1563,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5838926174496644e-05,
+      "loss": 1.0781,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.587248322147651e-05,
+      "loss": 1.0566,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.590604026845638e-05,
+      "loss": 1.0852,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5939597315436243e-05,
+      "loss": 1.0702,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5973154362416107e-05,
+      "loss": 1.086,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6006711409395974e-05,
+      "loss": 1.0231,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.604026845637584e-05,
+      "loss": 1.0523,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6073825503355706e-05,
+      "loss": 1.1293,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.610738255033557e-05,
+      "loss": 1.1381,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6140939597315437e-05,
+      "loss": 0.348,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6174496644295304e-05,
+      "loss": 1.0971,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.620805369127517e-05,
+      "loss": 1.0951,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6241610738255036e-05,
+      "loss": 1.0496,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.62751677852349e-05,
+      "loss": 1.0202,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6308724832214767e-05,
+      "loss": 1.0851,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.634228187919463e-05,
+      "loss": 1.0943,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.63758389261745e-05,
+      "loss": 1.0292,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6409395973154363e-05,
+      "loss": 1.113,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.644295302013423e-05,
+      "loss": 1.0528,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6476510067114094e-05,
+      "loss": 1.0488,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6510067114093962e-05,
+      "loss": 1.1569,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.654362416107383e-05,
+      "loss": 1.0635,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6577181208053693e-05,
+      "loss": 0.9442,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6610738255033557e-05,
+      "loss": 1.1285,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6644295302013425e-05,
+      "loss": 1.0899,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6677852348993292e-05,
+      "loss": 1.0071,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6711409395973156e-05,
+      "loss": 0.3414,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.674496644295302e-05,
+      "loss": 1.163,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6778523489932888e-05,
+      "loss": 1.0309,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6812080536912755e-05,
+      "loss": 1.09,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.684563758389262e-05,
+      "loss": 0.3381,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6879194630872483e-05,
+      "loss": 1.1045,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.691275167785235e-05,
+      "loss": 1.0348,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6946308724832218e-05,
+      "loss": 1.0138,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6979865771812082e-05,
+      "loss": 0.3935,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7013422818791946e-05,
+      "loss": 0.9681,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7046979865771814e-05,
+      "loss": 1.0463,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.708053691275168e-05,
+      "loss": 1.1321,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7114093959731545e-05,
+      "loss": 1.0647,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.714765100671141e-05,
+      "loss": 1.0995,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7181208053691277e-05,
+      "loss": 1.0242,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7214765100671144e-05,
+      "loss": 0.4593,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7248322147651008e-05,
+      "loss": 1.0072,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7281879194630872e-05,
+      "loss": 0.3299,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.731543624161074e-05,
+      "loss": 1.1048,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7348993288590607e-05,
+      "loss": 1.0921,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.738255033557047e-05,
+      "loss": 1.1166,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.741610738255034e-05,
+      "loss": 1.0742,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7449664429530202e-05,
+      "loss": 1.0802,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.748322147651007e-05,
+      "loss": 1.1035,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7516778523489934e-05,
+      "loss": 1.0864,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.75503355704698e-05,
+      "loss": 0.9796,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7583892617449665e-05,
+      "loss": 1.0814,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7617449664429533e-05,
+      "loss": 1.0958,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7651006711409397e-05,
+      "loss": 1.0844,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7684563758389264e-05,
+      "loss": 1.1239,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.771812080536913e-05,
+      "loss": 1.1326,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7751677852348996e-05,
+      "loss": 1.025,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.778523489932886e-05,
+      "loss": 1.167,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7818791946308727e-05,
+      "loss": 1.1244,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7852348993288595e-05,
+      "loss": 1.1183,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.788590604026846e-05,
+      "loss": 1.0518,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7919463087248323e-05,
+      "loss": 1.0344,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.795302013422819e-05,
+      "loss": 1.0331,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7986577181208054e-05,
+      "loss": 1.0644,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.802013422818792e-05,
+      "loss": 0.9597,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8053691275167786e-05,
+      "loss": 1.1213,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8087248322147653e-05,
+      "loss": 1.0946,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8120805369127517e-05,
+      "loss": 1.0434,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8154362416107385e-05,
+      "loss": 1.0615,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.818791946308725e-05,
+      "loss": 1.0553,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8221476510067116e-05,
+      "loss": 1.0009,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.825503355704698e-05,
+      "loss": 1.0472,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8288590604026847e-05,
+      "loss": 1.124,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.832214765100671e-05,
+      "loss": 1.0803,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.835570469798658e-05,
+      "loss": 1.0847,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8389261744966443e-05,
+      "loss": 1.0612,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.842281879194631e-05,
+      "loss": 1.1382,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8456375838926174e-05,
+      "loss": 0.9648,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8489932885906042e-05,
+      "loss": 1.0876,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8523489932885906e-05,
+      "loss": 1.0807,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8557046979865773e-05,
+      "loss": 1.1188,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.859060402684564e-05,
+      "loss": 1.0378,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8624161073825505e-05,
+      "loss": 0.9868,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.865771812080537e-05,
+      "loss": 1.0131,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8691275167785236e-05,
+      "loss": 1.0329,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8724832214765104e-05,
+      "loss": 1.1003,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8758389261744968e-05,
+      "loss": 1.0941,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8791946308724832e-05,
+      "loss": 1.0761,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.88255033557047e-05,
+      "loss": 1.042,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8859060402684567e-05,
+      "loss": 1.0572,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.889261744966443e-05,
+      "loss": 1.1119,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8926174496644295e-05,
+      "loss": 1.0798,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8959731543624162e-05,
+      "loss": 1.001,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.899328859060403e-05,
+      "loss": 1.0611,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9026845637583894e-05,
+      "loss": 1.0372,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9060402684563758e-05,
+      "loss": 1.0138,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9093959731543625e-05,
+      "loss": 1.083,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9127516778523493e-05,
+      "loss": 1.0331,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9161073825503357e-05,
+      "loss": 1.0998,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.919463087248322e-05,
+      "loss": 1.1196,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9228187919463088e-05,
+      "loss": 1.1123,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9261744966442955e-05,
+      "loss": 0.3488,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.929530201342282e-05,
+      "loss": 1.0397,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9328859060402687e-05,
+      "loss": 1.0372,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.936241610738255e-05,
+      "loss": 0.3856,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.939597315436242e-05,
+      "loss": 1.1361,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9429530201342282e-05,
+      "loss": 1.0285,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.946308724832215e-05,
+      "loss": 1.066,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9496644295302014e-05,
+      "loss": 1.0955,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.953020134228188e-05,
+      "loss": 1.0777,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9563758389261745e-05,
+      "loss": 1.0786,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9597315436241613e-05,
+      "loss": 1.014,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.963087248322148e-05,
+      "loss": 1.0713,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9664429530201344e-05,
+      "loss": 1.1096,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.969798657718121e-05,
+      "loss": 0.9774,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9731543624161076e-05,
+      "loss": 1.0916,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9765100671140943e-05,
+      "loss": 1.0419,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9798657718120807e-05,
+      "loss": 1.038,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983221476510067e-05,
+      "loss": 1.023,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.986577181208054e-05,
+      "loss": 1.0566,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9899328859060406e-05,
+      "loss": 1.0275,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.993288590604027e-05,
+      "loss": 1.0581,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9966442953020134e-05,
+      "loss": 1.0814,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.0201,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999867105713e-05,
+      "loss": 1.0831,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999468422846e-05,
+      "loss": 1.0969,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998803951418e-05,
+      "loss": 1.0468,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999997873691442e-05,
+      "loss": 1.0942,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999667764294e-05,
+      "loss": 1.0409,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999521580595e-05,
+      "loss": 1.0202,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999348818051e-05,
+      "loss": 1.0252,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999991494766663e-05,
+      "loss": 1.1162,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999989235564462e-05,
+      "loss": 1.0656,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998671057397e-05,
+      "loss": 1.0182,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998391979525e-05,
+      "loss": 1.0516,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998086322838e-05,
+      "loss": 1.0425,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999997754087344e-05,
+      "loss": 1.0765,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999973952730517e-05,
+      "loss": 1.0329,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999997009879971e-05,
+      "loss": 1.0344,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999965979081114e-05,
+      "loss": 1.1677,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996159357485e-05,
+      "loss": 1.0917,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999995694228102e-05,
+      "loss": 1.0213,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999952025199762e-05,
+      "loss": 1.0499,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999946842331197e-05,
+      "loss": 1.0601,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999994139367547e-05,
+      "loss": 0.943,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999935679232716e-05,
+      "loss": 1.0599,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999929699003095e-05,
+      "loss": 1.1103,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999923452986763e-05,
+      "loss": 1.0419,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999916941183887e-05,
+      "loss": 0.9919,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999991016359464e-05,
+      "loss": 1.132,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999903120219203e-05,
+      "loss": 1.0834,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999895811057765e-05,
+      "loss": 0.9727,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988823611051e-05,
+      "loss": 1.108,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988039537765e-05,
+      "loss": 1.0467,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999872288859392e-05,
+      "loss": 1.0924,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999863916555945e-05,
+      "loss": 1.0216,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999855278467538e-05,
+      "loss": 0.9697,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99998463745944e-05,
+      "loss": 1.0557,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999837204936768e-05,
+      "loss": 0.9948,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982776949488e-05,
+      "loss": 1.0146,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999818068268993e-05,
+      "loss": 1.0233,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999980810125936e-05,
+      "loss": 1.0614,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999979786846625e-05,
+      "loss": 1.0892,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999787369889936e-05,
+      "loss": 1.0873,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999776605530693e-05,
+      "loss": 1.0478,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999976557538881e-05,
+      "loss": 0.9679,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999754279464577e-05,
+      "loss": 1.0771,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999742717758294e-05,
+      "loss": 1.0507,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999730890270272e-05,
+      "loss": 1.0631,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999718797000825e-05,
+      "loss": 0.4141,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999970643795027e-05,
+      "loss": 1.0085,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999693813118943e-05,
+      "loss": 1.0008,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999680922507174e-05,
+      "loss": 1.0022,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999667766115304e-05,
+      "loss": 1.132,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999965434394369e-05,
+      "loss": 0.4256,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999964065599268e-05,
+      "loss": 1.0603,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999626702262645e-05,
+      "loss": 1.2125,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999961248275395e-05,
+      "loss": 1.0391,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999597997466977e-05,
+      "loss": 1.12,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999583246402112e-05,
+      "loss": 1.0972,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999568229559742e-05,
+      "loss": 1.1149,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999955294694027e-05,
+      "loss": 0.96,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99995373985441e-05,
+      "loss": 1.0082,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999952158437165e-05,
+      "loss": 1.0823,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999505504423337e-05,
+      "loss": 1.0502,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999489158699585e-05,
+      "loss": 1.0334,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999947254720083e-05,
+      "loss": 1.0835,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999945566992752e-05,
+      "loss": 1.0675,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999438526880098e-05,
+      "loss": 1.0669,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999942111805902e-05,
+      "loss": 1.045,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999403443464745e-05,
+      "loss": 1.0276,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999385503097753e-05,
+      "loss": 1.0093,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999936729695851e-05,
+      "loss": 1.0981,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999348825047507e-05,
+      "loss": 1.0263,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999330087365233e-05,
+      "loss": 1.0348,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999311083912186e-05,
+      "loss": 1.0428,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999291814688868e-05,
+      "loss": 1.0547,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999272279695795e-05,
+      "loss": 1.0305,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999252478933484e-05,
+      "loss": 1.0363,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999232412402464e-05,
+      "loss": 0.9945,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999212080103268e-05,
+      "loss": 1.1211,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999191482036434e-05,
+      "loss": 1.0798,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999917061820251e-05,
+      "loss": 0.9445,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999149488602052e-05,
+      "loss": 1.0591,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999128093235616e-05,
+      "loss": 1.1137,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999106432103785e-05,
+      "loss": 1.0676,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999084505207115e-05,
+      "loss": 0.8944,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999062312546205e-05,
+      "loss": 1.0476,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999039854121637e-05,
+      "loss": 1.0726,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999901712993401e-05,
+      "loss": 1.1354,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998994139983927e-05,
+      "loss": 1.0479,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998970884272e-05,
+      "loss": 1.0139,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998947362798846e-05,
+      "loss": 1.0933,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998923575565094e-05,
+      "loss": 1.064,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999889952257137e-05,
+      "loss": 1.1029,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998875203818317e-05,
+      "loss": 1.061,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998850619306583e-05,
+      "loss": 1.0013,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999882576903682e-05,
+      "loss": 1.0394,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998800653009682e-05,
+      "loss": 1.0739,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998775271225846e-05,
+      "loss": 1.1288,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998749623685985e-05,
+      "loss": 1.0389,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998723710390776e-05,
+      "loss": 1.0872,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998697531340912e-05,
+      "loss": 1.0296,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998671086537083e-05,
+      "loss": 1.0603,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999864437598e-05,
+      "loss": 1.0134,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998617399670367e-05,
+      "loss": 1.034,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999859015760891e-05,
+      "loss": 0.9933,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999856264979634e-05,
+      "loss": 1.0984,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999853487623339e-05,
+      "loss": 1.0635,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999850683692081e-05,
+      "loss": 1.1021,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998478531859335e-05,
+      "loss": 1.039,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999844996104972e-05,
+      "loss": 1.0483,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998421124492724e-05,
+      "loss": 1.0442,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998392022189112e-05,
+      "loss": 1.0715,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998362654139663e-05,
+      "loss": 1.119,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999833302034515e-05,
+      "loss": 1.0808,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999830312080637e-05,
+      "loss": 1.0466,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998272955524107e-05,
+      "loss": 1.0262,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998242524499168e-05,
+      "loss": 1.0492,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998211827732364e-05,
+      "loss": 1.0194,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999818086522451e-05,
+      "loss": 1.0182,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998149636976425e-05,
+      "loss": 1.0681,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998118142988947e-05,
+      "loss": 1.1173,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99980863832629e-05,
+      "loss": 0.9849,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999805435779914e-05,
+      "loss": 1.0231,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998022066598512e-05,
+      "loss": 0.994,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999798950966188e-05,
+      "loss": 1.0871,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99979566869901e-05,
+      "loss": 0.9793,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997923598584052e-05,
+      "loss": 1.0533,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997890244444615e-05,
+      "loss": 1.0353,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997856624572676e-05,
+      "loss": 1.0389,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999782273896912e-05,
+      "loss": 1.1565,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999778858763486e-05,
+      "loss": 1.0334,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997754170570797e-05,
+      "loss": 1.1067,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997719487777846e-05,
+      "loss": 1.0498,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999768453925693e-05,
+      "loss": 1.0628,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997649325008975e-05,
+      "loss": 1.088,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999761384503492e-05,
+      "loss": 1.0445,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997578099335714e-05,
+      "loss": 1.0526,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999754208791229e-05,
+      "loss": 1.0344,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997505810765622e-05,
+      "loss": 1.1225,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997469267896668e-05,
+      "loss": 1.0803,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997432459306398e-05,
+      "loss": 1.0477,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999739538499579e-05,
+      "loss": 1.1134,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997358044965833e-05,
+      "loss": 1.0307,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997320439217514e-05,
+      "loss": 1.008,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997282567751836e-05,
+      "loss": 0.9928,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997244430569808e-05,
+      "loss": 1.0356,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999720602767244e-05,
+      "loss": 1.0742,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997167359060755e-05,
+      "loss": 0.9753,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997128424735776e-05,
+      "loss": 1.0927,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997089224698544e-05,
+      "loss": 1.2066,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99970497589501e-05,
+      "loss": 1.1156,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997010027491487e-05,
+      "loss": 1.0227,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996970030323766e-05,
+      "loss": 1.0322,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996929767448e-05,
+      "loss": 0.9812,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999688923886526e-05,
+      "loss": 1.1674,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999684844457662e-05,
+      "loss": 1.0448,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996807384583167e-05,
+      "loss": 1.0758,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996766058885986e-05,
+      "loss": 1.0196,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996724467486184e-05,
+      "loss": 1.0517,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996682610384866e-05,
+      "loss": 1.0712,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999664048758314e-05,
+      "loss": 1.0747,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996598099082125e-05,
+      "loss": 1.0778,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996555444882952e-05,
+      "loss": 1.0392,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996512524986752e-05,
+      "loss": 0.4052,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996469339394663e-05,
+      "loss": 1.0525,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999642588810784e-05,
+      "loss": 1.0375,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999638217112743e-05,
+      "loss": 1.0508,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996338188454603e-05,
+      "loss": 1.0592,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996293940090522e-05,
+      "loss": 1.0791,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996249426036364e-05,
+      "loss": 0.9336,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996204646293316e-05,
+      "loss": 1.0857,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996159600862565e-05,
+      "loss": 1.0868,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996114289745304e-05,
+      "loss": 1.0253,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996068712942745e-05,
+      "loss": 1.0449,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996022870456095e-05,
+      "loss": 1.1565,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995976762286576e-05,
+      "loss": 0.9508,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999593038843541e-05,
+      "loss": 1.0776,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999588374890383e-05,
+      "loss": 1.03,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995836843693077e-05,
+      "loss": 1.0156,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995789672804395e-05,
+      "loss": 1.0634,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995742236239045e-05,
+      "loss": 1.0978,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995694533998277e-05,
+      "loss": 1.0359,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995646566083365e-05,
+      "loss": 0.9231,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999559833249559e-05,
+      "loss": 1.1069,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999554983323622e-05,
+      "loss": 1.0352,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995501068306554e-05,
+      "loss": 0.9805,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995452037707885e-05,
+      "loss": 1.0584,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995402741441516e-05,
+      "loss": 0.9946,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999535317950876e-05,
+      "loss": 1.0699,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995303351910934e-05,
+      "loss": 1.0859,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995253258649357e-05,
+      "loss": 1.066,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995202899725363e-05,
+      "loss": 1.0696,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995152275140296e-05,
+      "loss": 1.1562,
+      "step": 787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995101384895495e-05,
+      "loss": 1.0042,
+      "step": 788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995050228992316e-05,
+      "loss": 1.0767,
+      "step": 789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994998807432113e-05,
+      "loss": 1.0599,
+      "step": 790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999494712021626e-05,
+      "loss": 1.1098,
+      "step": 791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994895167346128e-05,
+      "loss": 1.0969,
+      "step": 792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994842948823097e-05,
+      "loss": 1.1164,
+      "step": 793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994790464648556e-05,
+      "loss": 0.9823,
+      "step": 794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994737714823903e-05,
+      "loss": 0.9286,
+      "step": 795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999468469935053e-05,
+      "loss": 1.0664,
+      "step": 796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994631418229858e-05,
+      "loss": 0.3896,
+      "step": 797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994577871463296e-05,
+      "loss": 1.0553,
+      "step": 798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994524059052266e-05,
+      "loss": 1.0748,
+      "step": 799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994469980998206e-05,
+      "loss": 1.144,
+      "step": 800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994415637302545e-05,
+      "loss": 1.0423,
+      "step": 801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994361027966737e-05,
+      "loss": 1.0504,
+      "step": 802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999430615299222e-05,
+      "loss": 1.1214,
+      "step": 803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994251012380466e-05,
+      "loss": 1.009,
+      "step": 804
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994195606132934e-05,
+      "loss": 1.017,
+      "step": 805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994139934251095e-05,
+      "loss": 1.0289,
+      "step": 806
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999408399673643e-05,
+      "loss": 0.9597,
+      "step": 807
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999402779359043e-05,
+      "loss": 0.9902,
+      "step": 808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993971324814583e-05,
+      "loss": 1.0514,
+      "step": 809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993914590410394e-05,
+      "loss": 1.033,
+      "step": 810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999385759037937e-05,
+      "loss": 1.0778,
+      "step": 811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999380032472302e-05,
+      "loss": 1.0803,
+      "step": 812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993742793442876e-05,
+      "loss": 1.0175,
+      "step": 813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999368499654046e-05,
+      "loss": 0.9983,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993626934017314e-05,
+      "loss": 1.0681,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993568605874976e-05,
+      "loss": 0.9885,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993510012115e-05,
+      "loss": 1.1101,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999345115273894e-05,
+      "loss": 1.0367,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999339202774836e-05,
+      "loss": 1.0657,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999333263714484e-05,
+      "loss": 1.1041,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993272980929944e-05,
+      "loss": 1.028,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999321305910527e-05,
+      "loss": 1.0515,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993152871672403e-05,
+      "loss": 1.0436,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993092418632947e-05,
+      "loss": 1.0835,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999303169998851e-05,
+      "loss": 1.0128,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992970715740704e-05,
+      "loss": 1.1188,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992909465891148e-05,
+      "loss": 1.0904,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999284795044147e-05,
+      "loss": 1.0949,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999278616939331e-05,
+      "loss": 0.9916,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992724122748303e-05,
+      "loss": 1.1302,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99926618105081e-05,
+      "loss": 1.1244,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992599232674363e-05,
+      "loss": 1.0523,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999253638924875e-05,
+      "loss": 0.9532,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999247328023293e-05,
+      "loss": 1.0228,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992409905628586e-05,
+      "loss": 0.9201,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992346265437396e-05,
+      "loss": 1.0338,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999228235966106e-05,
+      "loss": 1.0761,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992218188301266e-05,
+      "loss": 1.1043,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992153751359725e-05,
+      "loss": 1.075,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992089048838152e-05,
+      "loss": 1.0421,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992024080738262e-05,
+      "loss": 1.0885,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991958847061786e-05,
+      "loss": 0.9933,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991893347810453e-05,
+      "loss": 1.0035,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999182758298601e-05,
+      "loss": 0.4552,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991761552590197e-05,
+      "loss": 0.9977,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999169525662478e-05,
+      "loss": 1.1236,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999162869509151e-05,
+      "loss": 1.0455,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999156186799216e-05,
+      "loss": 1.0124,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991494775328512e-05,
+      "loss": 1.0354,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991427417102337e-05,
+      "loss": 1.0621,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999135979331544e-05,
+      "loss": 1.0026,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991291903969606e-05,
+      "loss": 1.0157,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991223749066647e-05,
+      "loss": 1.0251,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991155328608373e-05,
+      "loss": 1.1233,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991086642596596e-05,
+      "loss": 1.0696,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999101769103315e-05,
+      "loss": 1.1226,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990948473919864e-05,
+      "loss": 1.0542,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999087899125858e-05,
+      "loss": 1.0016,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999080924305114e-05,
+      "loss": 0.958,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990739229299406e-05,
+      "loss": 1.0156,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990668950005232e-05,
+      "loss": 1.0117,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990598405170485e-05,
+      "loss": 0.9629,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990527594797044e-05,
+      "loss": 1.074,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999045651888679e-05,
+      "loss": 1.0662,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999038517744161e-05,
+      "loss": 0.9806,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990313570463408e-05,
+      "loss": 1.2466,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990241697954077e-05,
+      "loss": 1.1504,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999016955991553e-05,
+      "loss": 1.0576,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990097156349687e-05,
+      "loss": 1.0867,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990024487258473e-05,
+      "loss": 1.0411,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989951552643817e-05,
+      "loss": 1.1196,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998987835250766e-05,
+      "loss": 1.1269,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989804886851943e-05,
+      "loss": 1.0607,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989731155678626e-05,
+      "loss": 1.1447,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998965715898966e-05,
+      "loss": 1.0503,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989582896787015e-05,
+      "loss": 1.0734,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998950836907267e-05,
+      "loss": 1.0199,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99894335758486e-05,
+      "loss": 1.0628,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989358517116796e-05,
+      "loss": 1.1017,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998928319287925e-05,
+      "loss": 1.0641,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989207603137966e-05,
+      "loss": 0.9889,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998913174789495e-05,
+      "loss": 0.994,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989055627152222e-05,
+      "loss": 1.0219,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988979240911807e-05,
+      "loss": 1.0502,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988902589175728e-05,
+      "loss": 1.0365,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988825671946032e-05,
+      "loss": 1.0419,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988748489224755e-05,
+      "loss": 1.0913,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998867104101395e-05,
+      "loss": 0.9906,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988593327315677e-05,
+      "loss": 1.0537,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988515348132003e-05,
+      "loss": 1.0445,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988437103464998e-05,
+      "loss": 1.0781,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988358593316745e-05,
+      "loss": 0.9856,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988279817689326e-05,
+      "loss": 1.107,
+      "step": 893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988200776584837e-05,
+      "loss": 0.9454,
+      "step": 894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988121470005382e-05,
+      "loss": 1.0565,
+      "step": 895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988041897953063e-05,
+      "loss": 1.0098,
+      "step": 896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998796206043e-05,
+      "loss": 1.0398,
+      "step": 897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998788195743831e-05,
+      "loss": 0.9791,
+      "step": 898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987801588980126e-05,
+      "loss": 0.3914,
+      "step": 899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987720955057583e-05,
+      "loss": 1.0567,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987640055672824e-05,
+      "loss": 1.0677,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987558890828e-05,
+      "loss": 1.0963,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987477460525263e-05,
+      "loss": 1.0975,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987395764766787e-05,
+      "loss": 1.0341,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998731380355474e-05,
+      "loss": 1.0588,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987231576891296e-05,
+      "loss": 1.0141,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998714908477864e-05,
+      "loss": 1.0519,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987066327218972e-05,
+      "loss": 1.0551,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986983304214485e-05,
+      "loss": 1.0178,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998690001576739e-05,
+      "loss": 1.0777,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986816461879897e-05,
+      "loss": 1.0647,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998673264255423e-05,
+      "loss": 1.0778,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986648557792613e-05,
+      "loss": 1.0427,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986564207597287e-05,
+      "loss": 1.0855,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986479591970488e-05,
+      "loss": 1.101,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986394710914467e-05,
+      "loss": 0.9713,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998630956443148e-05,
+      "loss": 1.0607,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998622415252379e-05,
+      "loss": 1.0246,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986138475193665e-05,
+      "loss": 1.1102,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998605253244339e-05,
+      "loss": 1.0785,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998596632427524e-05,
+      "loss": 1.0196,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985879850691513e-05,
+      "loss": 1.0252,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985793111694503e-05,
+      "loss": 1.0965,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985706107286515e-05,
+      "loss": 1.0956,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985618837469864e-05,
+      "loss": 1.0984,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985531302246873e-05,
+      "loss": 1.059,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998544350161986e-05,
+      "loss": 1.0744,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985355435591166e-05,
+      "loss": 1.0442,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985267104163128e-05,
+      "loss": 0.9939,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985178507338095e-05,
+      "loss": 1.0398,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998508964511842e-05,
+      "loss": 1.069,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985000517506464e-05,
+      "loss": 1.0533,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984911124504603e-05,
+      "loss": 1.0907,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998482146611521e-05,
+      "loss": 1.0528,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998473154234066e-05,
+      "loss": 1.0583,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998464135318335e-05,
+      "loss": 1.0816,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984550898645682e-05,
+      "loss": 1.1612,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998446017873005e-05,
+      "loss": 1.0508,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998436919343887e-05,
+      "loss": 1.015,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998427794277456e-05,
+      "loss": 1.1049,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998418642673955e-05,
+      "loss": 1.0432,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984094645336263e-05,
+      "loss": 1.0327,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998400259856714e-05,
+      "loss": 0.9683,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998391028643464e-05,
+      "loss": 1.0216,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99838177089412e-05,
+      "loss": 1.0121,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998372486608929e-05,
+      "loss": 1.1354,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998363175788138e-05,
+      "loss": 1.0499,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983538384319932e-05,
+      "loss": 1.0492,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998344474540744e-05,
+      "loss": 1.0337,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998335084114639e-05,
+      "loss": 1.0054,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983256671539276e-05,
+      "loss": 1.0606,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99831622365886e-05,
+      "loss": 0.9802,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983067536296876e-05,
+      "loss": 1.0037,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998297257066662e-05,
+      "loss": 1.1339,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998287733970035e-05,
+      "loss": 1.0718,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982781843400605e-05,
+      "loss": 0.9587,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998268608176992e-05,
+      "loss": 0.9947,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982590054810838e-05,
+      "loss": 1.0691,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982493762525915e-05,
+      "loss": 0.9592,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998239720491771e-05,
+      "loss": 1.0265,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998230038198879e-05,
+      "loss": 1.1022,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998220329374172e-05,
+      "loss": 1.0589,
+      "step": 962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982105940179094e-05,
+      "loss": 1.0403,
+      "step": 963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998200832130349e-05,
+      "loss": 1.0699,
+      "step": 964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981910437117502e-05,
+      "loss": 1.1972,
+      "step": 965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981812287623737e-05,
+      "loss": 1.0447,
+      "step": 966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981713872824807e-05,
+      "loss": 1.0018,
+      "step": 967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981615192723316e-05,
+      "loss": 1.1162,
+      "step": 968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981516247321896e-05,
+      "loss": 1.1505,
+      "step": 969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981417036623175e-05,
+      "loss": 1.0436,
+      "step": 970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998131756062979e-05,
+      "loss": 1.121,
+      "step": 971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998121781934438e-05,
+      "loss": 1.0141,
+      "step": 972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981117812769605e-05,
+      "loss": 1.0937,
+      "step": 973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981017540908112e-05,
+      "loss": 1.0174,
+      "step": 974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980917003762578e-05,
+      "loss": 1.0137,
+      "step": 975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980816201335663e-05,
+      "loss": 1.0379,
+      "step": 976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980715133630057e-05,
+      "loss": 1.0542,
+      "step": 977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998061380064844e-05,
+      "loss": 1.0894,
+      "step": 978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998051220239351e-05,
+      "loss": 1.1963,
+      "step": 979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998041033886796e-05,
+      "loss": 1.0135,
+      "step": 980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980308210074506e-05,
+      "loss": 1.1145,
+      "step": 981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980205816015857e-05,
+      "loss": 1.0396,
+      "step": 982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980103156694736e-05,
+      "loss": 1.0201,
+      "step": 983
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998000023211387e-05,
+      "loss": 0.961,
+      "step": 984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979897042275997e-05,
+      "loss": 1.0902,
+      "step": 985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997979358718386e-05,
+      "loss": 1.0809,
+      "step": 986
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979689866840207e-05,
+      "loss": 1.065,
+      "step": 987
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979585881247793e-05,
+      "loss": 0.4291,
+      "step": 988
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979481630409386e-05,
+      "loss": 1.0994,
+      "step": 989
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979377114327755e-05,
+      "loss": 1.0354,
+      "step": 990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979272333005677e-05,
+      "loss": 0.9651,
+      "step": 991
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997916728644594e-05,
+      "loss": 1.0485,
+      "step": 992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979061974651332e-05,
+      "loss": 0.9916,
+      "step": 993
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978956397624654e-05,
+      "loss": 1.1009,
+      "step": 994
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997885055536871e-05,
+      "loss": 1.0938,
+      "step": 995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997874444788632e-05,
+      "loss": 1.1423,
+      "step": 996
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978638075180296e-05,
+      "loss": 1.0455,
+      "step": 997
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997853143725347e-05,
+      "loss": 1.0203,
+      "step": 998
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978424534108674e-05,
+      "loss": 1.1002,
+      "step": 999
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997831736574875e-05,
+      "loss": 1.0728,
+      "step": 1000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997820993217655e-05,
+      "loss": 1.0625,
+      "step": 1001
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997810223339492e-05,
+      "loss": 1.0645,
+      "step": 1002
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977994269406734e-05,
+      "loss": 0.9614,
+      "step": 1003
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977886040214853e-05,
+      "loss": 0.9468,
+      "step": 1004
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997777754582216e-05,
+      "loss": 0.9482,
+      "step": 1005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977668786231536e-05,
+      "loss": 1.0298,
+      "step": 1006
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977559761445866e-05,
+      "loss": 1.0646,
+      "step": 1007
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997745047146806e-05,
+      "loss": 1.1289,
+      "step": 1008
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997734091630101e-05,
+      "loss": 1.0645,
+      "step": 1009
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977231095947634e-05,
+      "loss": 1.0214,
+      "step": 1010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977121010410854e-05,
+      "loss": 1.0726,
+      "step": 1011
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997701065969359e-05,
+      "loss": 1.0356,
+      "step": 1012
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976900043798778e-05,
+      "loss": 1.0609,
+      "step": 1013
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997678916272936e-05,
+      "loss": 0.9831,
+      "step": 1014
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976678016488277e-05,
+      "loss": 1.0282,
+      "step": 1015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976566605078485e-05,
+      "loss": 1.0404,
+      "step": 1016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976454928502953e-05,
+      "loss": 1.1242,
+      "step": 1017
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976342986764637e-05,
+      "loss": 0.9955,
+      "step": 1018
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976230779866527e-05,
+      "loss": 1.0583,
+      "step": 1019
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997611830781159e-05,
+      "loss": 1.1092,
+      "step": 1020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976005570602824e-05,
+      "loss": 1.1039,
+      "step": 1021
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975892568243224e-05,
+      "loss": 1.0097,
+      "step": 1022
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997577930073579e-05,
+      "loss": 1.0088,
+      "step": 1023
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997566576808354e-05,
+      "loss": 0.4018,
+      "step": 1024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975551970289487e-05,
+      "loss": 1.0948,
+      "step": 1025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975437907356655e-05,
+      "loss": 1.035,
+      "step": 1026
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975323579288073e-05,
+      "loss": 1.0523,
+      "step": 1027
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975208986086785e-05,
+      "loss": 1.0848,
+      "step": 1028
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975094127755836e-05,
+      "loss": 1.012,
+      "step": 1029
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997497900429828e-05,
+      "loss": 0.3978,
+      "step": 1030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997486361571717e-05,
+      "loss": 1.093,
+      "step": 1031
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997474796201558e-05,
+      "loss": 0.9845,
+      "step": 1032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997463204319658e-05,
+      "loss": 1.0269,
+      "step": 1033
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974515859263254e-05,
+      "loss": 1.1383,
+      "step": 1034
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997439941021869e-05,
+      "loss": 1.1029,
+      "step": 1035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974282696065977e-05,
+      "loss": 1.1221,
+      "step": 1036
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997416571680823e-05,
+      "loss": 0.9308,
+      "step": 1037
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997404847244854e-05,
+      "loss": 1.067,
+      "step": 1038
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973930962990037e-05,
+      "loss": 1.0399,
+      "step": 1039
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997381318843584e-05,
+      "loss": 1.0075,
+      "step": 1040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997369514878908e-05,
+      "loss": 1.0186,
+      "step": 1041
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973576844052895e-05,
+      "loss": 1.0719,
+      "step": 1042
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973458274230426e-05,
+      "loss": 1.0356,
+      "step": 1043
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997333943932483e-05,
+      "loss": 0.9656,
+      "step": 1044
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997322033933926e-05,
+      "loss": 1.0098,
+      "step": 1045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973100974276883e-05,
+      "loss": 1.0058,
+      "step": 1046
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972981344140875e-05,
+      "loss": 1.07,
+      "step": 1047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997286144893441e-05,
+      "loss": 0.9126,
+      "step": 1048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997274128866068e-05,
+      "loss": 1.0324,
+      "step": 1049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997262086332288e-05,
+      "loss": 0.9593,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99725001729242e-05,
+      "loss": 1.0834,
+      "step": 1051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997237921746786e-05,
+      "loss": 1.0867,
+      "step": 1052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997225799695707e-05,
+      "loss": 1.0568,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972136511395047e-05,
+      "loss": 1.0092,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997201476078503e-05,
+      "loss": 0.3966,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971892745130244e-05,
+      "loss": 1.065,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971770464433944e-05,
+      "loss": 1.0733,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971647918699374e-05,
+      "loss": 1.0002,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971525107929787e-05,
+      "loss": 1.1104,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971402032128453e-05,
+      "loss": 1.107,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997127869129864e-05,
+      "loss": 1.0582,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971155085443632e-05,
+      "loss": 0.9711,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971031214566707e-05,
+      "loss": 1.0256,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997090707867116e-05,
+      "loss": 1.0495,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970782677760293e-05,
+      "loss": 1.0138,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970658011837404e-05,
+      "loss": 1.1201,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970533080905817e-05,
+      "loss": 1.1462,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997040788496885e-05,
+      "loss": 1.0285,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970282424029825e-05,
+      "loss": 0.9471,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970156698092084e-05,
+      "loss": 1.0189,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970030707158963e-05,
+      "loss": 0.9918,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969904451233812e-05,
+      "loss": 1.0373,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996977793031999e-05,
+      "loss": 0.9987,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969651144420854e-05,
+      "loss": 1.1357,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969524093539778e-05,
+      "loss": 0.995,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996939677768014e-05,
+      "loss": 1.0479,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996926919684532e-05,
+      "loss": 1.0595,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996914135103871e-05,
+      "loss": 0.9928,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996901324026371e-05,
+      "loss": 1.0964,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996888486452372e-05,
+      "loss": 1.0884,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968756223822162e-05,
+      "loss": 1.0843,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968627318162446e-05,
+      "loss": 1.055,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968498147548e-05,
+      "loss": 1.0083,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996836871198226e-05,
+      "loss": 1.0565,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968239011468663e-05,
+      "loss": 1.0807,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996810904601066e-05,
+      "loss": 1.0768,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967978815611703e-05,
+      "loss": 0.9492,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967848320275253e-05,
+      "loss": 1.034,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996771756000478e-05,
+      "loss": 1.0005,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967586534803755e-05,
+      "loss": 0.3746,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967455244675667e-05,
+      "loss": 1.0659,
+      "step": 1091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967323689624004e-05,
+      "loss": 1.1133,
+      "step": 1092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967191869652257e-05,
+      "loss": 1.0506,
+      "step": 1093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967059784763935e-05,
+      "loss": 0.9555,
+      "step": 1094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996692743496255e-05,
+      "loss": 1.136,
+      "step": 1095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966794820251616e-05,
+      "loss": 0.9406,
+      "step": 1096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966661940634658e-05,
+      "loss": 1.0231,
+      "step": 1097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966528796115208e-05,
+      "loss": 1.0257,
+      "step": 1098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966395386696807e-05,
+      "loss": 1.0427,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966261712383e-05,
+      "loss": 1.0112,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966127773177335e-05,
+      "loss": 0.4045,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996599356908338e-05,
+      "loss": 1.0333,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965859100104698e-05,
+      "loss": 1.0614,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996572436624486e-05,
+      "loss": 1.0086,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965589367507454e-05,
+      "loss": 1.0738,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965454103896063e-05,
+      "loss": 1.0387,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996531857541428e-05,
+      "loss": 1.0593,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965182782065717e-05,
+      "loss": 1.088,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965046723853976e-05,
+      "loss": 0.9698,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996491040078267e-05,
+      "loss": 0.981,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996477381285543e-05,
+      "loss": 0.3728,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996463696007588e-05,
+      "loss": 1.0486,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964499842447665e-05,
+      "loss": 1.0468,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964362459974423e-05,
+      "loss": 1.1164,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964224812659805e-05,
+      "loss": 0.9823,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964086900507475e-05,
+      "loss": 1.0209,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996394872352109e-05,
+      "loss": 0.9981,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996381028170433e-05,
+      "loss": 1.0654,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963671575060877e-05,
+      "loss": 0.9152,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963532603594408e-05,
+      "loss": 1.0366,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963393367308624e-05,
+      "loss": 1.0389,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963253866207222e-05,
+      "loss": 1.0745,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963114100293916e-05,
+      "loss": 1.0241,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996297406957241e-05,
+      "loss": 0.9306,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962833774046437e-05,
+      "loss": 1.0417,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962693213719716e-05,
+      "loss": 1.0197,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996255238859599e-05,
+      "loss": 1.0294,
+      "step": 1127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962411298679003e-05,
+      "loss": 0.9952,
+      "step": 1128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99622699439725e-05,
+      "loss": 1.1031,
+      "step": 1129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996212832448024e-05,
+      "loss": 1.0641,
+      "step": 1130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961986440205985e-05,
+      "loss": 1.0433,
+      "step": 1131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961844291153508e-05,
+      "loss": 1.0557,
+      "step": 1132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961701877326588e-05,
+      "loss": 1.0723,
+      "step": 1133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996155919872901e-05,
+      "loss": 1.026,
+      "step": 1134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961416255364566e-05,
+      "loss": 1.0746,
+      "step": 1135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961273047237057e-05,
+      "loss": 1.0237,
+      "step": 1136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961129574350284e-05,
+      "loss": 0.9985,
+      "step": 1137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960985836708066e-05,
+      "loss": 0.9504,
+      "step": 1138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960841834314215e-05,
+      "loss": 1.0352,
+      "step": 1139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960697567172572e-05,
+      "loss": 1.0404,
+      "step": 1140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960553035286957e-05,
+      "loss": 1.0902,
+      "step": 1141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960408238661224e-05,
+      "loss": 0.9883,
+      "step": 1142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960263177299213e-05,
+      "loss": 1.0434,
+      "step": 1143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960117851204783e-05,
+      "loss": 1.0502,
+      "step": 1144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99599722603818e-05,
+      "loss": 1.0276,
+      "step": 1145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959826404834125e-05,
+      "loss": 1.14,
+      "step": 1146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959680284565638e-05,
+      "loss": 1.083,
+      "step": 1147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995953389958023e-05,
+      "loss": 1.0218,
+      "step": 1148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959387249881782e-05,
+      "loss": 1.0146,
+      "step": 1149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959240335474195e-05,
+      "loss": 0.9536,
+      "step": 1150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959093156361375e-05,
+      "loss": 1.1129,
+      "step": 1151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958945712547235e-05,
+      "loss": 1.0397,
+      "step": 1152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958798004035692e-05,
+      "loss": 0.989,
+      "step": 1153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995865003083067e-05,
+      "loss": 0.9854,
+      "step": 1154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958501792936105e-05,
+      "loss": 1.0204,
+      "step": 1155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958353290355935e-05,
+      "loss": 0.9997,
+      "step": 1156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995820452309411e-05,
+      "loss": 1.0328,
+      "step": 1157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995805549115458e-05,
+      "loss": 1.02,
+      "step": 1158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957906194541314e-05,
+      "loss": 0.3867,
+      "step": 1159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957756633258264e-05,
+      "loss": 1.0761,
+      "step": 1160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957606807309423e-05,
+      "loss": 1.0715,
+      "step": 1161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957456716698767e-05,
+      "loss": 0.9896,
+      "step": 1162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995730636143028e-05,
+      "loss": 0.966,
+      "step": 1163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957155741507963e-05,
+      "loss": 1.0352,
+      "step": 1164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995700485693582e-05,
+      "loss": 1.0391,
+      "step": 1165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956853707717857e-05,
+      "loss": 1.0206,
+      "step": 1166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956702293858098e-05,
+      "loss": 1.0887,
+      "step": 1167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995655061536056e-05,
+      "loss": 1.0346,
+      "step": 1168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995639867222928e-05,
+      "loss": 1.0432,
+      "step": 1169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956246464468294e-05,
+      "loss": 1.0432,
+      "step": 1170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995609399208165e-05,
+      "loss": 1.0898,
+      "step": 1171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955941255073396e-05,
+      "loss": 0.981,
+      "step": 1172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955788253447596e-05,
+      "loss": 1.0195,
+      "step": 1173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955634987208313e-05,
+      "loss": 1.0597,
+      "step": 1174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955481456359624e-05,
+      "loss": 1.0426,
+      "step": 1175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955327660905608e-05,
+      "loss": 1.1357,
+      "step": 1176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955173600850354e-05,
+      "loss": 0.4414,
+      "step": 1177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955019276197952e-05,
+      "loss": 0.9794,
+      "step": 1178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995486468695251e-05,
+      "loss": 1.043,
+      "step": 1179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954709833118136e-05,
+      "loss": 1.0158,
+      "step": 1180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995455471469894e-05,
+      "loss": 0.9857,
+      "step": 1181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995439933169905e-05,
+      "loss": 1.011,
+      "step": 1182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99542436841226e-05,
+      "loss": 1.0703,
+      "step": 1183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954087771973713e-05,
+      "loss": 1.1261,
+      "step": 1184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953931595256548e-05,
+      "loss": 0.9884,
+      "step": 1185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995377515397525e-05,
+      "loss": 1.1579,
+      "step": 1186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953618448133973e-05,
+      "loss": 1.0449,
+      "step": 1187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953461477736887e-05,
+      "loss": 0.9473,
+      "step": 1188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953304242788162e-05,
+      "loss": 1.0147,
+      "step": 1189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953146743291977e-05,
+      "loss": 1.0723,
+      "step": 1190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995298897925252e-05,
+      "loss": 0.9864,
+      "step": 1191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952830950673986e-05,
+      "loss": 0.3168,
+      "step": 1192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995267265756057e-05,
+      "loss": 1.0664,
+      "step": 1193
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952514099916485e-05,
+      "loss": 1.0629,
+      "step": 1194
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995235527774594e-05,
+      "loss": 1.0551,
+      "step": 1195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952196191053158e-05,
+      "loss": 1.0733,
+      "step": 1196
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952036839842366e-05,
+      "loss": 1.056,
+      "step": 1197
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951877224117804e-05,
+      "loss": 0.9817,
+      "step": 1198
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951717343883712e-05,
+      "loss": 0.9899,
+      "step": 1199
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951557199144337e-05,
+      "loss": 1.0544,
+      "step": 1200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995139678990394e-05,
+      "loss": 1.0554,
+      "step": 1201
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951236116166776e-05,
+      "loss": 1.0138,
+      "step": 1202
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951075177937124e-05,
+      "loss": 1.0201,
+      "step": 1203
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950913975219262e-05,
+      "loss": 1.0464,
+      "step": 1204
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995075250801747e-05,
+      "loss": 1.0708,
+      "step": 1205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995059077633604e-05,
+      "loss": 1.1008,
+      "step": 1206
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950428780179274e-05,
+      "loss": 1.0459,
+      "step": 1207
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950266519551475e-05,
+      "loss": 0.9197,
+      "step": 1208
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950103994456954e-05,
+      "loss": 0.9444,
+      "step": 1209
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949941204900037e-05,
+      "loss": 1.1198,
+      "step": 1210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994977815088504e-05,
+      "loss": 1.1016,
+      "step": 1211
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949614832416307e-05,
+      "loss": 1.1516,
+      "step": 1212
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949451249498178e-05,
+      "loss": 0.9927,
+      "step": 1213
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949287402134997e-05,
+      "loss": 1.0282,
+      "step": 1214
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994912329033112e-05,
+      "loss": 1.0395,
+      "step": 1215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994895891409091e-05,
+      "loss": 1.0758,
+      "step": 1216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994879427341873e-05,
+      "loss": 1.0368,
+      "step": 1217
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948629368318965e-05,
+      "loss": 1.0704,
+      "step": 1218
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948464198795992e-05,
+      "loss": 1.0268,
+      "step": 1219
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99482987648542e-05,
+      "loss": 1.042,
+      "step": 1220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948133066497995e-05,
+      "loss": 1.1341,
+      "step": 1221
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947967103731774e-05,
+      "loss": 1.0747,
+      "step": 1222
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947800876559946e-05,
+      "loss": 0.9994,
+      "step": 1223
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947634384986935e-05,
+      "loss": 1.0655,
+      "step": 1224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947467629017158e-05,
+      "loss": 1.1576,
+      "step": 1225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947300608655058e-05,
+      "loss": 0.9052,
+      "step": 1226
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947133323905068e-05,
+      "loss": 1.0056,
+      "step": 1227
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946965774771633e-05,
+      "loss": 1.0052,
+      "step": 1228
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994679796125921e-05,
+      "loss": 1.0541,
+      "step": 1229
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946629883372256e-05,
+      "loss": 1.0599,
+      "step": 1230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994646154111524e-05,
+      "loss": 0.9676,
+      "step": 1231
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946292934492637e-05,
+      "loss": 1.0041,
+      "step": 1232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946124063508925e-05,
+      "loss": 0.9844,
+      "step": 1233
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99459549281686e-05,
+      "loss": 0.9765,
+      "step": 1234
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945785528476148e-05,
+      "loss": 0.9804,
+      "step": 1235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945615864436078e-05,
+      "loss": 0.9761,
+      "step": 1236
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99454459360529e-05,
+      "loss": 0.9852,
+      "step": 1237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945275743331122e-05,
+      "loss": 1.0086,
+      "step": 1238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945105286275276e-05,
+      "loss": 1.075,
+      "step": 1239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994493456488989e-05,
+      "loss": 1.0395,
+      "step": 1240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99447635791795e-05,
+      "loss": 1.2157,
+      "step": 1241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944592329148652e-05,
+      "loss": 0.9698,
+      "step": 1242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99444208148019e-05,
+      "loss": 1.0596,
+      "step": 1243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99442490361438e-05,
+      "loss": 0.9825,
+      "step": 1244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994407699317892e-05,
+      "loss": 0.9373,
+      "step": 1245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943904685911826e-05,
+      "loss": 0.3805,
+      "step": 1246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943732114347104e-05,
+      "loss": 1.1104,
+      "step": 1247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994355927848934e-05,
+      "loss": 0.9854,
+      "step": 1248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943386178343127e-05,
+      "loss": 1.0634,
+      "step": 1249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943212813913066e-05,
+      "loss": 1.0634,
+      "step": 1250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943039185203763e-05,
+      "loss": 1.0584,
+      "step": 1251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942865292219837e-05,
+      "loss": 1.0998,
+      "step": 1252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942691134965905e-05,
+      "loss": 1.1387,
+      "step": 1253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99425167134466e-05,
+      "loss": 1.1308,
+      "step": 1254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942342027666558e-05,
+      "loss": 0.8932,
+      "step": 1255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942167077630417e-05,
+      "loss": 1.0031,
+      "step": 1256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994199186334283e-05,
+      "loss": 0.9934,
+      "step": 1257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941816384808454e-05,
+      "loss": 1.0559,
+      "step": 1258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941640642031955e-05,
+      "loss": 0.9824,
+      "step": 1259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941464635018002e-05,
+      "loss": 1.0478,
+      "step": 1260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994128836377127e-05,
+      "loss": 1.0006,
+      "step": 1261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994111182829645e-05,
+      "loss": 1.0404,
+      "step": 1262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940935028598234e-05,
+      "loss": 1.0566,
+      "step": 1263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940757964681317e-05,
+      "loss": 0.9766,
+      "step": 1264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940580636550405e-05,
+      "loss": 1.0335,
+      "step": 1265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940403044210212e-05,
+      "loss": 1.1233,
+      "step": 1266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994022518766546e-05,
+      "loss": 1.1219,
+      "step": 1267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994004706692088e-05,
+      "loss": 1.0394,
+      "step": 1268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939868681981192e-05,
+      "loss": 1.0139,
+      "step": 1269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939690032851154e-05,
+      "loss": 0.9725,
+      "step": 1270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939511119535507e-05,
+      "loss": 0.9832,
+      "step": 1271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939331942039004e-05,
+      "loss": 0.9934,
+      "step": 1272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939152500366413e-05,
+      "loss": 1.0752,
+      "step": 1273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938972794522495e-05,
+      "loss": 1.0146,
+      "step": 1274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938792824512036e-05,
+      "loss": 1.0935,
+      "step": 1275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938612590339815e-05,
+      "loss": 0.9517,
+      "step": 1276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993843209201062e-05,
+      "loss": 1.0028,
+      "step": 1277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938251329529252e-05,
+      "loss": 1.0252,
+      "step": 1278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938070302900513e-05,
+      "loss": 1.0501,
+      "step": 1279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937889012129218e-05,
+      "loss": 1.0382,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937707457220182e-05,
+      "loss": 1.0665,
+      "step": 1281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993752563817823e-05,
+      "loss": 1.1034,
+      "step": 1282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99373435550082e-05,
+      "loss": 1.0166,
+      "step": 1283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993716120771492e-05,
+      "loss": 0.9577,
+      "step": 1284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936978596303252e-05,
+      "loss": 0.9971,
+      "step": 1285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936795720778043e-05,
+      "loss": 1.0584,
+      "step": 1286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993661258114415e-05,
+      "loss": 0.9737,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993642917740644e-05,
+      "loss": 1.0657,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936245509569794e-05,
+      "loss": 0.9176,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993606157763909e-05,
+      "loss": 1.0586,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9935877381619216e-05,
+      "loss": 0.9653,
+      "step": 1291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935692921515073e-05,
+      "loss": 1.0108,
+      "step": 1292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935508197331556e-05,
+      "loss": 1.0042,
+      "step": 1293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993532320907358e-05,
+      "loss": 1.0104,
+      "step": 1294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993513795674606e-05,
+      "loss": 1.0563,
+      "step": 1295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934952440353923e-05,
+      "loss": 1.0442,
+      "step": 1296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934766659902096e-05,
+      "loss": 1.003,
+      "step": 1297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934580615395514e-05,
+      "loss": 0.9634,
+      "step": 1298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993439430683913e-05,
+      "loss": 1.056,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993420773423789e-05,
+      "loss": 0.9877,
+      "step": 1300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934020897596752e-05,
+      "loss": 1.0872,
+      "step": 1301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933833796920688e-05,
+      "loss": 0.9525,
+      "step": 1302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933646432214663e-05,
+      "loss": 1.0253,
+      "step": 1303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933458803483666e-05,
+      "loss": 1.0174,
+      "step": 1304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933270910732677e-05,
+      "loss": 1.0428,
+      "step": 1305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993308275396669e-05,
+      "loss": 1.0523,
+      "step": 1306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993289433319071e-05,
+      "loss": 1.0248,
+      "step": 1307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932705648409742e-05,
+      "loss": 0.4037,
+      "step": 1308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932516699628804e-05,
+      "loss": 1.0406,
+      "step": 1309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932327486852914e-05,
+      "loss": 1.0495,
+      "step": 1310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932138010087104e-05,
+      "loss": 0.9523,
+      "step": 1311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993194826933641e-05,
+      "loss": 1.0731,
+      "step": 1312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931758264605872e-05,
+      "loss": 1.0159,
+      "step": 1313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931567995900547e-05,
+      "loss": 1.0195,
+      "step": 1314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931377463225484e-05,
+      "loss": 1.0945,
+      "step": 1315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931186666585752e-05,
+      "loss": 0.4233,
+      "step": 1316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993099560598642e-05,
+      "loss": 1.0342,
+      "step": 1317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930804281432568e-05,
+      "loss": 1.0577,
+      "step": 1318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993061269292928e-05,
+      "loss": 1.04,
+      "step": 1319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930420840481652e-05,
+      "loss": 1.0573,
+      "step": 1320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930228724094773e-05,
+      "loss": 1.0484,
+      "step": 1321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993003634377376e-05,
+      "loss": 1.0558,
+      "step": 1322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992984369952372e-05,
+      "loss": 1.0246,
+      "step": 1323
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929650791349777e-05,
+      "loss": 1.0429,
+      "step": 1324
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929457619257055e-05,
+      "loss": 1.0822,
+      "step": 1325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929264183250693e-05,
+      "loss": 1.0203,
+      "step": 1326
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929070483335827e-05,
+      "loss": 1.0221,
+      "step": 1327
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928876519517608e-05,
+      "loss": 1.0686,
+      "step": 1328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928682291801192e-05,
+      "loss": 1.0626,
+      "step": 1329
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992848780019174e-05,
+      "loss": 1.0535,
+      "step": 1330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928293044694423e-05,
+      "loss": 1.1463,
+      "step": 1331
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928098025314414e-05,
+      "loss": 1.0231,
+      "step": 1332
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927902742056896e-05,
+      "loss": 0.9726,
+      "step": 1333
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927707194927067e-05,
+      "loss": 1.0763,
+      "step": 1334
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927511383930116e-05,
+      "loss": 1.1435,
+      "step": 1335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927315309071253e-05,
+      "loss": 1.0512,
+      "step": 1336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927118970355686e-05,
+      "loss": 1.0702,
+      "step": 1337
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926922367788632e-05,
+      "loss": 1.0635,
+      "step": 1338
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926725501375324e-05,
+      "loss": 1.0764,
+      "step": 1339
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926528371120987e-05,
+      "loss": 1.0636,
+      "step": 1340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926330977030863e-05,
+      "loss": 1.0691,
+      "step": 1341
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926133319110196e-05,
+      "loss": 1.0654,
+      "step": 1342
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925935397364248e-05,
+      "loss": 0.9992,
+      "step": 1343
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925737211798267e-05,
+      "loss": 0.9935,
+      "step": 1344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925538762417532e-05,
+      "loss": 1.0059,
+      "step": 1345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925340049227312e-05,
+      "loss": 1.0152,
+      "step": 1346
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925141072232888e-05,
+      "loss": 1.1097,
+      "step": 1347
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992494183143955e-05,
+      "loss": 1.0986,
+      "step": 1348
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924742326852597e-05,
+      "loss": 0.9998,
+      "step": 1349
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924542558477323e-05,
+      "loss": 1.1226,
+      "step": 1350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924342526319046e-05,
+      "loss": 1.0436,
+      "step": 1351
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924142230383078e-05,
+      "loss": 1.0353,
+      "step": 1352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923941670674748e-05,
+      "loss": 1.0729,
+      "step": 1353
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992374084719938e-05,
+      "loss": 1.0632,
+      "step": 1354
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992353975996231e-05,
+      "loss": 1.0175,
+      "step": 1355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992333840896889e-05,
+      "loss": 1.0565,
+      "step": 1356
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923136794224466e-05,
+      "loss": 0.9765,
+      "step": 1357
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922934915734406e-05,
+      "loss": 0.934,
+      "step": 1358
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922732773504063e-05,
+      "loss": 0.9562,
+      "step": 1359
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922530367538817e-05,
+      "loss": 1.1043,
+      "step": 1360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992232769784405e-05,
+      "loss": 1.0971,
+      "step": 1361
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992212476442514e-05,
+      "loss": 0.9753,
+      "step": 1362
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992192156728749e-05,
+      "loss": 0.9858,
+      "step": 1363
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921718106436495e-05,
+      "loss": 1.0216,
+      "step": 1364
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921514381877564e-05,
+      "loss": 1.0419,
+      "step": 1365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992131039361611e-05,
+      "loss": 0.3429,
+      "step": 1366
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992110614165756e-05,
+      "loss": 1.0179,
+      "step": 1367
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992090162600734e-05,
+      "loss": 0.9788,
+      "step": 1368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920696846670888e-05,
+      "loss": 1.0424,
+      "step": 1369
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920491803653638e-05,
+      "loss": 1.058,
+      "step": 1370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992028649696105e-05,
+      "loss": 1.0431,
+      "step": 1371
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992008092659858e-05,
+      "loss": 1.0525,
+      "step": 1372
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919875092571682e-05,
+      "loss": 1.0589,
+      "step": 1373
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991966899488584e-05,
+      "loss": 1.0449,
+      "step": 1374
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991946263354652e-05,
+      "loss": 1.018,
+      "step": 1375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919256008559214e-05,
+      "loss": 0.9923,
+      "step": 1376
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919049119929414e-05,
+      "loss": 0.4402,
+      "step": 1377
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918841967662614e-05,
+      "loss": 1.0445,
+      "step": 1378
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918634551764325e-05,
+      "loss": 1.0371,
+      "step": 1379
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918426872240058e-05,
+      "loss": 1.0424,
+      "step": 1380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918218929095333e-05,
+      "loss": 1.0199,
+      "step": 1381
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918010722335676e-05,
+      "loss": 1.1202,
+      "step": 1382
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991780225196662e-05,
+      "loss": 1.0672,
+      "step": 1383
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917593517993707e-05,
+      "loss": 0.4235,
+      "step": 1384
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991738452042249e-05,
+      "loss": 1.05,
+      "step": 1385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917175259258514e-05,
+      "loss": 0.9923,
+      "step": 1386
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916965734507348e-05,
+      "loss": 1.0132,
+      "step": 1387
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916755946174558e-05,
+      "loss": 0.992,
+      "step": 1388
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991654589426572e-05,
+      "loss": 0.9308,
+      "step": 1389
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916335578786422e-05,
+      "loss": 1.0129,
+      "step": 1390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916124999742247e-05,
+      "loss": 1.0697,
+      "step": 1391
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915914157138796e-05,
+      "loss": 1.0032,
+      "step": 1392
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991570305098167e-05,
+      "loss": 0.9806,
+      "step": 1393
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915491681276485e-05,
+      "loss": 1.0522,
+      "step": 1394
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915280048028853e-05,
+      "loss": 0.9926,
+      "step": 1395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915068151244405e-05,
+      "loss": 1.033,
+      "step": 1396
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914855990928766e-05,
+      "loss": 1.0462,
+      "step": 1397
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991464356708758e-05,
+      "loss": 0.9547,
+      "step": 1398
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991443087972649e-05,
+      "loss": 1.1574,
+      "step": 1399
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914217928851153e-05,
+      "loss": 1.0018,
+      "step": 1400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914004714467223e-05,
+      "loss": 1.1054,
+      "step": 1401
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913791236580376e-05,
+      "loss": 0.9141,
+      "step": 1402
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913577495196277e-05,
+      "loss": 1.0827,
+      "step": 1403
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991336349032061e-05,
+      "loss": 1.1135,
+      "step": 1404
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913149221959066e-05,
+      "loss": 0.4146,
+      "step": 1405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912934690117333e-05,
+      "loss": 0.3873,
+      "step": 1406
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991271989480112e-05,
+      "loss": 0.9364,
+      "step": 1407
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912504836016136e-05,
+      "loss": 1.0139,
+      "step": 1408
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912289513768094e-05,
+      "loss": 1.0379,
+      "step": 1409
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912073928062716e-05,
+      "loss": 1.0173,
+      "step": 1410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911858078905734e-05,
+      "loss": 0.9372,
+      "step": 1411
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911641966302887e-05,
+      "loss": 1.0314,
+      "step": 1412
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911425590259912e-05,
+      "loss": 0.9849,
+      "step": 1413
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991120895078257e-05,
+      "loss": 0.3623,
+      "step": 1414
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991099204787661e-05,
+      "loss": 1.0582,
+      "step": 1415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910774881547803e-05,
+      "loss": 1.0432,
+      "step": 1416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910557451801917e-05,
+      "loss": 1.0424,
+      "step": 1417
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910339758644735e-05,
+      "loss": 0.9766,
+      "step": 1418
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991012180208204e-05,
+      "loss": 1.0455,
+      "step": 1419
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909903582119624e-05,
+      "loss": 1.0453,
+      "step": 1420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909685098763292e-05,
+      "loss": 1.0734,
+      "step": 1421
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909466352018846e-05,
+      "loss": 0.4091,
+      "step": 1422
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99092473418921e-05,
+      "loss": 1.0183,
+      "step": 1423
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909028068388883e-05,
+      "loss": 0.9635,
+      "step": 1424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908808531515014e-05,
+      "loss": 1.0624,
+      "step": 1425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908588731276332e-05,
+      "loss": 1.0414,
+      "step": 1426
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908368667678677e-05,
+      "loss": 1.0009,
+      "step": 1427
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908148340727898e-05,
+      "loss": 1.0799,
+      "step": 1428
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907927750429853e-05,
+      "loss": 1.0229,
+      "step": 1429
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907706896790406e-05,
+      "loss": 0.9925,
+      "step": 1430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907485779815426e-05,
+      "loss": 1.0915,
+      "step": 1431
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907264399510787e-05,
+      "loss": 1.1897,
+      "step": 1432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990704275588238e-05,
+      "loss": 1.0043,
+      "step": 1433
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906820848936083e-05,
+      "loss": 0.9223,
+      "step": 1434
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906598678677808e-05,
+      "loss": 1.0409,
+      "step": 1435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906376245113453e-05,
+      "loss": 0.9608,
+      "step": 1436
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906153548248933e-05,
+      "loss": 1.0551,
+      "step": 1437
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905930588090165e-05,
+      "loss": 1.0584,
+      "step": 1438
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905707364643077e-05,
+      "loss": 1.0195,
+      "step": 1439
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99054838779136e-05,
+      "loss": 1.0242,
+      "step": 1440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905260127907672e-05,
+      "loss": 1.014,
+      "step": 1441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905036114631247e-05,
+      "loss": 0.9869,
+      "step": 1442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990481183809027e-05,
+      "loss": 0.99,
+      "step": 1443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990458729829071e-05,
+      "loss": 1.0803,
+      "step": 1444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904362495238532e-05,
+      "loss": 1.0436,
+      "step": 1445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990413742893971e-05,
+      "loss": 1.0373,
+      "step": 1446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903912099400228e-05,
+      "loss": 1.1557,
+      "step": 1447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990368650662607e-05,
+      "loss": 1.0185,
+      "step": 1448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990346065062324e-05,
+      "loss": 0.9933,
+      "step": 1449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903234531397733e-05,
+      "loss": 1.0365,
+      "step": 1450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903008148955566e-05,
+      "loss": 0.9851,
+      "step": 1451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990278150330275e-05,
+      "loss": 1.0454,
+      "step": 1452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902554594445312e-05,
+      "loss": 0.9628,
+      "step": 1453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902327422389284e-05,
+      "loss": 1.0907,
+      "step": 1454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99020999871407e-05,
+      "loss": 1.0021,
+      "step": 1455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990187228870561e-05,
+      "loss": 1.0629,
+      "step": 1456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901644327090063e-05,
+      "loss": 0.99,
+      "step": 1457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990141610230012e-05,
+      "loss": 1.0393,
+      "step": 1458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901187614341842e-05,
+      "loss": 1.0254,
+      "step": 1459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900958863221312e-05,
+      "loss": 1.1084,
+      "step": 1460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900729848944596e-05,
+      "loss": 1.0699,
+      "step": 1461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900500571517792e-05,
+      "loss": 1.0724,
+      "step": 1462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900271030946993e-05,
+      "loss": 1.0379,
+      "step": 1463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990004122723829e-05,
+      "loss": 1.0138,
+      "step": 1464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899811160397803e-05,
+      "loss": 1.0669,
+      "step": 1465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989958083043164e-05,
+      "loss": 1.0563,
+      "step": 1466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899350237345928e-05,
+      "loss": 0.9918,
+      "step": 1467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989911938114679e-05,
+      "loss": 1.1013,
+      "step": 1468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898888261840365e-05,
+      "loss": 1.0145,
+      "step": 1469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898656879432797e-05,
+      "loss": 1.1228,
+      "step": 1470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898425233930233e-05,
+      "loss": 0.9937,
+      "step": 1471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898193325338832e-05,
+      "loss": 1.0658,
+      "step": 1472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897961153664758e-05,
+      "loss": 1.0139,
+      "step": 1473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897728718914183e-05,
+      "loss": 1.0893,
+      "step": 1474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897496021093278e-05,
+      "loss": 1.0301,
+      "step": 1475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897263060208236e-05,
+      "loss": 1.0484,
+      "step": 1476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897029836265243e-05,
+      "loss": 1.0533,
+      "step": 1477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896796349270503e-05,
+      "loss": 1.0056,
+      "step": 1478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989656259923022e-05,
+      "loss": 0.4135,
+      "step": 1479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896328586150606e-05,
+      "loss": 1.0119,
+      "step": 1480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989609431003788e-05,
+      "loss": 1.08,
+      "step": 1481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895859770898267e-05,
+      "loss": 0.984,
+      "step": 1482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895624968738004e-05,
+      "loss": 1.0003,
+      "step": 1483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895389903563332e-05,
+      "loss": 1.0909,
+      "step": 1484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98951545753805e-05,
+      "loss": 1.0132,
+      "step": 1485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894918984195758e-05,
+      "loss": 1.0219,
+      "step": 1486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894683130015374e-05,
+      "loss": 0.9669,
+      "step": 1487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989444701284561e-05,
+      "loss": 1.0688,
+      "step": 1488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894210632692745e-05,
+      "loss": 1.0574,
+      "step": 1489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893973989563062e-05,
+      "loss": 0.9417,
+      "step": 1490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893737083462852e-05,
+      "loss": 1.1123,
+      "step": 1491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989349991439841e-05,
+      "loss": 1.0461,
+      "step": 1492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989326248237604e-05,
+      "loss": 1.0194,
+      "step": 1493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893024787402048e-05,
+      "loss": 1.0406,
+      "step": 1494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989278682948276e-05,
+      "loss": 1.0264,
+      "step": 1495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892548608624497e-05,
+      "loss": 0.9818,
+      "step": 1496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892310124833588e-05,
+      "loss": 0.9549,
+      "step": 1497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892071378116378e-05,
+      "loss": 0.9876,
+      "step": 1498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891832368479206e-05,
+      "loss": 1.0362,
+      "step": 1499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891593095928425e-05,
+      "loss": 0.4436,
+      "step": 1500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98913535604704e-05,
+      "loss": 1.0057,
+      "step": 1501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891113762111493e-05,
+      "loss": 1.074,
+      "step": 1502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989087370085808e-05,
+      "loss": 1.0548,
+      "step": 1503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989063337671654e-05,
+      "loss": 1.013,
+      "step": 1504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890392789693262e-05,
+      "loss": 1.0288,
+      "step": 1505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890151939794636e-05,
+      "loss": 1.0556,
+      "step": 1506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988991082702707e-05,
+      "loss": 1.0312,
+      "step": 1507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988966945139697e-05,
+      "loss": 0.9982,
+      "step": 1508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988942781291075e-05,
+      "loss": 1.0622,
+      "step": 1509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988918591157483e-05,
+      "loss": 0.9696,
+      "step": 1510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988894374739565e-05,
+      "loss": 1.0184,
+      "step": 1511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888701320379635e-05,
+      "loss": 1.0971,
+      "step": 1512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888458630533233e-05,
+      "loss": 0.9607,
+      "step": 1513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888215677862895e-05,
+      "loss": 1.0186,
+      "step": 1514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988797246237508e-05,
+      "loss": 0.9399,
+      "step": 1515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988772898407625e-05,
+      "loss": 0.9811,
+      "step": 1516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887485242972873e-05,
+      "loss": 0.9985,
+      "step": 1517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988724123907143e-05,
+      "loss": 0.981,
+      "step": 1518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988699697237841e-05,
+      "loss": 1.0469,
+      "step": 1519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886752442900304e-05,
+      "loss": 0.9673,
+      "step": 1520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886507650643608e-05,
+      "loss": 0.9741,
+      "step": 1521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886262595614828e-05,
+      "loss": 0.9421,
+      "step": 1522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886017277820483e-05,
+      "loss": 1.0404,
+      "step": 1523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885771697267086e-05,
+      "loss": 1.0943,
+      "step": 1524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885525853961168e-05,
+      "loss": 0.9596,
+      "step": 1525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988527974790927e-05,
+      "loss": 1.0397,
+      "step": 1526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988503337911792e-05,
+      "loss": 1.1139,
+      "step": 1527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884786747593673e-05,
+      "loss": 1.0548,
+      "step": 1528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884539853343084e-05,
+      "loss": 0.9825,
+      "step": 1529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988429269637272e-05,
+      "loss": 0.999,
+      "step": 1530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884045276689138e-05,
+      "loss": 1.0542,
+      "step": 1531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883797594298925e-05,
+      "loss": 1.0531,
+      "step": 1532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883549649208658e-05,
+      "loss": 0.9847,
+      "step": 1533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988330144142493e-05,
+      "loss": 0.9895,
+      "step": 1534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988305297095434e-05,
+      "loss": 1.021,
+      "step": 1535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882804237803487e-05,
+      "loss": 1.104,
+      "step": 1536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988255524197899e-05,
+      "loss": 1.0422,
+      "step": 1537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882305983487453e-05,
+      "loss": 1.0254,
+      "step": 1538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882056462335513e-05,
+      "loss": 1.0126,
+      "step": 1539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98818066785298e-05,
+      "loss": 1.1231,
+      "step": 1540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988155663207695e-05,
+      "loss": 1.025,
+      "step": 1541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881306322983612e-05,
+      "loss": 1.0321,
+      "step": 1542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881055751256437e-05,
+      "loss": 0.9869,
+      "step": 1543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880804916902085e-05,
+      "loss": 1.0172,
+      "step": 1544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988055381992722e-05,
+      "loss": 0.9747,
+      "step": 1545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880302460338526e-05,
+      "loss": 1.0491,
+      "step": 1546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880050838142675e-05,
+      "loss": 1.0856,
+      "step": 1547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879798953346353e-05,
+      "loss": 1.011,
+      "step": 1548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987954680595626e-05,
+      "loss": 1.0189,
+      "step": 1549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879294395979102e-05,
+      "loss": 1.0331,
+      "step": 1550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879041723421576e-05,
+      "loss": 1.0068,
+      "step": 1551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878788788290407e-05,
+      "loss": 0.9681,
+      "step": 1552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987853559059231e-05,
+      "loss": 0.9021,
+      "step": 1553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878282130334023e-05,
+      "loss": 0.91,
+      "step": 1554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878028407522277e-05,
+      "loss": 0.9748,
+      "step": 1555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877774422163823e-05,
+      "loss": 0.9718,
+      "step": 1556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877520174265403e-05,
+      "loss": 1.1437,
+      "step": 1557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877265663833776e-05,
+      "loss": 1.0987,
+      "step": 1558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877010890875712e-05,
+      "loss": 1.0557,
+      "step": 1559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876755855397976e-05,
+      "loss": 0.94,
+      "step": 1560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876500557407353e-05,
+      "loss": 1.0016,
+      "step": 1561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876244996910624e-05,
+      "loss": 1.0324,
+      "step": 1562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987598917391458e-05,
+      "loss": 1.0096,
+      "step": 1563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875733088426025e-05,
+      "loss": 0.988,
+      "step": 1564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875476740451765e-05,
+      "loss": 1.0045,
+      "step": 1565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875220129998612e-05,
+      "loss": 1.0613,
+      "step": 1566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874963257073382e-05,
+      "loss": 1.0505,
+      "step": 1567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874706121682915e-05,
+      "loss": 1.0574,
+      "step": 1568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987444872383403e-05,
+      "loss": 0.9277,
+      "step": 1569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987419106353358e-05,
+      "loss": 1.1463,
+      "step": 1570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987393314078841e-05,
+      "loss": 1.0142,
+      "step": 1571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873674955605368e-05,
+      "loss": 0.9761,
+      "step": 1572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987341650799133e-05,
+      "loss": 1.0085,
+      "step": 1573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987315779795315e-05,
+      "loss": 0.9422,
+      "step": 1574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872898825497716e-05,
+      "loss": 1.0168,
+      "step": 1575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872639590631907e-05,
+      "loss": 0.9889,
+      "step": 1576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872380093362612e-05,
+      "loss": 1.0699,
+      "step": 1577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872120333696734e-05,
+      "loss": 1.0429,
+      "step": 1578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871860311641167e-05,
+      "loss": 0.9987,
+      "step": 1579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987160002720283e-05,
+      "loss": 1.0648,
+      "step": 1580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987133948038864e-05,
+      "loss": 1.0224,
+      "step": 1581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987107867120552e-05,
+      "loss": 1.0513,
+      "step": 1582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98708175996604e-05,
+      "loss": 1.0275,
+      "step": 1583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870556265760225e-05,
+      "loss": 0.9516,
+      "step": 1584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870294669511937e-05,
+      "loss": 1.0471,
+      "step": 1585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870032810922488e-05,
+      "loss": 1.1268,
+      "step": 1586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986977068999884e-05,
+      "loss": 0.9803,
+      "step": 1587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869508306747966e-05,
+      "loss": 1.0251,
+      "step": 1588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869245661176825e-05,
+      "loss": 1.021,
+      "step": 1589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986898275329241e-05,
+      "loss": 1.0308,
+      "step": 1590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868719583101703e-05,
+      "loss": 0.9905,
+      "step": 1591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98684561506117e-05,
+      "loss": 1.0427,
+      "step": 1592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868192455829404e-05,
+      "loss": 1.0623,
+      "step": 1593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867928498761822e-05,
+      "loss": 1.0765,
+      "step": 1594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867664279415974e-05,
+      "loss": 1.031,
+      "step": 1595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986739979779888e-05,
+      "loss": 1.0605,
+      "step": 1596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867135053917564e-05,
+      "loss": 1.0327,
+      "step": 1597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986687004777907e-05,
+      "loss": 1.0475,
+      "step": 1598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866604779390443e-05,
+      "loss": 1.0382,
+      "step": 1599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866339248758723e-05,
+      "loss": 0.9576,
+      "step": 1600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986607345589098e-05,
+      "loss": 1.0308,
+      "step": 1601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865807400794267e-05,
+      "loss": 1.1206,
+      "step": 1602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865541083475666e-05,
+      "loss": 0.9881,
+      "step": 1603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986527450394225e-05,
+      "loss": 1.0271,
+      "step": 1604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865007662201105e-05,
+      "loss": 0.9698,
+      "step": 1605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986474055825932e-05,
+      "loss": 1.0639,
+      "step": 1606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864473192123998e-05,
+      "loss": 0.9603,
+      "step": 1607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986420556380225e-05,
+      "loss": 0.9425,
+      "step": 1608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986393767330118e-05,
+      "loss": 1.0227,
+      "step": 1609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863669520627913e-05,
+      "loss": 1.008,
+      "step": 1610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863401105789573e-05,
+      "loss": 1.0899,
+      "step": 1611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863132428793303e-05,
+      "loss": 1.031,
+      "step": 1612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986286348964623e-05,
+      "loss": 0.9825,
+      "step": 1613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862594288355516e-05,
+      "loss": 1.0522,
+      "step": 1614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862324824928308e-05,
+      "loss": 0.8725,
+      "step": 1615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986205509937177e-05,
+      "loss": 0.9991,
+      "step": 1616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861785111693067e-05,
+      "loss": 1.082,
+      "step": 1617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861514861899386e-05,
+      "loss": 1.1321,
+      "step": 1618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98612443499979e-05,
+      "loss": 1.0459,
+      "step": 1619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860973575995802e-05,
+      "loss": 1.0386,
+      "step": 1620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860702539900288e-05,
+      "loss": 0.9784,
+      "step": 1621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986043124171856e-05,
+      "loss": 0.9628,
+      "step": 1622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860159681457835e-05,
+      "loss": 1.0751,
+      "step": 1623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859887859125326e-05,
+      "loss": 1.0086,
+      "step": 1624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985961577472826e-05,
+      "loss": 1.0117,
+      "step": 1625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859343428273864e-05,
+      "loss": 1.0087,
+      "step": 1626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859070819769385e-05,
+      "loss": 1.031,
+      "step": 1627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985879794922206e-05,
+      "loss": 1.0726,
+      "step": 1628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858524816639147e-05,
+      "loss": 1.1271,
+      "step": 1629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858251422027903e-05,
+      "loss": 1.0119,
+      "step": 1630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857977765395596e-05,
+      "loss": 0.9393,
+      "step": 1631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98577038467495e-05,
+      "loss": 1.0997,
+      "step": 1632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985742966609689e-05,
+      "loss": 0.9901,
+      "step": 1633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857155223445063e-05,
+      "loss": 1.0779,
+      "step": 1634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856880518801305e-05,
+      "loss": 1.0276,
+      "step": 1635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985660555217292e-05,
+      "loss": 1.1186,
+      "step": 1636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985633032356722e-05,
+      "loss": 1.0114,
+      "step": 1637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856054832991515e-05,
+      "loss": 1.0124,
+      "step": 1638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855779080453127e-05,
+      "loss": 1.0376,
+      "step": 1639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985550306595939e-05,
+      "loss": 1.0127,
+      "step": 1640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985522678951764e-05,
+      "loss": 1.0471,
+      "step": 1641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854950251135215e-05,
+      "loss": 1.0893,
+      "step": 1642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854673450819467e-05,
+      "loss": 1.0239,
+      "step": 1643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854396388577754e-05,
+      "loss": 1.0502,
+      "step": 1644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854119064417437e-05,
+      "loss": 1.0574,
+      "step": 1645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853841478345894e-05,
+      "loss": 1.1076,
+      "step": 1646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98535636303705e-05,
+      "loss": 1.035,
+      "step": 1647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853285520498637e-05,
+      "loss": 1.0287,
+      "step": 1648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853007148737695e-05,
+      "loss": 1.0182,
+      "step": 1649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852728515095078e-05,
+      "loss": 1.0746,
+      "step": 1650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985244961957819e-05,
+      "loss": 0.9572,
+      "step": 1651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985217046219445e-05,
+      "loss": 1.1028,
+      "step": 1652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851891042951263e-05,
+      "loss": 1.1167,
+      "step": 1653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851611361856068e-05,
+      "loss": 0.9889,
+      "step": 1654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985133141891629e-05,
+      "loss": 0.4312,
+      "step": 1655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985105121413938e-05,
+      "loss": 1.0422,
+      "step": 1656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850770747532774e-05,
+      "loss": 1.0244,
+      "step": 1657
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985049001910394e-05,
+      "loss": 0.9623,
+      "step": 1658
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850209028860327e-05,
+      "loss": 0.9907,
+      "step": 1659
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849927776809408e-05,
+      "loss": 1.0857,
+      "step": 1660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849646262958657e-05,
+      "loss": 1.0031,
+      "step": 1661
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984936448731556e-05,
+      "loss": 0.4217,
+      "step": 1662
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98490824498876e-05,
+      "loss": 0.9804,
+      "step": 1663
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984880015068228e-05,
+      "loss": 0.9887,
+      "step": 1664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848517589707104e-05,
+      "loss": 0.9853,
+      "step": 1665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848234766969578e-05,
+      "loss": 0.9984,
+      "step": 1666
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847951682477214e-05,
+      "loss": 1.1258,
+      "step": 1667
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847668336237544e-05,
+      "loss": 0.9873,
+      "step": 1668
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98473847282581e-05,
+      "loss": 1.0114,
+      "step": 1669
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847100858546414e-05,
+      "loss": 1.0356,
+      "step": 1670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984681672711003e-05,
+      "loss": 0.3647,
+      "step": 1671
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984653233395651e-05,
+      "loss": 1.0663,
+      "step": 1672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846247679093404e-05,
+      "loss": 1.0616,
+      "step": 1673
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984596276252828e-05,
+      "loss": 0.9869,
+      "step": 1674
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984567758426871e-05,
+      "loss": 0.9465,
+      "step": 1675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845392144322274e-05,
+      "loss": 1.0453,
+      "step": 1676
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845106442696563e-05,
+      "loss": 1.0947,
+      "step": 1677
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844820479399166e-05,
+      "loss": 1.0451,
+      "step": 1678
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984453425443768e-05,
+      "loss": 1.0424,
+      "step": 1679
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984424776781972e-05,
+      "loss": 1.0105,
+      "step": 1680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98439610195529e-05,
+      "loss": 1.0556,
+      "step": 1681
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984367400964484e-05,
+      "loss": 0.4005,
+      "step": 1682
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843386738103164e-05,
+      "loss": 1.116,
+      "step": 1683
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984309920493551e-05,
+      "loss": 1.0789,
+      "step": 1684
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984281141014952e-05,
+      "loss": 0.991,
+      "step": 1685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842523353752846e-05,
+      "loss": 1.0966,
+      "step": 1686
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842235035753142e-05,
+      "loss": 1.0811,
+      "step": 1687
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841946456158074e-05,
+      "loss": 1.0693,
+      "step": 1688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984165761497531e-05,
+      "loss": 1.0862,
+      "step": 1689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984136851221252e-05,
+      "loss": 1.0608,
+      "step": 1690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841079147877402e-05,
+      "loss": 0.9698,
+      "step": 1691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840789521977638e-05,
+      "loss": 0.9891,
+      "step": 1692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984049963452093e-05,
+      "loss": 0.9823,
+      "step": 1693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840209485514977e-05,
+      "loss": 1.0036,
+      "step": 1694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839919074967496e-05,
+      "loss": 0.9476,
+      "step": 1695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839628402886206e-05,
+      "loss": 0.994,
+      "step": 1696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839337469278827e-05,
+      "loss": 0.3978,
+      "step": 1697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839046274153096e-05,
+      "loss": 1.0176,
+      "step": 1698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838754817516754e-05,
+      "loss": 0.9958,
+      "step": 1699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838463099377546e-05,
+      "loss": 1.0911,
+      "step": 1700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838171119743225e-05,
+      "loss": 1.0344,
+      "step": 1701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837878878621552e-05,
+      "loss": 1.0156,
+      "step": 1702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837586376020293e-05,
+      "loss": 1.0184,
+      "step": 1703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837293611947225e-05,
+      "loss": 1.0115,
+      "step": 1704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837000586410132e-05,
+      "loss": 0.4257,
+      "step": 1705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836707299416794e-05,
+      "loss": 1.0644,
+      "step": 1706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983641375097501e-05,
+      "loss": 1.0736,
+      "step": 1707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836119941092585e-05,
+      "loss": 1.0561,
+      "step": 1708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835825869777326e-05,
+      "loss": 1.0071,
+      "step": 1709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835531537037047e-05,
+      "loss": 1.0607,
+      "step": 1710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983523694287957e-05,
+      "loss": 1.0031,
+      "step": 1711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834942087312733e-05,
+      "loss": 1.0451,
+      "step": 1712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834646970344367e-05,
+      "loss": 1.0386,
+      "step": 1713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834351591982318e-05,
+      "loss": 0.9869,
+      "step": 1714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834055952234433e-05,
+      "loss": 1.0512,
+      "step": 1715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833760051108576e-05,
+      "loss": 1.0511,
+      "step": 1716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833463888612603e-05,
+      "loss": 1.0177,
+      "step": 1717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833167464754393e-05,
+      "loss": 0.9767,
+      "step": 1718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832870779541822e-05,
+      "loss": 0.9796,
+      "step": 1719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832573832982778e-05,
+      "loss": 1.0066,
+      "step": 1720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983227662508515e-05,
+      "loss": 0.9886,
+      "step": 1721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983197915585684e-05,
+      "loss": 1.072,
+      "step": 1722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983168142530575e-05,
+      "loss": 0.9757,
+      "step": 1723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831383433439798e-05,
+      "loss": 0.9882,
+      "step": 1724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831085180266903e-05,
+      "loss": 1.0324,
+      "step": 1725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830786665794993e-05,
+      "loss": 1.0812,
+      "step": 1726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830487890032e-05,
+      "loss": 1.0211,
+      "step": 1727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830188852985867e-05,
+      "loss": 0.9585,
+      "step": 1728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982988955466454e-05,
+      "loss": 1.0656,
+      "step": 1729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829589995075977e-05,
+      "loss": 1.0797,
+      "step": 1730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982929017422814e-05,
+      "loss": 1.0313,
+      "step": 1731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828990092128994e-05,
+      "loss": 1.0,
+      "step": 1732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828689748786516e-05,
+      "loss": 1.0841,
+      "step": 1733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828389144208692e-05,
+      "loss": 0.9573,
+      "step": 1734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982808827840351e-05,
+      "loss": 1.0129,
+      "step": 1735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827787151378965e-05,
+      "loss": 1.057,
+      "step": 1736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982748576314306e-05,
+      "loss": 1.1139,
+      "step": 1737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827184113703814e-05,
+      "loss": 1.0119,
+      "step": 1738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826882203069233e-05,
+      "loss": 1.0063,
+      "step": 1739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826580031247346e-05,
+      "loss": 1.048,
+      "step": 1740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826277598246188e-05,
+      "loss": 1.0479,
+      "step": 1741
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982597490407379e-05,
+      "loss": 1.0295,
+      "step": 1742
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825671948738205e-05,
+      "loss": 0.4571,
+      "step": 1743
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982536873224748e-05,
+      "loss": 0.9572,
+      "step": 1744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825065254609676e-05,
+      "loss": 1.0398,
+      "step": 1745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824761515832857e-05,
+      "loss": 1.0302,
+      "step": 1746
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98244575159251e-05,
+      "loss": 1.0027,
+      "step": 1747
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824153254894484e-05,
+      "loss": 1.0318,
+      "step": 1748
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823848732749093e-05,
+      "loss": 0.9953,
+      "step": 1749
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823543949497024e-05,
+      "loss": 1.06,
+      "step": 1750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823238905146374e-05,
+      "loss": 0.9742,
+      "step": 1751
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822933599705256e-05,
+      "loss": 1.1689,
+      "step": 1752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982262803318178e-05,
+      "loss": 0.9318,
+      "step": 1753
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822322205584066e-05,
+      "loss": 1.0623,
+      "step": 1754
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822016116920252e-05,
+      "loss": 0.4075,
+      "step": 1755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821709767198464e-05,
+      "loss": 1.0871,
+      "step": 1756
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821403156426848e-05,
+      "loss": 1.0355,
+      "step": 1757
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821096284613555e-05,
+      "loss": 1.0644,
+      "step": 1758
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982078915176674e-05,
+      "loss": 1.1111,
+      "step": 1759
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820481757894563e-05,
+      "loss": 0.9981,
+      "step": 1760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98201741030052e-05,
+      "loss": 0.9453,
+      "step": 1761
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819866187106824e-05,
+      "loss": 1.062,
+      "step": 1762
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819558010207615e-05,
+      "loss": 0.3916,
+      "step": 1763
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819249572315774e-05,
+      "loss": 0.9833,
+      "step": 1764
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818940873439495e-05,
+      "loss": 1.0446,
+      "step": 1765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981863191358698e-05,
+      "loss": 1.0635,
+      "step": 1766
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981832269276644e-05,
+      "loss": 1.0111,
+      "step": 1767
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818013210986103e-05,
+      "loss": 0.3776,
+      "step": 1768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817703468254182e-05,
+      "loss": 1.0496,
+      "step": 1769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981739346457892e-05,
+      "loss": 1.1414,
+      "step": 1770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817083199968552e-05,
+      "loss": 0.9786,
+      "step": 1771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816772674431322e-05,
+      "loss": 1.0447,
+      "step": 1772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981646188797549e-05,
+      "loss": 1.0341,
+      "step": 1773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981615084060931e-05,
+      "loss": 0.9955,
+      "step": 1774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981583953234105e-05,
+      "loss": 1.0292,
+      "step": 1775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815527963178987e-05,
+      "loss": 0.9915,
+      "step": 1776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815216133131404e-05,
+      "loss": 0.9992,
+      "step": 1777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981490404220658e-05,
+      "loss": 1.0505,
+      "step": 1778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814591690412822e-05,
+      "loss": 0.9818,
+      "step": 1779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981427907775842e-05,
+      "loss": 0.9553,
+      "step": 1780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98139662042517e-05,
+      "loss": 1.0167,
+      "step": 1781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981365306990096e-05,
+      "loss": 1.0831,
+      "step": 1782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981333967471453e-05,
+      "loss": 1.0519,
+      "step": 1783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981302601870074e-05,
+      "loss": 1.0324,
+      "step": 1784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812712101867923e-05,
+      "loss": 0.9919,
+      "step": 1785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981239792422443e-05,
+      "loss": 1.0714,
+      "step": 1786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812083485778603e-05,
+      "loss": 0.9961,
+      "step": 1787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811768786538804e-05,
+      "loss": 1.0109,
+      "step": 1788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98114538265134e-05,
+      "loss": 0.9666,
+      "step": 1789
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811138605710755e-05,
+      "loss": 1.0024,
+      "step": 1790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810823124139254e-05,
+      "loss": 1.0591,
+      "step": 1791
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981050738180728e-05,
+      "loss": 1.1102,
+      "step": 1792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810191378723224e-05,
+      "loss": 1.056,
+      "step": 1793
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809875114895486e-05,
+      "loss": 1.0142,
+      "step": 1794
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809558590332466e-05,
+      "loss": 0.966,
+      "step": 1795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980924180504259e-05,
+      "loss": 1.0162,
+      "step": 1796
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808924759034266e-05,
+      "loss": 1.0536,
+      "step": 1797
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808607452315924e-05,
+      "loss": 0.9014,
+      "step": 1798
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808289884896e-05,
+      "loss": 1.0852,
+      "step": 1799
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807972056782935e-05,
+      "loss": 1.1338,
+      "step": 1800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980765396798517e-05,
+      "loss": 1.0213,
+      "step": 1801
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807335618511168e-05,
+      "loss": 0.9607,
+      "step": 1802
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807017008369382e-05,
+      "loss": 1.0293,
+      "step": 1803
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980669813756829e-05,
+      "loss": 1.0495,
+      "step": 1804
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806379006116357e-05,
+      "loss": 0.9553,
+      "step": 1805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806059614022072e-05,
+      "loss": 1.0218,
+      "step": 1806
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980573996129392e-05,
+      "loss": 1.0309,
+      "step": 1807
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98054200479404e-05,
+      "loss": 1.0043,
+      "step": 1808
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805099873970016e-05,
+      "loss": 1.0362,
+      "step": 1809
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804779439391273e-05,
+      "loss": 1.021,
+      "step": 1810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804458744212697e-05,
+      "loss": 1.0669,
+      "step": 1811
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804137788442798e-05,
+      "loss": 0.924,
+      "step": 1812
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803816572090117e-05,
+      "loss": 0.9645,
+      "step": 1813
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980349509516319e-05,
+      "loss": 0.9868,
+      "step": 1814
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803173357670558e-05,
+      "loss": 0.9431,
+      "step": 1815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802851359620776e-05,
+      "loss": 0.961,
+      "step": 1816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98025291010224e-05,
+      "loss": 1.0422,
+      "step": 1817
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802206581883992e-05,
+      "loss": 1.0424,
+      "step": 1818
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980188380221413e-05,
+      "loss": 1.0775,
+      "step": 1819
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801560762021398e-05,
+      "loss": 1.0075,
+      "step": 1820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980123746131437e-05,
+      "loss": 1.0268,
+      "step": 1821
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800913900101646e-05,
+      "loss": 1.0818,
+      "step": 1822
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800590078391823e-05,
+      "loss": 0.9217,
+      "step": 1823
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980026599619351e-05,
+      "loss": 0.9641,
+      "step": 1824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799941653515316e-05,
+      "loss": 1.0479,
+      "step": 1825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979961705036587e-05,
+      "loss": 0.9871,
+      "step": 1826
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799292186753794e-05,
+      "loss": 1.014,
+      "step": 1827
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979896706268772e-05,
+      "loss": 1.0598,
+      "step": 1828
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97986416781763e-05,
+      "loss": 1.038,
+      "step": 1829
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979831603322817e-05,
+      "loss": 0.9845,
+      "step": 1830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797990127851992e-05,
+      "loss": 1.0204,
+      "step": 1831
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797663962056427e-05,
+      "loss": 1.0046,
+      "step": 1832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797337535850145e-05,
+      "loss": 0.9666,
+      "step": 1833
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797010849241817e-05,
+      "loss": 1.0814,
+      "step": 1834
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796683902240135e-05,
+      "loss": 0.9465,
+      "step": 1835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979635669485378e-05,
+      "loss": 0.9724,
+      "step": 1836
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796029227091452e-05,
+      "loss": 0.987,
+      "step": 1837
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795701498961858e-05,
+      "loss": 1.0694,
+      "step": 1838
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795373510473706e-05,
+      "loss": 0.9762,
+      "step": 1839
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795045261635713e-05,
+      "loss": 0.9839,
+      "step": 1840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794716752456602e-05,
+      "loss": 1.0933,
+      "step": 1841
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794387982945105e-05,
+      "loss": 1.0363,
+      "step": 1842
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979405895310997e-05,
+      "loss": 0.9799,
+      "step": 1843
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793729662959927e-05,
+      "loss": 1.0125,
+      "step": 1844
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793400112503735e-05,
+      "loss": 0.9947,
+      "step": 1845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793070301750154e-05,
+      "loss": 0.979,
+      "step": 1846
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979274023070795e-05,
+      "loss": 0.9945,
+      "step": 1847
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792409899385897e-05,
+      "loss": 0.9559,
+      "step": 1848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792079307792768e-05,
+      "loss": 0.9836,
+      "step": 1849
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791748455937358e-05,
+      "loss": 1.0192,
+      "step": 1850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791417343828456e-05,
+      "loss": 1.0806,
+      "step": 1851
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791085971474864e-05,
+      "loss": 0.9778,
+      "step": 1852
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790754338885388e-05,
+      "loss": 0.9985,
+      "step": 1853
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790422446068847e-05,
+      "loss": 0.9152,
+      "step": 1854
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979009029303406e-05,
+      "loss": 0.9643,
+      "step": 1855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978975787978985e-05,
+      "loss": 1.0436,
+      "step": 1856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978942520634506e-05,
+      "loss": 0.9452,
+      "step": 1857
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789092272708527e-05,
+      "loss": 1.0361,
+      "step": 1858
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788759078889105e-05,
+      "loss": 0.9734,
+      "step": 1859
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788425624895643e-05,
+      "loss": 1.0674,
+      "step": 1860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788091910737008e-05,
+      "loss": 1.0546,
+      "step": 1861
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787757936422065e-05,
+      "loss": 1.0216,
+      "step": 1862
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97874237019597e-05,
+      "loss": 1.0456,
+      "step": 1863
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978708920735879e-05,
+      "loss": 1.0151,
+      "step": 1864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786754452628226e-05,
+      "loss": 0.999,
+      "step": 1865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786419437776906e-05,
+      "loss": 1.0403,
+      "step": 1866
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786084162813735e-05,
+      "loss": 1.0406,
+      "step": 1867
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785748627747624e-05,
+      "loss": 0.9667,
+      "step": 1868
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785412832587493e-05,
+      "loss": 0.927,
+      "step": 1869
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785076777342264e-05,
+      "loss": 1.0456,
+      "step": 1870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784740462020867e-05,
+      "loss": 1.0601,
+      "step": 1871
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784403886632244e-05,
+      "loss": 1.01,
+      "step": 1872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784067051185343e-05,
+      "loss": 1.0195,
+      "step": 1873
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978372995568911e-05,
+      "loss": 1.0186,
+      "step": 1874
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783392600152516e-05,
+      "loss": 0.9931,
+      "step": 1875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783054984584516e-05,
+      "loss": 1.0701,
+      "step": 1876
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782717108994088e-05,
+      "loss": 1.0092,
+      "step": 1877
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782378973390214e-05,
+      "loss": 0.4452,
+      "step": 1878
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782040577781876e-05,
+      "loss": 1.0231,
+      "step": 1879
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781701922178075e-05,
+      "loss": 0.9679,
+      "step": 1880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781363006587804e-05,
+      "loss": 1.0702,
+      "step": 1881
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978102383102008e-05,
+      "loss": 1.0203,
+      "step": 1882
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780684395483913e-05,
+      "loss": 1.0823,
+      "step": 1883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780344699988324e-05,
+      "loss": 1.0044,
+      "step": 1884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780004744542345e-05,
+      "loss": 1.0089,
+      "step": 1885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.977966452915501e-05,
+      "loss": 1.0417,
+      "step": 1886
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.977932405383536e-05,
+      "loss": 0.9823,
+      "step": 1887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778983318592444e-05,
+      "loss": 0.9782,
+      "step": 1888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778642323435325e-05,
+      "loss": 1.0123,
+      "step": 1889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778301068373055e-05,
+      "loss": 1.046,
+      "step": 1890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777959553414713e-05,
+      "loss": 1.0474,
+      "step": 1891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777617778569376e-05,
+      "loss": 1.0215,
+      "step": 1892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777275743846123e-05,
+      "loss": 1.1001,
+      "step": 1893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776933449254048e-05,
+      "loss": 0.9986,
+      "step": 1894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776590894802247e-05,
+      "loss": 1.0199,
+      "step": 1895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776248080499826e-05,
+      "loss": 1.0627,
+      "step": 1896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97759050063559e-05,
+      "loss": 1.0599,
+      "step": 1897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775561672379582e-05,
+      "loss": 0.8704,
+      "step": 1898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977521807858e-05,
+      "loss": 1.066,
+      "step": 1899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774874224966284e-05,
+      "loss": 0.9903,
+      "step": 1900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977453011154758e-05,
+      "loss": 1.0294,
+      "step": 1901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774185738333023e-05,
+      "loss": 1.0466,
+      "step": 1902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773841105331775e-05,
+      "loss": 1.0001,
+      "step": 1903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773496212552993e-05,
+      "loss": 1.0076,
+      "step": 1904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773151060005845e-05,
+      "loss": 1.0921,
+      "step": 1905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772805647699503e-05,
+      "loss": 1.0375,
+      "step": 1906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977245997564315e-05,
+      "loss": 1.085,
+      "step": 1907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772114043845968e-05,
+      "loss": 0.9721,
+      "step": 1908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771767852317158e-05,
+      "loss": 1.0946,
+      "step": 1909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977142140106592e-05,
+      "loss": 1.0486,
+      "step": 1910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977107469010146e-05,
+      "loss": 1.1129,
+      "step": 1911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770727719432994e-05,
+      "loss": 1.0088,
+      "step": 1912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770380489069744e-05,
+      "loss": 1.0777,
+      "step": 1913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770032999020944e-05,
+      "loss": 1.0137,
+      "step": 1914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976968524929582e-05,
+      "loss": 1.0357,
+      "step": 1915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769337239903622e-05,
+      "loss": 1.0043,
+      "step": 1916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768988970853596e-05,
+      "loss": 1.0288,
+      "step": 1917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768640442155002e-05,
+      "loss": 1.0374,
+      "step": 1918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768291653817102e-05,
+      "loss": 1.0964,
+      "step": 1919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767942605849168e-05,
+      "loss": 1.0386,
+      "step": 1920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767593298260475e-05,
+      "loss": 0.9766,
+      "step": 1921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767243731060307e-05,
+      "loss": 0.99,
+      "step": 1922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766893904257954e-05,
+      "loss": 1.0927,
+      "step": 1923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766543817862717e-05,
+      "loss": 1.0198,
+      "step": 1924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97661934718839e-05,
+      "loss": 1.0717,
+      "step": 1925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765842866330815e-05,
+      "loss": 1.0772,
+      "step": 1926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976549200121278e-05,
+      "loss": 1.0533,
+      "step": 1927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976514087653912e-05,
+      "loss": 1.0054,
+      "step": 1928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976478949231917e-05,
+      "loss": 1.0776,
+      "step": 1929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764437848562267e-05,
+      "loss": 0.9327,
+      "step": 1930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976408594527776e-05,
+      "loss": 1.0641,
+      "step": 1931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763733782475e-05,
+      "loss": 0.9968,
+      "step": 1932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976338136016334e-05,
+      "loss": 1.0518,
+      "step": 1933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763028678352164e-05,
+      "loss": 0.9743,
+      "step": 1934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762675737050833e-05,
+      "loss": 0.9038,
+      "step": 1935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762322536268728e-05,
+      "loss": 1.0602,
+      "step": 1936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976196907601524e-05,
+      "loss": 1.0174,
+      "step": 1937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976161535629977e-05,
+      "loss": 1.0049,
+      "step": 1938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761261377131703e-05,
+      "loss": 0.9957,
+      "step": 1939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760907138520463e-05,
+      "loss": 0.4577,
+      "step": 1940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976055264047546e-05,
+      "loss": 0.9684,
+      "step": 1941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976019788300611e-05,
+      "loss": 1.0102,
+      "step": 1942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759842866121853e-05,
+      "loss": 1.0473,
+      "step": 1943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759487589832115e-05,
+      "loss": 1.0446,
+      "step": 1944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759132054146347e-05,
+      "loss": 0.9959,
+      "step": 1945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758776259073994e-05,
+      "loss": 1.0619,
+      "step": 1946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975842020462451e-05,
+      "loss": 1.0358,
+      "step": 1947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758063890807368e-05,
+      "loss": 0.888,
+      "step": 1948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975770731763203e-05,
+      "loss": 1.0649,
+      "step": 1949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757350485107974e-05,
+      "loss": 0.8425,
+      "step": 1950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756993393244687e-05,
+      "loss": 0.9636,
+      "step": 1951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975663604205166e-05,
+      "loss": 1.0006,
+      "step": 1952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756278431538393e-05,
+      "loss": 0.4785,
+      "step": 1953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755920561714385e-05,
+      "loss": 0.9924,
+      "step": 1954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975556243258915e-05,
+      "loss": 1.0101,
+      "step": 1955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755204044172207e-05,
+      "loss": 0.9677,
+      "step": 1956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754845396473086e-05,
+      "loss": 1.0297,
+      "step": 1957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754486489501313e-05,
+      "loss": 1.0545,
+      "step": 1958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754127323266426e-05,
+      "loss": 1.053,
+      "step": 1959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975376789777798e-05,
+      "loss": 0.3819,
+      "step": 1960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753408213045522e-05,
+      "loss": 1.1327,
+      "step": 1961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753048269078613e-05,
+      "loss": 1.1123,
+      "step": 1962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975268806588682e-05,
+      "loss": 1.0538,
+      "step": 1963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752327603479717e-05,
+      "loss": 1.0374,
+      "step": 1964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751966881866883e-05,
+      "loss": 0.9586,
+      "step": 1965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751605901057907e-05,
+      "loss": 1.1321,
+      "step": 1966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751244661062387e-05,
+      "loss": 1.0127,
+      "step": 1967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975088316188992e-05,
+      "loss": 1.0216,
+      "step": 1968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750521403550112e-05,
+      "loss": 0.414,
+      "step": 1969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750159386052586e-05,
+      "loss": 1.0974,
+      "step": 1970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749797109406957e-05,
+      "loss": 1.0056,
+      "step": 1971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749434573622856e-05,
+      "loss": 1.0449,
+      "step": 1972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974907177870992e-05,
+      "loss": 1.0849,
+      "step": 1973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974870872467779e-05,
+      "loss": 1.0149,
+      "step": 1974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748345411536113e-05,
+      "loss": 0.992,
+      "step": 1975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974798183929455e-05,
+      "loss": 0.9504,
+      "step": 1976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747618007962767e-05,
+      "loss": 0.9687,
+      "step": 1977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747253917550423e-05,
+      "loss": 1.0907,
+      "step": 1978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746889568067205e-05,
+      "loss": 1.018,
+      "step": 1979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97465249595228e-05,
+      "loss": 1.0035,
+      "step": 1980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746160091926888e-05,
+      "loss": 0.9847,
+      "step": 1981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745794965289172e-05,
+      "loss": 1.0426,
+      "step": 1982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974542957961936e-05,
+      "loss": 0.9796,
+      "step": 1983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745063934927158e-05,
+      "loss": 0.9902,
+      "step": 1984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744698031222284e-05,
+      "loss": 1.0077,
+      "step": 1985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744331868514467e-05,
+      "loss": 0.4293,
+      "step": 1986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974396544681344e-05,
+      "loss": 0.9708,
+      "step": 1987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974359876612894e-05,
+      "loss": 0.3744,
+      "step": 1988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743231826470713e-05,
+      "loss": 1.0256,
+      "step": 1989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974286462784851e-05,
+      "loss": 1.0245,
+      "step": 1990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742497170272096e-05,
+      "loss": 0.9533,
+      "step": 1991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742129453751228e-05,
+      "loss": 1.0245,
+      "step": 1992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974176147829569e-05,
+      "loss": 1.067,
+      "step": 1993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741393243915258e-05,
+      "loss": 1.0146,
+      "step": 1994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974102475061972e-05,
+      "loss": 0.9606,
+      "step": 1995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974065599841887e-05,
+      "loss": 1.0338,
+      "step": 1996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740286987322504e-05,
+      "loss": 1.0034,
+      "step": 1997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739917717340436e-05,
+      "loss": 1.0038,
+      "step": 1998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739548188482484e-05,
+      "loss": 1.006,
+      "step": 1999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739178400758462e-05,
+      "loss": 1.1152,
+      "step": 2000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738808354178197e-05,
+      "loss": 0.9827,
+      "step": 2001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738438048751535e-05,
+      "loss": 1.0135,
+      "step": 2002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738067484488307e-05,
+      "loss": 1.0505,
+      "step": 2003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973769666139837e-05,
+      "loss": 1.0156,
+      "step": 2004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737325579491577e-05,
+      "loss": 0.98,
+      "step": 2005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736954238777793e-05,
+      "loss": 1.0431,
+      "step": 2006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736582639266883e-05,
+      "loss": 1.0064,
+      "step": 2007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736210780968728e-05,
+      "loss": 0.9298,
+      "step": 2008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735838663893208e-05,
+      "loss": 1.0689,
+      "step": 2009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735466288050218e-05,
+      "loss": 0.4423,
+      "step": 2010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735093653449657e-05,
+      "loss": 1.0949,
+      "step": 2011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734720760101418e-05,
+      "loss": 1.0451,
+      "step": 2012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734347608015422e-05,
+      "loss": 1.02,
+      "step": 2013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733974197201583e-05,
+      "loss": 1.0454,
+      "step": 2014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973360052766983e-05,
+      "loss": 0.9746,
+      "step": 2015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973322659943009e-05,
+      "loss": 1.0394,
+      "step": 2016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732852412492303e-05,
+      "loss": 1.0505,
+      "step": 2017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732477966866413e-05,
+      "loss": 0.9514,
+      "step": 2018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973210326256238e-05,
+      "loss": 0.4521,
+      "step": 2019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731728299590152e-05,
+      "loss": 1.0259,
+      "step": 2020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731353077959707e-05,
+      "loss": 1.0167,
+      "step": 2021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730977597681005e-05,
+      "loss": 1.0024,
+      "step": 2022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730601858764035e-05,
+      "loss": 1.0486,
+      "step": 2023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730225861218782e-05,
+      "loss": 1.0865,
+      "step": 2024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729849605055237e-05,
+      "loss": 1.0202,
+      "step": 2025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97294730902834e-05,
+      "loss": 1.0139,
+      "step": 2026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729096316913284e-05,
+      "loss": 1.0842,
+      "step": 2027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728719284954896e-05,
+      "loss": 0.9452,
+      "step": 2028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728341994418263e-05,
+      "loss": 0.9735,
+      "step": 2029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972796444531341e-05,
+      "loss": 0.9694,
+      "step": 2030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727586637650373e-05,
+      "loss": 0.9969,
+      "step": 2031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727208571439192e-05,
+      "loss": 1.0103,
+      "step": 2032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726830246689917e-05,
+      "loss": 1.0459,
+      "step": 2033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726451663412604e-05,
+      "loss": 1.0273,
+      "step": 2034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726072821617315e-05,
+      "loss": 1.0595,
+      "step": 2035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972569372131412e-05,
+      "loss": 0.988,
+      "step": 2036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972531436251309e-05,
+      "loss": 1.0408,
+      "step": 2037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724934745224315e-05,
+      "loss": 0.9485,
+      "step": 2038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972455486945788e-05,
+      "loss": 0.9966,
+      "step": 2039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724174735223882e-05,
+      "loss": 1.0916,
+      "step": 2040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723794342532425e-05,
+      "loss": 1.0062,
+      "step": 2041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723413691393622e-05,
+      "loss": 0.9777,
+      "step": 2042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972303278181759e-05,
+      "loss": 1.0159,
+      "step": 2043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722651613814446e-05,
+      "loss": 1.107,
+      "step": 2044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972227018739433e-05,
+      "loss": 1.0773,
+      "step": 2045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721888502567374e-05,
+      "loss": 0.983,
+      "step": 2046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972150655934373e-05,
+      "loss": 1.0117,
+      "step": 2047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972112435773354e-05,
+      "loss": 1.0335,
+      "step": 2048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720741897746966e-05,
+      "loss": 1.0012,
+      "step": 2049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720359179394178e-05,
+      "loss": 1.1064,
+      "step": 2050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719976202685344e-05,
+      "loss": 1.0152,
+      "step": 2051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719592967630644e-05,
+      "loss": 1.0545,
+      "step": 2052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719209474240263e-05,
+      "loss": 1.061,
+      "step": 2053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718825722524396e-05,
+      "loss": 1.081,
+      "step": 2054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718441712493242e-05,
+      "loss": 1.0579,
+      "step": 2055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718057444157002e-05,
+      "loss": 0.9933,
+      "step": 2056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717672917525898e-05,
+      "loss": 0.9181,
+      "step": 2057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717288132610146e-05,
+      "loss": 1.0331,
+      "step": 2058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716903089419973e-05,
+      "loss": 1.1007,
+      "step": 2059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716517787965613e-05,
+      "loss": 0.8817,
+      "step": 2060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971613222825731e-05,
+      "loss": 1.0464,
+      "step": 2061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971574641030531e-05,
+      "loss": 1.0664,
+      "step": 2062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715360334119862e-05,
+      "loss": 0.9555,
+      "step": 2063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714973999711235e-05,
+      "loss": 1.0942,
+      "step": 2064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714587407089696e-05,
+      "loss": 0.9532,
+      "step": 2065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971420055626552e-05,
+      "loss": 1.0266,
+      "step": 2066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971381344724898e-05,
+      "loss": 0.4579,
+      "step": 2067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971342608005038e-05,
+      "loss": 1.1008,
+      "step": 2068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713038454680008e-05,
+      "loss": 0.4741,
+      "step": 2069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712650571148165e-05,
+      "loss": 1.0708,
+      "step": 2070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712262429465166e-05,
+      "loss": 0.9943,
+      "step": 2071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971187402964132e-05,
+      "loss": 0.4225,
+      "step": 2072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711485371686956e-05,
+      "loss": 0.9336,
+      "step": 2073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711096455612403e-05,
+      "loss": 1.0555,
+      "step": 2074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710707281427998e-05,
+      "loss": 1.0041,
+      "step": 2075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710317849144082e-05,
+      "loss": 0.9691,
+      "step": 2076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970992815877101e-05,
+      "loss": 0.9358,
+      "step": 2077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709538210319133e-05,
+      "loss": 0.9945,
+      "step": 2078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709148003798826e-05,
+      "loss": 1.0533,
+      "step": 2079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708757539220448e-05,
+      "loss": 1.0596,
+      "step": 2080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708366816594388e-05,
+      "loss": 1.0451,
+      "step": 2081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707975835931022e-05,
+      "loss": 0.4152,
+      "step": 2082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970758459724075e-05,
+      "loss": 1.0585,
+      "step": 2083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707193100533962e-05,
+      "loss": 1.1147,
+      "step": 2084
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970680134582107e-05,
+      "loss": 1.0169,
+      "step": 2085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706409333112487e-05,
+      "loss": 1.0426,
+      "step": 2086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706017062418628e-05,
+      "loss": 0.9747,
+      "step": 2087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970562453374992e-05,
+      "loss": 1.0161,
+      "step": 2088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97052317471168e-05,
+      "loss": 1.0252,
+      "step": 2089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97048387025297e-05,
+      "loss": 0.9821,
+      "step": 2090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704445399999072e-05,
+      "loss": 1.0879,
+      "step": 2091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970405183953537e-05,
+      "loss": 0.9879,
+      "step": 2092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703658021149055e-05,
+      "loss": 1.0466,
+      "step": 2093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970326394485059e-05,
+      "loss": 1.0173,
+      "step": 2094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702869610650452e-05,
+      "loss": 1.0191,
+      "step": 2095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970247501855912e-05,
+      "loss": 1.0633,
+      "step": 2096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702080168587086e-05,
+      "loss": 1.0243,
+      "step": 2097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970168506074484e-05,
+      "loss": 0.3626,
+      "step": 2098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701289695042887e-05,
+      "loss": 1.0296,
+      "step": 2099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700894071491736e-05,
+      "loss": 1.0006,
+      "step": 2100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700498190101895e-05,
+      "loss": 1.0309,
+      "step": 2101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700102050883895e-05,
+      "loss": 0.981,
+      "step": 2102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699705653848257e-05,
+      "loss": 0.9515,
+      "step": 2103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699308999005523e-05,
+      "loss": 0.964,
+      "step": 2104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698912086366232e-05,
+      "loss": 0.9445,
+      "step": 2105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698514915940937e-05,
+      "loss": 0.9891,
+      "step": 2106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698117487740192e-05,
+      "loss": 0.9699,
+      "step": 2107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697719801774562e-05,
+      "loss": 0.9212,
+      "step": 2108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697321858054614e-05,
+      "loss": 1.017,
+      "step": 2109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969692365659093e-05,
+      "loss": 1.0629,
+      "step": 2110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969652519739408e-05,
+      "loss": 0.9977,
+      "step": 2111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696126480474673e-05,
+      "loss": 1.0954,
+      "step": 2112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695727505843298e-05,
+      "loss": 1.0441,
+      "step": 2113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695328273510558e-05,
+      "loss": 1.0259,
+      "step": 2114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694928783487065e-05,
+      "loss": 1.0268,
+      "step": 2115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694529035783437e-05,
+      "loss": 1.0661,
+      "step": 2116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96941290304103e-05,
+      "loss": 1.0028,
+      "step": 2117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693728767378285e-05,
+      "loss": 1.0259,
+      "step": 2118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693328246698034e-05,
+      "loss": 1.0217,
+      "step": 2119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692927468380187e-05,
+      "loss": 0.951,
+      "step": 2120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692526432435395e-05,
+      "loss": 0.9808,
+      "step": 2121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692125138874324e-05,
+      "loss": 0.9585,
+      "step": 2122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691723587707633e-05,
+      "loss": 1.1208,
+      "step": 2123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691321778946003e-05,
+      "loss": 1.0091,
+      "step": 2124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690919712600104e-05,
+      "loss": 1.02,
+      "step": 2125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969051738868063e-05,
+      "loss": 0.973,
+      "step": 2126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690114807198273e-05,
+      "loss": 1.034,
+      "step": 2127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968971196816373e-05,
+      "loss": 1.0022,
+      "step": 2128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689308871587707e-05,
+      "loss": 0.9494,
+      "step": 2129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688905517480923e-05,
+      "loss": 0.9556,
+      "step": 2130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688501905854096e-05,
+      "loss": 0.9392,
+      "step": 2131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688098036717953e-05,
+      "loss": 0.9325,
+      "step": 2132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968769391008323e-05,
+      "loss": 0.9452,
+      "step": 2133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687289525960664e-05,
+      "loss": 0.9808,
+      "step": 2134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686884884361007e-05,
+      "loss": 0.9633,
+      "step": 2135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686479985295017e-05,
+      "loss": 1.0226,
+      "step": 2136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686074828773448e-05,
+      "loss": 0.9807,
+      "step": 2137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685669414807076e-05,
+      "loss": 1.0963,
+      "step": 2138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685263743406674e-05,
+      "loss": 1.002,
+      "step": 2139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968485781458302e-05,
+      "loss": 1.0706,
+      "step": 2140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684451628346906e-05,
+      "loss": 0.9537,
+      "step": 2141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968404518470913e-05,
+      "loss": 1.0066,
+      "step": 2142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683638483680495e-05,
+      "loss": 1.0642,
+      "step": 2143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683231525271805e-05,
+      "loss": 0.9992,
+      "step": 2144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682824309493882e-05,
+      "loss": 1.0573,
+      "step": 2145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682416836357547e-05,
+      "loss": 0.8886,
+      "step": 2146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682009105873633e-05,
+      "loss": 1.0888,
+      "step": 2147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968160111805297e-05,
+      "loss": 0.4449,
+      "step": 2148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681192872906414e-05,
+      "loss": 1.0436,
+      "step": 2149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680784370444807e-05,
+      "loss": 1.0281,
+      "step": 2150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680375610679005e-05,
+      "loss": 0.9353,
+      "step": 2151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679966593619878e-05,
+      "loss": 1.0479,
+      "step": 2152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967955731927829e-05,
+      "loss": 1.0398,
+      "step": 2153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679147787665128e-05,
+      "loss": 1.0455,
+      "step": 2154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967873799879127e-05,
+      "loss": 1.0416,
+      "step": 2155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967832795266761e-05,
+      "loss": 0.4127,
+      "step": 2156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967791764930505e-05,
+      "loss": 1.1457,
+      "step": 2157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677507088714485e-05,
+      "loss": 1.0007,
+      "step": 2158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677096270906843e-05,
+      "loss": 1.0365,
+      "step": 2159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676685195893028e-05,
+      "loss": 1.0111,
+      "step": 2160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676273863683978e-05,
+      "loss": 0.9732,
+      "step": 2161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675862274290612e-05,
+      "loss": 1.0211,
+      "step": 2162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675450427723884e-05,
+      "loss": 0.9947,
+      "step": 2163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967503832399473e-05,
+      "loss": 1.0017,
+      "step": 2164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674625963114115e-05,
+      "loss": 0.9914,
+      "step": 2165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674213345092984e-05,
+      "loss": 1.0322,
+      "step": 2166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673800469942313e-05,
+      "loss": 0.9985,
+      "step": 2167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673387337673074e-05,
+      "loss": 0.9471,
+      "step": 2168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967297394829625e-05,
+      "loss": 0.973,
+      "step": 2169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672560301822826e-05,
+      "loss": 0.9723,
+      "step": 2170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672146398263792e-05,
+      "loss": 1.0675,
+      "step": 2171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671732237630158e-05,
+      "loss": 0.9657,
+      "step": 2172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671317819932923e-05,
+      "loss": 0.9646,
+      "step": 2173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967090314518311e-05,
+      "loss": 1.0645,
+      "step": 2174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670488213391734e-05,
+      "loss": 1.0281,
+      "step": 2175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967007302456983e-05,
+      "loss": 1.0521,
+      "step": 2176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669657578728425e-05,
+      "loss": 0.9779,
+      "step": 2177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669241875878566e-05,
+      "loss": 0.9998,
+      "step": 2178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96688259160313e-05,
+      "loss": 0.9761,
+      "step": 2179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668409699197686e-05,
+      "loss": 1.0789,
+      "step": 2180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667993225388785e-05,
+      "loss": 0.9209,
+      "step": 2181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966757649461567e-05,
+      "loss": 1.0458,
+      "step": 2182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667159506889403e-05,
+      "loss": 0.9653,
+      "step": 2183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666742262221086e-05,
+      "loss": 0.9489,
+      "step": 2184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666324760621794e-05,
+      "loss": 1.0646,
+      "step": 2185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665907002102634e-05,
+      "loss": 1.0796,
+      "step": 2186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665488986674705e-05,
+      "loss": 1.0552,
+      "step": 2187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665070714349116e-05,
+      "loss": 1.0028,
+      "step": 2188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664652185136988e-05,
+      "loss": 1.1244,
+      "step": 2189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664233399049443e-05,
+      "loss": 1.0031,
+      "step": 2190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966381435609761e-05,
+      "loss": 0.4045,
+      "step": 2191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663395056292632e-05,
+      "loss": 0.9457,
+      "step": 2192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662975499645646e-05,
+      "loss": 0.9703,
+      "step": 2193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662555686167808e-05,
+      "loss": 0.9672,
+      "step": 2194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966213561587028e-05,
+      "loss": 0.9805,
+      "step": 2195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661715288764217e-05,
+      "loss": 0.9473,
+      "step": 2196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661294704860796e-05,
+      "loss": 0.9797,
+      "step": 2197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96608738641712e-05,
+      "loss": 1.0367,
+      "step": 2198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660452766706612e-05,
+      "loss": 1.0215,
+      "step": 2199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660031412478218e-05,
+      "loss": 1.0123,
+      "step": 2200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659609801497224e-05,
+      "loss": 1.0449,
+      "step": 2201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965918793377483e-05,
+      "loss": 0.3872,
+      "step": 2202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658765809322258e-05,
+      "loss": 1.0431,
+      "step": 2203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965834342815072e-05,
+      "loss": 1.0376,
+      "step": 2204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657920790271445e-05,
+      "loss": 1.1057,
+      "step": 2205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657497895695666e-05,
+      "loss": 1.1054,
+      "step": 2206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657074744434622e-05,
+      "loss": 1.1382,
+      "step": 2207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965665133649956e-05,
+      "loss": 1.0925,
+      "step": 2208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656227671901737e-05,
+      "loss": 0.9141,
+      "step": 2209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965580375065241e-05,
+      "loss": 1.092,
+      "step": 2210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655379572762846e-05,
+      "loss": 1.0856,
+      "step": 2211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965495513824432e-05,
+      "loss": 1.0046,
+      "step": 2212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654530447108115e-05,
+      "loss": 1.0538,
+      "step": 2213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654105499365515e-05,
+      "loss": 1.0337,
+      "step": 2214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965368029502782e-05,
+      "loss": 1.0039,
+      "step": 2215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653254834106324e-05,
+      "loss": 1.059,
+      "step": 2216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652829116612346e-05,
+      "loss": 0.9756,
+      "step": 2217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965240314255719e-05,
+      "loss": 1.1086,
+      "step": 2218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651976911952183e-05,
+      "loss": 1.0378,
+      "step": 2219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651550424808653e-05,
+      "loss": 0.9857,
+      "step": 2220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651123681137938e-05,
+      "loss": 1.0208,
+      "step": 2221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965069668095138e-05,
+      "loss": 0.9598,
+      "step": 2222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650269424260323e-05,
+      "loss": 1.0191,
+      "step": 2223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649841911076124e-05,
+      "loss": 1.0716,
+      "step": 2224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649414141410152e-05,
+      "loss": 0.9875,
+      "step": 2225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964898611527377e-05,
+      "loss": 1.1199,
+      "step": 2226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964855783267836e-05,
+      "loss": 0.9407,
+      "step": 2227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648129293635303e-05,
+      "loss": 1.019,
+      "step": 2228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647700498155985e-05,
+      "loss": 0.9945,
+      "step": 2229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647271446251812e-05,
+      "loss": 1.021,
+      "step": 2230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964684213793418e-05,
+      "loss": 0.8492,
+      "step": 2231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96464125732145e-05,
+      "loss": 1.0387,
+      "step": 2232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964598275210419e-05,
+      "loss": 1.0344,
+      "step": 2233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645552674614677e-05,
+      "loss": 1.0064,
+      "step": 2234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964512234075739e-05,
+      "loss": 1.0397,
+      "step": 2235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964469175054377e-05,
+      "loss": 1.0103,
+      "step": 2236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644260903985252e-05,
+      "loss": 0.9303,
+      "step": 2237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643829801093298e-05,
+      "loss": 0.9746,
+      "step": 2238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643398441879362e-05,
+      "loss": 0.9914,
+      "step": 2239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964296682635491e-05,
+      "loss": 1.0265,
+      "step": 2240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964253495453141e-05,
+      "loss": 0.9762,
+      "step": 2241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642102826420348e-05,
+      "loss": 0.4516,
+      "step": 2242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641670442033203e-05,
+      "loss": 1.0023,
+      "step": 2243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964123780138147e-05,
+      "loss": 0.9653,
+      "step": 2244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640804904476646e-05,
+      "loss": 1.0353,
+      "step": 2245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640371751330242e-05,
+      "loss": 0.9659,
+      "step": 2246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639938341953762e-05,
+      "loss": 1.0161,
+      "step": 2247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639504676358734e-05,
+      "loss": 1.0445,
+      "step": 2248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963907075455668e-05,
+      "loss": 1.0336,
+      "step": 2249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638636576559136e-05,
+      "loss": 1.0027,
+      "step": 2250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963820214237764e-05,
+      "loss": 0.9571,
+      "step": 2251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637767452023733e-05,
+      "loss": 0.9944,
+      "step": 2252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637332505508985e-05,
+      "loss": 1.0,
+      "step": 2253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636897302844938e-05,
+      "loss": 1.0631,
+      "step": 2254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636461844043172e-05,
+      "loss": 0.8045,
+      "step": 2255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636026129115252e-05,
+      "loss": 1.0075,
+      "step": 2256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635590158072765e-05,
+      "loss": 0.9637,
+      "step": 2257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635153930927294e-05,
+      "loss": 1.0555,
+      "step": 2258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634717447690438e-05,
+      "loss": 1.0943,
+      "step": 2259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634280708373797e-05,
+      "loss": 1.1135,
+      "step": 2260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633843712988976e-05,
+      "loss": 0.9944,
+      "step": 2261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633406461547593e-05,
+      "loss": 1.1122,
+      "step": 2262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632968954061268e-05,
+      "loss": 0.9338,
+      "step": 2263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963253119054163e-05,
+      "loss": 0.9648,
+      "step": 2264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632093171000318e-05,
+      "loss": 0.9325,
+      "step": 2265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963165489544897e-05,
+      "loss": 0.955,
+      "step": 2266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631216363899232e-05,
+      "loss": 1.0943,
+      "step": 2267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963077757636276e-05,
+      "loss": 0.4571,
+      "step": 2268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630338532851227e-05,
+      "loss": 1.047,
+      "step": 2269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962989923337629e-05,
+      "loss": 1.0057,
+      "step": 2270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962945967794963e-05,
+      "loss": 1.0184,
+      "step": 2271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629019866582932e-05,
+      "loss": 0.9977,
+      "step": 2272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962857979928788e-05,
+      "loss": 0.9827,
+      "step": 2273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628139476076175e-05,
+      "loss": 1.1147,
+      "step": 2274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962769889695952e-05,
+      "loss": 1.0042,
+      "step": 2275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627258061949624e-05,
+      "loss": 1.0618,
+      "step": 2276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626816971058205e-05,
+      "loss": 1.0293,
+      "step": 2277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626375624296984e-05,
+      "loss": 0.9467,
+      "step": 2278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625934021677695e-05,
+      "loss": 0.9775,
+      "step": 2279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625492163212073e-05,
+      "loss": 1.108,
+      "step": 2280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962505004891186e-05,
+      "loss": 0.9548,
+      "step": 2281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624607678788813e-05,
+      "loss": 1.0368,
+      "step": 2282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624165052854685e-05,
+      "loss": 1.0723,
+      "step": 2283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962372217112124e-05,
+      "loss": 0.9347,
+      "step": 2284
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623279033600257e-05,
+      "loss": 1.0491,
+      "step": 2285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622835640303505e-05,
+      "loss": 0.9473,
+      "step": 2286
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962239199124277e-05,
+      "loss": 0.9395,
+      "step": 2287
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621948086429847e-05,
+      "loss": 1.0148,
+      "step": 2288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962150392587653e-05,
+      "loss": 0.9771,
+      "step": 2289
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962105950959463e-05,
+      "loss": 0.9953,
+      "step": 2290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620614837595958e-05,
+      "loss": 1.0065,
+      "step": 2291
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962016990989233e-05,
+      "loss": 0.9206,
+      "step": 2292
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961972472649557e-05,
+      "loss": 1.0522,
+      "step": 2293
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961927928741752e-05,
+      "loss": 1.0506,
+      "step": 2294
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961883359267001e-05,
+      "loss": 1.0619,
+      "step": 2295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961838764226489e-05,
+      "loss": 1.0841,
+      "step": 2296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961794143621401e-05,
+      "loss": 1.0239,
+      "step": 2297
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961749497452923e-05,
+      "loss": 1.0847,
+      "step": 2298
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961704825722242e-05,
+      "loss": 0.9748,
+      "step": 2299
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961660128430545e-05,
+      "loss": 1.0261,
+      "step": 2300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616154055790206e-05,
+      "loss": 1.0089,
+      "step": 2301
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615706571688563e-05,
+      "loss": 0.9002,
+      "step": 2302
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615258832012424e-05,
+      "loss": 1.0308,
+      "step": 2303
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961481083677369e-05,
+      "loss": 0.4259,
+      "step": 2304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961436258598426e-05,
+      "loss": 1.0307,
+      "step": 2305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613914079656053e-05,
+      "loss": 1.0464,
+      "step": 2306
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613465317800994e-05,
+      "loss": 0.4659,
+      "step": 2307
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613016300431003e-05,
+      "loss": 0.9896,
+      "step": 2308
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961256702755802e-05,
+      "loss": 0.9268,
+      "step": 2309
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612117499193983e-05,
+      "loss": 1.085,
+      "step": 2310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961166771535084e-05,
+      "loss": 1.0082,
+      "step": 2311
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961121767604055e-05,
+      "loss": 1.0245,
+      "step": 2312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961076738127507e-05,
+      "loss": 1.0016,
+      "step": 2313
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961031683106637e-05,
+      "loss": 1.0351,
+      "step": 2314
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609866025426422e-05,
+      "loss": 1.0242,
+      "step": 2315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960941496436721e-05,
+      "loss": 0.4674,
+      "step": 2316
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608963647900726e-05,
+      "loss": 1.0434,
+      "step": 2317
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608512076038964e-05,
+      "loss": 1.0239,
+      "step": 2318
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608060248793922e-05,
+      "loss": 1.0023,
+      "step": 2319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607608166177615e-05,
+      "loss": 0.4018,
+      "step": 2320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607155828202054e-05,
+      "loss": 1.0743,
+      "step": 2321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606703234879263e-05,
+      "loss": 0.9778,
+      "step": 2322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606250386221273e-05,
+      "loss": 0.9403,
+      "step": 2323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960579728224012e-05,
+      "loss": 1.0224,
+      "step": 2324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605343922947843e-05,
+      "loss": 0.3928,
+      "step": 2325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604890308356495e-05,
+      "loss": 0.9952,
+      "step": 2326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604436438478135e-05,
+      "loss": 1.0866,
+      "step": 2327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603982313324824e-05,
+      "loss": 0.9159,
+      "step": 2328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603527932908627e-05,
+      "loss": 0.9822,
+      "step": 2329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960307329724163e-05,
+      "loss": 0.9949,
+      "step": 2330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602618406335915e-05,
+      "loss": 1.1391,
+      "step": 2331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602163260203567e-05,
+      "loss": 0.9767,
+      "step": 2332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601707858856685e-05,
+      "loss": 1.0044,
+      "step": 2333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601252202307375e-05,
+      "loss": 1.033,
+      "step": 2334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600796290567747e-05,
+      "loss": 1.0583,
+      "step": 2335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600340123649923e-05,
+      "loss": 0.415,
+      "step": 2336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599883701566018e-05,
+      "loss": 0.919,
+      "step": 2337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959942702432817e-05,
+      "loss": 1.0091,
+      "step": 2338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598970091948514e-05,
+      "loss": 0.429,
+      "step": 2339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598512904439197e-05,
+      "loss": 1.1326,
+      "step": 2340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959805546181237e-05,
+      "loss": 1.0499,
+      "step": 2341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959759776408019e-05,
+      "loss": 0.9807,
+      "step": 2342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597139811254827e-05,
+      "loss": 0.9547,
+      "step": 2343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596681603348446e-05,
+      "loss": 1.0546,
+      "step": 2344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596223140373224e-05,
+      "loss": 1.06,
+      "step": 2345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595764422341356e-05,
+      "loss": 1.0252,
+      "step": 2346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595305449265027e-05,
+      "loss": 1.0536,
+      "step": 2347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959484622115644e-05,
+      "loss": 1.0759,
+      "step": 2348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95943867380278e-05,
+      "loss": 1.0765,
+      "step": 2349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593926999891315e-05,
+      "loss": 1.02,
+      "step": 2350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959346700675921e-05,
+      "loss": 0.4278,
+      "step": 2351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593006758643707e-05,
+      "loss": 1.0047,
+      "step": 2352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592546255557038e-05,
+      "loss": 0.9331,
+      "step": 2353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959208549751145e-05,
+      "loss": 1.0716,
+      "step": 2354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591624484519183e-05,
+      "loss": 1.0082,
+      "step": 2355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959116321659249e-05,
+      "loss": 1.0816,
+      "step": 2356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590701693743638e-05,
+      "loss": 1.05,
+      "step": 2357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590239915984886e-05,
+      "loss": 1.0227,
+      "step": 2358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589777883328506e-05,
+      "loss": 0.9923,
+      "step": 2359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958931559578679e-05,
+      "loss": 1.0135,
+      "step": 2360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958885305337201e-05,
+      "loss": 0.9946,
+      "step": 2361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958839025609647e-05,
+      "loss": 0.9488,
+      "step": 2362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958792720397247e-05,
+      "loss": 1.0381,
+      "step": 2363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587463897012317e-05,
+      "loss": 1.0922,
+      "step": 2364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587000335228316e-05,
+      "loss": 1.0517,
+      "step": 2365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586536518632803e-05,
+      "loss": 1.0141,
+      "step": 2366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586072447238094e-05,
+      "loss": 1.0021,
+      "step": 2367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958560812105653e-05,
+      "loss": 0.9746,
+      "step": 2368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958514354010045e-05,
+      "loss": 0.3988,
+      "step": 2369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95846787043822e-05,
+      "loss": 1.086,
+      "step": 2370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958421361391414e-05,
+      "loss": 0.9953,
+      "step": 2371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583748268708624e-05,
+      "loss": 1.0775,
+      "step": 2372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583282668778025e-05,
+      "loss": 0.9851,
+      "step": 2373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958281681413472e-05,
+      "loss": 1.0284,
+      "step": 2374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958235070479109e-05,
+      "loss": 1.1259,
+      "step": 2375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958188434075952e-05,
+      "loss": 1.0136,
+      "step": 2376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958141772205241e-05,
+      "loss": 0.9637,
+      "step": 2377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958095084868216e-05,
+      "loss": 0.9923,
+      "step": 2378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580483720661175e-05,
+      "loss": 1.0217,
+      "step": 2379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580016338001877e-05,
+      "loss": 0.9316,
+      "step": 2380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579548700716685e-05,
+      "loss": 1.0102,
+      "step": 2381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579080808818035e-05,
+      "loss": 1.0168,
+      "step": 2382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957861266231835e-05,
+      "loss": 1.0553,
+      "step": 2383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578144261230086e-05,
+      "loss": 0.914,
+      "step": 2384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577675605565683e-05,
+      "loss": 1.0228,
+      "step": 2385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577206695337603e-05,
+      "loss": 1.0179,
+      "step": 2386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576737530558304e-05,
+      "loss": 0.9566,
+      "step": 2387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957626811124026e-05,
+      "loss": 0.9737,
+      "step": 2388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575798437395953e-05,
+      "loss": 0.8801,
+      "step": 2389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575328509037855e-05,
+      "loss": 0.9236,
+      "step": 2390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574858326178462e-05,
+      "loss": 1.117,
+      "step": 2391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574387888830267e-05,
+      "loss": 0.9055,
+      "step": 2392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957391719700578e-05,
+      "loss": 0.9927,
+      "step": 2393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957344625071751e-05,
+      "loss": 1.0017,
+      "step": 2394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957297504997797e-05,
+      "loss": 1.0219,
+      "step": 2395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957250359479969e-05,
+      "loss": 1.0252,
+      "step": 2396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572031885195197e-05,
+      "loss": 0.9914,
+      "step": 2397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571559921177026e-05,
+      "loss": 1.0009,
+      "step": 2398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571087702757725e-05,
+      "loss": 1.0474,
+      "step": 2399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570615229949844e-05,
+      "loss": 0.9973,
+      "step": 2400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957014250276594e-05,
+      "loss": 0.9721,
+      "step": 2401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956966952121858e-05,
+      "loss": 1.0357,
+      "step": 2402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569196285320332e-05,
+      "loss": 1.0294,
+      "step": 2403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568722795083778e-05,
+      "loss": 0.9876,
+      "step": 2404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568249050521498e-05,
+      "loss": 1.0227,
+      "step": 2405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567775051646087e-05,
+      "loss": 1.0514,
+      "step": 2406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567300798470144e-05,
+      "loss": 1.0573,
+      "step": 2407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956682629100627e-05,
+      "loss": 1.0272,
+      "step": 2408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566351529267082e-05,
+      "loss": 1.0745,
+      "step": 2409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565876513265198e-05,
+      "loss": 0.9864,
+      "step": 2410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565401243013234e-05,
+      "loss": 0.9749,
+      "step": 2411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564925718523837e-05,
+      "loss": 1.0373,
+      "step": 2412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564449939809638e-05,
+      "loss": 1.0079,
+      "step": 2413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956397390688328e-05,
+      "loss": 1.052,
+      "step": 2414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956349761975742e-05,
+      "loss": 0.9695,
+      "step": 2415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563021078444717e-05,
+      "loss": 0.9955,
+      "step": 2416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562544282957832e-05,
+      "loss": 1.018,
+      "step": 2417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956206723330945e-05,
+      "loss": 0.9965,
+      "step": 2418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561589929512234e-05,
+      "loss": 1.0354,
+      "step": 2419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956111237157888e-05,
+      "loss": 0.9738,
+      "step": 2420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956063455952208e-05,
+      "loss": 1.0479,
+      "step": 2421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956015649335453e-05,
+      "loss": 0.9816,
+      "step": 2422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955967817308894e-05,
+      "loss": 1.0699,
+      "step": 2423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559199598738027e-05,
+      "loss": 0.997,
+      "step": 2424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95587207703145e-05,
+      "loss": 0.9721,
+      "step": 2425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558241687831096e-05,
+      "loss": 0.9419,
+      "step": 2426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557762351300546e-05,
+      "loss": 0.9447,
+      "step": 2427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557282760735585e-05,
+      "loss": 1.0456,
+      "step": 2428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556802916148963e-05,
+      "loss": 1.0374,
+      "step": 2429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556322817553437e-05,
+      "loss": 0.9885,
+      "step": 2430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555842464961763e-05,
+      "loss": 0.9597,
+      "step": 2431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955536185838671e-05,
+      "loss": 1.0095,
+      "step": 2432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554880997841055e-05,
+      "loss": 0.9991,
+      "step": 2433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554399883337573e-05,
+      "loss": 0.4191,
+      "step": 2434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553918514889054e-05,
+      "loss": 1.0176,
+      "step": 2435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553436892508296e-05,
+      "loss": 0.9522,
+      "step": 2436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955295501620809e-05,
+      "loss": 0.9557,
+      "step": 2437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552472886001257e-05,
+      "loss": 0.9867,
+      "step": 2438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551990501900602e-05,
+      "loss": 0.3885,
+      "step": 2439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955150786391895e-05,
+      "loss": 1.0469,
+      "step": 2440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551024972069127e-05,
+      "loss": 1.0231,
+      "step": 2441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550541826363965e-05,
+      "loss": 0.9918,
+      "step": 2442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550058426816314e-05,
+      "loss": 1.0255,
+      "step": 2443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549574773439017e-05,
+      "loss": 1.0282,
+      "step": 2444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549090866244927e-05,
+      "loss": 0.954,
+      "step": 2445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548606705246907e-05,
+      "loss": 0.4208,
+      "step": 2446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954812229045783e-05,
+      "loss": 1.0114,
+      "step": 2447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547637621890563e-05,
+      "loss": 0.9837,
+      "step": 2448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547152699557996e-05,
+      "loss": 1.0537,
+      "step": 2449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546667523473012e-05,
+      "loss": 1.0556,
+      "step": 2450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954618209364851e-05,
+      "loss": 0.9919,
+      "step": 2451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954569641009739e-05,
+      "loss": 0.9423,
+      "step": 2452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545210472832563e-05,
+      "loss": 0.9548,
+      "step": 2453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544724281866942e-05,
+      "loss": 0.9793,
+      "step": 2454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954423783721345e-05,
+      "loss": 1.1646,
+      "step": 2455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543751138885018e-05,
+      "loss": 1.132,
+      "step": 2456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543264186894582e-05,
+      "loss": 0.9999,
+      "step": 2457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542776981255082e-05,
+      "loss": 1.0587,
+      "step": 2458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954228952197947e-05,
+      "loss": 1.0757,
+      "step": 2459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541801809080697e-05,
+      "loss": 0.9995,
+      "step": 2460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954131384257173e-05,
+      "loss": 0.9766,
+      "step": 2461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540825622465546e-05,
+      "loss": 1.0419,
+      "step": 2462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540337148775108e-05,
+      "loss": 0.9547,
+      "step": 2463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95398484215134e-05,
+      "loss": 1.0157,
+      "step": 2464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539359440693423e-05,
+      "loss": 1.0473,
+      "step": 2465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538870206328168e-05,
+      "loss": 0.9462,
+      "step": 2466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538380718430633e-05,
+      "loss": 1.0462,
+      "step": 2467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537890977013833e-05,
+      "loss": 0.9692,
+      "step": 2468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537400982090786e-05,
+      "loss": 1.025,
+      "step": 2469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536910733674514e-05,
+      "loss": 1.0236,
+      "step": 2470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536420231778046e-05,
+      "loss": 1.0149,
+      "step": 2471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953592947641442e-05,
+      "loss": 1.0648,
+      "step": 2472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535438467596678e-05,
+      "loss": 1.0164,
+      "step": 2473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534947205337873e-05,
+      "loss": 1.0172,
+      "step": 2474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534455689651062e-05,
+      "loss": 0.9863,
+      "step": 2475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533963920549307e-05,
+      "loss": 0.9877,
+      "step": 2476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533471898045682e-05,
+      "loss": 0.9782,
+      "step": 2477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953297962215326e-05,
+      "loss": 1.0697,
+      "step": 2478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953248709288513e-05,
+      "loss": 1.0073,
+      "step": 2479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531994310254374e-05,
+      "loss": 1.0588,
+      "step": 2480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95315012742741e-05,
+      "loss": 0.952,
+      "step": 2481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531007984957408e-05,
+      "loss": 1.0346,
+      "step": 2482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953051444231741e-05,
+      "loss": 0.9808,
+      "step": 2483
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530020646367226e-05,
+      "loss": 1.0356,
+      "step": 2484
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529526597119974e-05,
+      "loss": 0.9276,
+      "step": 2485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952903229458879e-05,
+      "loss": 1.0171,
+      "step": 2486
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528537738786804e-05,
+      "loss": 0.9802,
+      "step": 2487
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528042929727176e-05,
+      "loss": 1.1119,
+      "step": 2488
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527547867423044e-05,
+      "loss": 1.0553,
+      "step": 2489
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527052551887576e-05,
+      "loss": 1.0107,
+      "step": 2490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952655698313393e-05,
+      "loss": 1.0699,
+      "step": 2491
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526061161175275e-05,
+      "loss": 1.0388,
+      "step": 2492
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525565086024804e-05,
+      "loss": 1.0323,
+      "step": 2493
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952506875769568e-05,
+      "loss": 1.0326,
+      "step": 2494
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524572176201118e-05,
+      "loss": 0.9546,
+      "step": 2495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95240753415543e-05,
+      "loss": 0.9833,
+      "step": 2496
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952357825376844e-05,
+      "loss": 1.0051,
+      "step": 2497
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523080912856743e-05,
+      "loss": 0.4626,
+      "step": 2498
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522583318832436e-05,
+      "loss": 1.0505,
+      "step": 2499
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522085471708738e-05,
+      "loss": 1.0391,
+      "step": 2500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521587371498883e-05,
+      "loss": 0.8402,
+      "step": 2501
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952108901821611e-05,
+      "loss": 1.0761,
+      "step": 2502
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520590411873667e-05,
+      "loss": 0.9556,
+      "step": 2503
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520091552484804e-05,
+      "loss": 1.0097,
+      "step": 2504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951959244006278e-05,
+      "loss": 1.0273,
+      "step": 2505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951909307462086e-05,
+      "loss": 0.9885,
+      "step": 2506
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518593456172322e-05,
+      "loss": 1.087,
+      "step": 2507
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518093584730438e-05,
+      "loss": 1.068,
+      "step": 2508
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517593460308493e-05,
+      "loss": 1.0616,
+      "step": 2509
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951709308291979e-05,
+      "loss": 0.8806,
+      "step": 2510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516592452577622e-05,
+      "loss": 0.8266,
+      "step": 2511
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516091569295296e-05,
+      "loss": 1.0041,
+      "step": 2512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515590433086122e-05,
+      "loss": 1.0685,
+      "step": 2513
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515089043963424e-05,
+      "loss": 1.0195,
+      "step": 2514
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514587401940524e-05,
+      "loss": 0.9866,
+      "step": 2515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951408550703076e-05,
+      "loss": 1.0657,
+      "step": 2516
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951358335924747e-05,
+      "loss": 0.9192,
+      "step": 2517
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513080958603998e-05,
+      "loss": 0.9587,
+      "step": 2518
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512578305113702e-05,
+      "loss": 0.4586,
+      "step": 2519
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512075398789933e-05,
+      "loss": 1.0091,
+      "step": 2520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951157223964607e-05,
+      "loss": 1.0439,
+      "step": 2521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511068827695476e-05,
+      "loss": 1.0131,
+      "step": 2522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510565162951538e-05,
+      "loss": 0.9951,
+      "step": 2523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951006124542764e-05,
+      "loss": 1.0527,
+      "step": 2524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950955707513717e-05,
+      "loss": 1.0843,
+      "step": 2525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950905265209354e-05,
+      "loss": 1.0657,
+      "step": 2526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950854797631015e-05,
+      "loss": 0.9483,
+      "step": 2527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508043047800415e-05,
+      "loss": 0.9818,
+      "step": 2528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507537866577752e-05,
+      "loss": 0.9906,
+      "step": 2529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507032432655592e-05,
+      "loss": 0.9625,
+      "step": 2530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950652674604737e-05,
+      "loss": 1.0952,
+      "step": 2531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506020806766526e-05,
+      "loss": 0.9461,
+      "step": 2532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505514614826505e-05,
+      "loss": 1.061,
+      "step": 2533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505008170240757e-05,
+      "loss": 1.1228,
+      "step": 2534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950450147302275e-05,
+      "loss": 1.0455,
+      "step": 2535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950399452318595e-05,
+      "loss": 1.0082,
+      "step": 2536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950348732074383e-05,
+      "loss": 1.0289,
+      "step": 2537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502979865709873e-05,
+      "loss": 0.9987,
+      "step": 2538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502472158097563e-05,
+      "loss": 1.0092,
+      "step": 2539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501964197920394e-05,
+      "loss": 1.043,
+      "step": 2540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501455985191874e-05,
+      "loss": 1.0066,
+      "step": 2541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500947519925503e-05,
+      "loss": 0.9502,
+      "step": 2542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500438802134795e-05,
+      "loss": 1.012,
+      "step": 2543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949992983183328e-05,
+      "loss": 1.009,
+      "step": 2544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499420609034473e-05,
+      "loss": 1.0065,
+      "step": 2545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498911133751917e-05,
+      "loss": 1.0366,
+      "step": 2546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498401405999154e-05,
+      "loss": 0.9725,
+      "step": 2547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949789142578973e-05,
+      "loss": 1.0681,
+      "step": 2548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94973811931372e-05,
+      "loss": 0.9194,
+      "step": 2549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496870708055123e-05,
+      "loss": 1.0242,
+      "step": 2550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496359970557073e-05,
+      "loss": 1.0252,
+      "step": 2551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495848980656618e-05,
+      "loss": 1.0504,
+      "step": 2552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495337738367342e-05,
+      "loss": 0.9794,
+      "step": 2553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494826243702834e-05,
+      "loss": 1.0671,
+      "step": 2554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494314496676687e-05,
+      "loss": 0.9624,
+      "step": 2555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493802497302508e-05,
+      "loss": 0.9852,
+      "step": 2556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493290245593897e-05,
+      "loss": 1.0031,
+      "step": 2557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949277774156448e-05,
+      "loss": 1.0527,
+      "step": 2558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492264985227868e-05,
+      "loss": 1.0359,
+      "step": 2559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491751976597693e-05,
+      "loss": 0.4651,
+      "step": 2560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949123871568759e-05,
+      "loss": 0.9586,
+      "step": 2561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490725202511207e-05,
+      "loss": 0.9909,
+      "step": 2562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490211437082186e-05,
+      "loss": 1.0779,
+      "step": 2563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948969741941418e-05,
+      "loss": 1.028,
+      "step": 2564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948918314952086e-05,
+      "loss": 1.0001,
+      "step": 2565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948866862741589e-05,
+      "loss": 0.9366,
+      "step": 2566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488153853112942e-05,
+      "loss": 1.0158,
+      "step": 2567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487638826625698e-05,
+      "loss": 0.9733,
+      "step": 2568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487123547967855e-05,
+      "loss": 0.9998,
+      "step": 2569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94866080171531e-05,
+      "loss": 0.9694,
+      "step": 2570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948609223419514e-05,
+      "loss": 1.0198,
+      "step": 2571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485576199107684e-05,
+      "loss": 0.9965,
+      "step": 2572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485059911904448e-05,
+      "loss": 0.9576,
+      "step": 2573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948454337259915e-05,
+      "loss": 1.0941,
+      "step": 2574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948402658120552e-05,
+      "loss": 0.397,
+      "step": 2575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483509537737297e-05,
+      "loss": 0.9807,
+      "step": 2576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482992242208222e-05,
+      "loss": 1.0026,
+      "step": 2577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482474694632043e-05,
+      "loss": 1.0092,
+      "step": 2578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481956895022514e-05,
+      "loss": 1.0798,
+      "step": 2579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481438843393406e-05,
+      "loss": 1.0265,
+      "step": 2580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480920539758478e-05,
+      "loss": 0.4358,
+      "step": 2581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948040198413151e-05,
+      "loss": 0.9603,
+      "step": 2582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479883176526287e-05,
+      "loss": 1.0982,
+      "step": 2583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479364116956595e-05,
+      "loss": 1.0575,
+      "step": 2584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478844805436234e-05,
+      "loss": 0.9958,
+      "step": 2585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478325241979e-05,
+      "loss": 1.1148,
+      "step": 2586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477805426598707e-05,
+      "loss": 1.0507,
+      "step": 2587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477285359309174e-05,
+      "loss": 0.4221,
+      "step": 2588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476765040124216e-05,
+      "loss": 0.9664,
+      "step": 2589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947624446905767e-05,
+      "loss": 0.9398,
+      "step": 2590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475723646123366e-05,
+      "loss": 1.0276,
+      "step": 2591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475202571335156e-05,
+      "loss": 0.9896,
+      "step": 2592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474681244706876e-05,
+      "loss": 1.0327,
+      "step": 2593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947415966625239e-05,
+      "loss": 0.93,
+      "step": 2594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473637835985565e-05,
+      "loss": 0.9492,
+      "step": 2595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473115753920268e-05,
+      "loss": 1.0612,
+      "step": 2596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472593420070366e-05,
+      "loss": 1.0648,
+      "step": 2597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472070834449755e-05,
+      "loss": 1.0219,
+      "step": 2598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471547997072318e-05,
+      "loss": 1.0386,
+      "step": 2599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471024907951953e-05,
+      "loss": 1.0093,
+      "step": 2600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947050156710256e-05,
+      "loss": 1.0197,
+      "step": 2601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469977974538056e-05,
+      "loss": 0.9877,
+      "step": 2602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946945413027235e-05,
+      "loss": 1.0403,
+      "step": 2603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946893003431937e-05,
+      "loss": 1.0172,
+      "step": 2604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468405686693044e-05,
+      "loss": 0.416,
+      "step": 2605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946788108740731e-05,
+      "loss": 1.0286,
+      "step": 2606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946735623647611e-05,
+      "loss": 1.0115,
+      "step": 2607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946683113391339e-05,
+      "loss": 0.969,
+      "step": 2608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466305779733114e-05,
+      "loss": 1.0814,
+      "step": 2609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465780173949242e-05,
+      "loss": 1.0368,
+      "step": 2610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465254316575744e-05,
+      "loss": 1.0036,
+      "step": 2611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464728207626597e-05,
+      "loss": 1.0194,
+      "step": 2612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946420184711578e-05,
+      "loss": 0.9413,
+      "step": 2613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946367523505729e-05,
+      "loss": 0.9946,
+      "step": 2614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946314837146512e-05,
+      "loss": 0.9864,
+      "step": 2615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462621256353274e-05,
+      "loss": 0.9169,
+      "step": 2616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462093889735766e-05,
+      "loss": 0.9712,
+      "step": 2617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461566271626607e-05,
+      "loss": 0.9629,
+      "step": 2618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461038402039824e-05,
+      "loss": 1.0547,
+      "step": 2619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460510280989443e-05,
+      "loss": 1.0335,
+      "step": 2620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459981908489505e-05,
+      "loss": 1.0052,
+      "step": 2621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945945328455405e-05,
+      "loss": 1.0002,
+      "step": 2622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458924409197134e-05,
+      "loss": 1.0196,
+      "step": 2623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458395282432805e-05,
+      "loss": 1.0288,
+      "step": 2624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945786590427514e-05,
+      "loss": 0.9789,
+      "step": 2625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457336274738195e-05,
+      "loss": 1.0102,
+      "step": 2626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456806393836056e-05,
+      "loss": 1.0318,
+      "step": 2627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94562762615828e-05,
+      "loss": 1.0503,
+      "step": 2628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455745877992527e-05,
+      "loss": 1.0465,
+      "step": 2629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945521524307932e-05,
+      "loss": 1.0431,
+      "step": 2630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94546843568573e-05,
+      "loss": 0.3814,
+      "step": 2631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454153219340563e-05,
+      "loss": 0.8803,
+      "step": 2632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453621830543232e-05,
+      "loss": 1.0858,
+      "step": 2633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945309019047943e-05,
+      "loss": 0.9751,
+      "step": 2634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452558299163285e-05,
+      "loss": 1.0422,
+      "step": 2635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452026156608935e-05,
+      "loss": 0.9453,
+      "step": 2636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9451493762830533e-05,
+      "loss": 0.9582,
+      "step": 2637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450961117842216e-05,
+      "loss": 0.517,
+      "step": 2638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450428221658147e-05,
+      "loss": 0.9625,
+      "step": 2639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449895074292485e-05,
+      "loss": 1.0667,
+      "step": 2640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944936167575941e-05,
+      "loss": 1.047,
+      "step": 2641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448828026073093e-05,
+      "loss": 0.9735,
+      "step": 2642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448294125247718e-05,
+      "loss": 1.0011,
+      "step": 2643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447759973297477e-05,
+      "loss": 0.9618,
+      "step": 2644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447225570236567e-05,
+      "loss": 0.3506,
+      "step": 2645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944669091607919e-05,
+      "loss": 1.0588,
+      "step": 2646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944615601083956e-05,
+      "loss": 0.9195,
+      "step": 2647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445620854531886e-05,
+      "loss": 1.0088,
+      "step": 2648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94450854471704e-05,
+      "loss": 1.0838,
+      "step": 2649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444549788769333e-05,
+      "loss": 1.0691,
+      "step": 2650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444013879342915e-05,
+      "loss": 0.9833,
+      "step": 2651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94434777189054e-05,
+      "loss": 0.9382,
+      "step": 2652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442941307471027e-05,
+      "loss": 1.0459,
+      "step": 2653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944240464505406e-05,
+      "loss": 1.016,
+      "step": 2654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441867731668768e-05,
+      "loss": 0.9996,
+      "step": 2655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441330567329405e-05,
+      "loss": 0.9586,
+      "step": 2656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440793152050267e-05,
+      "loss": 1.0327,
+      "step": 2657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440255485845628e-05,
+      "loss": 0.9692,
+      "step": 2658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439717568729775e-05,
+      "loss": 0.9509,
+      "step": 2659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439179400717014e-05,
+      "loss": 0.9998,
+      "step": 2660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438640981821646e-05,
+      "loss": 0.9812,
+      "step": 2661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438102312057978e-05,
+      "loss": 1.0372,
+      "step": 2662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943756339144033e-05,
+      "loss": 1.0253,
+      "step": 2663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943702421998303e-05,
+      "loss": 1.0165,
+      "step": 2664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94364847977004e-05,
+      "loss": 0.9596,
+      "step": 2665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435945124606784e-05,
+      "loss": 0.97,
+      "step": 2666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435405200716523e-05,
+      "loss": 1.0469,
+      "step": 2667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434865026043966e-05,
+      "loss": 1.0281,
+      "step": 2668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434324600603474e-05,
+      "loss": 1.0289,
+      "step": 2669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433783924409407e-05,
+      "loss": 1.0183,
+      "step": 2670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943324299747614e-05,
+      "loss": 1.0881,
+      "step": 2671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943270181981805e-05,
+      "loss": 1.0143,
+      "step": 2672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432160391449515e-05,
+      "loss": 0.9933,
+      "step": 2673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943161871238493e-05,
+      "loss": 1.0225,
+      "step": 2674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9431076782638693e-05,
+      "loss": 0.9754,
+      "step": 2675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430534602225205e-05,
+      "loss": 1.015,
+      "step": 2676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942999217115888e-05,
+      "loss": 0.972,
+      "step": 2677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942944948945413e-05,
+      "loss": 1.0148,
+      "step": 2678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428906557125383e-05,
+      "loss": 0.9771,
+      "step": 2679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942836337418707e-05,
+      "loss": 0.9623,
+      "step": 2680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9427819940653628e-05,
+      "loss": 1.0297,
+      "step": 2681
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9427276256539497e-05,
+      "loss": 1.0104,
+      "step": 2682
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942673232185913e-05,
+      "loss": 1.1443,
+      "step": 2683
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426188136626988e-05,
+      "loss": 1.0696,
+      "step": 2684
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425643700857524e-05,
+      "loss": 1.1005,
+      "step": 2685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942509901456522e-05,
+      "loss": 1.0206,
+      "step": 2686
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424554077764548e-05,
+      "loss": 0.9668,
+      "step": 2687
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424008890469994e-05,
+      "loss": 1.0722,
+      "step": 2688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423463452696044e-05,
+      "loss": 0.9813,
+      "step": 2689
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94229177644572e-05,
+      "loss": 0.9769,
+      "step": 2690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422371825767964e-05,
+      "loss": 0.9308,
+      "step": 2691
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421825636642848e-05,
+      "loss": 0.9728,
+      "step": 2692
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421279197096363e-05,
+      "loss": 1.0188,
+      "step": 2693
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942073250714304e-05,
+      "loss": 0.4044,
+      "step": 2694
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420185566797406e-05,
+      "loss": 1.0683,
+      "step": 2695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419638376074e-05,
+      "loss": 1.0436,
+      "step": 2696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941909093498736e-05,
+      "loss": 1.0097,
+      "step": 2697
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418543243552043e-05,
+      "loss": 1.116,
+      "step": 2698
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417995301782605e-05,
+      "loss": 1.0068,
+      "step": 2699
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417447109693606e-05,
+      "loss": 0.9927,
+      "step": 2700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941689866729962e-05,
+      "loss": 1.0515,
+      "step": 2701
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941634997461522e-05,
+      "loss": 0.9899,
+      "step": 2702
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415801031654994e-05,
+      "loss": 1.0246,
+      "step": 2703
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941525183843353e-05,
+      "loss": 1.0199,
+      "step": 2704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414702394965427e-05,
+      "loss": 1.0319,
+      "step": 2705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941415270126529e-05,
+      "loss": 0.3952,
+      "step": 2706
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413602757347722e-05,
+      "loss": 1.0086,
+      "step": 2707
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413052563227342e-05,
+      "loss": 1.0278,
+      "step": 2708
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412502118918777e-05,
+      "loss": 1.0454,
+      "step": 2709
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411951424436655e-05,
+      "loss": 1.0288,
+      "step": 2710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411400479795618e-05,
+      "loss": 0.956,
+      "step": 2711
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94108492850103e-05,
+      "loss": 0.9491,
+      "step": 2712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941029784009536e-05,
+      "loss": 0.9172,
+      "step": 2713
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409746145065452e-05,
+      "loss": 1.0176,
+      "step": 2714
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409194199935233e-05,
+      "loss": 0.961,
+      "step": 2715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408642004719384e-05,
+      "loss": 0.9667,
+      "step": 2716
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940808955943257e-05,
+      "loss": 1.002,
+      "step": 2717
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407536864089486e-05,
+      "loss": 1.0822,
+      "step": 2718
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406983918704812e-05,
+      "loss": 0.96,
+      "step": 2719
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406430723293255e-05,
+      "loss": 0.9667,
+      "step": 2720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940587727786951e-05,
+      "loss": 1.053,
+      "step": 2721
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940532358244829e-05,
+      "loss": 0.998,
+      "step": 2722
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940476963704431e-05,
+      "loss": 0.9721,
+      "step": 2723
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404215441672294e-05,
+      "loss": 0.443,
+      "step": 2724
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403660996346972e-05,
+      "loss": 1.0161,
+      "step": 2725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403106301083083e-05,
+      "loss": 1.0249,
+      "step": 2726
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402551355895366e-05,
+      "loss": 0.9644,
+      "step": 2727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401996160798574e-05,
+      "loss": 0.8882,
+      "step": 2728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940144071580746e-05,
+      "loss": 1.0099,
+      "step": 2729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940088502093679e-05,
+      "loss": 1.0764,
+      "step": 2730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400329076201334e-05,
+      "loss": 1.0449,
+      "step": 2731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399772881615866e-05,
+      "loss": 1.0142,
+      "step": 2732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939921643719517e-05,
+      "loss": 1.0563,
+      "step": 2733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398659742954037e-05,
+      "loss": 0.8856,
+      "step": 2734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398102798907265e-05,
+      "loss": 0.9935,
+      "step": 2735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397545605069653e-05,
+      "loss": 0.9214,
+      "step": 2736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939698816145601e-05,
+      "loss": 1.036,
+      "step": 2737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396430468081152e-05,
+      "loss": 0.9811,
+      "step": 2738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939587252495991e-05,
+      "loss": 1.0468,
+      "step": 2739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395314332107103e-05,
+      "loss": 1.029,
+      "step": 2740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394755889537572e-05,
+      "loss": 1.072,
+      "step": 2741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939419719726616e-05,
+      "loss": 0.9845,
+      "step": 2742
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393638255307715e-05,
+      "loss": 0.9537,
+      "step": 2743
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393079063677093e-05,
+      "loss": 1.0622,
+      "step": 2744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392519622389157e-05,
+      "loss": 0.977,
+      "step": 2745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391959931458782e-05,
+      "loss": 0.9422,
+      "step": 2746
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391399990900835e-05,
+      "loss": 1.0345,
+      "step": 2747
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390839800730204e-05,
+      "loss": 0.9789,
+      "step": 2748
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939027936096177e-05,
+      "loss": 1.0087,
+      "step": 2749
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389718671610446e-05,
+      "loss": 0.9698,
+      "step": 2750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389157732691116e-05,
+      "loss": 0.9834,
+      "step": 2751
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388596544218697e-05,
+      "loss": 1.0269,
+      "step": 2752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388035106208106e-05,
+      "loss": 1.0309,
+      "step": 2753
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387473418674264e-05,
+      "loss": 1.0119,
+      "step": 2754
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93869114816321e-05,
+      "loss": 1.0475,
+      "step": 2755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938634929509655e-05,
+      "loss": 1.0413,
+      "step": 2756
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938578685908256e-05,
+      "loss": 1.0055,
+      "step": 2757
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385224173605072e-05,
+      "loss": 0.9662,
+      "step": 2758
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384661238679042e-05,
+      "loss": 0.9203,
+      "step": 2759
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384098054319434e-05,
+      "loss": 1.0036,
+      "step": 2760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938353462054122e-05,
+      "loss": 0.923,
+      "step": 2761
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382970937359374e-05,
+      "loss": 1.0326,
+      "step": 2762
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382407004788875e-05,
+      "loss": 1.0599,
+      "step": 2763
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938184282284471e-05,
+      "loss": 0.9739,
+      "step": 2764
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381278391541885e-05,
+      "loss": 0.948,
+      "step": 2765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938071371089539e-05,
+      "loss": 0.936,
+      "step": 2766
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938014878092024e-05,
+      "loss": 0.974,
+      "step": 2767
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937958360163145e-05,
+      "loss": 1.017,
+      "step": 2768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379018173044038e-05,
+      "loss": 0.9358,
+      "step": 2769
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378452495173038e-05,
+      "loss": 1.0313,
+      "step": 2770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377886568033476e-05,
+      "loss": 0.9592,
+      "step": 2771
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377320391640404e-05,
+      "loss": 0.9982,
+      "step": 2772
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376753966008864e-05,
+      "loss": 1.0791,
+      "step": 2773
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376187291153915e-05,
+      "loss": 1.067,
+      "step": 2774
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375620367090614e-05,
+      "loss": 0.9804,
+      "step": 2775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375053193834032e-05,
+      "loss": 0.9975,
+      "step": 2776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374485771399245e-05,
+      "loss": 1.0233,
+      "step": 2777
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373918099801332e-05,
+      "loss": 1.0917,
+      "step": 2778
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373350179055382e-05,
+      "loss": 0.4156,
+      "step": 2779
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937278200917649e-05,
+      "loss": 0.9914,
+      "step": 2780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372213590179757e-05,
+      "loss": 0.9756,
+      "step": 2781
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937164492208029e-05,
+      "loss": 0.9587,
+      "step": 2782
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371076004893207e-05,
+      "loss": 1.0175,
+      "step": 2783
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370506838633623e-05,
+      "loss": 1.0446,
+      "step": 2784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369937423316672e-05,
+      "loss": 0.9554,
+      "step": 2785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369367758957483e-05,
+      "loss": 0.984,
+      "step": 2786
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93687978455712e-05,
+      "loss": 0.9654,
+      "step": 2787
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368227683172973e-05,
+      "loss": 0.9394,
+      "step": 2788
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367657271777955e-05,
+      "loss": 0.967,
+      "step": 2789
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93670866114013e-05,
+      "loss": 1.0134,
+      "step": 2790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366515702058186e-05,
+      "loss": 0.9799,
+      "step": 2791
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365944543763782e-05,
+      "loss": 0.9541,
+      "step": 2792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365373136533267e-05,
+      "loss": 0.9848,
+      "step": 2793
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936480148038183e-05,
+      "loss": 1.0722,
+      "step": 2794
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364229575324665e-05,
+      "loss": 1.0224,
+      "step": 2795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363657421376975e-05,
+      "loss": 0.9963,
+      "step": 2796
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363085018553965e-05,
+      "loss": 0.932,
+      "step": 2797
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362512366870846e-05,
+      "loss": 0.9847,
+      "step": 2798
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361939466342844e-05,
+      "loss": 1.0574,
+      "step": 2799
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361366316985185e-05,
+      "loss": 0.9544,
+      "step": 2800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93607929188131e-05,
+      "loss": 0.9838,
+      "step": 2801
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360219271841828e-05,
+      "loss": 0.9879,
+      "step": 2802
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359645376086616e-05,
+      "loss": 1.0198,
+      "step": 2803
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359071231562724e-05,
+      "loss": 1.0053,
+      "step": 2804
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358496838285408e-05,
+      "loss": 1.0446,
+      "step": 2805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935792219626993e-05,
+      "loss": 0.9462,
+      "step": 2806
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357347305531574e-05,
+      "loss": 1.0328,
+      "step": 2807
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356772166085608e-05,
+      "loss": 0.9871,
+      "step": 2808
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356196777947326e-05,
+      "loss": 1.0079,
+      "step": 2809
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355621141132022e-05,
+      "loss": 1.0244,
+      "step": 2810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935504525565499e-05,
+      "loss": 0.3824,
+      "step": 2811
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354469121531543e-05,
+      "loss": 0.927,
+      "step": 2812
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353892738776986e-05,
+      "loss": 0.9462,
+      "step": 2813
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353316107406644e-05,
+      "loss": 1.0762,
+      "step": 2814
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352739227435846e-05,
+      "loss": 0.961,
+      "step": 2815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935216209887992e-05,
+      "loss": 0.9986,
+      "step": 2816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351584721754205e-05,
+      "loss": 0.9987,
+      "step": 2817
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935100709607405e-05,
+      "loss": 1.0858,
+      "step": 2818
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350429221854805e-05,
+      "loss": 0.9341,
+      "step": 2819
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349851099111832e-05,
+      "loss": 0.9406,
+      "step": 2820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349272727860495e-05,
+      "loss": 0.984,
+      "step": 2821
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348694108116166e-05,
+      "loss": 1.0303,
+      "step": 2822
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348115239894225e-05,
+      "loss": 1.0017,
+      "step": 2823
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934753612321006e-05,
+      "loss": 0.9033,
+      "step": 2824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346956758079057e-05,
+      "loss": 0.9514,
+      "step": 2825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934637714451662e-05,
+      "loss": 1.0263,
+      "step": 2826
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345797282538152e-05,
+      "loss": 0.9667,
+      "step": 2827
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345217172159066e-05,
+      "loss": 1.0522,
+      "step": 2828
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344636813394785e-05,
+      "loss": 1.0421,
+      "step": 2829
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344056206260725e-05,
+      "loss": 0.9796,
+      "step": 2830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343475350772326e-05,
+      "loss": 0.896,
+      "step": 2831
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934289424694502e-05,
+      "loss": 1.0175,
+      "step": 2832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934231289479426e-05,
+      "loss": 0.9848,
+      "step": 2833
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341731294335492e-05,
+      "loss": 1.1237,
+      "step": 2834
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341149445584176e-05,
+      "loss": 0.997,
+      "step": 2835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340567348555774e-05,
+      "loss": 0.9492,
+      "step": 2836
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339985003265763e-05,
+      "loss": 1.0748,
+      "step": 2837
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339402409729615e-05,
+      "loss": 0.9466,
+      "step": 2838
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338819567962815e-05,
+      "loss": 0.4185,
+      "step": 2839
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338236477980864e-05,
+      "loss": 1.114,
+      "step": 2840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337653139799245e-05,
+      "loss": 1.0347,
+      "step": 2841
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337069553433476e-05,
+      "loss": 0.9273,
+      "step": 2842
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933648571889906e-05,
+      "loss": 1.0606,
+      "step": 2843
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335901636211517e-05,
+      "loss": 1.0351,
+      "step": 2844
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335317305386373e-05,
+      "loss": 1.0256,
+      "step": 2845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334732726439153e-05,
+      "loss": 0.9695,
+      "step": 2846
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93341478993854e-05,
+      "loss": 0.9883,
+      "step": 2847
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933356282424066e-05,
+      "loss": 0.8954,
+      "step": 2848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933297750102048e-05,
+      "loss": 0.9421,
+      "step": 2849
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332391929740412e-05,
+      "loss": 0.9632,
+      "step": 2850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331806110416027e-05,
+      "loss": 1.0321,
+      "step": 2851
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331220043062894e-05,
+      "loss": 0.9164,
+      "step": 2852
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933063372769659e-05,
+      "loss": 1.0396,
+      "step": 2853
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330047164332698e-05,
+      "loss": 0.9933,
+      "step": 2854
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329460352986807e-05,
+      "loss": 1.0823,
+      "step": 2855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932887329367452e-05,
+      "loss": 1.0022,
+      "step": 2856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328285986411427e-05,
+      "loss": 0.9992,
+      "step": 2857
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327698431213153e-05,
+      "loss": 0.9918,
+      "step": 2858
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327110628095307e-05,
+      "loss": 0.9661,
+      "step": 2859
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326522577073514e-05,
+      "loss": 0.9776,
+      "step": 2860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325934278163398e-05,
+      "loss": 0.9792,
+      "step": 2861
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325345731380606e-05,
+      "loss": 1.0181,
+      "step": 2862
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932475693674077e-05,
+      "loss": 0.9725,
+      "step": 2863
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324167894259547e-05,
+      "loss": 1.0463,
+      "step": 2864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932357860395259e-05,
+      "loss": 1.1156,
+      "step": 2865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322989065835562e-05,
+      "loss": 1.054,
+      "step": 2866
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322399279924132e-05,
+      "loss": 1.0163,
+      "step": 2867
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932180924623398e-05,
+      "loss": 0.9664,
+      "step": 2868
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321218964780783e-05,
+      "loss": 1.0312,
+      "step": 2869
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320628435580228e-05,
+      "loss": 0.9976,
+      "step": 2870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320037658648018e-05,
+      "loss": 0.99,
+      "step": 2871
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931944663399985e-05,
+      "loss": 1.0721,
+      "step": 2872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318855361651436e-05,
+      "loss": 1.0407,
+      "step": 2873
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931826384161849e-05,
+      "loss": 0.9986,
+      "step": 2874
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931767207391673e-05,
+      "loss": 0.9397,
+      "step": 2875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931708005856189e-05,
+      "loss": 0.9938,
+      "step": 2876
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316487795569704e-05,
+      "loss": 1.0161,
+      "step": 2877
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315895284955913e-05,
+      "loss": 1.0007,
+      "step": 2878
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315302526736264e-05,
+      "loss": 0.9398,
+      "step": 2879
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314709520926514e-05,
+      "loss": 0.9828,
+      "step": 2880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9314116267542423e-05,
+      "loss": 0.9704,
+      "step": 2881
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931352276659976e-05,
+      "loss": 1.0479,
+      "step": 2882
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312929018114296e-05,
+      "loss": 1.0441,
+      "step": 2883
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312335022101818e-05,
+      "loss": 1.0654,
+      "step": 2884
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931174077857811e-05,
+      "loss": 1.0423,
+      "step": 2885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931114628755897e-05,
+      "loss": 0.9835,
+      "step": 2886
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931055154906019e-05,
+      "loss": 0.9984,
+      "step": 2887
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309956563097586e-05,
+      "loss": 1.0043,
+      "step": 2888
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930936132968697e-05,
+      "loss": 1.0026,
+      "step": 2889
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308765848844163e-05,
+      "loss": 0.9144,
+      "step": 2890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930817012058499e-05,
+      "loss": 0.9993,
+      "step": 2891
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307574144925288e-05,
+      "loss": 0.9561,
+      "step": 2892
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306977921880895e-05,
+      "loss": 1.0124,
+      "step": 2893
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306381451467656e-05,
+      "loss": 1.0241,
+      "step": 2894
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930578473370143e-05,
+      "loss": 0.9625,
+      "step": 2895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930518776859807e-05,
+      "loss": 1.0285,
+      "step": 2896
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304590556173452e-05,
+      "loss": 0.9943,
+      "step": 2897
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930399309644344e-05,
+      "loss": 1.0347,
+      "step": 2898
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303395389423918e-05,
+      "loss": 1.0592,
+      "step": 2899
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302797435130772e-05,
+      "loss": 1.0157,
+      "step": 2900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302199233579896e-05,
+      "loss": 0.9707,
+      "step": 2901
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930160078478719e-05,
+      "loss": 1.0605,
+      "step": 2902
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301002088768557e-05,
+      "loss": 1.007,
+      "step": 2903
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930040314553991e-05,
+      "loss": 0.9183,
+      "step": 2904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929980395511717e-05,
+      "loss": 1.0415,
+      "step": 2905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299204517516264e-05,
+      "loss": 0.9909,
+      "step": 2906
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929860483275312e-05,
+      "loss": 1.0173,
+      "step": 2907
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298004900843685e-05,
+      "loss": 0.9482,
+      "step": 2908
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297404721803893e-05,
+      "loss": 1.0015,
+      "step": 2909
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296804295649704e-05,
+      "loss": 1.0468,
+      "step": 2910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296203622397077e-05,
+      "loss": 0.997,
+      "step": 2911
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295602702061972e-05,
+      "loss": 1.0545,
+      "step": 2912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295001534660366e-05,
+      "loss": 0.4425,
+      "step": 2913
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294400120208232e-05,
+      "loss": 1.0076,
+      "step": 2914
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929379845872156e-05,
+      "loss": 0.9473,
+      "step": 2915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929319655021634e-05,
+      "loss": 1.066,
+      "step": 2916
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292594394708567e-05,
+      "loss": 1.0254,
+      "step": 2917
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291991992214252e-05,
+      "loss": 1.0215,
+      "step": 2918
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92913893427494e-05,
+      "loss": 1.0044,
+      "step": 2919
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290786446330032e-05,
+      "loss": 0.9871,
+      "step": 2920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929018330297217e-05,
+      "loss": 1.0235,
+      "step": 2921
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289579912691846e-05,
+      "loss": 1.0479,
+      "step": 2922
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288976275505102e-05,
+      "loss": 0.8729,
+      "step": 2923
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288372391427975e-05,
+      "loss": 1.045,
+      "step": 2924
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287768260476517e-05,
+      "loss": 0.985,
+      "step": 2925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287163882666786e-05,
+      "loss": 0.9975,
+      "step": 2926
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286559258014845e-05,
+      "loss": 0.8796,
+      "step": 2927
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928595438653677e-05,
+      "loss": 0.9973,
+      "step": 2928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285349268248626e-05,
+      "loss": 1.0232,
+      "step": 2929
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284743903166508e-05,
+      "loss": 1.023,
+      "step": 2930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284138291306504e-05,
+      "loss": 0.956,
+      "step": 2931
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283532432684704e-05,
+      "loss": 1.0712,
+      "step": 2932
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282926327317213e-05,
+      "loss": 0.988,
+      "step": 2933
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282319975220144e-05,
+      "loss": 0.9693,
+      "step": 2934
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281713376409615e-05,
+      "loss": 0.4028,
+      "step": 2935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928110653090174e-05,
+      "loss": 1.0357,
+      "step": 2936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280499438712655e-05,
+      "loss": 1.0157,
+      "step": 2937
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279892099858496e-05,
+      "loss": 0.9353,
+      "step": 2938
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279284514355404e-05,
+      "loss": 0.393,
+      "step": 2939
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278676682219523e-05,
+      "loss": 0.9454,
+      "step": 2940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278068603467017e-05,
+      "loss": 1.0071,
+      "step": 2941
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277460278114043e-05,
+      "loss": 1.045,
+      "step": 2942
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927685170617677e-05,
+      "loss": 1.0009,
+      "step": 2943
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276242887671374e-05,
+      "loss": 0.9398,
+      "step": 2944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9275633822614037e-05,
+      "loss": 1.0105,
+      "step": 2945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927502451102095e-05,
+      "loss": 1.0633,
+      "step": 2946
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92744149529083e-05,
+      "loss": 0.9862,
+      "step": 2947
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273805148292296e-05,
+      "loss": 1.0757,
+      "step": 2948
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927319509718914e-05,
+      "loss": 0.9548,
+      "step": 2949
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9272584799615052e-05,
+      "loss": 1.0024,
+      "step": 2950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271974255586246e-05,
+      "loss": 0.9865,
+      "step": 2951
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271363465118956e-05,
+      "loss": 0.9597,
+      "step": 2952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270752428229416e-05,
+      "loss": 0.8898,
+      "step": 2953
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270141144933865e-05,
+      "loss": 1.0552,
+      "step": 2954
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926952961524855e-05,
+      "loss": 1.0375,
+      "step": 2955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926891783918972e-05,
+      "loss": 0.377,
+      "step": 2956
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268305816773644e-05,
+      "loss": 1.0411,
+      "step": 2957
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267693548016584e-05,
+      "loss": 0.4382,
+      "step": 2958
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267081032934815e-05,
+      "loss": 0.9517,
+      "step": 2959
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266468271544612e-05,
+      "loss": 1.0316,
+      "step": 2960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926585526386227e-05,
+      "loss": 1.0171,
+      "step": 2961
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265242009904076e-05,
+      "loss": 1.1052,
+      "step": 2962
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264628509686333e-05,
+      "loss": 1.0596,
+      "step": 2963
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264014763225345e-05,
+      "loss": 0.9855,
+      "step": 2964
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263400770537425e-05,
+      "loss": 1.0036,
+      "step": 2965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926278653163889e-05,
+      "loss": 1.0226,
+      "step": 2966
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926217204654607e-05,
+      "loss": 1.045,
+      "step": 2967
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261557315275293e-05,
+      "loss": 1.0344,
+      "step": 2968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260942337842905e-05,
+      "loss": 1.0167,
+      "step": 2969
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260327114265246e-05,
+      "loss": 1.0572,
+      "step": 2970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259711644558664e-05,
+      "loss": 0.4234,
+      "step": 2971
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259095928739527e-05,
+      "loss": 1.0201,
+      "step": 2972
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925847996682419e-05,
+      "loss": 1.0987,
+      "step": 2973
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257863758829038e-05,
+      "loss": 0.9725,
+      "step": 2974
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925724730477043e-05,
+      "loss": 0.9561,
+      "step": 2975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925663060466477e-05,
+      "loss": 0.8998,
+      "step": 2976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256013658528437e-05,
+      "loss": 1.0012,
+      "step": 2977
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255396466377833e-05,
+      "loss": 0.9498,
+      "step": 2978
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254779028229362e-05,
+      "loss": 0.9901,
+      "step": 2979
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925416134409943e-05,
+      "loss": 1.1257,
+      "step": 2980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253543414004465e-05,
+      "loss": 1.029,
+      "step": 2981
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925292523796088e-05,
+      "loss": 0.9673,
+      "step": 2982
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252306815985116e-05,
+      "loss": 0.991,
+      "step": 2983
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92516881480936e-05,
+      "loss": 1.0871,
+      "step": 2984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251069234302776e-05,
+      "loss": 1.0907,
+      "step": 2985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250450074629103e-05,
+      "loss": 0.9568,
+      "step": 2986
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924983066908903e-05,
+      "loss": 1.073,
+      "step": 2987
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9249211017699023e-05,
+      "loss": 1.0005,
+      "step": 2988
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248591120475552e-05,
+      "loss": 1.062,
+      "step": 2989
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924797097743509e-05,
+      "loss": 0.987,
+      "step": 2990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924735058859412e-05,
+      "loss": 0.9266,
+      "step": 2991
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246729953969135e-05,
+      "loss": 1.0249,
+      "step": 2992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924610907357663e-05,
+      "loss": 1.0075,
+      "step": 2993
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245487947433103e-05,
+      "loss": 0.9484,
+      "step": 2994
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244866575555066e-05,
+      "loss": 1.0353,
+      "step": 2995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244244957959032e-05,
+      "loss": 0.9831,
+      "step": 2996
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924362309466153e-05,
+      "loss": 1.0107,
+      "step": 2997
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924300098567908e-05,
+      "loss": 0.9201,
+      "step": 2998
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242378631028223e-05,
+      "loss": 1.0032,
+      "step": 2999
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241756030725496e-05,
+      "loss": 0.9834,
+      "step": 3000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924113318478745e-05,
+      "loss": 0.9799,
+      "step": 3001
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240510093230637e-05,
+      "loss": 0.9686,
+      "step": 3002
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923988675607162e-05,
+      "loss": 1.0736,
+      "step": 3003
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239263173326966e-05,
+      "loss": 1.0001,
+      "step": 3004
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923863934501325e-05,
+      "loss": 1.0212,
+      "step": 3005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923801527114705e-05,
+      "loss": 0.9386,
+      "step": 3006
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237390951744955e-05,
+      "loss": 0.9253,
+      "step": 3007
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236766386823554e-05,
+      "loss": 1.0014,
+      "step": 3008
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923614157639946e-05,
+      "loss": 0.9905,
+      "step": 3009
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235516520489267e-05,
+      "loss": 1.0731,
+      "step": 3010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234891219109598e-05,
+      "loss": 1.0316,
+      "step": 3011
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923426567227706e-05,
+      "loss": 0.8782,
+      "step": 3012
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233639880008294e-05,
+      "loss": 0.9874,
+      "step": 3013
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923301384231992e-05,
+      "loss": 0.9888,
+      "step": 3014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232387559228587e-05,
+      "loss": 0.9868,
+      "step": 3015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231761030750936e-05,
+      "loss": 0.4121,
+      "step": 3016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923113425690362e-05,
+      "loss": 0.3943,
+      "step": 3017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92305072377033e-05,
+      "loss": 1.0041,
+      "step": 3018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922987997316664e-05,
+      "loss": 1.0037,
+      "step": 3019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922925246331031e-05,
+      "loss": 1.0283,
+      "step": 3020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922862470815099e-05,
+      "loss": 0.9884,
+      "step": 3021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227996707705364e-05,
+      "loss": 1.0018,
+      "step": 3022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922736846199013e-05,
+      "loss": 1.029,
+      "step": 3023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922673997102198e-05,
+      "loss": 1.0815,
+      "step": 3024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226111234817617e-05,
+      "loss": 1.0156,
+      "step": 3025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225482253393753e-05,
+      "loss": 0.9856,
+      "step": 3026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922485302676711e-05,
+      "loss": 1.0952,
+      "step": 3027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922422355495441e-05,
+      "loss": 0.9457,
+      "step": 3028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223593837972382e-05,
+      "loss": 1.0199,
+      "step": 3029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222963875837764e-05,
+      "loss": 1.0855,
+      "step": 3030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92223336685673e-05,
+      "loss": 1.0317,
+      "step": 3031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221703216177737e-05,
+      "loss": 1.0993,
+      "step": 3032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922107251868584e-05,
+      "loss": 1.0704,
+      "step": 3033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220441576108363e-05,
+      "loss": 0.9633,
+      "step": 3034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219810388462082e-05,
+      "loss": 1.0206,
+      "step": 3035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921917895576377e-05,
+      "loss": 1.0093,
+      "step": 3036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921854727803021e-05,
+      "loss": 0.9258,
+      "step": 3037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217915355278194e-05,
+      "loss": 0.945,
+      "step": 3038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217283187524512e-05,
+      "loss": 0.9304,
+      "step": 3039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216650774785975e-05,
+      "loss": 1.0077,
+      "step": 3040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216018117079384e-05,
+      "loss": 0.4307,
+      "step": 3041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215385214421557e-05,
+      "loss": 1.0554,
+      "step": 3042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214752066829314e-05,
+      "loss": 1.0244,
+      "step": 3043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921411867431949e-05,
+      "loss": 1.002,
+      "step": 3044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213485036908912e-05,
+      "loss": 1.0368,
+      "step": 3045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212851154614426e-05,
+      "loss": 0.9537,
+      "step": 3046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921221702745288e-05,
+      "loss": 1.0126,
+      "step": 3047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921158265544112e-05,
+      "loss": 0.9887,
+      "step": 3048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921094803859602e-05,
+      "loss": 1.0029,
+      "step": 3049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921031317693444e-05,
+      "loss": 0.9161,
+      "step": 3050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209678070473253e-05,
+      "loss": 0.4601,
+      "step": 3051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209042719229343e-05,
+      "loss": 1.0304,
+      "step": 3052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208407123219594e-05,
+      "loss": 0.9855,
+      "step": 3053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92077712824609e-05,
+      "loss": 0.9805,
+      "step": 3054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207135196970163e-05,
+      "loss": 1.0273,
+      "step": 3055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920649886676429e-05,
+      "loss": 1.076,
+      "step": 3056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920586229186019e-05,
+      "loss": 0.9527,
+      "step": 3057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205225472274782e-05,
+      "loss": 0.9672,
+      "step": 3058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204588408024996e-05,
+      "loss": 1.019,
+      "step": 3059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203951099127765e-05,
+      "loss": 0.9621,
+      "step": 3060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203313545600025e-05,
+      "loss": 0.9005,
+      "step": 3061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202675747458722e-05,
+      "loss": 0.9688,
+      "step": 3062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202037704720805e-05,
+      "loss": 1.0233,
+      "step": 3063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920139941740324e-05,
+      "loss": 1.0344,
+      "step": 3064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200760885522983e-05,
+      "loss": 1.0152,
+      "step": 3065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200122109097013e-05,
+      "loss": 0.4105,
+      "step": 3066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199483088142304e-05,
+      "loss": 0.9176,
+      "step": 3067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198843822675845e-05,
+      "loss": 1.0476,
+      "step": 3068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919820431271462e-05,
+      "loss": 1.0077,
+      "step": 3069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197564558275626e-05,
+      "loss": 1.0002,
+      "step": 3070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196924559375874e-05,
+      "loss": 0.9436,
+      "step": 3071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919628431603237e-05,
+      "loss": 1.016,
+      "step": 3072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195643828262135e-05,
+      "loss": 0.9831,
+      "step": 3073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919500309608219e-05,
+      "loss": 0.9462,
+      "step": 3074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194362119509562e-05,
+      "loss": 0.9499,
+      "step": 3075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919372089856129e-05,
+      "loss": 0.9689,
+      "step": 3076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193079433254415e-05,
+      "loss": 0.9104,
+      "step": 3077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919243772360599e-05,
+      "loss": 0.95,
+      "step": 3078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919179576963307e-05,
+      "loss": 1.02,
+      "step": 3079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191153571352713e-05,
+      "loss": 1.059,
+      "step": 3080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190511128781997e-05,
+      "loss": 0.9603,
+      "step": 3081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918986844193799e-05,
+      "loss": 1.0348,
+      "step": 3082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189225510837772e-05,
+      "loss": 0.9951,
+      "step": 3083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918858233549844e-05,
+      "loss": 0.9934,
+      "step": 3084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918793891593708e-05,
+      "loss": 0.9646,
+      "step": 3085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187295252170797e-05,
+      "loss": 0.8842,
+      "step": 3086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186651344216703e-05,
+      "loss": 1.014,
+      "step": 3087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186007192091905e-05,
+      "loss": 0.9745,
+      "step": 3088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185362795813526e-05,
+      "loss": 1.083,
+      "step": 3089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91847181553987e-05,
+      "loss": 1.0032,
+      "step": 3090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9184073270864553e-05,
+      "loss": 0.9154,
+      "step": 3091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183428142228223e-05,
+      "loss": 1.0221,
+      "step": 3092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182782769506867e-05,
+      "loss": 0.9901,
+      "step": 3093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918213715271763e-05,
+      "loss": 0.4251,
+      "step": 3094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181491291877675e-05,
+      "loss": 0.9942,
+      "step": 3095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180845187004167e-05,
+      "loss": 0.9518,
+      "step": 3096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180198838114284e-05,
+      "loss": 0.9939,
+      "step": 3097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179552245225197e-05,
+      "loss": 1.0595,
+      "step": 3098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178905408354097e-05,
+      "loss": 1.0173,
+      "step": 3099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917825832751817e-05,
+      "loss": 0.9238,
+      "step": 3100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177611002734623e-05,
+      "loss": 0.9811,
+      "step": 3101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917696343402066e-05,
+      "loss": 1.0076,
+      "step": 3102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9176315621393486e-05,
+      "loss": 0.9955,
+      "step": 3103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175667564870324e-05,
+      "loss": 0.9761,
+      "step": 3104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175019264468396e-05,
+      "loss": 0.4102,
+      "step": 3105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174370720204937e-05,
+      "loss": 0.9425,
+      "step": 3106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173721932097183e-05,
+      "loss": 1.0468,
+      "step": 3107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917307290016238e-05,
+      "loss": 1.0351,
+      "step": 3108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917242362441777e-05,
+      "loss": 1.0262,
+      "step": 3109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917177410488062e-05,
+      "loss": 0.987,
+      "step": 3110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171124341568187e-05,
+      "loss": 1.0001,
+      "step": 3111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170474334497744e-05,
+      "loss": 0.9464,
+      "step": 3112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169824083686567e-05,
+      "loss": 1.071,
+      "step": 3113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916917358915194e-05,
+      "loss": 1.0424,
+      "step": 3114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916852285091115e-05,
+      "loss": 0.9366,
+      "step": 3115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167871868981493e-05,
+      "loss": 0.9731,
+      "step": 3116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167220643380277e-05,
+      "loss": 0.9556,
+      "step": 3117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91665691741248e-05,
+      "loss": 1.1154,
+      "step": 3118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165917461232386e-05,
+      "loss": 1.0361,
+      "step": 3119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165265504720355e-05,
+      "loss": 0.9424,
+      "step": 3120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164613304606033e-05,
+      "loss": 1.0276,
+      "step": 3121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916396086090676e-05,
+      "loss": 0.9564,
+      "step": 3122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163308173639872e-05,
+      "loss": 1.0366,
+      "step": 3123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162655242822715e-05,
+      "loss": 0.9036,
+      "step": 3124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162002068472652e-05,
+      "loss": 0.9378,
+      "step": 3125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161348650607034e-05,
+      "loss": 1.0548,
+      "step": 3126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160694989243235e-05,
+      "loss": 1.0084,
+      "step": 3127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160041084398622e-05,
+      "loss": 0.9669,
+      "step": 3128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915938693609058e-05,
+      "loss": 0.8701,
+      "step": 3129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91587325443365e-05,
+      "loss": 1.0746,
+      "step": 3130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915807790915376e-05,
+      "loss": 1.0147,
+      "step": 3131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157423030559773e-05,
+      "loss": 0.9365,
+      "step": 3132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915676790857194e-05,
+      "loss": 1.0359,
+      "step": 3133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156112543207674e-05,
+      "loss": 0.9852,
+      "step": 3134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155456934484396e-05,
+      "loss": 1.0579,
+      "step": 3135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915480108241953e-05,
+      "loss": 0.8255,
+      "step": 3136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154144987030502e-05,
+      "loss": 0.9724,
+      "step": 3137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915348864833476e-05,
+      "loss": 0.9997,
+      "step": 3138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915283206634974e-05,
+      "loss": 0.4572,
+      "step": 3139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91521752410929e-05,
+      "loss": 1.1113,
+      "step": 3140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151518172581693e-05,
+      "loss": 0.9104,
+      "step": 3141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150860860833588e-05,
+      "loss": 0.9998,
+      "step": 3142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915020330586605e-05,
+      "loss": 1.0307,
+      "step": 3143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914954550769656e-05,
+      "loss": 0.9412,
+      "step": 3144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91488874663426e-05,
+      "loss": 1.0284,
+      "step": 3145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148229181821656e-05,
+      "loss": 1.0099,
+      "step": 3146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147570654151237e-05,
+      "loss": 0.9975,
+      "step": 3147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914691188334883e-05,
+      "loss": 1.005,
+      "step": 3148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146252869431956e-05,
+      "loss": 0.9552,
+      "step": 3149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145593612418125e-05,
+      "loss": 1.0043,
+      "step": 3150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144934112324863e-05,
+      "loss": 0.4452,
+      "step": 3151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144274369169692e-05,
+      "loss": 1.0839,
+      "step": 3152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143614382970157e-05,
+      "loss": 1.0331,
+      "step": 3153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142954153743792e-05,
+      "loss": 0.9845,
+      "step": 3154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914229368150815e-05,
+      "loss": 0.9845,
+      "step": 3155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141632966280784e-05,
+      "loss": 1.006,
+      "step": 3156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914097200807925e-05,
+      "loss": 0.9951,
+      "step": 3157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140310806921124e-05,
+      "loss": 0.9137,
+      "step": 3158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139649362823974e-05,
+      "loss": 1.0563,
+      "step": 3159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138987675805385e-05,
+      "loss": 1.0467,
+      "step": 3160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913832574588294e-05,
+      "loss": 0.9295,
+      "step": 3161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137663573074232e-05,
+      "loss": 1.0315,
+      "step": 3162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137001157396864e-05,
+      "loss": 1.0595,
+      "step": 3163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136338498868442e-05,
+      "loss": 1.0355,
+      "step": 3164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135675597506574e-05,
+      "loss": 1.0493,
+      "step": 3165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135012453328884e-05,
+      "loss": 0.9866,
+      "step": 3166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134349066352998e-05,
+      "loss": 1.0149,
+      "step": 3167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9133685436596545e-05,
+      "loss": 0.9038,
+      "step": 3168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9133021564077165e-05,
+      "loss": 1.0444,
+      "step": 3169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132357448812502e-05,
+      "loss": 1.0691,
+      "step": 3170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913169309082021e-05,
+      "loss": 1.0606,
+      "step": 3171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131028490117943e-05,
+      "loss": 0.9535,
+      "step": 3172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130363646723368e-05,
+      "loss": 1.0514,
+      "step": 3173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129698560654152e-05,
+      "loss": 0.9739,
+      "step": 3174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912903323192798e-05,
+      "loss": 0.9865,
+      "step": 3175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128367660562528e-05,
+      "loss": 1.0425,
+      "step": 3176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912770184657549e-05,
+      "loss": 1.0272,
+      "step": 3177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127035789984564e-05,
+      "loss": 0.9884,
+      "step": 3178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912636949080745e-05,
+      "loss": 0.9975,
+      "step": 3179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125702949061855e-05,
+      "loss": 0.4261,
+      "step": 3180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125036164765502e-05,
+      "loss": 1.0875,
+      "step": 3181
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124369137936108e-05,
+      "loss": 1.0154,
+      "step": 3182
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123701868591403e-05,
+      "loss": 0.9935,
+      "step": 3183
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123034356749125e-05,
+      "loss": 0.9355,
+      "step": 3184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912236660242701e-05,
+      "loss": 0.9718,
+      "step": 3185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912169860564281e-05,
+      "loss": 0.9483,
+      "step": 3186
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912103036641428e-05,
+      "loss": 0.4122,
+      "step": 3187
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912036188475918e-05,
+      "loss": 1.0106,
+      "step": 3188
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119693160695276e-05,
+      "loss": 1.0831,
+      "step": 3189
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911902419424035e-05,
+      "loss": 1.0488,
+      "step": 3190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9118354985412165e-05,
+      "loss": 1.1255,
+      "step": 3191
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117685534228525e-05,
+      "loss": 0.9366,
+      "step": 3192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117015840707215e-05,
+      "loss": 1.0248,
+      "step": 3193
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911634590486604e-05,
+      "loss": 0.9984,
+      "step": 3194
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115675726722802e-05,
+      "loss": 0.9739,
+      "step": 3195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115005306295312e-05,
+      "loss": 1.041,
+      "step": 3196
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911433464360139e-05,
+      "loss": 0.9625,
+      "step": 3197
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113663738658868e-05,
+      "loss": 1.0256,
+      "step": 3198
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911299259148557e-05,
+      "loss": 0.9551,
+      "step": 3199
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112321202099335e-05,
+      "loss": 1.0729,
+      "step": 3200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911164957051801e-05,
+      "loss": 1.0095,
+      "step": 3201
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110977696759447e-05,
+      "loss": 0.8891,
+      "step": 3202
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110305580841503e-05,
+      "loss": 0.9956,
+      "step": 3203
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109633222782043e-05,
+      "loss": 0.9907,
+      "step": 3204
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108960622598934e-05,
+      "loss": 1.0233,
+      "step": 3205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108287780310054e-05,
+      "loss": 1.0405,
+      "step": 3206
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910761469593329e-05,
+      "loss": 1.038,
+      "step": 3207
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910694136948653e-05,
+      "loss": 0.9486,
+      "step": 3208
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106267800987667e-05,
+      "loss": 0.9545,
+      "step": 3209
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105593990454605e-05,
+      "loss": 0.9975,
+      "step": 3210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104919937905257e-05,
+      "loss": 0.9985,
+      "step": 3211
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104245643357535e-05,
+      "loss": 0.9835,
+      "step": 3212
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103571106829363e-05,
+      "loss": 1.0694,
+      "step": 3213
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102896328338663e-05,
+      "loss": 0.9319,
+      "step": 3214
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102221307903384e-05,
+      "loss": 1.0449,
+      "step": 3215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101546045541454e-05,
+      "loss": 0.9932,
+      "step": 3216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100870541270825e-05,
+      "loss": 1.0229,
+      "step": 3217
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910019479510945e-05,
+      "loss": 0.9771,
+      "step": 3218
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909951880707529e-05,
+      "loss": 0.9454,
+      "step": 3219
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098842577186315e-05,
+      "loss": 1.0726,
+      "step": 3220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098166105460498e-05,
+      "loss": 0.935,
+      "step": 3221
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909748939191581e-05,
+      "loss": 1.0003,
+      "step": 3222
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909681243657025e-05,
+      "loss": 0.9695,
+      "step": 3223
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096135239441805e-05,
+      "loss": 1.0652,
+      "step": 3224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095457800548473e-05,
+      "loss": 0.949,
+      "step": 3225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909478011990826e-05,
+      "loss": 0.9431,
+      "step": 3226
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094102197539182e-05,
+      "loss": 0.9546,
+      "step": 3227
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909342403345925e-05,
+      "loss": 1.0033,
+      "step": 3228
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092745627686495e-05,
+      "loss": 0.9877,
+      "step": 3229
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092066980238944e-05,
+      "loss": 0.9846,
+      "step": 3230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909138809113464e-05,
+      "loss": 1.0219,
+      "step": 3231
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909070896039162e-05,
+      "loss": 1.0073,
+      "step": 3232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090029588027942e-05,
+      "loss": 0.9643,
+      "step": 3233
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089349974061656e-05,
+      "loss": 1.0247,
+      "step": 3234
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9088670118510833e-05,
+      "loss": 0.9018,
+      "step": 3235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087990021393535e-05,
+      "loss": 1.0579,
+      "step": 3236
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908730968272784e-05,
+      "loss": 1.0208,
+      "step": 3237
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086629102531838e-05,
+      "loss": 1.0397,
+      "step": 3238
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908594828082361e-05,
+      "loss": 0.9966,
+      "step": 3239
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908526721762125e-05,
+      "loss": 1.0084,
+      "step": 3240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908458591294287e-05,
+      "loss": 1.0757,
+      "step": 3241
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908390436680657e-05,
+      "loss": 1.0251,
+      "step": 3242
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083222579230464e-05,
+      "loss": 0.8865,
+      "step": 3243
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908254055023268e-05,
+      "loss": 1.0353,
+      "step": 3244
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081858279831337e-05,
+      "loss": 1.048,
+      "step": 3245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081175768044575e-05,
+      "loss": 1.0204,
+      "step": 3246
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080493014890537e-05,
+      "loss": 1.0374,
+      "step": 3247
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907981002038736e-05,
+      "loss": 0.9606,
+      "step": 3248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079126784553205e-05,
+      "loss": 1.0276,
+      "step": 3249
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907844330740623e-05,
+      "loss": 0.9777,
+      "step": 3250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077759588964597e-05,
+      "loss": 1.047,
+      "step": 3251
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077075629246488e-05,
+      "loss": 0.9676,
+      "step": 3252
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076391428270068e-05,
+      "loss": 0.9476,
+      "step": 3253
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075706986053538e-05,
+      "loss": 0.9735,
+      "step": 3254
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075022302615076e-05,
+      "loss": 1.106,
+      "step": 3255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907433737797289e-05,
+      "loss": 0.976,
+      "step": 3256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073652212145177e-05,
+      "loss": 0.9806,
+      "step": 3257
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072966805150152e-05,
+      "loss": 1.0383,
+      "step": 3258
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072281157006036e-05,
+      "loss": 1.0195,
+      "step": 3259
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071595267731043e-05,
+      "loss": 0.9553,
+      "step": 3260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907090913734341e-05,
+      "loss": 1.0369,
+      "step": 3261
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070222765861373e-05,
+      "loss": 1.052,
+      "step": 3262
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906953615330317e-05,
+      "loss": 1.0381,
+      "step": 3263
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906884929968706e-05,
+      "loss": 0.9987,
+      "step": 3264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068162205031292e-05,
+      "loss": 0.8965,
+      "step": 3265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067474869354128e-05,
+      "loss": 1.1262,
+      "step": 3266
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066787292673837e-05,
+      "loss": 1.0281,
+      "step": 3267
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90660994750087e-05,
+      "loss": 1.0571,
+      "step": 3268
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065411416376988e-05,
+      "loss": 0.9794,
+      "step": 3269
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064723116796994e-05,
+      "loss": 1.0763,
+      "step": 3270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064034576287016e-05,
+      "loss": 0.9754,
+      "step": 3271
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906334579486535e-05,
+      "loss": 0.9579,
+      "step": 3272
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062656772550304e-05,
+      "loss": 1.0202,
+      "step": 3273
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061967509360194e-05,
+      "loss": 0.9846,
+      "step": 3274
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906127800531333e-05,
+      "loss": 0.9746,
+      "step": 3275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906058826042805e-05,
+      "loss": 1.0376,
+      "step": 3276
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9059898274722687e-05,
+      "loss": 1.0033,
+      "step": 3277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905920804821557e-05,
+      "loss": 0.9909,
+      "step": 3278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058517580925052e-05,
+      "loss": 0.9168,
+      "step": 3279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905782687286948e-05,
+      "loss": 0.9324,
+      "step": 3280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057135924067212e-05,
+      "loss": 0.9945,
+      "step": 3281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056444734536624e-05,
+      "loss": 1.0313,
+      "step": 3282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905575330429607e-05,
+      "loss": 0.9222,
+      "step": 3283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905506163336394e-05,
+      "loss": 0.3743,
+      "step": 3284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905436972175861e-05,
+      "loss": 0.454,
+      "step": 3285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905367756949848e-05,
+      "loss": 1.0238,
+      "step": 3286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052985176601935e-05,
+      "loss": 1.0364,
+      "step": 3287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905229254308739e-05,
+      "loss": 1.0503,
+      "step": 3288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051599668973245e-05,
+      "loss": 0.9058,
+      "step": 3289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050906554277914e-05,
+      "loss": 1.1662,
+      "step": 3290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905021319901983e-05,
+      "loss": 1.0127,
+      "step": 3291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049519603217413e-05,
+      "loss": 0.4144,
+      "step": 3292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90488257668891e-05,
+      "loss": 1.0486,
+      "step": 3293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048131690053334e-05,
+      "loss": 1.0076,
+      "step": 3294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047437372728562e-05,
+      "loss": 0.8797,
+      "step": 3295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046742814933234e-05,
+      "loss": 1.0044,
+      "step": 3296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904604801668582e-05,
+      "loss": 1.0467,
+      "step": 3297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045352978004777e-05,
+      "loss": 0.4764,
+      "step": 3298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044657698908586e-05,
+      "loss": 0.471,
+      "step": 3299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904396217941572e-05,
+      "loss": 0.9881,
+      "step": 3300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904326641954467e-05,
+      "loss": 1.0322,
+      "step": 3301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042570419313927e-05,
+      "loss": 1.0259,
+      "step": 3302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041874178741987e-05,
+      "loss": 1.0113,
+      "step": 3303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041177697847362e-05,
+      "loss": 1.1234,
+      "step": 3304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904048097664856e-05,
+      "loss": 1.0174,
+      "step": 3305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039784015164095e-05,
+      "loss": 1.0296,
+      "step": 3306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90390868134125e-05,
+      "loss": 1.0032,
+      "step": 3307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038389371412296e-05,
+      "loss": 1.0568,
+      "step": 3308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903769168918203e-05,
+      "loss": 0.9273,
+      "step": 3309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036993766740236e-05,
+      "loss": 1.0492,
+      "step": 3310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036295604105476e-05,
+      "loss": 0.9598,
+      "step": 3311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035597201296295e-05,
+      "loss": 0.9235,
+      "step": 3312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034898558331258e-05,
+      "loss": 0.9686,
+      "step": 3313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034199675228938e-05,
+      "loss": 1.0589,
+      "step": 3314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033500552007912e-05,
+      "loss": 0.9517,
+      "step": 3315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032801188686755e-05,
+      "loss": 1.0052,
+      "step": 3316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903210158528406e-05,
+      "loss": 0.4131,
+      "step": 3317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903140174181842e-05,
+      "loss": 0.9502,
+      "step": 3318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030701658308437e-05,
+      "loss": 0.9526,
+      "step": 3319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030001334772715e-05,
+      "loss": 0.9772,
+      "step": 3320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029300771229876e-05,
+      "loss": 1.0167,
+      "step": 3321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9028599967698533e-05,
+      "loss": 0.9802,
+      "step": 3322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027898924197316e-05,
+      "loss": 1.099,
+      "step": 3323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902719764074485e-05,
+      "loss": 0.7981,
+      "step": 3324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026496117359787e-05,
+      "loss": 0.9872,
+      "step": 3325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025794354060765e-05,
+      "loss": 1.0423,
+      "step": 3326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025092350866437e-05,
+      "loss": 0.9647,
+      "step": 3327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024390107795462e-05,
+      "loss": 1.1103,
+      "step": 3328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9023687624866505e-05,
+      "loss": 0.9823,
+      "step": 3329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022984902098237e-05,
+      "loss": 0.9819,
+      "step": 3330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022281939509332e-05,
+      "loss": 0.9738,
+      "step": 3331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021578737118483e-05,
+      "loss": 0.9565,
+      "step": 3332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902087529494437e-05,
+      "loss": 0.9597,
+      "step": 3333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020171613005698e-05,
+      "loss": 0.9945,
+      "step": 3334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019467691321166e-05,
+      "loss": 1.0429,
+      "step": 3335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018763529909484e-05,
+      "loss": 0.9561,
+      "step": 3336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901805912878937e-05,
+      "loss": 1.0164,
+      "step": 3337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901735448797954e-05,
+      "loss": 0.9501,
+      "step": 3338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016649607498726e-05,
+      "loss": 0.9694,
+      "step": 3339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015944487365664e-05,
+      "loss": 0.982,
+      "step": 3340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90152391275991e-05,
+      "loss": 0.996,
+      "step": 3341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901453352821777e-05,
+      "loss": 1.0243,
+      "step": 3342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013827689240434e-05,
+      "loss": 0.9869,
+      "step": 3343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901312161068586e-05,
+      "loss": 0.9472,
+      "step": 3344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012415292572802e-05,
+      "loss": 0.9679,
+      "step": 3345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901170873492004e-05,
+      "loss": 1.0599,
+      "step": 3346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901100193774635e-05,
+      "loss": 1.0795,
+      "step": 3347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010294901070522e-05,
+      "loss": 1.0035,
+      "step": 3348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900958762491135e-05,
+      "loss": 1.0534,
+      "step": 3349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9008880109287625e-05,
+      "loss": 0.9871,
+      "step": 3350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9008172354218156e-05,
+      "loss": 0.979,
+      "step": 3351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007464359721752e-05,
+      "loss": 0.8968,
+      "step": 3352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006756125817236e-05,
+      "loss": 1.0622,
+      "step": 3353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900604765252343e-05,
+      "loss": 1.0378,
+      "step": 3354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900533893985916e-05,
+      "loss": 0.9556,
+      "step": 3355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004629987843268e-05,
+      "loss": 1.002,
+      "step": 3356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003920796494592e-05,
+      "loss": 0.9826,
+      "step": 3357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003211365831986e-05,
+      "loss": 0.9713,
+      "step": 3358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002501695874306e-05,
+      "loss": 1.0567,
+      "step": 3359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001791786640417e-05,
+      "loss": 0.9764,
+      "step": 3360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001081638149176e-05,
+      "loss": 1.0844,
+      "step": 3361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900037125041947e-05,
+      "loss": 0.9399,
+      "step": 3362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999660623470172e-05,
+      "loss": 1.0196,
+      "step": 3363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998949757320174e-05,
+      "loss": 1.0471,
+      "step": 3364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899823865198837e-05,
+      "loss": 1.0255,
+      "step": 3365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899752730749366e-05,
+      "loss": 0.9623,
+      "step": 3366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996815723854948e-05,
+      "loss": 0.9662,
+      "step": 3367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996103901091153e-05,
+      "loss": 0.9796,
+      "step": 3368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899539183922119e-05,
+      "loss": 0.9918,
+      "step": 3369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994679538263985e-05,
+      "loss": 1.0244,
+      "step": 3370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993966998238467e-05,
+      "loss": 1.078,
+      "step": 3371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993254219163586e-05,
+      "loss": 1.0114,
+      "step": 3372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992541201058273e-05,
+      "loss": 0.8566,
+      "step": 3373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899182794394149e-05,
+      "loss": 0.9986,
+      "step": 3374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991114447832188e-05,
+      "loss": 1.0315,
+      "step": 3375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899040071274933e-05,
+      "loss": 0.98,
+      "step": 3376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898968673871189e-05,
+      "loss": 1.0485,
+      "step": 3377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988972525738846e-05,
+      "loss": 0.9823,
+      "step": 3378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988258073849178e-05,
+      "loss": 1.0266,
+      "step": 3379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987543383061873e-05,
+      "loss": 0.8518,
+      "step": 3380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986828453395936e-05,
+      "loss": 0.9761,
+      "step": 3381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986113284870357e-05,
+      "loss": 0.9618,
+      "step": 3382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8985397877504153e-05,
+      "loss": 0.9766,
+      "step": 3383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984682231316335e-05,
+      "loss": 1.0141,
+      "step": 3384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983966346325924e-05,
+      "loss": 1.0156,
+      "step": 3385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898325022255195e-05,
+      "loss": 0.9623,
+      "step": 3386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982533860013443e-05,
+      "loss": 0.9632,
+      "step": 3387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898181725872944e-05,
+      "loss": 0.9624,
+      "step": 3388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981100418719e-05,
+      "loss": 0.9961,
+      "step": 3389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898038334000117e-05,
+      "loss": 0.9944,
+      "step": 3390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979666022595e-05,
+      "loss": 0.9592,
+      "step": 3391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8978948466519567e-05,
+      "loss": 1.0764,
+      "step": 3392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897823067179394e-05,
+      "loss": 1.071,
+      "step": 3393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977512638437195e-05,
+      "loss": 0.9142,
+      "step": 3394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976794366468416e-05,
+      "loss": 0.9765,
+      "step": 3395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976075855906693e-05,
+      "loss": 1.0256,
+      "step": 3396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897535710677113e-05,
+      "loss": 0.9949,
+      "step": 3397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974638119080825e-05,
+      "loss": 1.0296,
+      "step": 3398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973918892854886e-05,
+      "loss": 1.0379,
+      "step": 3399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973199428112434e-05,
+      "loss": 0.9484,
+      "step": 3400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897247972487259e-05,
+      "loss": 0.9891,
+      "step": 3401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897175978315448e-05,
+      "loss": 0.979,
+      "step": 3402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971039602977244e-05,
+      "loss": 1.1412,
+      "step": 3403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970319184360023e-05,
+      "loss": 1.0087,
+      "step": 3404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896959852732196e-05,
+      "loss": 1.0134,
+      "step": 3405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968877631882212e-05,
+      "loss": 0.852,
+      "step": 3406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968156498059942e-05,
+      "loss": 0.8972,
+      "step": 3407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967435125874317e-05,
+      "loss": 1.0531,
+      "step": 3408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8966713515344503e-05,
+      "loss": 0.9583,
+      "step": 3409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965991666489687e-05,
+      "loss": 0.9072,
+      "step": 3410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965269579329053e-05,
+      "loss": 0.9803,
+      "step": 3411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964547253881795e-05,
+      "loss": 0.3625,
+      "step": 3412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963824690167106e-05,
+      "loss": 1.0861,
+      "step": 3413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963101888204197e-05,
+      "loss": 1.0289,
+      "step": 3414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962378848012273e-05,
+      "loss": 0.9622,
+      "step": 3415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961655569610557e-05,
+      "loss": 1.0082,
+      "step": 3416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960932053018273e-05,
+      "loss": 1.0399,
+      "step": 3417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960208298254646e-05,
+      "loss": 1.0562,
+      "step": 3418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895948430533892e-05,
+      "loss": 0.9526,
+      "step": 3419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895876007429033e-05,
+      "loss": 0.4181,
+      "step": 3420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895803560512813e-05,
+      "loss": 0.9955,
+      "step": 3421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895731089787158e-05,
+      "loss": 1.02,
+      "step": 3422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895658595253993e-05,
+      "loss": 0.986,
+      "step": 3423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955860769152455e-05,
+      "loss": 1.0084,
+      "step": 3424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955135347728434e-05,
+      "loss": 0.4308,
+      "step": 3425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8954409688287142e-05,
+      "loss": 1.0014,
+      "step": 3426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953683790847864e-05,
+      "loss": 1.0388,
+      "step": 3427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89529576554299e-05,
+      "loss": 0.3967,
+      "step": 3428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952231282052545e-05,
+      "loss": 0.9508,
+      "step": 3429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895150467073511e-05,
+      "loss": 0.8952,
+      "step": 3430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8950777821496898e-05,
+      "loss": 0.9685,
+      "step": 3431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895005073435724e-05,
+      "loss": 1.042,
+      "step": 3432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949323409335453e-05,
+      "loss": 1.0201,
+      "step": 3433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948595846450873e-05,
+      "loss": 1.0639,
+      "step": 3434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947868045722836e-05,
+      "loss": 0.9768,
+      "step": 3435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947140007170682e-05,
+      "loss": 1.049,
+      "step": 3436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8946411730813768e-05,
+      "loss": 0.9196,
+      "step": 3437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945683216671445e-05,
+      "loss": 1.0331,
+      "step": 3438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894495446476308e-05,
+      "loss": 1.0099,
+      "step": 3439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944225475108043e-05,
+      "loss": 1.0166,
+      "step": 3440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894349624772571e-05,
+      "loss": 1.0041,
+      "step": 3441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894276678263546e-05,
+      "loss": 1.0294,
+      "step": 3442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894203707985668e-05,
+      "loss": 0.9931,
+      "step": 3443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894130713940877e-05,
+      "loss": 1.1431,
+      "step": 3444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940576961311126e-05,
+      "loss": 1.0592,
+      "step": 3445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939846545583158e-05,
+      "loss": 0.9646,
+      "step": 3446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939115892244284e-05,
+      "loss": 0.9894,
+      "step": 3447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893838500131391e-05,
+      "loss": 0.9281,
+      "step": 3448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937653872811478e-05,
+      "loss": 0.988,
+      "step": 3449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893692250675641e-05,
+      "loss": 1.0701,
+      "step": 3450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893619090316815e-05,
+      "loss": 0.9091,
+      "step": 3451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935459062066142e-05,
+      "loss": 1.0219,
+      "step": 3452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893472698346984e-05,
+      "loss": 0.8951,
+      "step": 3453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933994667398695e-05,
+      "loss": 0.9768,
+      "step": 3454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893326211387218e-05,
+      "loss": 1.0239,
+      "step": 3455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932529322909757e-05,
+      "loss": 0.3951,
+      "step": 3456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893179629453091e-05,
+      "loss": 0.9873,
+      "step": 3457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931063028755113e-05,
+      "loss": 0.9999,
+      "step": 3458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930329525601867e-05,
+      "loss": 0.9822,
+      "step": 3459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929595785090657e-05,
+      "loss": 1.0348,
+      "step": 3460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928861807240994e-05,
+      "loss": 0.9889,
+      "step": 3461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928127592072378e-05,
+      "loss": 1.0288,
+      "step": 3462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8927393139604327e-05,
+      "loss": 1.0515,
+      "step": 3463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8926658449856364e-05,
+      "loss": 1.0029,
+      "step": 3464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925923522848015e-05,
+      "loss": 1.1093,
+      "step": 3465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925188358598815e-05,
+      "loss": 0.9865,
+      "step": 3466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924452957128298e-05,
+      "loss": 0.9804,
+      "step": 3467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923717318456016e-05,
+      "loss": 0.9948,
+      "step": 3468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892298144260152e-05,
+      "loss": 1.0172,
+      "step": 3469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892224532958437e-05,
+      "loss": 1.0505,
+      "step": 3470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921508979424126e-05,
+      "loss": 1.0534,
+      "step": 3471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920772392140364e-05,
+      "loss": 1.0323,
+      "step": 3472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892003556775266e-05,
+      "loss": 1.0043,
+      "step": 3473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89192985062806e-05,
+      "loss": 1.0367,
+      "step": 3474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918561207743767e-05,
+      "loss": 1.0089,
+      "step": 3475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891782367216177e-05,
+      "loss": 1.009,
+      "step": 3476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917085899554202e-05,
+      "loss": 1.035,
+      "step": 3477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8916347889940674e-05,
+      "loss": 0.9148,
+      "step": 3478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915609643340806e-05,
+      "loss": 1.0086,
+      "step": 3479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891487115977421e-05,
+      "loss": 1.0553,
+      "step": 3480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914132439260526e-05,
+      "loss": 0.9863,
+      "step": 3481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891339348181938e-05,
+      "loss": 0.9704,
+      "step": 3482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891265428747042e-05,
+      "loss": 1.0256,
+      "step": 3483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891191485623328e-05,
+      "loss": 0.9734,
+      "step": 3484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911175188127625e-05,
+      "loss": 1.0097,
+      "step": 3485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910435283173113e-05,
+      "loss": 0.9961,
+      "step": 3486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909695141389405e-05,
+      "loss": 0.9344,
+      "step": 3487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890895476279618e-05,
+      "loss": 1.0402,
+      "step": 3488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890821414741311e-05,
+      "loss": 0.9565,
+      "step": 3489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890747329525988e-05,
+      "loss": 1.0144,
+      "step": 3490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890673220635619e-05,
+      "loss": 0.968,
+      "step": 3491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905990880721722e-05,
+      "loss": 1.0074,
+      "step": 3492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905249318376192e-05,
+      "loss": 1.0124,
+      "step": 3493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890450751933931e-05,
+      "loss": 1.1217,
+      "step": 3494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890376548363078e-05,
+      "loss": 1.1196,
+      "step": 3495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903023211270337e-05,
+      "loss": 1.0067,
+      "step": 3496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902280702277706e-05,
+      "loss": 1.0333,
+      "step": 3497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890153795667262e-05,
+      "loss": 1.048,
+      "step": 3498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890079497447482e-05,
+      "loss": 0.9844,
+      "step": 3499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900051755704057e-05,
+      "loss": 0.9338,
+      "step": 3500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8899308300380085e-05,
+      "loss": 1.0176,
+      "step": 3501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889856460852266e-05,
+      "loss": 1.0091,
+      "step": 3502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8897820680151548e-05,
+      "loss": 1.033,
+      "step": 3503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889707651528653e-05,
+      "loss": 1.0285,
+      "step": 3504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896332113947376e-05,
+      "loss": 1.0231,
+      "step": 3505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889558747615388e-05,
+      "loss": 0.9661,
+      "step": 3506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894842601925823e-05,
+      "loss": 1.0629,
+      "step": 3507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889409749128301e-05,
+      "loss": 0.8848,
+      "step": 3508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893352144245245e-05,
+      "loss": 0.9842,
+      "step": 3509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892606560832335e-05,
+      "loss": 1.0137,
+      "step": 3510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88918607410641e-05,
+      "loss": 1.0481,
+      "step": 3511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891114684960367e-05,
+      "loss": 0.9617,
+      "step": 3512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890368392540953e-05,
+      "loss": 1.0569,
+      "step": 3513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88896218638257e-05,
+      "loss": 0.9904,
+      "step": 3514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888875098834457e-05,
+      "loss": 0.9818,
+      "step": 3515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888128097587065e-05,
+      "loss": 0.9814,
+      "step": 3516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887380860103377e-05,
+      "loss": 1.0331,
+      "step": 3517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886633386403257e-05,
+      "loss": 0.4075,
+      "step": 3518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888588567650657e-05,
+      "loss": 1.0307,
+      "step": 3519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888513773043319e-05,
+      "loss": 1.024,
+      "step": 3520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8884389548202998e-05,
+      "loss": 0.9423,
+      "step": 3521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883641129835882e-05,
+      "loss": 0.9794,
+      "step": 3522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882892475351725e-05,
+      "loss": 1.043,
+      "step": 3523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882143584770436e-05,
+      "loss": 0.8819,
+      "step": 3524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881394458111914e-05,
+      "loss": 0.941,
+      "step": 3525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888064509539607e-05,
+      "loss": 1.0443,
+      "step": 3526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887989549664282e-05,
+      "loss": 1.0515,
+      "step": 3527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887914566187209e-05,
+      "loss": 1.0568,
+      "step": 3528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878395591103814e-05,
+      "loss": 1.0604,
+      "step": 3529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887764528435792e-05,
+      "loss": 1.0131,
+      "step": 3530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887689474165435e-05,
+      "loss": 1.0382,
+      "step": 3531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887614396301306e-05,
+      "loss": 0.9994,
+      "step": 3532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875392948454e-05,
+      "loss": 1.0192,
+      "step": 3533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874641697997137e-05,
+      "loss": 0.9929,
+      "step": 3534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873890211662426e-05,
+      "loss": 0.9769,
+      "step": 3535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887313848946985e-05,
+      "loss": 1.0622,
+      "step": 3536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872386531439392e-05,
+      "loss": 1.1338,
+      "step": 3537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871634337591026e-05,
+      "loss": 0.9517,
+      "step": 3538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870881907944757e-05,
+      "loss": 1.0502,
+      "step": 3539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870129242520576e-05,
+      "loss": 0.9452,
+      "step": 3540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869376341338493e-05,
+      "loss": 0.9894,
+      "step": 3541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868623204418514e-05,
+      "loss": 1.0327,
+      "step": 3542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886786983178066e-05,
+      "loss": 0.9204,
+      "step": 3543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867116223444952e-05,
+      "loss": 1.0032,
+      "step": 3544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866362379431423e-05,
+      "loss": 0.9495,
+      "step": 3545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886560829976011e-05,
+      "loss": 0.8948,
+      "step": 3546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864853984451054e-05,
+      "loss": 0.9787,
+      "step": 3547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864099433524302e-05,
+      "loss": 0.9584,
+      "step": 3548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863344646999913e-05,
+      "loss": 0.9055,
+      "step": 3549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8862589624897945e-05,
+      "loss": 0.9616,
+      "step": 3550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861834367238465e-05,
+      "loss": 1.0265,
+      "step": 3551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861078874041553e-05,
+      "loss": 1.0736,
+      "step": 3552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886032314532728e-05,
+      "loss": 0.9674,
+      "step": 3553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885956718111574e-05,
+      "loss": 1.0128,
+      "step": 3554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885881098142702e-05,
+      "loss": 1.0724,
+      "step": 3555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858054546281223e-05,
+      "loss": 1.026,
+      "step": 3556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857297875698455e-05,
+      "loss": 0.9223,
+      "step": 3557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856540969698826e-05,
+      "loss": 0.9462,
+      "step": 3558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855783828302454e-05,
+      "loss": 1.022,
+      "step": 3559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855026451529455e-05,
+      "loss": 0.9511,
+      "step": 3560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854268839399973e-05,
+      "loss": 1.0244,
+      "step": 3561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853510991934137e-05,
+      "loss": 0.4281,
+      "step": 3562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885275290915209e-05,
+      "loss": 0.9756,
+      "step": 3563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851994591073982e-05,
+      "loss": 1.0232,
+      "step": 3564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851236037719968e-05,
+      "loss": 1.0379,
+      "step": 3565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850477249110203e-05,
+      "loss": 0.9868,
+      "step": 3566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849718225264868e-05,
+      "loss": 0.9523,
+      "step": 3567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848958966204127e-05,
+      "loss": 0.9952,
+      "step": 3568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884819947194816e-05,
+      "loss": 0.9118,
+      "step": 3569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847439742517168e-05,
+      "loss": 0.9205,
+      "step": 3570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884667977793132e-05,
+      "loss": 1.0478,
+      "step": 3571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845919578210833e-05,
+      "loss": 0.9605,
+      "step": 3572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845159143375907e-05,
+      "loss": 1.0208,
+      "step": 3573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884439847344675e-05,
+      "loss": 1.0245,
+      "step": 3574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884363756844359e-05,
+      "loss": 0.9871,
+      "step": 3575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8842876428386635e-05,
+      "loss": 0.9368,
+      "step": 3576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884211505329613e-05,
+      "loss": 0.9864,
+      "step": 3577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841353443192304e-05,
+      "loss": 1.0164,
+      "step": 3578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840591598095398e-05,
+      "loss": 1.025,
+      "step": 3579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839829518025667e-05,
+      "loss": 1.0491,
+      "step": 3580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839067203003366e-05,
+      "loss": 0.9604,
+      "step": 3581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838304653048748e-05,
+      "loss": 0.9283,
+      "step": 3582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837541868182092e-05,
+      "loss": 1.015,
+      "step": 3583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836778848423662e-05,
+      "loss": 1.0514,
+      "step": 3584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836015593793747e-05,
+      "loss": 0.3406,
+      "step": 3585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883525210431263e-05,
+      "loss": 1.0869,
+      "step": 3586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834488380000598e-05,
+      "loss": 1.0085,
+      "step": 3587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833724420877957e-05,
+      "loss": 0.9591,
+      "step": 3588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883296022696501e-05,
+      "loss": 0.3904,
+      "step": 3589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883219579828207e-05,
+      "loss": 1.0588,
+      "step": 3590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831431134849453e-05,
+      "loss": 0.963,
+      "step": 3591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883066623668748e-05,
+      "loss": 1.0125,
+      "step": 3592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829901103816485e-05,
+      "loss": 0.9806,
+      "step": 3593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829135736256804e-05,
+      "loss": 0.9464,
+      "step": 3594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828370134028777e-05,
+      "loss": 1.0889,
+      "step": 3595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827604297152758e-05,
+      "loss": 0.4584,
+      "step": 3596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826838225649097e-05,
+      "loss": 0.9795,
+      "step": 3597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826071919538158e-05,
+      "loss": 0.9206,
+      "step": 3598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825305378840305e-05,
+      "loss": 0.9892,
+      "step": 3599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882453860357592e-05,
+      "loss": 1.0023,
+      "step": 3600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823771593765375e-05,
+      "loss": 0.883,
+      "step": 3601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823004349429057e-05,
+      "loss": 0.966,
+      "step": 3602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882223687058736e-05,
+      "loss": 0.372,
+      "step": 3603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821469157260687e-05,
+      "loss": 0.9482,
+      "step": 3604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820701209469436e-05,
+      "loss": 0.9666,
+      "step": 3605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881993302723402e-05,
+      "loss": 0.9848,
+      "step": 3606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819164610574856e-05,
+      "loss": 0.9884,
+      "step": 3607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818395959512372e-05,
+      "loss": 0.4023,
+      "step": 3608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817627074067e-05,
+      "loss": 1.0152,
+      "step": 3609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881685795425916e-05,
+      "loss": 1.0305,
+      "step": 3610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816088600109316e-05,
+      "loss": 1.0035,
+      "step": 3611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88153190116379e-05,
+      "loss": 0.9989,
+      "step": 3612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881454918886537e-05,
+      "loss": 0.9332,
+      "step": 3613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813779131812196e-05,
+      "loss": 0.9961,
+      "step": 3614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813008840498836e-05,
+      "loss": 1.0156,
+      "step": 3615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881223831494577e-05,
+      "loss": 1.0716,
+      "step": 3616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881146755517347e-05,
+      "loss": 1.0216,
+      "step": 3617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810696561202423e-05,
+      "loss": 0.9798,
+      "step": 3618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880992533305313e-05,
+      "loss": 1.0182,
+      "step": 3619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809153870746083e-05,
+      "loss": 1.016,
+      "step": 3620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808382174301788e-05,
+      "loss": 1.0274,
+      "step": 3621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880761024374075e-05,
+      "loss": 0.9925,
+      "step": 3622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806838079083497e-05,
+      "loss": 0.9387,
+      "step": 3623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880606568035054e-05,
+      "loss": 1.0035,
+      "step": 3624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880529304756242e-05,
+      "loss": 1.0114,
+      "step": 3625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804520180739664e-05,
+      "loss": 0.9621,
+      "step": 3626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880374707990282e-05,
+      "loss": 0.9509,
+      "step": 3627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802973745072432e-05,
+      "loss": 0.9827,
+      "step": 3628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802200176269055e-05,
+      "loss": 0.9948,
+      "step": 3629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880142637351325e-05,
+      "loss": 1.041,
+      "step": 3630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800652336825585e-05,
+      "loss": 0.4113,
+      "step": 3631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799878066226632e-05,
+      "loss": 1.0516,
+      "step": 3632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799103561736973e-05,
+      "loss": 0.9698,
+      "step": 3633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879832882337719e-05,
+      "loss": 0.9979,
+      "step": 3634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797553851167873e-05,
+      "loss": 1.0294,
+      "step": 3635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796778645129623e-05,
+      "loss": 0.9496,
+      "step": 3636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796003205283045e-05,
+      "loss": 0.9565,
+      "step": 3637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795227531648747e-05,
+      "loss": 0.4274,
+      "step": 3638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794451624247348e-05,
+      "loss": 0.9827,
+      "step": 3639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793675483099466e-05,
+      "loss": 0.9949,
+      "step": 3640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792899108225737e-05,
+      "loss": 1.0773,
+      "step": 3641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792122499646788e-05,
+      "loss": 0.9953,
+      "step": 3642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791345657383267e-05,
+      "loss": 0.9928,
+      "step": 3643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790568581455818e-05,
+      "loss": 0.8796,
+      "step": 3644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789791271885097e-05,
+      "loss": 1.004,
+      "step": 3645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878901372869176e-05,
+      "loss": 0.9916,
+      "step": 3646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878823595189648e-05,
+      "loss": 0.9653,
+      "step": 3647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878745794151992e-05,
+      "loss": 0.9881,
+      "step": 3648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878667969758277e-05,
+      "loss": 1.0373,
+      "step": 3649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785901220105708e-05,
+      "loss": 0.9642,
+      "step": 3650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785122509109425e-05,
+      "loss": 1.0528,
+      "step": 3651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878434356461462e-05,
+      "loss": 1.0111,
+      "step": 3652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783564386641994e-05,
+      "loss": 0.9493,
+      "step": 3653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878278497521226e-05,
+      "loss": 0.9913,
+      "step": 3654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878200533034613e-05,
+      "loss": 0.9689,
+      "step": 3655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878122545206433e-05,
+      "loss": 1.0102,
+      "step": 3656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780445340387587e-05,
+      "loss": 1.0001,
+      "step": 3657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779664995336636e-05,
+      "loss": 1.042,
+      "step": 3658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778884416932216e-05,
+      "loss": 0.9645,
+      "step": 3659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778103605195073e-05,
+      "loss": 1.0082,
+      "step": 3660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777322560145962e-05,
+      "loss": 1.0009,
+      "step": 3661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776541281805643e-05,
+      "loss": 1.012,
+      "step": 3662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877575977019488e-05,
+      "loss": 0.9189,
+      "step": 3663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774978025334446e-05,
+      "loss": 0.9786,
+      "step": 3664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774196047245116e-05,
+      "loss": 1.0269,
+      "step": 3665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877341383594768e-05,
+      "loss": 1.0085,
+      "step": 3666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772631391462923e-05,
+      "loss": 1.0179,
+      "step": 3667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877184871381164e-05,
+      "loss": 0.9169,
+      "step": 3668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877106580301464e-05,
+      "loss": 1.0255,
+      "step": 3669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770282659092728e-05,
+      "loss": 1.0119,
+      "step": 3670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769499282066716e-05,
+      "loss": 0.9552,
+      "step": 3671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768715671957434e-05,
+      "loss": 0.9688,
+      "step": 3672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767931828785698e-05,
+      "loss": 0.9863,
+      "step": 3673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767147752572352e-05,
+      "loss": 0.4713,
+      "step": 3674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766363443338233e-05,
+      "loss": 0.9991,
+      "step": 3675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765578901104185e-05,
+      "loss": 1.0185,
+      "step": 3676
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876479412589106e-05,
+      "loss": 1.1327,
+      "step": 3677
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764009117719717e-05,
+      "loss": 0.9443,
+      "step": 3678
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763223876611022e-05,
+      "loss": 0.9858,
+      "step": 3679
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762438402585845e-05,
+      "loss": 1.0978,
+      "step": 3680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761652695665065e-05,
+      "loss": 0.9832,
+      "step": 3681
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760866755869558e-05,
+      "loss": 0.9103,
+      "step": 3682
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876008058322022e-05,
+      "loss": 0.9786,
+      "step": 3683
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875929417773795e-05,
+      "loss": 1.0169,
+      "step": 3684
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875850753944364e-05,
+      "loss": 0.9896,
+      "step": 3685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757720668358204e-05,
+      "loss": 1.0093,
+      "step": 3686
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756933564502557e-05,
+      "loss": 1.0423,
+      "step": 3687
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756146227897613e-05,
+      "loss": 0.9676,
+      "step": 3688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755358658564303e-05,
+      "loss": 0.9263,
+      "step": 3689
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754570856523565e-05,
+      "loss": 0.9972,
+      "step": 3690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753782821796326e-05,
+      "loss": 1.011,
+      "step": 3691
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875299455440354e-05,
+      "loss": 1.0057,
+      "step": 3692
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875220605436616e-05,
+      "loss": 0.9549,
+      "step": 3693
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875141732170513e-05,
+      "loss": 1.0226,
+      "step": 3694
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750628356441428e-05,
+      "loss": 0.9894,
+      "step": 3695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749839158596018e-05,
+      "loss": 0.9925,
+      "step": 3696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749049728189878e-05,
+      "loss": 1.0066,
+      "step": 3697
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748260065243985e-05,
+      "loss": 0.9216,
+      "step": 3698
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747470169779332e-05,
+      "loss": 1.0328,
+      "step": 3699
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746680041816915e-05,
+      "loss": 1.0043,
+      "step": 3700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874588968137773e-05,
+      "loss": 0.9689,
+      "step": 3701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745099088482788e-05,
+      "loss": 0.8442,
+      "step": 3702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744308263153097e-05,
+      "loss": 1.0096,
+      "step": 3703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743517205409682e-05,
+      "loss": 1.0808,
+      "step": 3704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874272591527356e-05,
+      "loss": 1.0828,
+      "step": 3705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741934392765777e-05,
+      "loss": 0.943,
+      "step": 3706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741142637907357e-05,
+      "loss": 1.0241,
+      "step": 3707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874035065071935e-05,
+      "loss": 1.0323,
+      "step": 3708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739558431222808e-05,
+      "loss": 1.072,
+      "step": 3709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738765979438778e-05,
+      "loss": 1.0179,
+      "step": 3710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737973295388333e-05,
+      "loss": 0.9446,
+      "step": 3711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737180379092536e-05,
+      "loss": 1.0189,
+      "step": 3712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736387230572465e-05,
+      "loss": 0.4221,
+      "step": 3713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735593849849198e-05,
+      "loss": 0.9149,
+      "step": 3714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734800236943825e-05,
+      "loss": 1.0151,
+      "step": 3715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873400639187744e-05,
+      "loss": 0.9646,
+      "step": 3716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733212314671136e-05,
+      "loss": 0.8856,
+      "step": 3717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873241800534602e-05,
+      "loss": 1.0237,
+      "step": 3718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731623463923213e-05,
+      "loss": 0.9696,
+      "step": 3719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730828690423826e-05,
+      "loss": 0.9393,
+      "step": 3720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730033684868983e-05,
+      "loss": 0.9463,
+      "step": 3721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729238447279814e-05,
+      "loss": 1.0319,
+      "step": 3722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872844297767746e-05,
+      "loss": 0.9781,
+      "step": 3723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727647276083056e-05,
+      "loss": 1.0142,
+      "step": 3724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726851342517757e-05,
+      "loss": 0.9578,
+      "step": 3725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726055177002717e-05,
+      "loss": 1.0085,
+      "step": 3726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725258779559096e-05,
+      "loss": 1.0619,
+      "step": 3727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724462150208062e-05,
+      "loss": 0.9538,
+      "step": 3728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872366528897079e-05,
+      "loss": 1.0293,
+      "step": 3729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722868195868457e-05,
+      "loss": 1.0708,
+      "step": 3730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872207087092225e-05,
+      "loss": 0.9983,
+      "step": 3731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721273314153362e-05,
+      "loss": 0.9671,
+      "step": 3732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720475525582985e-05,
+      "loss": 0.9805,
+      "step": 3733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871967750523233e-05,
+      "loss": 0.9939,
+      "step": 3734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871887925312261e-05,
+      "loss": 0.9891,
+      "step": 3735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718080769275033e-05,
+      "loss": 0.9547,
+      "step": 3736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871728205371083e-05,
+      "loss": 0.9621,
+      "step": 3737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716483106451224e-05,
+      "loss": 1.0341,
+      "step": 3738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715683927517453e-05,
+      "loss": 1.0127,
+      "step": 3739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714884516930752e-05,
+      "loss": 1.0046,
+      "step": 3740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714084874712382e-05,
+      "loss": 0.9784,
+      "step": 3741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713285000883586e-05,
+      "loss": 0.9772,
+      "step": 3742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712484895465624e-05,
+      "loss": 1.0085,
+      "step": 3743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711684558479765e-05,
+      "loss": 1.0039,
+      "step": 3744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710883989947278e-05,
+      "loss": 1.082,
+      "step": 3745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871008318988945e-05,
+      "loss": 1.0149,
+      "step": 3746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709282158327552e-05,
+      "loss": 0.9187,
+      "step": 3747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708480895282882e-05,
+      "loss": 0.9744,
+      "step": 3748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870767940077674e-05,
+      "loss": 0.9426,
+      "step": 3749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706877674830423e-05,
+      "loss": 0.973,
+      "step": 3750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870607571746524e-05,
+      "loss": 0.8968,
+      "step": 3751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705273528702507e-05,
+      "loss": 0.4378,
+      "step": 3752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870447110856355e-05,
+      "loss": 0.9259,
+      "step": 3753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870366845706969e-05,
+      "loss": 0.3804,
+      "step": 3754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870286557424226e-05,
+      "loss": 0.9711,
+      "step": 3755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702062460102606e-05,
+      "loss": 0.5202,
+      "step": 3756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870125911467207e-05,
+      "loss": 0.9445,
+      "step": 3757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700455537972006e-05,
+      "loss": 0.9752,
+      "step": 3758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699651730023768e-05,
+      "loss": 1.0006,
+      "step": 3759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869884769084872e-05,
+      "loss": 1.0113,
+      "step": 3760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698043420468243e-05,
+      "loss": 0.9766,
+      "step": 3761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86972389189037e-05,
+      "loss": 0.9779,
+      "step": 3762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696434186176486e-05,
+      "loss": 1.0167,
+      "step": 3763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869562922230798e-05,
+      "loss": 0.9164,
+      "step": 3764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869482402731958e-05,
+      "loss": 0.9569,
+      "step": 3765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694018601232687e-05,
+      "loss": 0.9763,
+      "step": 3766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869321294406871e-05,
+      "loss": 0.9907,
+      "step": 3767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692407055849063e-05,
+      "loss": 1.0053,
+      "step": 3768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869160093659516e-05,
+      "loss": 1.0449,
+      "step": 3769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690794586328435e-05,
+      "loss": 0.9228,
+      "step": 3770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689988005070315e-05,
+      "loss": 1.0052,
+      "step": 3771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689181192842238e-05,
+      "loss": 0.9925,
+      "step": 3772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688374149665647e-05,
+      "loss": 1.03,
+      "step": 3773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687566875562e-05,
+      "loss": 1.0207,
+      "step": 3774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686759370552743e-05,
+      "loss": 0.9845,
+      "step": 3775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685951634659344e-05,
+      "loss": 1.003,
+      "step": 3776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685143667903273e-05,
+      "loss": 0.985,
+      "step": 3777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684335470306e-05,
+      "loss": 1.0165,
+      "step": 3778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868352704188901e-05,
+      "loss": 0.9594,
+      "step": 3779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682718382673787e-05,
+      "loss": 0.9206,
+      "step": 3780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681909492681826e-05,
+      "loss": 1.0107,
+      "step": 3781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868110037193463e-05,
+      "loss": 0.9681,
+      "step": 3782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680291020453694e-05,
+      "loss": 0.9482,
+      "step": 3783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679481438260543e-05,
+      "loss": 0.9785,
+      "step": 3784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867867162537669e-05,
+      "loss": 0.9641,
+      "step": 3785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677861581823648e-05,
+      "loss": 1.0566,
+      "step": 3786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677051307622963e-05,
+      "loss": 1.0166,
+      "step": 3787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676240802796162e-05,
+      "loss": 1.0338,
+      "step": 3788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675430067364794e-05,
+      "loss": 1.0048,
+      "step": 3789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86746191013504e-05,
+      "loss": 0.8923,
+      "step": 3790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673807904774536e-05,
+      "loss": 0.9994,
+      "step": 3791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672996477658767e-05,
+      "loss": 1.0118,
+      "step": 3792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672184820024655e-05,
+      "loss": 0.9236,
+      "step": 3793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671372931893775e-05,
+      "loss": 1.0805,
+      "step": 3794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670560813287708e-05,
+      "loss": 1.0441,
+      "step": 3795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8669748464228035e-05,
+      "loss": 0.972,
+      "step": 3796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866893588473635e-05,
+      "loss": 0.4477,
+      "step": 3797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668123074834254e-05,
+      "loss": 0.8868,
+      "step": 3798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667310034543343e-05,
+      "loss": 0.4878,
+      "step": 3799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666496763885233e-05,
+      "loss": 0.9998,
+      "step": 3800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866568326288153e-05,
+      "loss": 0.958,
+      "step": 3801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866486953155387e-05,
+      "loss": 1.0471,
+      "step": 3802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664055569923872e-05,
+      "loss": 1.0311,
+      "step": 3803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663241378013172e-05,
+      "loss": 0.9887,
+      "step": 3804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866242695584341e-05,
+      "loss": 0.9261,
+      "step": 3805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661612303436235e-05,
+      "loss": 0.9449,
+      "step": 3806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660797420813298e-05,
+      "loss": 0.9329,
+      "step": 3807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659982307996255e-05,
+      "loss": 0.9548,
+      "step": 3808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659166965006774e-05,
+      "loss": 0.9471,
+      "step": 3809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658351391866522e-05,
+      "loss": 1.0406,
+      "step": 3810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8657535588597186e-05,
+      "loss": 1.0668,
+      "step": 3811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8656719555220433e-05,
+      "loss": 1.0669,
+      "step": 3812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655903291757968e-05,
+      "loss": 0.9602,
+      "step": 3813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655086798231475e-05,
+      "loss": 0.9461,
+      "step": 3814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865427007466266e-05,
+      "loss": 0.9909,
+      "step": 3815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653453121073234e-05,
+      "loss": 1.0369,
+      "step": 3816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652635937484903e-05,
+      "loss": 0.8921,
+      "step": 3817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651818523919394e-05,
+      "loss": 0.9313,
+      "step": 3818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865100088039843e-05,
+      "loss": 0.9758,
+      "step": 3819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650183006943742e-05,
+      "loss": 0.9496,
+      "step": 3820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649364903577068e-05,
+      "loss": 0.9733,
+      "step": 3821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648546570320157e-05,
+      "loss": 0.8941,
+      "step": 3822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647728007194753e-05,
+      "loss": 1.0036,
+      "step": 3823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646909214222615e-05,
+      "loss": 1.0668,
+      "step": 3824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864609019142551e-05,
+      "loss": 0.9703,
+      "step": 3825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645270938825197e-05,
+      "loss": 0.9852,
+      "step": 3826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644451456443457e-05,
+      "loss": 1.0265,
+      "step": 3827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643631744302074e-05,
+      "loss": 1.0226,
+      "step": 3828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642811802422826e-05,
+      "loss": 0.4497,
+      "step": 3829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641991630827517e-05,
+      "loss": 0.9669,
+      "step": 3830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641171229537937e-05,
+      "loss": 1.0252,
+      "step": 3831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640350598575895e-05,
+      "loss": 1.0016,
+      "step": 3832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86395297379632e-05,
+      "loss": 0.9315,
+      "step": 3833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638708647721677e-05,
+      "loss": 0.9394,
+      "step": 3834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863788732787314e-05,
+      "loss": 1.0112,
+      "step": 3835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637065778439425e-05,
+      "loss": 1.0796,
+      "step": 3836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636243999442368e-05,
+      "loss": 0.9118,
+      "step": 3837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863542199090381e-05,
+      "loss": 0.9793,
+      "step": 3838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634599752845594e-05,
+      "loss": 1.0449,
+      "step": 3839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633777285289582e-05,
+      "loss": 0.9511,
+      "step": 3840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632954588257626e-05,
+      "loss": 0.4994,
+      "step": 3841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632131661771597e-05,
+      "loss": 0.9542,
+      "step": 3842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631308505853373e-05,
+      "loss": 1.0731,
+      "step": 3843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630485120524822e-05,
+      "loss": 0.9822,
+      "step": 3844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629661505807834e-05,
+      "loss": 1.0418,
+      "step": 3845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628837661724298e-05,
+      "loss": 0.9092,
+      "step": 3846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628013588296116e-05,
+      "loss": 1.0638,
+      "step": 3847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627189285545184e-05,
+      "loss": 0.9677,
+      "step": 3848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626364753493415e-05,
+      "loss": 0.9826,
+      "step": 3849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625539992162722e-05,
+      "loss": 1.0595,
+      "step": 3850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862471500157503e-05,
+      "loss": 1.0353,
+      "step": 3851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862388978175226e-05,
+      "loss": 0.938,
+      "step": 3852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623064332716354e-05,
+      "loss": 1.0255,
+      "step": 3853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622238654489243e-05,
+      "loss": 1.0109,
+      "step": 3854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621412747092875e-05,
+      "loss": 1.0776,
+      "step": 3855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620586610549206e-05,
+      "loss": 1.078,
+      "step": 3856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619760244880188e-05,
+      "loss": 0.9734,
+      "step": 3857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618933650107788e-05,
+      "loss": 0.9792,
+      "step": 3858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618106826253976e-05,
+      "loss": 1.0495,
+      "step": 3859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617279773340726e-05,
+      "loss": 0.9381,
+      "step": 3860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616452491390022e-05,
+      "loss": 1.0216,
+      "step": 3861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615624980423856e-05,
+      "loss": 0.9741,
+      "step": 3862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614797240464212e-05,
+      "loss": 1.0152,
+      "step": 3863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613969271533098e-05,
+      "loss": 1.0247,
+      "step": 3864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613141073652522e-05,
+      "loss": 0.9692,
+      "step": 3865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612312646844487e-05,
+      "loss": 1.052,
+      "step": 3866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611483991131022e-05,
+      "loss": 1.0062,
+      "step": 3867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610655106534147e-05,
+      "loss": 0.9684,
+      "step": 3868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609825993075896e-05,
+      "loss": 1.0335,
+      "step": 3869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86089966507783e-05,
+      "loss": 1.0438,
+      "step": 3870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8608167079663407e-05,
+      "loss": 1.0771,
+      "step": 3871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607337279753266e-05,
+      "loss": 0.4889,
+      "step": 3872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8606507251069928e-05,
+      "loss": 0.8907,
+      "step": 3873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860567699363546e-05,
+      "loss": 0.953,
+      "step": 3874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604846507471924e-05,
+      "loss": 0.9963,
+      "step": 3875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604015792601395e-05,
+      "loss": 0.9669,
+      "step": 3876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8603184849045953e-05,
+      "loss": 0.9166,
+      "step": 3877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860235367682769e-05,
+      "loss": 0.9706,
+      "step": 3878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860152227596868e-05,
+      "loss": 0.9948,
+      "step": 3879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860069064649104e-05,
+      "loss": 0.5138,
+      "step": 3880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859985878841686e-05,
+      "loss": 1.035,
+      "step": 3881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599026701768258e-05,
+      "loss": 0.9511,
+      "step": 3882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598194386567346e-05,
+      "loss": 0.9565,
+      "step": 3883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859736184283625e-05,
+      "loss": 1.0306,
+      "step": 3884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596529070597093e-05,
+      "loss": 0.9731,
+      "step": 3885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595696069872013e-05,
+      "loss": 0.9957,
+      "step": 3886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594862840683148e-05,
+      "loss": 0.9509,
+      "step": 3887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594029383052644e-05,
+      "loss": 0.9624,
+      "step": 3888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8593195697002656e-05,
+      "loss": 1.044,
+      "step": 3889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859236178255534e-05,
+      "loss": 0.8952,
+      "step": 3890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591527639732862e-05,
+      "loss": 0.9727,
+      "step": 3891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590693268557392e-05,
+      "loss": 0.9933,
+      "step": 3892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589858669051106e-05,
+      "loss": 1.0778,
+      "step": 3893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589023841236183e-05,
+      "loss": 0.9691,
+      "step": 3894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588188785134822e-05,
+      "loss": 0.985,
+      "step": 3895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587353500769208e-05,
+      "loss": 1.0169,
+      "step": 3896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586517988161547e-05,
+      "loss": 1.0086,
+      "step": 3897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585682247334046e-05,
+      "loss": 0.9811,
+      "step": 3898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584846278308915e-05,
+      "loss": 1.0881,
+      "step": 3899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584010081108376e-05,
+      "loss": 0.9013,
+      "step": 3900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858317365575465e-05,
+      "loss": 1.025,
+      "step": 3901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582337002269975e-05,
+      "loss": 1.0144,
+      "step": 3902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581500120676582e-05,
+      "loss": 1.0234,
+      "step": 3903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858066301099672e-05,
+      "loss": 0.9063,
+      "step": 3904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857982567325263e-05,
+      "loss": 1.0201,
+      "step": 3905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578988107466577e-05,
+      "loss": 0.9674,
+      "step": 3906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578150313660814e-05,
+      "loss": 0.9081,
+      "step": 3907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577312291857615e-05,
+      "loss": 0.9943,
+      "step": 3908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576474042079255e-05,
+      "loss": 0.9439,
+      "step": 3909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575635564348005e-05,
+      "loss": 0.9847,
+      "step": 3910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857479685868616e-05,
+      "loss": 0.94,
+      "step": 3911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8573957925116003e-05,
+      "loss": 0.9963,
+      "step": 3912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8573118763659843e-05,
+      "loss": 1.0345,
+      "step": 3913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857227937433997e-05,
+      "loss": 0.4433,
+      "step": 3914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857143975717871e-05,
+      "loss": 1.0052,
+      "step": 3915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8570599912198365e-05,
+      "loss": 0.9017,
+      "step": 3916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569759839421263e-05,
+      "loss": 0.9572,
+      "step": 3917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8568919538869737e-05,
+      "loss": 1.0665,
+      "step": 3918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856807901056611e-05,
+      "loss": 0.9146,
+      "step": 3919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567238254532735e-05,
+      "loss": 0.9493,
+      "step": 3920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566397270791948e-05,
+      "loss": 0.9392,
+      "step": 3921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565556059366107e-05,
+      "loss": 0.9716,
+      "step": 3922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564714620277565e-05,
+      "loss": 1.0618,
+      "step": 3923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563872953548693e-05,
+      "loss": 0.9976,
+      "step": 3924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856303105920186e-05,
+      "loss": 0.958,
+      "step": 3925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562188937259443e-05,
+      "loss": 0.9991,
+      "step": 3926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856134658774382e-05,
+      "loss": 1.0053,
+      "step": 3927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856050401067738e-05,
+      "loss": 0.9399,
+      "step": 3928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559661206082524e-05,
+      "loss": 0.9823,
+      "step": 3929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558818173981647e-05,
+      "loss": 1.1063,
+      "step": 3930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557974914397162e-05,
+      "loss": 1.006,
+      "step": 3931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557131427351473e-05,
+      "loss": 1.0254,
+      "step": 3932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556287712867006e-05,
+      "loss": 1.0578,
+      "step": 3933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555443770966186e-05,
+      "loss": 1.0687,
+      "step": 3934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855459960167144e-05,
+      "loss": 0.9877,
+      "step": 3935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553755205005208e-05,
+      "loss": 1.0343,
+      "step": 3936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552910580989928e-05,
+      "loss": 0.9254,
+      "step": 3937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552065729648058e-05,
+      "loss": 0.9693,
+      "step": 3938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551220651002043e-05,
+      "loss": 1.0535,
+      "step": 3939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550375345074354e-05,
+      "loss": 0.9919,
+      "step": 3940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854952981188745e-05,
+      "loss": 1.0304,
+      "step": 3941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854868405146381e-05,
+      "loss": 1.0053,
+      "step": 3942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547838063825914e-05,
+      "loss": 0.9862,
+      "step": 3943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546991848996243e-05,
+      "loss": 1.0028,
+      "step": 3944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854614540699729e-05,
+      "loss": 1.0494,
+      "step": 3945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545298737851554e-05,
+      "loss": 0.9722,
+      "step": 3946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544451841581536e-05,
+      "loss": 0.9619,
+      "step": 3947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543604718209747e-05,
+      "loss": 0.9872,
+      "step": 3948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85427573677587e-05,
+      "loss": 0.3971,
+      "step": 3949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541909790250922e-05,
+      "loss": 1.0119,
+      "step": 3950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854106198570894e-05,
+      "loss": 1.0211,
+      "step": 3951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854021395415528e-05,
+      "loss": 1.0302,
+      "step": 3952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539365695612487e-05,
+      "loss": 0.9574,
+      "step": 3953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853851721010311e-05,
+      "loss": 1.0129,
+      "step": 3954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537668497649695e-05,
+      "loss": 0.9817,
+      "step": 3955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536819558274805e-05,
+      "loss": 0.9284,
+      "step": 3956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535970392000998e-05,
+      "loss": 0.9463,
+      "step": 3957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853512099885085e-05,
+      "loss": 0.9298,
+      "step": 3958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853427137884693e-05,
+      "loss": 0.9997,
+      "step": 3959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533421532011825e-05,
+      "loss": 1.0024,
+      "step": 3960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8532571458368122e-05,
+      "loss": 0.923,
+      "step": 3961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853172115793842e-05,
+      "loss": 1.1168,
+      "step": 3962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8530870630745307e-05,
+      "loss": 0.9269,
+      "step": 3963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85300198768114e-05,
+      "loss": 0.92,
+      "step": 3964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529168896159304e-05,
+      "loss": 0.934,
+      "step": 3965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528317688811638e-05,
+      "loss": 0.9393,
+      "step": 3966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8527466254791033e-05,
+      "loss": 0.9648,
+      "step": 3967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526614594120115e-05,
+      "loss": 0.924,
+      "step": 3968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525762706821514e-05,
+      "loss": 0.947,
+      "step": 3969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524910592917883e-05,
+      "loss": 0.9756,
+      "step": 3970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852405825243186e-05,
+      "loss": 1.0456,
+      "step": 3971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8523205685386107e-05,
+      "loss": 1.0271,
+      "step": 3972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852235289180328e-05,
+      "loss": 1.0599,
+      "step": 3973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521499871706047e-05,
+      "loss": 1.0157,
+      "step": 3974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520646625117084e-05,
+      "loss": 0.9399,
+      "step": 3975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851979315205906e-05,
+      "loss": 0.9419,
+      "step": 3976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518939452554664e-05,
+      "loss": 0.9502,
+      "step": 3977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518085526626588e-05,
+      "loss": 1.0351,
+      "step": 3978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517231374297528e-05,
+      "loss": 1.0665,
+      "step": 3979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516376995590185e-05,
+      "loss": 1.092,
+      "step": 3980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851552239052727e-05,
+      "loss": 1.018,
+      "step": 3981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514667559131494e-05,
+      "loss": 1.0127,
+      "step": 3982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851381250142558e-05,
+      "loss": 1.003,
+      "step": 3983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512957217432253e-05,
+      "loss": 1.0242,
+      "step": 3984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512101707174246e-05,
+      "loss": 1.034,
+      "step": 3985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511245970674295e-05,
+      "loss": 0.9431,
+      "step": 3986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851039000795515e-05,
+      "loss": 1.0026,
+      "step": 3987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850953381903956e-05,
+      "loss": 0.9951,
+      "step": 3988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850867740395028e-05,
+      "loss": 1.0233,
+      "step": 3989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507820762710066e-05,
+      "loss": 0.9629,
+      "step": 3990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8506963895341696e-05,
+      "loss": 0.9785,
+      "step": 3991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8506106801867943e-05,
+      "loss": 1.0332,
+      "step": 3992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505249482311583e-05,
+      "loss": 0.9612,
+      "step": 3993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504391936695407e-05,
+      "loss": 0.9168,
+      "step": 3994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850353416504221e-05,
+      "loss": 0.4679,
+      "step": 3995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502676167374784e-05,
+      "loss": 0.9505,
+      "step": 3996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501817943715937e-05,
+      "loss": 1.0069,
+      "step": 3997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500959494088476e-05,
+      "loss": 1.0438,
+      "step": 3998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500100818515224e-05,
+      "loss": 1.0776,
+      "step": 3999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499241917019e-05,
+      "loss": 0.9237,
+      "step": 4000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498382789622633e-05,
+      "loss": 1.0218,
+      "step": 4001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497523436348955e-05,
+      "loss": 0.9942,
+      "step": 4002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849666385722081e-05,
+      "loss": 0.9784,
+      "step": 4003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849580405226105e-05,
+      "loss": 1.0177,
+      "step": 4004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494944021492514e-05,
+      "loss": 0.9036,
+      "step": 4005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849408376493807e-05,
+      "loss": 1.0174,
+      "step": 4006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849322328262058e-05,
+      "loss": 0.9399,
+      "step": 4007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492362574562918e-05,
+      "loss": 0.9579,
+      "step": 4008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491501640787957e-05,
+      "loss": 0.9967,
+      "step": 4009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849064048131858e-05,
+      "loss": 1.057,
+      "step": 4010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8489779096177677e-05,
+      "loss": 0.9339,
+      "step": 4011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848891748538814e-05,
+      "loss": 1.0906,
+      "step": 4012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488055648972873e-05,
+      "loss": 1.0006,
+      "step": 4013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848719358695478e-05,
+      "loss": 0.941,
+      "step": 4014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8486331299356776e-05,
+      "loss": 0.4074,
+      "step": 4015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848546878620178e-05,
+      "loss": 1.0116,
+      "step": 4016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484606047512713e-05,
+      "loss": 1.0012,
+      "step": 4017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483743083312505e-05,
+      "loss": 1.0222,
+      "step": 4018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84828798936241e-05,
+      "loss": 1.004,
+      "step": 4019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482016478470434e-05,
+      "loss": 0.9353,
+      "step": 4020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481152837874455e-05,
+      "loss": 0.9121,
+      "step": 4021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480288971859122e-05,
+      "loss": 0.9973,
+      "step": 4022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479424880447394e-05,
+      "loss": 0.9826,
+      "step": 4023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478560563662236e-05,
+      "loss": 1.0914,
+      "step": 4024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847769602152662e-05,
+      "loss": 0.962,
+      "step": 4025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476831254063528e-05,
+      "loss": 0.8834,
+      "step": 4026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475966261295947e-05,
+      "loss": 1.0193,
+      "step": 4027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475101043246858e-05,
+      "loss": 0.9359,
+      "step": 4028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474235599939267e-05,
+      "loss": 1.0161,
+      "step": 4029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847336993139617e-05,
+      "loss": 1.0045,
+      "step": 4030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8472504037640576e-05,
+      "loss": 0.9829,
+      "step": 4031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84716379186955e-05,
+      "loss": 0.979,
+      "step": 4032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8470771574583966e-05,
+      "loss": 0.9881,
+      "step": 4033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469905005329e-05,
+      "loss": 0.9429,
+      "step": 4034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469038210953633e-05,
+      "loss": 0.9769,
+      "step": 4035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84681711914809e-05,
+      "loss": 0.9815,
+      "step": 4036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8467303946933846e-05,
+      "loss": 1.0426,
+      "step": 4037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846643647733553e-05,
+      "loss": 0.4406,
+      "step": 4038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8465568782708998e-05,
+      "loss": 0.9297,
+      "step": 4039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464700863077313e-05,
+      "loss": 0.9711,
+      "step": 4040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463832718463553e-05,
+      "loss": 0.9745,
+      "step": 4041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8462964348890785e-05,
+      "loss": 1.0225,
+      "step": 4042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846209575438209e-05,
+      "loss": 1.0128,
+      "step": 4043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846122693496055e-05,
+      "loss": 0.9832,
+      "step": 4044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460357890649265e-05,
+      "loss": 1.0143,
+      "step": 4045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845948862147133e-05,
+      "loss": 1.1065,
+      "step": 4046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845861912744985e-05,
+      "loss": 1.0018,
+      "step": 4047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457749408607933e-05,
+      "loss": 0.9749,
+      "step": 4048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84568794649687e-05,
+      "loss": 1.0087,
+      "step": 4049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456009296555263e-05,
+      "loss": 0.9859,
+      "step": 4050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8455138903390758e-05,
+      "loss": 1.0272,
+      "step": 4051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454268285498323e-05,
+      "loss": 0.9879,
+      "step": 4052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453397442901087e-05,
+      "loss": 0.9431,
+      "step": 4053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452526375622203e-05,
+      "loss": 1.0222,
+      "step": 4054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451655083684824e-05,
+      "loss": 0.9265,
+      "step": 4055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450783567112105e-05,
+      "loss": 1.0083,
+      "step": 4056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449911825927208e-05,
+      "loss": 1.0162,
+      "step": 4057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449039860153307e-05,
+      "loss": 0.9354,
+      "step": 4058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8448167669813577e-05,
+      "loss": 1.0019,
+      "step": 4059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84472952549312e-05,
+      "loss": 1.02,
+      "step": 4060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446422615529356e-05,
+      "loss": 0.9865,
+      "step": 4061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8445549751631256e-05,
+      "loss": 1.1352,
+      "step": 4062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844467666326008e-05,
+      "loss": 0.9814,
+      "step": 4063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443803350439048e-05,
+      "loss": 0.9976,
+      "step": 4064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8442929813191367e-05,
+      "loss": 1.0202,
+      "step": 4065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8442056051540254e-05,
+      "loss": 1.0003,
+      "step": 4066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844118206550893e-05,
+      "loss": 1.0708,
+      "step": 4067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844030785512063e-05,
+      "loss": 0.4676,
+      "step": 4068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439433420398588e-05,
+      "loss": 0.9766,
+      "step": 4069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8438558761366042e-05,
+      "loss": 0.9877,
+      "step": 4070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437683878046244e-05,
+      "loss": 0.9343,
+      "step": 4071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843680877046244e-05,
+      "loss": 0.966,
+      "step": 4072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435933438637903e-05,
+      "loss": 0.9559,
+      "step": 4073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435057882595885e-05,
+      "loss": 0.9755,
+      "step": 4074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434182102359664e-05,
+      "loss": 1.0127,
+      "step": 4075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843330609795251e-05,
+      "loss": 0.9654,
+      "step": 4076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843242986939772e-05,
+      "loss": 1.057,
+      "step": 4077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843155341671857e-05,
+      "loss": 1.0556,
+      "step": 4078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843067673993836e-05,
+      "loss": 1.0584,
+      "step": 4079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8429799839080394e-05,
+      "loss": 0.9762,
+      "step": 4080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842892271416797e-05,
+      "loss": 1.0174,
+      "step": 4081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428045365224415e-05,
+      "loss": 0.914,
+      "step": 4082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427167792273035e-05,
+      "loss": 1.0326,
+      "step": 4083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426289995337158e-05,
+      "loss": 1.0164,
+      "step": 4084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842541197444012e-05,
+      "loss": 1.0032,
+      "step": 4085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424533729605256e-05,
+      "loss": 1.0604,
+      "step": 4086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423655260855903e-05,
+      "loss": 1.0079,
+      "step": 4087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422776568215414e-05,
+      "loss": 0.9648,
+      "step": 4088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842189765170715e-05,
+      "loss": 1.007,
+      "step": 4089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842101851135446e-05,
+      "loss": 0.9993,
+      "step": 4090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420139147180713e-05,
+      "loss": 0.9805,
+      "step": 4091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841925955920929e-05,
+      "loss": 1.04,
+      "step": 4092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418379747463563e-05,
+      "loss": 0.9665,
+      "step": 4093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841749971196691e-05,
+      "loss": 1.1414,
+      "step": 4094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416619452742733e-05,
+      "loss": 0.9938,
+      "step": 4095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415738969814426e-05,
+      "loss": 1.0119,
+      "step": 4096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841485826320539e-05,
+      "loss": 1.0705,
+      "step": 4097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841397733293903e-05,
+      "loss": 0.9649,
+      "step": 4098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413096179038757e-05,
+      "loss": 1.0019,
+      "step": 4099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412214801528e-05,
+      "loss": 0.9675,
+      "step": 4100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411333200430182e-05,
+      "loss": 1.0662,
+      "step": 4101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8410451375768734e-05,
+      "loss": 0.9736,
+      "step": 4102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409569327567095e-05,
+      "loss": 0.9767,
+      "step": 4103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408687055848707e-05,
+      "loss": 0.8583,
+      "step": 4104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840780456063702e-05,
+      "loss": 0.9424,
+      "step": 4105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840692184195549e-05,
+      "loss": 1.0392,
+      "step": 4106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840603889982758e-05,
+      "loss": 1.0261,
+      "step": 4107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840515573427676e-05,
+      "loss": 0.8598,
+      "step": 4108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404272345326496e-05,
+      "loss": 0.9727,
+      "step": 4109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840338873300027e-05,
+      "loss": 0.9275,
+      "step": 4110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402504897321574e-05,
+      "loss": 0.9349,
+      "step": 4111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401620838313893e-05,
+      "loss": 0.9873,
+      "step": 4112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8400736556000724e-05,
+      "loss": 1.0382,
+      "step": 4113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839985205040557e-05,
+      "loss": 1.0094,
+      "step": 4114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398967321551946e-05,
+      "loss": 0.9925,
+      "step": 4115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839808236946336e-05,
+      "loss": 0.9577,
+      "step": 4116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839719719416334e-05,
+      "loss": 1.0411,
+      "step": 4117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396311795675407e-05,
+      "loss": 0.906,
+      "step": 4118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395426174023096e-05,
+      "loss": 0.959,
+      "step": 4119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394540329229947e-05,
+      "loss": 0.4347,
+      "step": 4120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393654261319504e-05,
+      "loss": 0.9661,
+      "step": 4121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392767970315314e-05,
+      "loss": 1.0552,
+      "step": 4122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391881456240935e-05,
+      "loss": 0.95,
+      "step": 4123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390994719119938e-05,
+      "loss": 0.9412,
+      "step": 4124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839010775897588e-05,
+      "loss": 0.9104,
+      "step": 4125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389220575832342e-05,
+      "loss": 0.9659,
+      "step": 4126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388333169712902e-05,
+      "loss": 1.0259,
+      "step": 4127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387445540641146e-05,
+      "loss": 0.9657,
+      "step": 4128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838655768864067e-05,
+      "loss": 1.0458,
+      "step": 4129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385669613735063e-05,
+      "loss": 0.9475,
+      "step": 4130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384781315947943e-05,
+      "loss": 0.9484,
+      "step": 4131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8383892795302907e-05,
+      "loss": 0.9568,
+      "step": 4132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8383004051823577e-05,
+      "loss": 0.9707,
+      "step": 4133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382115085533572e-05,
+      "loss": 0.4217,
+      "step": 4134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838122589645652e-05,
+      "loss": 0.9897,
+      "step": 4135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380336484616063e-05,
+      "loss": 0.9311,
+      "step": 4136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8379446850035826e-05,
+      "loss": 1.0424,
+      "step": 4137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8378556992739467e-05,
+      "loss": 0.9991,
+      "step": 4138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377666912750628e-05,
+      "loss": 0.9985,
+      "step": 4139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376776610092974e-05,
+      "loss": 1.0023,
+      "step": 4140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8375886084790163e-05,
+      "loss": 0.9935,
+      "step": 4141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837499533686587e-05,
+      "loss": 1.0262,
+      "step": 4142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837410436634376e-05,
+      "loss": 1.0044,
+      "step": 4143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373213173247526e-05,
+      "loss": 1.009,
+      "step": 4144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8372321757600845e-05,
+      "loss": 1.002,
+      "step": 4145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371430119427414e-05,
+      "loss": 0.9289,
+      "step": 4146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370538258750934e-05,
+      "loss": 0.419,
+      "step": 4147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369646175595104e-05,
+      "loss": 1.0825,
+      "step": 4148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836875386998364e-05,
+      "loss": 1.0806,
+      "step": 4149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367861341940253e-05,
+      "loss": 1.069,
+      "step": 4150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366968591488675e-05,
+      "loss": 0.9639,
+      "step": 4151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366075618652622e-05,
+      "loss": 1.0047,
+      "step": 4152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8365182423455835e-05,
+      "loss": 0.9507,
+      "step": 4153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8364289005922054e-05,
+      "loss": 0.9716,
+      "step": 4154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363395366075027e-05,
+      "loss": 0.9808,
+      "step": 4155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83625015039385e-05,
+      "loss": 0.9724,
+      "step": 4156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361607419536235e-05,
+      "loss": 0.9844,
+      "step": 4157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836071311289199e-05,
+      "loss": 1.0364,
+      "step": 4158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359818584029545e-05,
+      "loss": 0.9518,
+      "step": 4159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835892383297267e-05,
+      "loss": 0.9933,
+      "step": 4160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835802885974514e-05,
+      "loss": 0.9593,
+      "step": 4161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357133664370757e-05,
+      "loss": 1.0758,
+      "step": 4162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8356238246873302e-05,
+      "loss": 0.9409,
+      "step": 4163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8355342607276577e-05,
+      "loss": 1.0176,
+      "step": 4164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8354446745604388e-05,
+      "loss": 0.9585,
+      "step": 4165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353550661880547e-05,
+      "loss": 0.9978,
+      "step": 4166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352654356128872e-05,
+      "loss": 0.9933,
+      "step": 4167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351757828373183e-05,
+      "loss": 0.9599,
+      "step": 4168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835086107863731e-05,
+      "loss": 1.0286,
+      "step": 4169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8349964106945083e-05,
+      "loss": 0.9687,
+      "step": 4170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834906691332035e-05,
+      "loss": 0.9626,
+      "step": 4171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348169497786955e-05,
+      "loss": 0.9703,
+      "step": 4172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347271860368746e-05,
+      "loss": 0.8497,
+      "step": 4173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834637400108959e-05,
+      "loss": 0.9091,
+      "step": 4174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834547591997334e-05,
+      "loss": 0.8883,
+      "step": 4175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8344577617043872e-05,
+      "loss": 0.9752,
+      "step": 4176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343679092325064e-05,
+      "loss": 1.001,
+      "step": 4177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8342780345840796e-05,
+      "loss": 0.9995,
+      "step": 4178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834188137761495e-05,
+      "loss": 0.9455,
+      "step": 4179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834098218767143e-05,
+      "loss": 1.038,
+      "step": 4180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340082776034125e-05,
+      "loss": 0.9947,
+      "step": 4181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8339183142726945e-05,
+      "loss": 1.0127,
+      "step": 4182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8338283287773806e-05,
+      "loss": 0.9996,
+      "step": 4183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337383211198618e-05,
+      "loss": 1.1293,
+      "step": 4184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336482913025305e-05,
+      "loss": 0.9155,
+      "step": 4185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83355823932778e-05,
+      "loss": 1.0768,
+      "step": 4186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833468165198003e-05,
+      "loss": 1.0574,
+      "step": 4187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333780689155946e-05,
+      "loss": 0.9856,
+      "step": 4188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833287950482949e-05,
+      "loss": 0.9808,
+      "step": 4189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833197809902461e-05,
+      "loss": 1.0534,
+      "step": 4190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833107647176527e-05,
+      "loss": 1.0472,
+      "step": 4191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833017462307543e-05,
+      "loss": 1.023,
+      "step": 4192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8329272552979064e-05,
+      "loss": 0.996,
+      "step": 4193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832837026150015e-05,
+      "loss": 0.8441,
+      "step": 4194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832746774866266e-05,
+      "loss": 1.0173,
+      "step": 4195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326565014490595e-05,
+      "loss": 0.9691,
+      "step": 4196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832566205900794e-05,
+      "loss": 0.9821,
+      "step": 4197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324758882238692e-05,
+      "loss": 0.9615,
+      "step": 4198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832385548420686e-05,
+      "loss": 0.9387,
+      "step": 4199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322951864936462e-05,
+      "loss": 0.9852,
+      "step": 4200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322048024451506e-05,
+      "loss": 0.9333,
+      "step": 4201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832114396277602e-05,
+      "loss": 0.975,
+      "step": 4202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832023967993403e-05,
+      "loss": 1.0458,
+      "step": 4203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831933517594957e-05,
+      "loss": 0.926,
+      "step": 4204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318430450846686e-05,
+      "loss": 1.0333,
+      "step": 4205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831752550464942e-05,
+      "loss": 0.9776,
+      "step": 4206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8316620337381823e-05,
+      "loss": 1.0047,
+      "step": 4207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831571494906796e-05,
+      "loss": 1.0536,
+      "step": 4208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831480933973189e-05,
+      "loss": 1.041,
+      "step": 4209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313903509397682e-05,
+      "loss": 0.9852,
+      "step": 4210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312997458089415e-05,
+      "loss": 0.9123,
+      "step": 4211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831209118583117e-05,
+      "loss": 0.9849,
+      "step": 4212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8311184692647035e-05,
+      "loss": 0.9471,
+      "step": 4213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8310277978561104e-05,
+      "loss": 0.9613,
+      "step": 4214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8309371043597472e-05,
+      "loss": 0.9872,
+      "step": 4215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830846388778025e-05,
+      "loss": 0.9523,
+      "step": 4216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830755651113355e-05,
+      "loss": 0.4259,
+      "step": 4217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306648913681484e-05,
+      "loss": 0.9061,
+      "step": 4218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305741095448177e-05,
+      "loss": 0.9635,
+      "step": 4219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304833056457756e-05,
+      "loss": 0.9179,
+      "step": 4220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830392479673436e-05,
+      "loss": 0.9762,
+      "step": 4221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303016316302124e-05,
+      "loss": 1.0178,
+      "step": 4222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83021076151852e-05,
+      "loss": 0.96,
+      "step": 4223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8301198693407735e-05,
+      "loss": 0.9792,
+      "step": 4224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830028955099389e-05,
+      "loss": 1.0429,
+      "step": 4225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829938018796783e-05,
+      "loss": 1.0123,
+      "step": 4226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8298470604353724e-05,
+      "loss": 0.9494,
+      "step": 4227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297560800175745e-05,
+      "loss": 1.008,
+      "step": 4228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829665077545808e-05,
+      "loss": 0.9617,
+      "step": 4229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829574053022491e-05,
+      "loss": 0.9379,
+      "step": 4230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8294830064500433e-05,
+      "loss": 0.9701,
+      "step": 4231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8293919378308844e-05,
+      "loss": 1.0772,
+      "step": 4232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8293008471674353e-05,
+      "loss": 0.9585,
+      "step": 4233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292097344621168e-05,
+      "loss": 0.9751,
+      "step": 4234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8291185997173506e-05,
+      "loss": 1.0184,
+      "step": 4235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829027442935559e-05,
+      "loss": 1.016,
+      "step": 4236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8289362641191646e-05,
+      "loss": 0.9967,
+      "step": 4237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828845063270591e-05,
+      "loss": 0.9666,
+      "step": 4238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8287538403922624e-05,
+      "loss": 1.0173,
+      "step": 4239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8286625954866032e-05,
+      "loss": 0.9314,
+      "step": 4240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8285713285560386e-05,
+      "loss": 1.068,
+      "step": 4241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8284800396029948e-05,
+      "loss": 0.9467,
+      "step": 4242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283887286298973e-05,
+      "loss": 0.9686,
+      "step": 4243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828297395639174e-05,
+      "loss": 0.9737,
+      "step": 4244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282060406332513e-05,
+      "loss": 1.0262,
+      "step": 4245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8281146636145585e-05,
+      "loss": 0.9283,
+      "step": 4246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280232645855235e-05,
+      "loss": 1.02,
+      "step": 4247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8279318435485756e-05,
+      "loss": 0.9865,
+      "step": 4248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827840400506145e-05,
+      "loss": 1.0126,
+      "step": 4249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277489354606622e-05,
+      "loss": 1.0008,
+      "step": 4250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827657448414558e-05,
+      "loss": 0.9466,
+      "step": 4251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827565939370264e-05,
+      "loss": 0.9294,
+      "step": 4252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8274744083302126e-05,
+      "loss": 0.8611,
+      "step": 4253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273828552968368e-05,
+      "loss": 0.9848,
+      "step": 4254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827291280272569e-05,
+      "loss": 1.0544,
+      "step": 4255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827199683259844e-05,
+      "loss": 0.9951,
+      "step": 4256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827108064261097e-05,
+      "loss": 1.0359,
+      "step": 4257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8270164232787613e-05,
+      "loss": 0.9785,
+      "step": 4258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826924760315274e-05,
+      "loss": 0.9404,
+      "step": 4259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826833075373071e-05,
+      "loss": 0.885,
+      "step": 4260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826741368454589e-05,
+      "loss": 1.1457,
+      "step": 4261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826649639562266e-05,
+      "loss": 0.9813,
+      "step": 4262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265578886985394e-05,
+      "loss": 0.9676,
+      "step": 4263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8264661158658485e-05,
+      "loss": 0.9951,
+      "step": 4264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263743210666317e-05,
+      "loss": 0.9887,
+      "step": 4265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8262825043033298e-05,
+      "loss": 0.9906,
+      "step": 4266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826190665578382e-05,
+      "loss": 1.022,
+      "step": 4267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260988048942303e-05,
+      "loss": 1.0121,
+      "step": 4268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260069222533157e-05,
+      "loss": 0.9604,
+      "step": 4269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259150176580806e-05,
+      "loss": 1.0562,
+      "step": 4270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8258230911109675e-05,
+      "loss": 1.0304,
+      "step": 4271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257311426144203e-05,
+      "loss": 0.9913,
+      "step": 4272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256391721708818e-05,
+      "loss": 1.0633,
+      "step": 4273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255471797827976e-05,
+      "loss": 0.9741,
+      "step": 4274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254551654526115e-05,
+      "loss": 0.9497,
+      "step": 4275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8253631291827705e-05,
+      "loss": 0.9541,
+      "step": 4276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82527107097572e-05,
+      "loss": 0.9935,
+      "step": 4277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251789908339068e-05,
+      "loss": 0.9903,
+      "step": 4278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250868887597786e-05,
+      "loss": 0.9765,
+      "step": 4279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824994764755783e-05,
+      "loss": 0.987,
+      "step": 4280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8249026188243696e-05,
+      "loss": 0.9865,
+      "step": 4281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824810450967986e-05,
+      "loss": 1.0707,
+      "step": 4282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824718261189083e-05,
+      "loss": 1.0081,
+      "step": 4283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8246260494901103e-05,
+      "loss": 0.9941,
+      "step": 4284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245338158735188e-05,
+      "loss": 1.0261,
+      "step": 4285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244415603417603e-05,
+      "loss": 1.0297,
+      "step": 4286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824349282897287e-05,
+      "loss": 1.0258,
+      "step": 4287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824256983542551e-05,
+      "loss": 0.936,
+      "step": 4288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824164662280006e-05,
+      "loss": 1.0145,
+      "step": 4289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8240723191121054e-05,
+      "loss": 0.9507,
+      "step": 4290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823979954041304e-05,
+      "loss": 1.0113,
+      "step": 4291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238875670700563e-05,
+      "loss": 1.0349,
+      "step": 4292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823795158200818e-05,
+      "loss": 0.9521,
+      "step": 4293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237027274360452e-05,
+      "loss": 1.0096,
+      "step": 4294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823610274778195e-05,
+      "loss": 0.989,
+      "step": 4295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823517800229724e-05,
+      "loss": 1.0762,
+      "step": 4296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234253037930902e-05,
+      "loss": 0.9131,
+      "step": 4297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233327854707527e-05,
+      "loss": 0.9858,
+      "step": 4298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82324024526517e-05,
+      "loss": 1.086,
+      "step": 4299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823147683178802e-05,
+      "loss": 0.473,
+      "step": 4300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230550992141083e-05,
+      "loss": 1.0232,
+      "step": 4301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229624933735502e-05,
+      "loss": 0.9102,
+      "step": 4302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822869865659589e-05,
+      "loss": 1.0245,
+      "step": 4303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227772160746865e-05,
+      "loss": 0.9742,
+      "step": 4304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226845446213056e-05,
+      "loss": 0.8978,
+      "step": 4305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225918513019088e-05,
+      "loss": 1.0266,
+      "step": 4306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224991361189602e-05,
+      "loss": 1.0172,
+      "step": 4307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822406399074924e-05,
+      "loss": 1.0309,
+      "step": 4308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223136401722648e-05,
+      "loss": 0.9318,
+      "step": 4309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222208594134485e-05,
+      "loss": 0.9484,
+      "step": 4310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221280568009405e-05,
+      "loss": 1.0491,
+      "step": 4311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220352323372078e-05,
+      "loss": 0.9704,
+      "step": 4312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219423860247176e-05,
+      "loss": 0.9591,
+      "step": 4313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218495178659376e-05,
+      "loss": 0.443,
+      "step": 4314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821756627863336e-05,
+      "loss": 1.0054,
+      "step": 4315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216637160193818e-05,
+      "loss": 0.9569,
+      "step": 4316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821570782336544e-05,
+      "loss": 0.9643,
+      "step": 4317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8214778268172935e-05,
+      "loss": 1.0102,
+      "step": 4318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213848494641007e-05,
+      "loss": 0.9017,
+      "step": 4319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212918502794365e-05,
+      "loss": 1.075,
+      "step": 4320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821198829265773e-05,
+      "loss": 1.013,
+      "step": 4321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211057864255826e-05,
+      "loss": 0.8922,
+      "step": 4322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821012721761338e-05,
+      "loss": 0.9696,
+      "step": 4323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8209196352755132e-05,
+      "loss": 0.9764,
+      "step": 4324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208265269705818e-05,
+      "loss": 0.9711,
+      "step": 4325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820733396849019e-05,
+      "loss": 0.9862,
+      "step": 4326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206402449132997e-05,
+      "loss": 0.9851,
+      "step": 4327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205470711659e-05,
+      "loss": 1.0036,
+      "step": 4328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204538756092965e-05,
+      "loss": 1.0169,
+      "step": 4329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203606582459658e-05,
+      "loss": 1.0266,
+      "step": 4330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202674190783858e-05,
+      "loss": 1.0015,
+      "step": 4331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201741581090346e-05,
+      "loss": 1.0546,
+      "step": 4332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8200808753403913e-05,
+      "loss": 1.0985,
+      "step": 4333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199875707749343e-05,
+      "loss": 0.8978,
+      "step": 4334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198942444151448e-05,
+      "loss": 0.9936,
+      "step": 4335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198008962635024e-05,
+      "loss": 0.8881,
+      "step": 4336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197075263224884e-05,
+      "loss": 1.036,
+      "step": 4337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819614134594585e-05,
+      "loss": 1.0315,
+      "step": 4338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8195207210822737e-05,
+      "loss": 0.9827,
+      "step": 4339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194272857880376e-05,
+      "loss": 1.0341,
+      "step": 4340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81933382871436e-05,
+      "loss": 1.0801,
+      "step": 4341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192403498637255e-05,
+      "loss": 1.0232,
+      "step": 4342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191468492386177e-05,
+      "loss": 0.9667,
+      "step": 4343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190533268415222e-05,
+      "loss": 0.928,
+      "step": 4344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818959782674925e-05,
+      "loss": 1.047,
+      "step": 4345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818866216741312e-05,
+      "loss": 0.9263,
+      "step": 4346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81877262904317e-05,
+      "loss": 0.9512,
+      "step": 4347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818679019582987e-05,
+      "loss": 0.9998,
+      "step": 4348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185853883632502e-05,
+      "loss": 0.9338,
+      "step": 4349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818491735386449e-05,
+      "loss": 1.0094,
+      "step": 4350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183980606550725e-05,
+      "loss": 1.0241,
+      "step": 4351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183043641716097e-05,
+      "loss": 0.8951,
+      "step": 4352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182106459385517e-05,
+      "loss": 0.951,
+      "step": 4353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181169059583896e-05,
+      "loss": 1.0389,
+      "step": 4354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180231442336144e-05,
+      "loss": 0.9979,
+      "step": 4355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8179293607667177e-05,
+      "loss": 0.958,
+      "step": 4356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8178355555601933e-05,
+      "loss": 0.8929,
+      "step": 4357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177417286165337e-05,
+      "loss": 0.9683,
+      "step": 4358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817647879938233e-05,
+      "loss": 1.0026,
+      "step": 4359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175540095277856e-05,
+      "loss": 1.0066,
+      "step": 4360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817460117387686e-05,
+      "loss": 0.4741,
+      "step": 4361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173662035204303e-05,
+      "loss": 0.8688,
+      "step": 4362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172722679285146e-05,
+      "loss": 1.0595,
+      "step": 4363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171783106144353e-05,
+      "loss": 0.9854,
+      "step": 4364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81708433158069e-05,
+      "loss": 0.9488,
+      "step": 4365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169903308297762e-05,
+      "loss": 0.9267,
+      "step": 4366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168963083641922e-05,
+      "loss": 0.9851,
+      "step": 4367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816802264186438e-05,
+      "loss": 0.9723,
+      "step": 4368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167081982990116e-05,
+      "loss": 0.9577,
+      "step": 4369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8166141107044145e-05,
+      "loss": 0.9953,
+      "step": 4370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8165200014051474e-05,
+      "loss": 0.9176,
+      "step": 4371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8164258704037106e-05,
+      "loss": 1.0069,
+      "step": 4372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816331717702607e-05,
+      "loss": 0.9379,
+      "step": 4373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162375433043388e-05,
+      "loss": 1.0111,
+      "step": 4374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8161433472114087e-05,
+      "loss": 1.0074,
+      "step": 4375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160491294263202e-05,
+      "loss": 1.0129,
+      "step": 4376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8159548899515785e-05,
+      "loss": 1.0147,
+      "step": 4377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158606287896872e-05,
+      "loss": 0.9983,
+      "step": 4378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157663459431524e-05,
+      "loss": 0.9082,
+      "step": 4379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81567204141448e-05,
+      "loss": 0.9559,
+      "step": 4380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815577715206176e-05,
+      "loss": 0.9497,
+      "step": 4381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154833673207483e-05,
+      "loss": 0.9672,
+      "step": 4382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8153889977607036e-05,
+      "loss": 1.0077,
+      "step": 4383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8152946065285508e-05,
+      "loss": 1.0226,
+      "step": 4384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815200193626798e-05,
+      "loss": 0.9141,
+      "step": 4385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815105759057956e-05,
+      "loss": 1.0251,
+      "step": 4386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8150113028245333e-05,
+      "loss": 0.9808,
+      "step": 4387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814916824929041e-05,
+      "loss": 1.0253,
+      "step": 4388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8148223253739907e-05,
+      "loss": 1.0625,
+      "step": 4389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814727804161893e-05,
+      "loss": 1.0456,
+      "step": 4390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146332612952607e-05,
+      "loss": 0.9665,
+      "step": 4391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8145386967766075e-05,
+      "loss": 1.0202,
+      "step": 4392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8144441106084457e-05,
+      "loss": 1.0014,
+      "step": 4393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143495027932893e-05,
+      "loss": 0.9312,
+      "step": 4394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142548733336534e-05,
+      "loss": 0.958,
+      "step": 4395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814160222232053e-05,
+      "loss": 0.9692,
+      "step": 4396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140655494910036e-05,
+      "loss": 0.9272,
+      "step": 4397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8139708551130217e-05,
+      "loss": 1.0595,
+      "step": 4398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813876139100624e-05,
+      "loss": 0.9938,
+      "step": 4399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137814014563284e-05,
+      "loss": 0.9635,
+      "step": 4400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136866421826524e-05,
+      "loss": 1.0125,
+      "step": 4401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135918612821147e-05,
+      "loss": 0.9876,
+      "step": 4402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134970587572345e-05,
+      "loss": 0.9651,
+      "step": 4403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134022346105316e-05,
+      "loss": 0.9709,
+      "step": 4404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8133073888445266e-05,
+      "loss": 0.9787,
+      "step": 4405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132125214617397e-05,
+      "loss": 0.436,
+      "step": 4406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813117632464693e-05,
+      "loss": 0.9756,
+      "step": 4407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130227218559083e-05,
+      "loss": 0.9546,
+      "step": 4408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129277896379077e-05,
+      "loss": 0.9103,
+      "step": 4409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8128328358132154e-05,
+      "loss": 1.1348,
+      "step": 4410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812737860384355e-05,
+      "loss": 0.9261,
+      "step": 4411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126428633538496e-05,
+      "loss": 1.0367,
+      "step": 4412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8125478447242257e-05,
+      "loss": 0.9881,
+      "step": 4413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124528044980078e-05,
+      "loss": 0.9924,
+      "step": 4414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123577426777226e-05,
+      "loss": 0.9963,
+      "step": 4415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122626592658964e-05,
+      "loss": 1.0302,
+      "step": 4416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121675542650558e-05,
+      "loss": 0.9806,
+      "step": 4417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81207242767773e-05,
+      "loss": 0.8645,
+      "step": 4418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811977279506446e-05,
+      "loss": 0.9771,
+      "step": 4419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8118821097537332e-05,
+      "loss": 0.9593,
+      "step": 4420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8117869184221216e-05,
+      "loss": 1.0355,
+      "step": 4421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116917055141407e-05,
+      "loss": 1.0142,
+      "step": 4422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811596471032321e-05,
+      "loss": 1.0484,
+      "step": 4423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115012149791944e-05,
+      "loss": 1.02,
+      "step": 4424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114059373572922e-05,
+      "loss": 1.0397,
+      "step": 4425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811310638169147e-05,
+      "loss": 1.0217,
+      "step": 4426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8112153174172916e-05,
+      "loss": 1.0164,
+      "step": 4427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111199751042597e-05,
+      "loss": 0.9452,
+      "step": 4428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811024611232585e-05,
+      "loss": 0.4403,
+      "step": 4429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109292258048023e-05,
+      "loss": 1.0143,
+      "step": 4430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810833818823447e-05,
+      "loss": 0.9456,
+      "step": 4431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107383902910554e-05,
+      "loss": 0.981,
+      "step": 4432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810642940210163e-05,
+      "loss": 0.9986,
+      "step": 4433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105474685833067e-05,
+      "loss": 1.0445,
+      "step": 4434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104519754130246e-05,
+      "loss": 0.947,
+      "step": 4435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810356460701855e-05,
+      "loss": 0.9525,
+      "step": 4436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810260924452336e-05,
+      "loss": 1.0176,
+      "step": 4437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810165366667007e-05,
+      "loss": 1.0227,
+      "step": 4438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8100697873484078e-05,
+      "loss": 0.9661,
+      "step": 4439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809974186499079e-05,
+      "loss": 1.0361,
+      "step": 4440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809878564121561e-05,
+      "loss": 1.0084,
+      "step": 4441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809782920218396e-05,
+      "loss": 0.9769,
+      "step": 4442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809687254792126e-05,
+      "loss": 0.9547,
+      "step": 4443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095915678452935e-05,
+      "loss": 1.0009,
+      "step": 4444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094958593804415e-05,
+      "loss": 0.9083,
+      "step": 4445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094001294001144e-05,
+      "loss": 0.9466,
+      "step": 4446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809304377906856e-05,
+      "loss": 0.9021,
+      "step": 4447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092086049032116e-05,
+      "loss": 0.9578,
+      "step": 4448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809112810391727e-05,
+      "loss": 1.0238,
+      "step": 4449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090169943749477e-05,
+      "loss": 0.9513,
+      "step": 4450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089211568554206e-05,
+      "loss": 1.0369,
+      "step": 4451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808825297835693e-05,
+      "loss": 1.0202,
+      "step": 4452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808729417318313e-05,
+      "loss": 1.0006,
+      "step": 4453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8086335153058287e-05,
+      "loss": 1.0041,
+      "step": 4454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085375918007894e-05,
+      "loss": 1.0006,
+      "step": 4455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084416468057436e-05,
+      "loss": 0.9963,
+      "step": 4456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083456803232427e-05,
+      "loss": 1.0076,
+      "step": 4457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082496923558367e-05,
+      "loss": 0.8625,
+      "step": 4458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808153682906077e-05,
+      "loss": 0.9403,
+      "step": 4459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8080576519765155e-05,
+      "loss": 1.0343,
+      "step": 4460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079615995697045e-05,
+      "loss": 0.9668,
+      "step": 4461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078655256881966e-05,
+      "loss": 0.9293,
+      "step": 4462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077694303345458e-05,
+      "loss": 1.0583,
+      "step": 4463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8076733135113064e-05,
+      "loss": 0.9105,
+      "step": 4464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075771752210327e-05,
+      "loss": 0.9919,
+      "step": 4465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074810154662798e-05,
+      "loss": 0.9552,
+      "step": 4466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073848342496042e-05,
+      "loss": 0.8753,
+      "step": 4467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8072886315735613e-05,
+      "loss": 0.9541,
+      "step": 4468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807192407440709e-05,
+      "loss": 0.9948,
+      "step": 4469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070961618536042e-05,
+      "loss": 1.0922,
+      "step": 4470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806999894814805e-05,
+      "loss": 0.9125,
+      "step": 4471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069036063268707e-05,
+      "loss": 0.9521,
+      "step": 4472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80680729639236e-05,
+      "loss": 0.8919,
+      "step": 4473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8067109650138326e-05,
+      "loss": 1.004,
+      "step": 4474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8066146121938493e-05,
+      "loss": 1.0286,
+      "step": 4475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065182379349708e-05,
+      "loss": 1.0013,
+      "step": 4476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806421842239759e-05,
+      "loss": 0.9438,
+      "step": 4477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806325425110775e-05,
+      "loss": 0.9329,
+      "step": 4478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062289865505826e-05,
+      "loss": 1.0408,
+      "step": 4479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806132526561744e-05,
+      "loss": 0.4535,
+      "step": 4480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806036045146824e-05,
+      "loss": 0.9404,
+      "step": 4481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805939542308387e-05,
+      "loss": 0.9852,
+      "step": 4482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8058430180489967e-05,
+      "loss": 0.942,
+      "step": 4483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057464723712193e-05,
+      "loss": 0.9618,
+      "step": 4484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805649905277621e-05,
+      "loss": 0.9631,
+      "step": 4485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8055533167707686e-05,
+      "loss": 0.975,
+      "step": 4486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805456706853229e-05,
+      "loss": 1.027,
+      "step": 4487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80536007552757e-05,
+      "loss": 0.928,
+      "step": 4488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80526342279636e-05,
+      "loss": 1.0105,
+      "step": 4489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8051667486621677e-05,
+      "loss": 0.9883,
+      "step": 4490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050700531275632e-05,
+      "loss": 0.9506,
+      "step": 4491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804973336195116e-05,
+      "loss": 0.9575,
+      "step": 4492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048765978673966e-05,
+      "loss": 0.9049,
+      "step": 4493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047798381469766e-05,
+      "loss": 0.9327,
+      "step": 4494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804683057036428e-05,
+      "loss": 0.9689,
+      "step": 4495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045862545383226e-05,
+      "loss": 1.0377,
+      "step": 4496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044894306552338e-05,
+      "loss": 0.9391,
+      "step": 4497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043925853897343e-05,
+      "loss": 0.9893,
+      "step": 4498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8042957187443986e-05,
+      "loss": 1.0119,
+      "step": 4499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8041988307218015e-05,
+      "loss": 0.9451,
+      "step": 4500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804101921324518e-05,
+      "loss": 0.8814,
+      "step": 4501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8040049905551232e-05,
+      "loss": 0.8875,
+      "step": 4502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039080384161947e-05,
+      "loss": 0.9572,
+      "step": 4503
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038110649103086e-05,
+      "loss": 0.9101,
+      "step": 4504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803714070040042e-05,
+      "loss": 0.9626,
+      "step": 4505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803617053807974e-05,
+      "loss": 1.0145,
+      "step": 4506
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035200162166822e-05,
+      "loss": 1.0182,
+      "step": 4507
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803422957268746e-05,
+      "loss": 1.0692,
+      "step": 4508
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8033258769667452e-05,
+      "loss": 0.4546,
+      "step": 4509
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80322877531326e-05,
+      "loss": 1.0006,
+      "step": 4510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031316523108715e-05,
+      "loss": 0.972,
+      "step": 4511
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030345079621608e-05,
+      "loss": 1.0262,
+      "step": 4512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80293734226971e-05,
+      "loss": 0.9214,
+      "step": 4513
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802840155236102e-05,
+      "loss": 0.9548,
+      "step": 4514
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802742946863919e-05,
+      "loss": 1.0803,
+      "step": 4515
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802645717155746e-05,
+      "loss": 1.016,
+      "step": 4516
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025484661141662e-05,
+      "loss": 0.9435,
+      "step": 4517
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802451193741765e-05,
+      "loss": 1.0268,
+      "step": 4518
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8023539000411273e-05,
+      "loss": 0.975,
+      "step": 4519
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8022565850148396e-05,
+      "loss": 0.4331,
+      "step": 4520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802159248665488e-05,
+      "loss": 0.9494,
+      "step": 4521
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8020618909956596e-05,
+      "loss": 0.996,
+      "step": 4522
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801964512007942e-05,
+      "loss": 1.0852,
+      "step": 4523
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8018671117049238e-05,
+      "loss": 0.9962,
+      "step": 4524
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8017696900891938e-05,
+      "loss": 0.9551,
+      "step": 4525
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801672247163341e-05,
+      "loss": 0.9531,
+      "step": 4526
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015747829299553e-05,
+      "loss": 0.4654,
+      "step": 4527
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8014772973916276e-05,
+      "loss": 0.9926,
+      "step": 4528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801379790550949e-05,
+      "loss": 0.9924,
+      "step": 4529
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80128226241051e-05,
+      "loss": 1.013,
+      "step": 4530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801184712972904e-05,
+      "loss": 0.9083,
+      "step": 4531
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010871422407238e-05,
+      "loss": 0.4327,
+      "step": 4532
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800989550216562e-05,
+      "loss": 0.8895,
+      "step": 4533
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008919369030122e-05,
+      "loss": 0.9602,
+      "step": 4534
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80079430230267e-05,
+      "loss": 0.9564,
+      "step": 4535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006966464181298e-05,
+      "loss": 0.8706,
+      "step": 4536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800598969251987e-05,
+      "loss": 0.9661,
+      "step": 4537
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005012708068384e-05,
+      "loss": 1.0178,
+      "step": 4538
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80040355108528e-05,
+      "loss": 0.9447,
+      "step": 4539
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8003058100899092e-05,
+      "loss": 0.9795,
+      "step": 4540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800208047823324e-05,
+      "loss": 0.9929,
+      "step": 4541
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800110264288123e-05,
+      "loss": 1.0055,
+      "step": 4542
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8000124594869053e-05,
+      "loss": 0.9239,
+      "step": 4543
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79991463342227e-05,
+      "loss": 0.9625,
+      "step": 4544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799816786096817e-05,
+      "loss": 0.9325,
+      "step": 4545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7997189175131473e-05,
+      "loss": 0.9976,
+      "step": 4546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996210276738623e-05,
+      "loss": 1.0209,
+      "step": 4547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995231165815636e-05,
+      "loss": 0.9742,
+      "step": 4548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994251842388538e-05,
+      "loss": 0.9667,
+      "step": 4549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7993272306483354e-05,
+      "loss": 0.4439,
+      "step": 4550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799229255812612e-05,
+      "loss": 0.826,
+      "step": 4551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799131259734288e-05,
+      "loss": 0.9967,
+      "step": 4552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799033242415968e-05,
+      "loss": 1.0138,
+      "step": 4553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7989352038602567e-05,
+      "loss": 1.0705,
+      "step": 4554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79883714406976e-05,
+      "loss": 0.9288,
+      "step": 4555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798739063047085e-05,
+      "loss": 1.0181,
+      "step": 4556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7986409607948373e-05,
+      "loss": 0.9737,
+      "step": 4557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7985428373156254e-05,
+      "loss": 0.9179,
+      "step": 4558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7984446926120565e-05,
+      "loss": 1.0397,
+      "step": 4559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983465266867402e-05,
+      "loss": 0.9388,
+      "step": 4560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7982483395422847e-05,
+      "loss": 0.9673,
+      "step": 4561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981501311812997e-05,
+      "loss": 1.0396,
+      "step": 4562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7980519016063963e-05,
+      "loss": 0.9682,
+      "step": 4563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979536508201847e-05,
+      "loss": 0.9157,
+      "step": 4564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7978553788252763e-05,
+      "loss": 0.9484,
+      "step": 4565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797757085624283e-05,
+      "loss": 0.9646,
+      "step": 4566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7976587712198178e-05,
+      "loss": 0.9748,
+      "step": 4567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7975604356144932e-05,
+      "loss": 0.9872,
+      "step": 4568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7974620788109236e-05,
+      "loss": 0.9344,
+      "step": 4569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973637008117223e-05,
+      "loss": 1.0083,
+      "step": 4570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972653016195046e-05,
+      "loss": 0.8821,
+      "step": 4571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971668812368858e-05,
+      "loss": 1.0867,
+      "step": 4572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970684396664814e-05,
+      "loss": 0.9064,
+      "step": 4573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969699769109086e-05,
+      "loss": 0.9971,
+      "step": 4574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968714929727843e-05,
+      "loss": 0.9908,
+      "step": 4575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967729878547252e-05,
+      "loss": 0.9754,
+      "step": 4576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7966744615593506e-05,
+      "loss": 0.9409,
+      "step": 4577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965759140892785e-05,
+      "loss": 1.0241,
+      "step": 4578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964773454471283e-05,
+      "loss": 0.9993,
+      "step": 4579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7963787556355197e-05,
+      "loss": 0.9919,
+      "step": 4580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962801446570738e-05,
+      "loss": 0.9422,
+      "step": 4581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796181512514411e-05,
+      "loss": 0.9548,
+      "step": 4582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796082859210153e-05,
+      "loss": 0.9449,
+      "step": 4583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7959841847469212e-05,
+      "loss": 0.9746,
+      "step": 4584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958854891273393e-05,
+      "loss": 0.9298,
+      "step": 4585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957867723540303e-05,
+      "loss": 1.0147,
+      "step": 4586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795688034429618e-05,
+      "loss": 0.9831,
+      "step": 4587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7955892753567256e-05,
+      "loss": 1.0114,
+      "step": 4588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954904951379794e-05,
+      "loss": 0.9824,
+      "step": 4589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7953916937760044e-05,
+      "loss": 0.9599,
+      "step": 4590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7952928712734266e-05,
+      "loss": 0.95,
+      "step": 4591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951940276328727e-05,
+      "loss": 1.0159,
+      "step": 4592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950951628569697e-05,
+      "loss": 1.028,
+      "step": 4593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794996276948345e-05,
+      "loss": 0.4417,
+      "step": 4594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948973699096275e-05,
+      "loss": 0.9882,
+      "step": 4595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7947984417434456e-05,
+      "loss": 1.0828,
+      "step": 4596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794699492452429e-05,
+      "loss": 1.0271,
+      "step": 4597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794600522039208e-05,
+      "loss": 0.4478,
+      "step": 4598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794501530506412e-05,
+      "loss": 0.945,
+      "step": 4599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794402517856673e-05,
+      "loss": 1.0554,
+      "step": 4600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794303484092622e-05,
+      "loss": 0.9831,
+      "step": 4601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794204429216892e-05,
+      "loss": 1.036,
+      "step": 4602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7941053532321154e-05,
+      "loss": 0.4095,
+      "step": 4603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794006256140925e-05,
+      "loss": 0.9536,
+      "step": 4604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939071379459556e-05,
+      "loss": 0.967,
+      "step": 4605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793807998649841e-05,
+      "loss": 1.0027,
+      "step": 4606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937088382552163e-05,
+      "loss": 0.9213,
+      "step": 4607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936096567647175e-05,
+      "loss": 1.0476,
+      "step": 4608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79351045418098e-05,
+      "loss": 1.0497,
+      "step": 4609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7934112305066415e-05,
+      "loss": 0.9254,
+      "step": 4610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793311985744338e-05,
+      "loss": 1.0275,
+      "step": 4611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793212719896708e-05,
+      "loss": 1.0503,
+      "step": 4612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7931134329663903e-05,
+      "loss": 1.0189,
+      "step": 4613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930141249560235e-05,
+      "loss": 1.0822,
+      "step": 4614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7929147958682463e-05,
+      "loss": 1.0479,
+      "step": 4615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928154457057e-05,
+      "loss": 0.9326,
+      "step": 4616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7927160744710243e-05,
+      "loss": 0.9287,
+      "step": 4617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7926166821668608e-05,
+      "loss": 0.8749,
+      "step": 4618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925172687958512e-05,
+      "loss": 1.0239,
+      "step": 4619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924178343606376e-05,
+      "loss": 0.9992,
+      "step": 4620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792318378863863e-05,
+      "loss": 0.4335,
+      "step": 4621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792218902308171e-05,
+      "loss": 1.0113,
+      "step": 4622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921194046962052e-05,
+      "loss": 1.0298,
+      "step": 4623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920198860306105e-05,
+      "loss": 1.0744,
+      "step": 4624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7919203463140317e-05,
+      "loss": 1.0074,
+      "step": 4625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7918207855491144e-05,
+      "loss": 0.9298,
+      "step": 4626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791721203738505e-05,
+      "loss": 0.9109,
+      "step": 4627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7916216008848503e-05,
+      "loss": 0.977,
+      "step": 4628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7915219769907978e-05,
+      "loss": 0.9602,
+      "step": 4629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791422332058995e-05,
+      "loss": 0.8893,
+      "step": 4630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7913226660920904e-05,
+      "loss": 0.9139,
+      "step": 4631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912229790927332e-05,
+      "loss": 0.9457,
+      "step": 4632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911232710635726e-05,
+      "loss": 0.9863,
+      "step": 4633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910235420072593e-05,
+      "loss": 0.9856,
+      "step": 4634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7909237919264436e-05,
+      "loss": 0.9201,
+      "step": 4635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790824020823777e-05,
+      "loss": 1.0387,
+      "step": 4636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7907242287019108e-05,
+      "loss": 1.0238,
+      "step": 4637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790624415563498e-05,
+      "loss": 0.9756,
+      "step": 4638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7905245814111913e-05,
+      "loss": 0.9558,
+      "step": 4639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7904247262476438e-05,
+      "loss": 0.9747,
+      "step": 4640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79032485007551e-05,
+      "loss": 0.4465,
+      "step": 4641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790224952897444e-05,
+      "loss": 1.0669,
+      "step": 4642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7901250347161018e-05,
+      "loss": 1.0721,
+      "step": 4643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900250955341382e-05,
+      "loss": 0.943,
+      "step": 4644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78992513535421e-05,
+      "loss": 1.0458,
+      "step": 4645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7898251541789738e-05,
+      "loss": 0.9325,
+      "step": 4646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7897251520110872e-05,
+      "loss": 0.9993,
+      "step": 4647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789625128853208e-05,
+      "loss": 0.8986,
+      "step": 4648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7895250847079945e-05,
+      "loss": 0.9175,
+      "step": 4649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894250195781063e-05,
+      "loss": 1.0404,
+      "step": 4650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893249334662024e-05,
+      "loss": 0.9652,
+      "step": 4651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7892248263749433e-05,
+      "loss": 1.0949,
+      "step": 4652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78912469830699e-05,
+      "loss": 0.9106,
+      "step": 4653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890245492650032e-05,
+      "loss": 1.0745,
+      "step": 4654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7889243792516452e-05,
+      "loss": 0.9808,
+      "step": 4655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7888241882695777e-05,
+      "loss": 1.0066,
+      "step": 4656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887239763214647e-05,
+      "loss": 1.058,
+      "step": 4657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7886237434099692e-05,
+      "loss": 0.9575,
+      "step": 4658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885234895377555e-05,
+      "loss": 0.9733,
+      "step": 4659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7884232147074876e-05,
+      "loss": 0.9436,
+      "step": 4660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883229189218315e-05,
+      "loss": 0.9745,
+      "step": 4661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882226021834525e-05,
+      "loss": 1.0082,
+      "step": 4662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7881222644950166e-05,
+      "loss": 0.976,
+      "step": 4663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788021905859191e-05,
+      "loss": 1.0331,
+      "step": 4664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879215262786437e-05,
+      "loss": 0.9772,
+      "step": 4665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787821125756042e-05,
+      "loss": 1.046,
+      "step": 4666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787720704294054e-05,
+      "loss": 0.4096,
+      "step": 4667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78762026189535e-05,
+      "loss": 0.9634,
+      "step": 4668
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7875197985625985e-05,
+      "loss": 1.0028,
+      "step": 4669
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874193142984704e-05,
+      "loss": 0.9872,
+      "step": 4670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7873188091056364e-05,
+      "loss": 0.9723,
+      "step": 4671
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787218282986767e-05,
+      "loss": 0.4397,
+      "step": 4672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7871177359445356e-05,
+      "loss": 0.9885,
+      "step": 4673
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870171679816132e-05,
+      "loss": 1.0042,
+      "step": 4674
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869165791006738e-05,
+      "loss": 1.0414,
+      "step": 4675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78681596930439e-05,
+      "loss": 0.9566,
+      "step": 4676
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786715338595437e-05,
+      "loss": 0.94,
+      "step": 4677
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786614686976488e-05,
+      "loss": 0.9614,
+      "step": 4678
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7865140144502193e-05,
+      "loss": 0.418,
+      "step": 4679
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864133210193063e-05,
+      "loss": 1.0133,
+      "step": 4680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786312606686426e-05,
+      "loss": 0.9423,
+      "step": 4681
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786211871454254e-05,
+      "loss": 0.9655,
+      "step": 4682
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786111115325469e-05,
+      "loss": 0.9237,
+      "step": 4683
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786010338302748e-05,
+      "loss": 0.9677,
+      "step": 4684
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859095403887697e-05,
+      "loss": 0.9379,
+      "step": 4685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858087215862134e-05,
+      "loss": 1.0013,
+      "step": 4686
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7857078818977592e-05,
+      "loss": 0.9904,
+      "step": 4687
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7856070213260868e-05,
+      "loss": 0.9985,
+      "step": 4688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785506139873877e-05,
+      "loss": 1.0148,
+      "step": 4689
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785405237543811e-05,
+      "loss": 1.0458,
+      "step": 4690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853043143385708e-05,
+      "loss": 0.9442,
+      "step": 4691
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852033702608386e-05,
+      "loss": 0.943,
+      "step": 4692
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785102405313298e-05,
+      "loss": 0.9517,
+      "step": 4693
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785001419498632e-05,
+      "loss": 0.9624,
+      "step": 4694
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849004128195247e-05,
+      "loss": 0.9617,
+      "step": 4695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7847993852786612e-05,
+      "loss": 1.0198,
+      "step": 4696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784698336878726e-05,
+      "loss": 1.0115,
+      "step": 4697
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7845972676224056e-05,
+      "loss": 0.9718,
+      "step": 4698
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844961775123858e-05,
+      "loss": 1.0319,
+      "step": 4699
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7843950665513533e-05,
+      "loss": 1.0089,
+      "step": 4700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784293934741996e-05,
+      "loss": 1.0322,
+      "step": 4701
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7841927820870014e-05,
+      "loss": 1.0058,
+      "step": 4702
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784091608589059e-05,
+      "loss": 0.4103,
+      "step": 4703
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839904142508563e-05,
+      "loss": 1.0394,
+      "step": 4704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7838891990750845e-05,
+      "loss": 0.4357,
+      "step": 4705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837879630644327e-05,
+      "loss": 0.9675,
+      "step": 4706
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783686706221592e-05,
+      "loss": 0.8777,
+      "step": 4707
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835854285492537e-05,
+      "loss": 1.0083,
+      "step": 4708
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78348413005011e-05,
+      "loss": 0.9426,
+      "step": 4709
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833828107268525e-05,
+      "loss": 0.9447,
+      "step": 4710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832814705821748e-05,
+      "loss": 0.8989,
+      "step": 4711
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7831801096187703e-05,
+      "loss": 0.9979,
+      "step": 4712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830787278393326e-05,
+      "loss": 0.9466,
+      "step": 4713
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7829773252465572e-05,
+      "loss": 0.8844,
+      "step": 4714
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782875901843138e-05,
+      "loss": 0.9626,
+      "step": 4715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782774457631772e-05,
+      "loss": 1.0128,
+      "step": 4716
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782672992615155e-05,
+      "loss": 0.926,
+      "step": 4717
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825715067959837e-05,
+      "loss": 0.466,
+      "step": 4718
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7824700001769556e-05,
+      "loss": 0.9879,
+      "step": 4719
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782368472760768e-05,
+      "loss": 1.0599,
+      "step": 4720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7822669245501204e-05,
+      "loss": 0.9476,
+      "step": 4721
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821653555477113e-05,
+      "loss": 0.8813,
+      "step": 4722
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782063765756241e-05,
+      "loss": 1.0149,
+      "step": 4723
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781962155178408e-05,
+      "loss": 0.9561,
+      "step": 4724
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818605238169144e-05,
+      "loss": 1.0014,
+      "step": 4725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781758871674461e-05,
+      "loss": 0.9528,
+      "step": 4726
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7816571987537497e-05,
+      "loss": 0.9823,
+      "step": 4727
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781555505057483e-05,
+      "loss": 0.9693,
+      "step": 4728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814537905883632e-05,
+      "loss": 0.9899,
+      "step": 4729
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7813520553490945e-05,
+      "loss": 0.9711,
+      "step": 4730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812502993423803e-05,
+      "loss": 1.0481,
+      "step": 4731
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811485225709255e-05,
+      "loss": 0.9424,
+      "step": 4732
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810467250374352e-05,
+      "loss": 1.0082,
+      "step": 4733
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780944906744615e-05,
+      "loss": 0.9676,
+      "step": 4734
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780843067695171e-05,
+      "loss": 0.9225,
+      "step": 4735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78074120789181e-05,
+      "loss": 0.9919,
+      "step": 4736
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806393273372396e-05,
+      "loss": 0.9372,
+      "step": 4737
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780537426034167e-05,
+      "loss": 0.9545,
+      "step": 4738
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804355039853017e-05,
+      "loss": 1.0132,
+      "step": 4739
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7803335611933517e-05,
+      "loss": 1.0687,
+      "step": 4740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780231597661027e-05,
+      "loss": 0.9688,
+      "step": 4741
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801296133910374e-05,
+      "loss": 1.0549,
+      "step": 4742
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7800276083860938e-05,
+      "loss": 0.9728,
+      "step": 4743
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779925582648907e-05,
+      "loss": 1.0377,
+      "step": 4744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7798235361821895e-05,
+      "loss": 1.0468,
+      "step": 4745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797214689886527e-05,
+      "loss": 0.9781,
+      "step": 4746
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77961938107101e-05,
+      "loss": 0.9929,
+      "step": 4747
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779517272431974e-05,
+      "loss": 1.0186,
+      "step": 4748
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794151430742596e-05,
+      "loss": 0.9659,
+      "step": 4749
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779312993000581e-05,
+      "loss": 0.9656,
+      "step": 4750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779210822213653e-05,
+      "loss": 0.9983,
+      "step": 4751
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779108630716191e-05,
+      "loss": 0.9915,
+      "step": 4752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790064185109116e-05,
+      "loss": 0.4862,
+      "step": 4753
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789041856005314e-05,
+      "loss": 1.0595,
+      "step": 4754
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7788019319877678e-05,
+      "loss": 1.0032,
+      "step": 4755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786996576753377e-05,
+      "loss": 1.0315,
+      "step": 4756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785973626659604e-05,
+      "loss": 1.1246,
+      "step": 4757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7784950469623546e-05,
+      "loss": 0.9868,
+      "step": 4758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7783927105672393e-05,
+      "loss": 1.1299,
+      "step": 4759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7782903534833347e-05,
+      "loss": 0.9863,
+      "step": 4760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7781879757133616e-05,
+      "loss": 0.9714,
+      "step": 4761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778085577260041e-05,
+      "loss": 1.0642,
+      "step": 4762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777983158126094e-05,
+      "loss": 1.0129,
+      "step": 4763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7778807183142437e-05,
+      "loss": 0.4262,
+      "step": 4764
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777778257827212e-05,
+      "loss": 0.9251,
+      "step": 4765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776757766677227e-05,
+      "loss": 0.948,
+      "step": 4766
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7775732748384994e-05,
+      "loss": 0.9791,
+      "step": 4767
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7774707523422665e-05,
+      "loss": 1.0632,
+      "step": 4768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7773682091817487e-05,
+      "loss": 0.831,
+      "step": 4769
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777265645359672e-05,
+      "loss": 1.0254,
+      "step": 4770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777163060878762e-05,
+      "loss": 1.0105,
+      "step": 4771
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770604557417458e-05,
+      "loss": 0.9874,
+      "step": 4772
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77695782995135e-05,
+      "loss": 0.9925,
+      "step": 4773
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768551835103024e-05,
+      "loss": 1.0203,
+      "step": 4774
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7767525164213312e-05,
+      "loss": 1.0177,
+      "step": 4775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7766498286871656e-05,
+      "loss": 0.9821,
+      "step": 4776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7765471203105343e-05,
+      "loss": 1.039,
+      "step": 4777
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764443912941675e-05,
+      "loss": 0.8482,
+      "step": 4778
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763416416407953e-05,
+      "loss": 0.9786,
+      "step": 4779
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776238871353149e-05,
+      "loss": 0.9804,
+      "step": 4780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761360804339603e-05,
+      "loss": 0.458,
+      "step": 4781
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776033268885961e-05,
+      "loss": 0.9544,
+      "step": 4782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7759304367118837e-05,
+      "loss": 0.8915,
+      "step": 4783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7758275839144615e-05,
+      "loss": 0.9963,
+      "step": 4784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757247104964284e-05,
+      "loss": 0.9266,
+      "step": 4785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775621816460518e-05,
+      "loss": 0.8964,
+      "step": 4786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755189018094662e-05,
+      "loss": 1.0264,
+      "step": 4787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754159665460073e-05,
+      "loss": 1.0776,
+      "step": 4788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7753130106728774e-05,
+      "loss": 1.0282,
+      "step": 4789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752100341928135e-05,
+      "loss": 1.1029,
+      "step": 4790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7751070371085525e-05,
+      "loss": 1.0256,
+      "step": 4791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750040194228315e-05,
+      "loss": 0.9859,
+      "step": 4792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749009811383885e-05,
+      "loss": 0.9755,
+      "step": 4793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7747979222579628e-05,
+      "loss": 1.0237,
+      "step": 4794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774694842784293e-05,
+      "loss": 1.0121,
+      "step": 4795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745917427201194e-05,
+      "loss": 0.9037,
+      "step": 4796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7744886220681815e-05,
+      "loss": 0.4431,
+      "step": 4797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743854808312207e-05,
+      "loss": 0.9526,
+      "step": 4798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7742823190119786e-05,
+      "loss": 0.9784,
+      "step": 4799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741791366131966e-05,
+      "loss": 0.9564,
+      "step": 4800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774075933637617e-05,
+      "loss": 1.0809,
+      "step": 4801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7739727100879838e-05,
+      "loss": 1.0384,
+      "step": 4802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738694659670395e-05,
+      "loss": 1.0442,
+      "step": 4803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737662012775287e-05,
+      "loss": 0.9804,
+      "step": 4804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773662916022196e-05,
+      "loss": 0.9705,
+      "step": 4805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735596102037864e-05,
+      "loss": 0.9941,
+      "step": 4806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734562838250463e-05,
+      "loss": 0.974,
+      "step": 4807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733529368887214e-05,
+      "loss": 1.0506,
+      "step": 4808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7732495693975584e-05,
+      "loss": 0.9784,
+      "step": 4809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773146181354305e-05,
+      "loss": 0.8894,
+      "step": 4810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7730427727617093e-05,
+      "loss": 0.926,
+      "step": 4811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7729393436225198e-05,
+      "loss": 0.968,
+      "step": 4812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772835893939485e-05,
+      "loss": 0.9601,
+      "step": 4813
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772732423715355e-05,
+      "loss": 1.0113,
+      "step": 4814
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7726289329528796e-05,
+      "loss": 1.0158,
+      "step": 4815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77252542165481e-05,
+      "loss": 0.9417,
+      "step": 4816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7724218898238963e-05,
+      "loss": 0.9499,
+      "step": 4817
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7723183374628914e-05,
+      "loss": 0.9492,
+      "step": 4818
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772214764574547e-05,
+      "loss": 0.9508,
+      "step": 4819
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7721111711616163e-05,
+      "loss": 1.0985,
+      "step": 4820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7720075572268522e-05,
+      "loss": 1.0548,
+      "step": 4821
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771903922773009e-05,
+      "loss": 0.9736,
+      "step": 4822
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771800267802841e-05,
+      "loss": 0.9621,
+      "step": 4823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771696592319104e-05,
+      "loss": 1.065,
+      "step": 4824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7715928963245526e-05,
+      "loss": 0.947,
+      "step": 4825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7714891798219432e-05,
+      "loss": 0.9758,
+      "step": 4826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713854428140326e-05,
+      "loss": 0.914,
+      "step": 4827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712816853035776e-05,
+      "loss": 1.0417,
+      "step": 4828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7711779072933367e-05,
+      "loss": 0.9509,
+      "step": 4829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710741087860676e-05,
+      "loss": 0.9524,
+      "step": 4830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709702897845295e-05,
+      "loss": 0.9309,
+      "step": 4831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708664502914815e-05,
+      "loss": 1.0135,
+      "step": 4832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707625903096836e-05,
+      "loss": 0.9924,
+      "step": 4833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706587098418966e-05,
+      "loss": 1.1152,
+      "step": 4834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770554808890881e-05,
+      "loss": 1.0191,
+      "step": 4835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704508874593988e-05,
+      "loss": 1.0285,
+      "step": 4836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7703469455502118e-05,
+      "loss": 0.4152,
+      "step": 4837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770242983166083e-05,
+      "loss": 0.9548,
+      "step": 4838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7701390003097754e-05,
+      "loss": 0.9178,
+      "step": 4839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700349969840525e-05,
+      "loss": 0.9496,
+      "step": 4840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769930973191679e-05,
+      "loss": 0.9682,
+      "step": 4841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7698269289354195e-05,
+      "loss": 0.9928,
+      "step": 4842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697228642180398e-05,
+      "loss": 0.9853,
+      "step": 4843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696187790423048e-05,
+      "loss": 0.997,
+      "step": 4844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695146734109823e-05,
+      "loss": 0.9836,
+      "step": 4845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7694105473268384e-05,
+      "loss": 0.9961,
+      "step": 4846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769306400792641e-05,
+      "loss": 0.9996,
+      "step": 4847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769202233811158e-05,
+      "loss": 1.105,
+      "step": 4848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690980463851582e-05,
+      "loss": 1.0021,
+      "step": 4849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689938385174103e-05,
+      "loss": 0.9959,
+      "step": 4850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768889610210685e-05,
+      "loss": 0.9703,
+      "step": 4851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7687853614677517e-05,
+      "loss": 1.0113,
+      "step": 4852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686810922913817e-05,
+      "loss": 1.0407,
+      "step": 4853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7685768026843462e-05,
+      "loss": 0.9576,
+      "step": 4854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768472492649417e-05,
+      "loss": 1.0968,
+      "step": 4855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683681621893666e-05,
+      "loss": 0.9484,
+      "step": 4856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7682638113069682e-05,
+      "loss": 0.916,
+      "step": 4857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768159440004995e-05,
+      "loss": 0.9294,
+      "step": 4858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7680550482862212e-05,
+      "loss": 0.9225,
+      "step": 4859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679506361534216e-05,
+      "loss": 1.092,
+      "step": 4860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767846203609371e-05,
+      "loss": 1.0248,
+      "step": 4861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767741750656845e-05,
+      "loss": 0.9974,
+      "step": 4862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7676372772986207e-05,
+      "loss": 0.7855,
+      "step": 4863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767532783537474e-05,
+      "loss": 1.0544,
+      "step": 4864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7674282693761822e-05,
+      "loss": 1.0154,
+      "step": 4865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7673237348175238e-05,
+      "loss": 0.4603,
+      "step": 4866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7672191798642767e-05,
+      "loss": 0.9545,
+      "step": 4867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7671146045192203e-05,
+      "loss": 1.012,
+      "step": 4868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670100087851342e-05,
+      "loss": 0.9829,
+      "step": 4869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669053926647974e-05,
+      "loss": 0.9835,
+      "step": 4870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7668007561609915e-05,
+      "loss": 1.0452,
+      "step": 4871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666960992764972e-05,
+      "loss": 0.4064,
+      "step": 4872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665914220140964e-05,
+      "loss": 0.9752,
+      "step": 4873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766486724376571e-05,
+      "loss": 0.8887,
+      "step": 4874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7663820063667035e-05,
+      "loss": 0.9778,
+      "step": 4875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766277267987278e-05,
+      "loss": 0.9292,
+      "step": 4876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766172509241078e-05,
+      "loss": 1.0255,
+      "step": 4877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766067730130887e-05,
+      "loss": 0.9427,
+      "step": 4878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659629306594914e-05,
+      "loss": 0.9259,
+      "step": 4879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7658581108296756e-05,
+      "loss": 1.0314,
+      "step": 4880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7657532706442262e-05,
+      "loss": 0.9509,
+      "step": 4881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656484101059293e-05,
+      "loss": 0.9814,
+      "step": 4882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765543529217572e-05,
+      "loss": 0.9178,
+      "step": 4883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765438627981942e-05,
+      "loss": 0.495,
+      "step": 4884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653337064018276e-05,
+      "loss": 0.3674,
+      "step": 4885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652287644800173e-05,
+      "loss": 1.0105,
+      "step": 4886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651238022193004e-05,
+      "loss": 1.0002,
+      "step": 4887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765018819622467e-05,
+      "loss": 1.0287,
+      "step": 4888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764913816692307e-05,
+      "loss": 0.9975,
+      "step": 4889
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764808793431611e-05,
+      "loss": 1.0563,
+      "step": 4890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7647037498431713e-05,
+      "loss": 0.4709,
+      "step": 4891
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645986859297786e-05,
+      "loss": 0.9932,
+      "step": 4892
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644936016942268e-05,
+      "loss": 0.9374,
+      "step": 4893
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764388497139308e-05,
+      "loss": 1.0294,
+      "step": 4894
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7642833722678157e-05,
+      "loss": 0.9915,
+      "step": 4895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641782270825444e-05,
+      "loss": 0.9257,
+      "step": 4896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7640730615862888e-05,
+      "loss": 1.0037,
+      "step": 4897
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7639678757818436e-05,
+      "loss": 1.0097,
+      "step": 4898
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638626696720048e-05,
+      "loss": 0.9675,
+      "step": 4899
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763757443259569e-05,
+      "loss": 0.9377,
+      "step": 4900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7636521965473324e-05,
+      "loss": 0.9512,
+      "step": 4901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635469295380923e-05,
+      "loss": 0.9606,
+      "step": 4902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7634416422346473e-05,
+      "loss": 1.0197,
+      "step": 4903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633363346397954e-05,
+      "loss": 0.9482,
+      "step": 4904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763231006756335e-05,
+      "loss": 0.9191,
+      "step": 4905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631256585870665e-05,
+      "loss": 1.1314,
+      "step": 4906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7630202901347896e-05,
+      "loss": 0.9791,
+      "step": 4907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762914901402305e-05,
+      "loss": 0.903,
+      "step": 4908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628094923924133e-05,
+      "loss": 0.9482,
+      "step": 4909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762704063107917e-05,
+      "loss": 0.8959,
+      "step": 4910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7625986135516173e-05,
+      "loss": 0.9241,
+      "step": 4911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7624931437263177e-05,
+      "loss": 0.9832,
+      "step": 4912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762387653634821e-05,
+      "loss": 1.0132,
+      "step": 4913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7622821432799312e-05,
+      "loss": 0.9302,
+      "step": 4914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762176612664453e-05,
+      "loss": 0.9178,
+      "step": 4915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7620710617911903e-05,
+      "loss": 0.9776,
+      "step": 4916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76196549066295e-05,
+      "loss": 1.0132,
+      "step": 4917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618598992825365e-05,
+      "loss": 1.0204,
+      "step": 4918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761754287652757e-05,
+      "loss": 0.8973,
+      "step": 4919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616486557764187e-05,
+      "loss": 0.9593,
+      "step": 4920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761543003656329e-05,
+      "loss": 0.9878,
+      "step": 4921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614373312952963e-05,
+      "loss": 0.8967,
+      "step": 4922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7613316386961286e-05,
+      "loss": 0.9459,
+      "step": 4923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761225925861636e-05,
+      "loss": 0.9403,
+      "step": 4924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611201927946272e-05,
+      "loss": 0.9511,
+      "step": 4925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761014439497913e-05,
+      "loss": 0.961,
+      "step": 4926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7609086659743045e-05,
+      "loss": 1.0128,
+      "step": 4927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760802872226612e-05,
+      "loss": 0.9511,
+      "step": 4928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606970582576485e-05,
+      "loss": 0.8837,
+      "step": 4929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7605912240702264e-05,
+      "loss": 0.9082,
+      "step": 4930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7604853696671578e-05,
+      "loss": 0.9867,
+      "step": 4931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7603794950512566e-05,
+      "loss": 1.039,
+      "step": 4932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760273600225337e-05,
+      "loss": 0.9931,
+      "step": 4933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7601676851922132e-05,
+      "loss": 0.9184,
+      "step": 4934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760061749954701e-05,
+      "loss": 0.9103,
+      "step": 4935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759955794515615e-05,
+      "loss": 0.9219,
+      "step": 4936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7598498188777724e-05,
+      "loss": 1.0183,
+      "step": 4937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759743823043989e-05,
+      "loss": 0.4523,
+      "step": 4938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7596378070170828e-05,
+      "loss": 0.9888,
+      "step": 4939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759531770799871e-05,
+      "loss": 1.0198,
+      "step": 4940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594257143951726e-05,
+      "loss": 1.0093,
+      "step": 4941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759319637805806e-05,
+      "loss": 0.9091,
+      "step": 4942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592135410345904e-05,
+      "loss": 0.9357,
+      "step": 4943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759107424084346e-05,
+      "loss": 1.0204,
+      "step": 4944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590012869578937e-05,
+      "loss": 0.9776,
+      "step": 4945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588951296580536e-05,
+      "loss": 1.0502,
+      "step": 4946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758788952187648e-05,
+      "loss": 1.0068,
+      "step": 4947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758682754549498e-05,
+      "loss": 1.0341,
+      "step": 4948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585765367464276e-05,
+      "loss": 0.97,
+      "step": 4949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758470298781259e-05,
+      "loss": 0.9976,
+      "step": 4950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583640406568163e-05,
+      "loss": 1.0799,
+      "step": 4951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758257762375923e-05,
+      "loss": 0.9541,
+      "step": 4952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7581514639414048e-05,
+      "loss": 1.0258,
+      "step": 4953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7580451453560866e-05,
+      "loss": 0.9055,
+      "step": 4954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757938806622794e-05,
+      "loss": 0.9339,
+      "step": 4955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7578324477443537e-05,
+      "loss": 0.9043,
+      "step": 4956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7577260687235925e-05,
+      "loss": 1.0376,
+      "step": 4957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757619669563338e-05,
+      "loss": 0.9658,
+      "step": 4958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7575132502664176e-05,
+      "loss": 1.0295,
+      "step": 4959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574068108356603e-05,
+      "loss": 1.0214,
+      "step": 4960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573003512738947e-05,
+      "loss": 0.9706,
+      "step": 4961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757193871583951e-05,
+      "loss": 1.001,
+      "step": 4962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757087371768659e-05,
+      "loss": 0.9649,
+      "step": 4963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7569808518308494e-05,
+      "loss": 1.0444,
+      "step": 4964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568743117733532e-05,
+      "loss": 0.9914,
+      "step": 4965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7567677515990023e-05,
+      "loss": 1.0514,
+      "step": 4966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566611713106287e-05,
+      "loss": 0.9497,
+      "step": 4967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7565545709110657e-05,
+      "loss": 0.9346,
+      "step": 4968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564479504031458e-05,
+      "loss": 0.9596,
+      "step": 4969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7563413097897037e-05,
+      "loss": 0.9812,
+      "step": 4970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7562346490735733e-05,
+      "loss": 0.892,
+      "step": 4971
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561279682575897e-05,
+      "loss": 1.0285,
+      "step": 4972
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756021267344588e-05,
+      "loss": 0.8882,
+      "step": 4973
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755914546337405e-05,
+      "loss": 0.9082,
+      "step": 4974
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755807805238876e-05,
+      "loss": 1.0428,
+      "step": 4975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7557010440518392e-05,
+      "loss": 0.9584,
+      "step": 4976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7555942627791317e-05,
+      "loss": 0.9511,
+      "step": 4977
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554874614235915e-05,
+      "loss": 0.9456,
+      "step": 4978
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7553806399880576e-05,
+      "loss": 0.964,
+      "step": 4979
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755273798475369e-05,
+      "loss": 0.938,
+      "step": 4980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551669368883653e-05,
+      "loss": 0.9337,
+      "step": 4981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7550600552298873e-05,
+      "loss": 1.0265,
+      "step": 4982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754953153502775e-05,
+      "loss": 1.0122,
+      "step": 4983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7548462317098702e-05,
+      "loss": 0.8425,
+      "step": 4984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7547392898540145e-05,
+      "loss": 0.9701,
+      "step": 4985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7546323279380507e-05,
+      "loss": 0.9795,
+      "step": 4986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7545253459648213e-05,
+      "loss": 1.0211,
+      "step": 4987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7544183439371705e-05,
+      "loss": 0.9135,
+      "step": 4988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543113218579414e-05,
+      "loss": 0.9805,
+      "step": 4989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7542042797299785e-05,
+      "loss": 0.9349,
+      "step": 4990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540972175561277e-05,
+      "loss": 1.0698,
+      "step": 4991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753990135339234e-05,
+      "loss": 0.8847,
+      "step": 4992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7538830330821438e-05,
+      "loss": 0.9685,
+      "step": 4993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7537759107877033e-05,
+      "loss": 1.0045,
+      "step": 4994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536687684587602e-05,
+      "loss": 0.94,
+      "step": 4995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753561606098162e-05,
+      "loss": 0.9812,
+      "step": 4996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534544237087568e-05,
+      "loss": 0.9712,
+      "step": 4997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533472212933936e-05,
+      "loss": 1.1648,
+      "step": 4998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7532399988549217e-05,
+      "loss": 0.9762,
+      "step": 4999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753132756396191e-05,
+      "loss": 1.0046,
+      "step": 5000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7530254939200518e-05,
+      "loss": 0.9232,
+      "step": 5001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529182114293546e-05,
+      "loss": 1.0243,
+      "step": 5002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7528109089269518e-05,
+      "loss": 1.0229,
+      "step": 5003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527035864156944e-05,
+      "loss": 1.0149,
+      "step": 5004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7525962438984353e-05,
+      "loss": 0.9554,
+      "step": 5005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752488881378028e-05,
+      "loss": 0.9909,
+      "step": 5006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523814988573255e-05,
+      "loss": 0.9554,
+      "step": 5007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752274096339182e-05,
+      "loss": 1.0008,
+      "step": 5008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7521666738264524e-05,
+      "loss": 0.999,
+      "step": 5009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520592313219917e-05,
+      "loss": 0.9774,
+      "step": 5010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7519517688286552e-05,
+      "loss": 0.9197,
+      "step": 5011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7518442863492996e-05,
+      "loss": 0.9172,
+      "step": 5012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7517367838867816e-05,
+      "loss": 0.9827,
+      "step": 5013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516292614439586e-05,
+      "loss": 1.0088,
+      "step": 5014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751521719023688e-05,
+      "loss": 0.8497,
+      "step": 5015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7514141566288286e-05,
+      "loss": 0.9828,
+      "step": 5016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513065742622392e-05,
+      "loss": 0.9628,
+      "step": 5017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751198971926779e-05,
+      "loss": 0.9481,
+      "step": 5018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751091349625308e-05,
+      "loss": 0.9728,
+      "step": 5019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750983707360687e-05,
+      "loss": 1.0066,
+      "step": 5020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508760451357764e-05,
+      "loss": 0.9867,
+      "step": 5021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507683629534384e-05,
+      "loss": 1.0618,
+      "step": 5022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7506606608165346e-05,
+      "loss": 0.9374,
+      "step": 5023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750552938727928e-05,
+      "loss": 0.9288,
+      "step": 5024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504451966904814e-05,
+      "loss": 1.0459,
+      "step": 5025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7503374347070586e-05,
+      "loss": 0.9984,
+      "step": 5026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502296527805236e-05,
+      "loss": 0.9412,
+      "step": 5027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7501218509137414e-05,
+      "loss": 1.0703,
+      "step": 5028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750014029109577e-05,
+      "loss": 1.0149,
+      "step": 5029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7499061873708962e-05,
+      "loss": 1.0064,
+      "step": 5030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497983257005657e-05,
+      "loss": 0.8992,
+      "step": 5031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749690444101452e-05,
+      "loss": 0.9742,
+      "step": 5032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7495825425764222e-05,
+      "loss": 0.8754,
+      "step": 5033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749474621128345e-05,
+      "loss": 0.4719,
+      "step": 5034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7493666797600882e-05,
+      "loss": 1.0722,
+      "step": 5035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492587184745208e-05,
+      "loss": 0.9968,
+      "step": 5036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7491507372745125e-05,
+      "loss": 0.9381,
+      "step": 5037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749042736162933e-05,
+      "loss": 0.9715,
+      "step": 5038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489347151426534e-05,
+      "loss": 0.9724,
+      "step": 5039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7488266742165444e-05,
+      "loss": 0.9516,
+      "step": 5040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487186133874773e-05,
+      "loss": 0.955,
+      "step": 5041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748610532658325e-05,
+      "loss": 0.866,
+      "step": 5042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485024320319595e-05,
+      "loss": 1.1672,
+      "step": 5043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748394311511254e-05,
+      "loss": 0.9767,
+      "step": 5044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482861710990828e-05,
+      "loss": 1.0457,
+      "step": 5045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481780107983197e-05,
+      "loss": 1.0123,
+      "step": 5046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480698306118396e-05,
+      "loss": 0.992,
+      "step": 5047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479616305425178e-05,
+      "loss": 0.9653,
+      "step": 5048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74785341059323e-05,
+      "loss": 1.009,
+      "step": 5049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747745170766853e-05,
+      "loss": 1.024,
+      "step": 5050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747636911066263e-05,
+      "loss": 0.4287,
+      "step": 5051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475286314943377e-05,
+      "loss": 0.9754,
+      "step": 5052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474203320539555e-05,
+      "loss": 0.9983,
+      "step": 5053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7473120127479944e-05,
+      "loss": 1.0017,
+      "step": 5054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472036735793333e-05,
+      "loss": 1.0306,
+      "step": 5055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747095314550852e-05,
+      "loss": 1.015,
+      "step": 5056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7469869356654307e-05,
+      "loss": 1.0149,
+      "step": 5057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468785369259497e-05,
+      "loss": 0.9603,
+      "step": 5058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7467701183352904e-05,
+      "loss": 0.9174,
+      "step": 5059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746661679896334e-05,
+      "loss": 0.9505,
+      "step": 5060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465532216119628e-05,
+      "loss": 1.0172,
+      "step": 5061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7464447434850598e-05,
+      "loss": 1.0268,
+      "step": 5062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746336245518508e-05,
+      "loss": 0.9887,
+      "step": 5063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7462277277151912e-05,
+      "loss": 0.9345,
+      "step": 5064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7461191900779936e-05,
+      "loss": 1.0001,
+      "step": 5065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7460106326098004e-05,
+      "loss": 1.1063,
+      "step": 5066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7459020553134962e-05,
+      "loss": 0.9538,
+      "step": 5067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7457934581919675e-05,
+      "loss": 1.0005,
+      "step": 5068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7456848412481003e-05,
+      "loss": 0.9838,
+      "step": 5069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745576204484782e-05,
+      "loss": 1.0127,
+      "step": 5070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454675479048996e-05,
+      "loss": 0.966,
+      "step": 5071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745358871511341e-05,
+      "loss": 1.0333,
+      "step": 5072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7452501753069948e-05,
+      "loss": 0.4691,
+      "step": 5073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7451414592947504e-05,
+      "loss": 0.9898,
+      "step": 5074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745032723477497e-05,
+      "loss": 1.0514,
+      "step": 5075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7449239678581244e-05,
+      "loss": 0.9754,
+      "step": 5076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7448151924395237e-05,
+      "loss": 1.0446,
+      "step": 5077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7447063972245857e-05,
+      "loss": 0.9762,
+      "step": 5078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445975822162024e-05,
+      "loss": 0.4552,
+      "step": 5079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744488747417266e-05,
+      "loss": 0.9635,
+      "step": 5080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443798928306685e-05,
+      "loss": 0.973,
+      "step": 5081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744271018459304e-05,
+      "loss": 0.8697,
+      "step": 5082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441621243060654e-05,
+      "loss": 1.005,
+      "step": 5083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440532103738477e-05,
+      "loss": 1.0595,
+      "step": 5084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439442766655456e-05,
+      "loss": 0.9156,
+      "step": 5085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7438353231840545e-05,
+      "loss": 0.9937,
+      "step": 5086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7437263499322696e-05,
+      "loss": 0.9043,
+      "step": 5087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743617356913088e-05,
+      "loss": 0.954,
+      "step": 5088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7435083441294064e-05,
+      "loss": 0.9321,
+      "step": 5089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743399311584122e-05,
+      "loss": 1.0375,
+      "step": 5090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432902592801327e-05,
+      "loss": 1.0011,
+      "step": 5091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7431811872203373e-05,
+      "loss": 0.8894,
+      "step": 5092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7430720954076353e-05,
+      "loss": 0.9374,
+      "step": 5093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742962983844925e-05,
+      "loss": 1.033,
+      "step": 5094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742853852535108e-05,
+      "loss": 0.9785,
+      "step": 5095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427447014810832e-05,
+      "loss": 0.9848,
+      "step": 5096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7426355306857526e-05,
+      "loss": 1.0335,
+      "step": 5097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742526340152018e-05,
+      "loss": 0.9383,
+      "step": 5098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7424171298827812e-05,
+      "loss": 0.8887,
+      "step": 5099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7423078998809453e-05,
+      "loss": 0.999,
+      "step": 5100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742198650149413e-05,
+      "loss": 0.9379,
+      "step": 5101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420893806910882e-05,
+      "loss": 0.9756,
+      "step": 5102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741980091508875e-05,
+      "loss": 1.0142,
+      "step": 5103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741870782605679e-05,
+      "loss": 0.9334,
+      "step": 5104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7417614539844044e-05,
+      "loss": 0.914,
+      "step": 5105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7416521056479577e-05,
+      "loss": 0.9696,
+      "step": 5106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741542737599245e-05,
+      "loss": 0.9153,
+      "step": 5107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414333498411734e-05,
+      "loss": 1.0089,
+      "step": 5108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413239423766502e-05,
+      "loss": 0.912,
+      "step": 5109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741214515208583e-05,
+      "loss": 0.974,
+      "step": 5110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411050683398806e-05,
+      "loss": 0.9352,
+      "step": 5111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740995601773452e-05,
+      "loss": 1.0128,
+      "step": 5112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740886115512207e-05,
+      "loss": 1.0182,
+      "step": 5113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407766095590545e-05,
+      "loss": 0.4241,
+      "step": 5114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740667083916906e-05,
+      "loss": 0.9546,
+      "step": 5115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405575385886723e-05,
+      "loss": 1.0582,
+      "step": 5116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7404479735772654e-05,
+      "loss": 0.9425,
+      "step": 5117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740338388885597e-05,
+      "loss": 1.0192,
+      "step": 5118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402287845165794e-05,
+      "loss": 1.0433,
+      "step": 5119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740119160473126e-05,
+      "loss": 0.9489,
+      "step": 5120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740009516758151e-05,
+      "loss": 1.0502,
+      "step": 5121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739899853374568e-05,
+      "loss": 0.9912,
+      "step": 5122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739790170325292e-05,
+      "loss": 1.0388,
+      "step": 5123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7396804676132385e-05,
+      "loss": 0.8296,
+      "step": 5124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395707452413226e-05,
+      "loss": 0.969,
+      "step": 5125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739461003212461e-05,
+      "loss": 0.9981,
+      "step": 5126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739351241529571e-05,
+      "loss": 0.8943,
+      "step": 5127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7392414601955688e-05,
+      "loss": 1.0544,
+      "step": 5128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391316592133728e-05,
+      "loss": 0.9452,
+      "step": 5129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739021838585902e-05,
+      "loss": 0.97,
+      "step": 5130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7389119983160742e-05,
+      "loss": 0.9685,
+      "step": 5131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73880213840681e-05,
+      "loss": 0.9995,
+      "step": 5132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738692258861028e-05,
+      "loss": 0.906,
+      "step": 5133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385823596816503e-05,
+      "loss": 0.9885,
+      "step": 5134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7384724408715964e-05,
+      "loss": 0.9638,
+      "step": 5135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383625024337888e-05,
+      "loss": 0.9518,
+      "step": 5136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382525443711487e-05,
+      "loss": 0.9665,
+      "step": 5137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7381425666865998e-05,
+      "loss": 1.0739,
+      "step": 5138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738032569383064e-05,
+      "loss": 1.0182,
+      "step": 5139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379225524634657e-05,
+      "loss": 1.0438,
+      "step": 5140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378125159307287e-05,
+      "loss": 0.9365,
+      "step": 5141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7377024597877774e-05,
+      "loss": 1.0274,
+      "step": 5142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375923840375377e-05,
+      "loss": 0.9614,
+      "step": 5143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7374822886829345e-05,
+      "loss": 1.0299,
+      "step": 5144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373721737268946e-05,
+      "loss": 0.9787,
+      "step": 5145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372620391723443e-05,
+      "loss": 0.9158,
+      "step": 5146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737151885022211e-05,
+      "loss": 0.9801,
+      "step": 5147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370417112794226e-05,
+      "loss": 0.9629,
+      "step": 5148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369315179469074e-05,
+      "loss": 0.9712,
+      "step": 5149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736821305027594e-05,
+      "loss": 0.94,
+      "step": 5150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7367110725244117e-05,
+      "loss": 1.0034,
+      "step": 5151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7366008204402906e-05,
+      "loss": 0.9686,
+      "step": 5152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364905487781613e-05,
+      "loss": 1.0129,
+      "step": 5153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736380257540954e-05,
+      "loss": 0.9814,
+      "step": 5154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362699467316005e-05,
+      "loss": 1.0405,
+      "step": 5155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7361596163530324e-05,
+      "loss": 0.9848,
+      "step": 5156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360492664081832e-05,
+      "loss": 1.0114,
+      "step": 5157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7359388968999846e-05,
+      "loss": 1.0452,
+      "step": 5158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735828507831371e-05,
+      "loss": 0.9523,
+      "step": 5159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735718099205276e-05,
+      "loss": 0.9982,
+      "step": 5160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7356076710246338e-05,
+      "loss": 0.4734,
+      "step": 5161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7354972232923804e-05,
+      "loss": 0.9952,
+      "step": 5162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7353867560114505e-05,
+      "loss": 0.924,
+      "step": 5163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7352762691847808e-05,
+      "loss": 0.93,
+      "step": 5164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7351657628153072e-05,
+      "loss": 1.038,
+      "step": 5165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350552369059673e-05,
+      "loss": 0.9952,
+      "step": 5166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734944691459699e-05,
+      "loss": 1.0676,
+      "step": 5167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7348341264794402e-05,
+      "loss": 1.0516,
+      "step": 5168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7347235419681293e-05,
+      "loss": 0.9775,
+      "step": 5169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734612937928706e-05,
+      "loss": 0.9164,
+      "step": 5170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345023143641094e-05,
+      "loss": 0.9353,
+      "step": 5171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734391671277281e-05,
+      "loss": 1.0024,
+      "step": 5172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73428100867116e-05,
+      "loss": 1.111,
+      "step": 5173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734170326548689e-05,
+      "loss": 0.9677,
+      "step": 5174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7340596249128084e-05,
+      "loss": 1.0035,
+      "step": 5175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733948903766462e-05,
+      "loss": 0.9797,
+      "step": 5176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733838163112592e-05,
+      "loss": 1.017,
+      "step": 5177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337274029541416e-05,
+      "loss": 0.942,
+      "step": 5178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7336166232940548e-05,
+      "loss": 1.0109,
+      "step": 5179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733505824135276e-05,
+      "loss": 0.9773,
+      "step": 5180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73339500548075e-05,
+      "loss": 0.9913,
+      "step": 5181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7332841673334225e-05,
+      "loss": 1.0581,
+      "step": 5182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7331733096962395e-05,
+      "loss": 1.0007,
+      "step": 5183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7330624325721474e-05,
+      "loss": 0.4411,
+      "step": 5184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7329515359640928e-05,
+      "loss": 1.0524,
+      "step": 5185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328406198750235e-05,
+      "loss": 0.9143,
+      "step": 5186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7327296843078874e-05,
+      "loss": 1.0213,
+      "step": 5187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7326187292656332e-05,
+      "loss": 0.9823,
+      "step": 5188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73250775475121e-05,
+      "loss": 0.9436,
+      "step": 5189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7323967607675675e-05,
+      "loss": 0.9452,
+      "step": 5190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732285747317655e-05,
+      "loss": 0.9415,
+      "step": 5191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321747144044244e-05,
+      "loss": 1.0119,
+      "step": 5192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7320636620308255e-05,
+      "loss": 0.9312,
+      "step": 5193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731952590199811e-05,
+      "loss": 0.891,
+      "step": 5194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318414989143322e-05,
+      "loss": 1.0662,
+      "step": 5195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7317303881773427e-05,
+      "loss": 1.0286,
+      "step": 5196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7316192579917946e-05,
+      "loss": 1.0196,
+      "step": 5197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315081083606425e-05,
+      "loss": 1.0082,
+      "step": 5198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313969392868405e-05,
+      "loss": 0.9881,
+      "step": 5199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7312857507733433e-05,
+      "loss": 0.9264,
+      "step": 5200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311745428231058e-05,
+      "loss": 0.9769,
+      "step": 5201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731063315439084e-05,
+      "loss": 0.9556,
+      "step": 5202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7309520686242342e-05,
+      "loss": 1.0205,
+      "step": 5203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308408023815135e-05,
+      "loss": 0.8897,
+      "step": 5204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7307295167138787e-05,
+      "loss": 0.9881,
+      "step": 5205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730618211624288e-05,
+      "loss": 0.9311,
+      "step": 5206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305068871157e-05,
+      "loss": 0.922,
+      "step": 5207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730395543191073e-05,
+      "loss": 1.006,
+      "step": 5208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302841798533662e-05,
+      "loss": 0.9868,
+      "step": 5209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301727971055404e-05,
+      "loss": 0.9552,
+      "step": 5210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7300613949505557e-05,
+      "loss": 0.9003,
+      "step": 5211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729949973391373e-05,
+      "loss": 0.4195,
+      "step": 5212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729838532430953e-05,
+      "loss": 0.9757,
+      "step": 5213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7297270720722594e-05,
+      "loss": 0.9292,
+      "step": 5214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7296155923182526e-05,
+      "loss": 0.9951,
+      "step": 5215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295040931718975e-05,
+      "loss": 0.9905,
+      "step": 5216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7293925746361563e-05,
+      "loss": 0.9968,
+      "step": 5217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292810367139935e-05,
+      "loss": 0.9133,
+      "step": 5218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729169479408374e-05,
+      "loss": 0.9544,
+      "step": 5219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290579027222625e-05,
+      "loss": 1.0243,
+      "step": 5220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7289463066586242e-05,
+      "loss": 0.9659,
+      "step": 5221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728834691220426e-05,
+      "loss": 0.9673,
+      "step": 5222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728723056410634e-05,
+      "loss": 0.9455,
+      "step": 5223
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286114022322153e-05,
+      "loss": 0.9485,
+      "step": 5224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728499728688138e-05,
+      "loss": 0.9115,
+      "step": 5225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7283880357813698e-05,
+      "loss": 1.0749,
+      "step": 5226
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7282763235148793e-05,
+      "loss": 0.9707,
+      "step": 5227
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281645918916358e-05,
+      "loss": 0.9959,
+      "step": 5228
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7280528409146097e-05,
+      "loss": 0.9244,
+      "step": 5229
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72794107058677e-05,
+      "loss": 1.0089,
+      "step": 5230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278292809110884e-05,
+      "loss": 0.9865,
+      "step": 5231
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277174718905353e-05,
+      "loss": 0.9896,
+      "step": 5232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7276056435280835e-05,
+      "loss": 0.974,
+      "step": 5233
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7274937958267046e-05,
+      "loss": 0.9209,
+      "step": 5234
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727381928789371e-05,
+      "loss": 0.9841,
+      "step": 5235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727270042419057e-05,
+      "loss": 1.0393,
+      "step": 5236
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271581367187357e-05,
+      "loss": 0.9452,
+      "step": 5237
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727046211691382e-05,
+      "loss": 0.9607,
+      "step": 5238
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72693426733997e-05,
+      "loss": 1.0555,
+      "step": 5239
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268223036674753e-05,
+      "loss": 0.9766,
+      "step": 5240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726710320676874e-05,
+      "loss": 0.9826,
+      "step": 5241
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7265983183711425e-05,
+      "loss": 0.9471,
+      "step": 5242
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264862967532573e-05,
+      "loss": 1.0096,
+      "step": 5243
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7263742558261964e-05,
+      "loss": 0.9897,
+      "step": 5244
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262621955929367e-05,
+      "loss": 0.972,
+      "step": 5245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726150116056458e-05,
+      "loss": 0.9262,
+      "step": 5246
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726038017219738e-05,
+      "loss": 0.9815,
+      "step": 5247
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7259258990857574e-05,
+      "loss": 0.9788,
+      "step": 5248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725813761657495e-05,
+      "loss": 0.4504,
+      "step": 5249
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725701604937932e-05,
+      "loss": 1.0065,
+      "step": 5250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725589428930049e-05,
+      "loss": 1.0669,
+      "step": 5251
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725477233636828e-05,
+      "loss": 1.057,
+      "step": 5252
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72536501906125e-05,
+      "loss": 0.9492,
+      "step": 5253
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725252785206299e-05,
+      "loss": 1.0001,
+      "step": 5254
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7251405320749567e-05,
+      "loss": 1.012,
+      "step": 5255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250282596702077e-05,
+      "loss": 1.0465,
+      "step": 5256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7249159679950357e-05,
+      "loss": 0.9974,
+      "step": 5257
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7248036570524246e-05,
+      "loss": 1.0336,
+      "step": 5258
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7246913268453606e-05,
+      "loss": 0.9945,
+      "step": 5259
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7245789773768287e-05,
+      "loss": 1.0252,
+      "step": 5260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244666086498152e-05,
+      "loss": 0.9408,
+      "step": 5261
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7243542206673064e-05,
+      "loss": 0.9786,
+      "step": 5262
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7242418134322897e-05,
+      "loss": 1.0069,
+      "step": 5263
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241293869477533e-05,
+      "loss": 1.009,
+      "step": 5264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240169412166842e-05,
+      "loss": 0.9881,
+      "step": 5265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723904476242072e-05,
+      "loss": 1.039,
+      "step": 5266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7237919920269054e-05,
+      "loss": 0.9842,
+      "step": 5267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236794885741747e-05,
+      "loss": 1.0073,
+      "step": 5268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7235669658868698e-05,
+      "loss": 0.9631,
+      "step": 5269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7234544239679807e-05,
+      "loss": 1.0072,
+      "step": 5270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233418628204997e-05,
+      "loss": 1.0902,
+      "step": 5271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723229282447418e-05,
+      "loss": 0.9638,
+      "step": 5272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231166828517277e-05,
+      "loss": 0.8957,
+      "step": 5273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7230040640364225e-05,
+      "loss": 0.993,
+      "step": 5274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7228914260044948e-05,
+      "loss": 0.9456,
+      "step": 5275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7227787687589382e-05,
+      "loss": 0.9406,
+      "step": 5276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7226660923027476e-05,
+      "loss": 0.9564,
+      "step": 5277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722553396638918e-05,
+      "loss": 0.9163,
+      "step": 5278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722440681770444e-05,
+      "loss": 0.8698,
+      "step": 5279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223279477003217e-05,
+      "loss": 0.4629,
+      "step": 5280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222151944315476e-05,
+      "loss": 0.9329,
+      "step": 5281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722102421967119e-05,
+      "loss": 0.9986,
+      "step": 5282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721989630310032e-05,
+      "loss": 0.9792,
+      "step": 5283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218768194632858e-05,
+      "loss": 1.008,
+      "step": 5284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217639894298778e-05,
+      "loss": 0.8879,
+      "step": 5285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216511402128076e-05,
+      "loss": 0.9456,
+      "step": 5286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721538271815074e-05,
+      "loss": 1.0389,
+      "step": 5287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214253842396774e-05,
+      "loss": 0.9633,
+      "step": 5288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721312477489618e-05,
+      "loss": 0.9798,
+      "step": 5289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211995515678968e-05,
+      "loss": 0.963,
+      "step": 5290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210866064775152e-05,
+      "loss": 1.0067,
+      "step": 5291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720973642221475e-05,
+      "loss": 0.3876,
+      "step": 5292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7208606588027788e-05,
+      "loss": 0.9467,
+      "step": 5293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207476562244297e-05,
+      "loss": 0.9984,
+      "step": 5294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7206346344894314e-05,
+      "loss": 0.9125,
+      "step": 5295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720521593600787e-05,
+      "loss": 0.9713,
+      "step": 5296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720408533561502e-05,
+      "loss": 0.9155,
+      "step": 5297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720295454374581e-05,
+      "loss": 0.9133,
+      "step": 5298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201823560430288e-05,
+      "loss": 0.9342,
+      "step": 5299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7200692385698527e-05,
+      "loss": 0.9566,
+      "step": 5300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7199561019580587e-05,
+      "loss": 1.0073,
+      "step": 5301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198429462106533e-05,
+      "loss": 1.0655,
+      "step": 5302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719729771330645e-05,
+      "loss": 0.9877,
+      "step": 5303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719616577321041e-05,
+      "loss": 0.889,
+      "step": 5304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195033641848507e-05,
+      "loss": 0.8508,
+      "step": 5305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193901319250822e-05,
+      "loss": 1.0772,
+      "step": 5306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719276880544746e-05,
+      "loss": 1.0072,
+      "step": 5307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7191636100468518e-05,
+      "loss": 1.0101,
+      "step": 5308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71905032043441e-05,
+      "loss": 0.9185,
+      "step": 5309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7189370117104324e-05,
+      "loss": 1.066,
+      "step": 5310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7188236838779297e-05,
+      "loss": 0.9979,
+      "step": 5311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187103369399144e-05,
+      "loss": 0.9793,
+      "step": 5312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185969708993993e-05,
+      "loss": 0.9699,
+      "step": 5313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184835857593975e-05,
+      "loss": 0.9622,
+      "step": 5314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183701815229224e-05,
+      "loss": 0.9655,
+      "step": 5315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7182567581929883e-05,
+      "loss": 0.9376,
+      "step": 5316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7181433157726105e-05,
+      "loss": 0.9086,
+      "step": 5317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7180298542648026e-05,
+      "loss": 0.9721,
+      "step": 5318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717916373672582e-05,
+      "loss": 1.0634,
+      "step": 5319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7178028739989637e-05,
+      "loss": 1.0853,
+      "step": 5320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7176893552469653e-05,
+      "loss": 0.9512,
+      "step": 5321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717575817419603e-05,
+      "loss": 0.973,
+      "step": 5322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7174622605198952e-05,
+      "loss": 0.4897,
+      "step": 5323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71734868455086e-05,
+      "loss": 0.91,
+      "step": 5324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717235089515516e-05,
+      "loss": 0.9736,
+      "step": 5325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171214754168826e-05,
+      "loss": 1.0148,
+      "step": 5326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717007842257979e-05,
+      "loss": 1.0039,
+      "step": 5327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168941900418262e-05,
+      "loss": 0.894,
+      "step": 5328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716780518771445e-05,
+      "loss": 0.9534,
+      "step": 5329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716666828449856e-05,
+      "loss": 0.9576,
+      "step": 5330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7165531190800807e-05,
+      "loss": 1.0018,
+      "step": 5331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7164393906651424e-05,
+      "loss": 0.9044,
+      "step": 5332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716325643208063e-05,
+      "loss": 0.972,
+      "step": 5333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162118767118664e-05,
+      "loss": 0.9367,
+      "step": 5334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716098091179576e-05,
+      "loss": 0.9954,
+      "step": 5335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7159842866142162e-05,
+      "loss": 1.0547,
+      "step": 5336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7158704630188118e-05,
+      "loss": 0.9748,
+      "step": 5337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715756620396388e-05,
+      "loss": 0.9923,
+      "step": 5338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715642758749971e-05,
+      "loss": 0.9944,
+      "step": 5339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7155288780825866e-05,
+      "loss": 0.8958,
+      "step": 5340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154149783972615e-05,
+      "loss": 0.9391,
+      "step": 5341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7153010596970238e-05,
+      "loss": 0.9111,
+      "step": 5342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715187121984901e-05,
+      "loss": 0.9426,
+      "step": 5343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715073165263921e-05,
+      "loss": 0.9675,
+      "step": 5344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714959189537113e-05,
+      "loss": 0.9469,
+      "step": 5345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7148451948075063e-05,
+      "loss": 0.9348,
+      "step": 5346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7147311810781307e-05,
+      "loss": 0.9835,
+      "step": 5347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714617148352017e-05,
+      "loss": 1.0003,
+      "step": 5348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7145030966321954e-05,
+      "loss": 0.944,
+      "step": 5349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143890259216974e-05,
+      "loss": 1.0332,
+      "step": 5350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7142749362235553e-05,
+      "loss": 0.9942,
+      "step": 5351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714160827540801e-05,
+      "loss": 1.0332,
+      "step": 5352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7140466998764673e-05,
+      "loss": 0.9841,
+      "step": 5353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139325532335883e-05,
+      "loss": 1.0396,
+      "step": 5354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138183876151973e-05,
+      "loss": 1.0221,
+      "step": 5355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7137042030243285e-05,
+      "loss": 1.0126,
+      "step": 5356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135899994640175e-05,
+      "loss": 0.9825,
+      "step": 5357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7134757769372992e-05,
+      "loss": 0.9654,
+      "step": 5358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7133615354472093e-05,
+      "loss": 0.9035,
+      "step": 5359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713247274996785e-05,
+      "loss": 0.9432,
+      "step": 5360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131329955890626e-05,
+      "loss": 0.924,
+      "step": 5361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7130186972270796e-05,
+      "loss": 0.9164,
+      "step": 5362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712904379913874e-05,
+      "loss": 0.8875,
+      "step": 5363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712790043652484e-05,
+      "loss": 0.9007,
+      "step": 5364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7126756884459487e-05,
+      "loss": 0.9853,
+      "step": 5365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712561314297308e-05,
+      "loss": 0.9424,
+      "step": 5366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712446921209601e-05,
+      "loss": 0.9462,
+      "step": 5367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7123325091858688e-05,
+      "loss": 0.946,
+      "step": 5368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7122180782291518e-05,
+      "loss": 0.9484,
+      "step": 5369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7121036283424918e-05,
+      "loss": 0.9596,
+      "step": 5370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119891595289308e-05,
+      "loss": 0.9432,
+      "step": 5371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711874671791511e-05,
+      "loss": 0.9465,
+      "step": 5372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7117601651332757e-05,
+      "loss": 0.8817,
+      "step": 5373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116456395572675e-05,
+      "loss": 0.9717,
+      "step": 5374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7115310950665317e-05,
+      "loss": 0.9709,
+      "step": 5375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114165316641113e-05,
+      "loss": 0.9918,
+      "step": 5376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113019493530525e-05,
+      "loss": 0.9814,
+      "step": 5377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7111873481364e-05,
+      "loss": 0.9734,
+      "step": 5378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7110727280172006e-05,
+      "loss": 0.8962,
+      "step": 5379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109580889984997e-05,
+      "loss": 0.9942,
+      "step": 5380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7108434310833448e-05,
+      "loss": 0.969,
+      "step": 5381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7107287542747836e-05,
+      "loss": 0.9399,
+      "step": 5382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7106140585758636e-05,
+      "loss": 1.008,
+      "step": 5383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7104993439896336e-05,
+      "loss": 0.8585,
+      "step": 5384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710384610519143e-05,
+      "loss": 1.0177,
+      "step": 5385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71026985816744e-05,
+      "loss": 0.9865,
+      "step": 5386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101550869375758e-05,
+      "loss": 1.0507,
+      "step": 5387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7100402968326e-05,
+      "loss": 0.909,
+      "step": 5388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099254878555646e-05,
+      "loss": 0.9704,
+      "step": 5389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7098106600095204e-05,
+      "loss": 1.0054,
+      "step": 5390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7096958132975195e-05,
+      "loss": 0.9765,
+      "step": 5391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095809477226146e-05,
+      "loss": 0.9592,
+      "step": 5392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709466063287858e-05,
+      "loss": 1.0,
+      "step": 5393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093511599963044e-05,
+      "loss": 0.8889,
+      "step": 5394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7092362378510068e-05,
+      "loss": 0.9433,
+      "step": 5395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70912129685502e-05,
+      "loss": 1.0265,
+      "step": 5396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709006337011399e-05,
+      "loss": 0.9957,
+      "step": 5397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7088913583231997e-05,
+      "loss": 0.9776,
+      "step": 5398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7087763607934772e-05,
+      "loss": 1.0475,
+      "step": 5399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7086613444252886e-05,
+      "loss": 0.9654,
+      "step": 5400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708546309221691e-05,
+      "loss": 0.9671,
+      "step": 5401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708431255185742e-05,
+      "loss": 0.9218,
+      "step": 5402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708316182320499e-05,
+      "loss": 1.0167,
+      "step": 5403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082010906290208e-05,
+      "loss": 0.4514,
+      "step": 5404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708085980114367e-05,
+      "loss": 0.9415,
+      "step": 5405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079708507795957e-05,
+      "loss": 0.931,
+      "step": 5406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707855702627768e-05,
+      "loss": 0.9528,
+      "step": 5407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7077405356619445e-05,
+      "loss": 1.0178,
+      "step": 5408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7076253498851858e-05,
+      "loss": 0.9978,
+      "step": 5409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7075101453005537e-05,
+      "loss": 1.0133,
+      "step": 5410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7073949219111096e-05,
+      "loss": 1.0357,
+      "step": 5411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7072796797199166e-05,
+      "loss": 1.0336,
+      "step": 5412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071644187300372e-05,
+      "loss": 0.9619,
+      "step": 5413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7070491389445357e-05,
+      "loss": 0.9691,
+      "step": 5414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706933840366475e-05,
+      "loss": 0.9833,
+      "step": 5415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068185229989207e-05,
+      "loss": 1.0073,
+      "step": 5416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706703186844937e-05,
+      "loss": 0.9104,
+      "step": 5417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70658783190759e-05,
+      "loss": 0.9684,
+      "step": 5418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7064724581899453e-05,
+      "loss": 0.9533,
+      "step": 5419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063570656950695e-05,
+      "loss": 0.8626,
+      "step": 5420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7062416544260297e-05,
+      "loss": 1.0448,
+      "step": 5421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7061262243858937e-05,
+      "loss": 1.057,
+      "step": 5422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7060107755777286e-05,
+      "loss": 1.042,
+      "step": 5423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705895308004604e-05,
+      "loss": 0.9671,
+      "step": 5424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057798216695872e-05,
+      "loss": 0.9799,
+      "step": 5425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7056643165757497e-05,
+      "loss": 0.9503,
+      "step": 5426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7055487927261608e-05,
+      "loss": 1.028,
+      "step": 5427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7054332501238904e-05,
+      "loss": 1.0262,
+      "step": 5428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7053176887720096e-05,
+      "loss": 0.4554,
+      "step": 5429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7052021086735906e-05,
+      "loss": 0.9557,
+      "step": 5430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705086509831705e-05,
+      "loss": 0.9868,
+      "step": 5431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704970892249425e-05,
+      "loss": 0.9695,
+      "step": 5432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704855255929824e-05,
+      "loss": 1.0155,
+      "step": 5433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7047396008759755e-05,
+      "loss": 0.952,
+      "step": 5434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046239270909532e-05,
+      "loss": 0.4578,
+      "step": 5435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045082345778315e-05,
+      "loss": 1.0411,
+      "step": 5436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043925233396855e-05,
+      "loss": 0.9298,
+      "step": 5437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042767933795906e-05,
+      "loss": 0.9339,
+      "step": 5438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7041610447006232e-05,
+      "loss": 0.9919,
+      "step": 5439
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040452773058595e-05,
+      "loss": 0.9657,
+      "step": 5440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7039294911983764e-05,
+      "loss": 1.0062,
+      "step": 5441
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038136863812515e-05,
+      "loss": 1.0051,
+      "step": 5442
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7036978628575625e-05,
+      "loss": 0.9673,
+      "step": 5443
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703582020630388e-05,
+      "loss": 0.9641,
+      "step": 5444
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034661597028067e-05,
+      "loss": 0.9233,
+      "step": 5445
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7033502800778986e-05,
+      "loss": 0.941,
+      "step": 5446
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703234381758743e-05,
+      "loss": 1.0858,
+      "step": 5447
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7031184647484213e-05,
+      "loss": 0.9023,
+      "step": 5448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7030025290500134e-05,
+      "loss": 1.0247,
+      "step": 5449
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7028865746666014e-05,
+      "loss": 0.9698,
+      "step": 5450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702770601601267e-05,
+      "loss": 1.0224,
+      "step": 5451
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026546098570924e-05,
+      "loss": 0.9899,
+      "step": 5452
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702538599437161e-05,
+      "loss": 0.9398,
+      "step": 5453
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702422570344556e-05,
+      "loss": 0.9992,
+      "step": 5454
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7023065225823616e-05,
+      "loss": 0.3954,
+      "step": 5455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021904561536614e-05,
+      "loss": 1.0005,
+      "step": 5456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7020743710615413e-05,
+      "loss": 0.9872,
+      "step": 5457
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701958267309086e-05,
+      "loss": 0.9656,
+      "step": 5458
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7018421448993816e-05,
+      "loss": 1.018,
+      "step": 5459
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7017260038355147e-05,
+      "loss": 0.9929,
+      "step": 5460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7016098441205722e-05,
+      "loss": 0.9298,
+      "step": 5461
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7014936657576414e-05,
+      "loss": 0.9035,
+      "step": 5462
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70137746874981e-05,
+      "loss": 1.0698,
+      "step": 5463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701261253100166e-05,
+      "loss": 0.9932,
+      "step": 5464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7011450188117996e-05,
+      "loss": 0.4152,
+      "step": 5465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701028765887799e-05,
+      "loss": 0.9511,
+      "step": 5466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009124943312548e-05,
+      "loss": 1.0178,
+      "step": 5467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007962041452566e-05,
+      "loss": 0.963,
+      "step": 5468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700679895332896e-05,
+      "loss": 0.9889,
+      "step": 5469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700563567897264e-05,
+      "loss": 0.9061,
+      "step": 5470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700447221841452e-05,
+      "loss": 0.9963,
+      "step": 5471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7003308571685534e-05,
+      "loss": 0.9834,
+      "step": 5472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7002144738816602e-05,
+      "loss": 1.0744,
+      "step": 5473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700098071983866e-05,
+      "loss": 0.938,
+      "step": 5474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6999816514782647e-05,
+      "loss": 0.9403,
+      "step": 5475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6998652123679505e-05,
+      "loss": 0.4708,
+      "step": 5476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699748754656018e-05,
+      "loss": 0.8645,
+      "step": 5477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6996322783455636e-05,
+      "loss": 0.9929,
+      "step": 5478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6995157834396817e-05,
+      "loss": 1.0288,
+      "step": 5479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993992699414693e-05,
+      "loss": 0.868,
+      "step": 5480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992827378540235e-05,
+      "loss": 1.0184,
+      "step": 5481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6991661871804404e-05,
+      "loss": 0.9308,
+      "step": 5482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699049617923819e-05,
+      "loss": 0.8691,
+      "step": 5483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989330300872576e-05,
+      "loss": 0.9688,
+      "step": 5484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698816423673854e-05,
+      "loss": 1.1106,
+      "step": 5485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698699798686708e-05,
+      "loss": 0.93,
+      "step": 5486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6985831551289192e-05,
+      "loss": 0.8846,
+      "step": 5487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6984664930035883e-05,
+      "loss": 0.8728,
+      "step": 5488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983498123138152e-05,
+      "loss": 0.9424,
+      "step": 5489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982331130627026e-05,
+      "loss": 0.417,
+      "step": 5490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6981163952533504e-05,
+      "loss": 1.0787,
+      "step": 5491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6979996588888624e-05,
+      "loss": 0.9532,
+      "step": 5492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978829039723403e-05,
+      "loss": 0.9962,
+      "step": 5493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697766130506888e-05,
+      "loss": 1.0058,
+      "step": 5494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6976493384956085e-05,
+      "loss": 0.9298,
+      "step": 5495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6975325279416062e-05,
+      "loss": 0.9622,
+      "step": 5496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6974156988479867e-05,
+      "loss": 0.8084,
+      "step": 5497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697298851217854e-05,
+      "loss": 0.9454,
+      "step": 5498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6971819850543144e-05,
+      "loss": 0.9055,
+      "step": 5499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6970651003604738e-05,
+      "loss": 0.9698,
+      "step": 5500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969481971394384e-05,
+      "loss": 0.9789,
+      "step": 5501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6968312753943167e-05,
+      "loss": 0.9857,
+      "step": 5502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6967143351282154e-05,
+      "loss": 0.9029,
+      "step": 5503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965973763442426e-05,
+      "loss": 0.9977,
+      "step": 5504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6964803990455072e-05,
+      "loss": 0.9908,
+      "step": 5505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696363403235118e-05,
+      "loss": 0.9994,
+      "step": 5506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6962463889161857e-05,
+      "loss": 0.9731,
+      "step": 5507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6961293560918188e-05,
+      "loss": 0.9792,
+      "step": 5508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696012304765129e-05,
+      "loss": 0.9431,
+      "step": 5509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6958952349392265e-05,
+      "loss": 0.9551,
+      "step": 5510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695778146617224e-05,
+      "loss": 0.9493,
+      "step": 5511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695661039802233e-05,
+      "loss": 1.0359,
+      "step": 5512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955439144973657e-05,
+      "loss": 0.9289,
+      "step": 5513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6954267707057362e-05,
+      "loss": 0.9915,
+      "step": 5514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6953096084304567e-05,
+      "loss": 1.0265,
+      "step": 5515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6951924276746425e-05,
+      "loss": 0.934,
+      "step": 5516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950752284414073e-05,
+      "loss": 0.9585,
+      "step": 5517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694958010733866e-05,
+      "loss": 0.9931,
+      "step": 5518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694840774555135e-05,
+      "loss": 0.9735,
+      "step": 5519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69472351990833e-05,
+      "loss": 0.9609,
+      "step": 5520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946062467965666e-05,
+      "loss": 1.0233,
+      "step": 5521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694488955222963e-05,
+      "loss": 0.4392,
+      "step": 5522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694371645190636e-05,
+      "loss": 0.9764,
+      "step": 5523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942543167027037e-05,
+      "loss": 1.0129,
+      "step": 5524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6941369697622846e-05,
+      "loss": 0.9335,
+      "step": 5525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6940196043724976e-05,
+      "loss": 0.9847,
+      "step": 5526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693902220536462e-05,
+      "loss": 0.9352,
+      "step": 5527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6937848182572982e-05,
+      "loss": 0.9952,
+      "step": 5528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693667397538126e-05,
+      "loss": 0.8698,
+      "step": 5529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6935499583820665e-05,
+      "loss": 1.03,
+      "step": 5530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6934325007922418e-05,
+      "loss": 0.9689,
+      "step": 5531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693315024771773e-05,
+      "loss": 1.0083,
+      "step": 5532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6931975303237827e-05,
+      "loss": 0.9855,
+      "step": 5533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930800174513934e-05,
+      "loss": 0.9351,
+      "step": 5534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692962486157729e-05,
+      "loss": 0.9513,
+      "step": 5535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928449364459134e-05,
+      "loss": 0.9897,
+      "step": 5536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927273683190706e-05,
+      "loss": 1.0228,
+      "step": 5537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926097817803252e-05,
+      "loss": 0.993,
+      "step": 5538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692492176832803e-05,
+      "loss": 1.0832,
+      "step": 5539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69237455347963e-05,
+      "loss": 0.9231,
+      "step": 5540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922569117239316e-05,
+      "loss": 0.9161,
+      "step": 5541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6921392515688354e-05,
+      "loss": 0.9136,
+      "step": 5542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692021573017468e-05,
+      "loss": 0.9565,
+      "step": 5543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919038760729583e-05,
+      "loss": 1.0021,
+      "step": 5544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917861607384334e-05,
+      "loss": 0.9629,
+      "step": 5545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691668427017022e-05,
+      "loss": 1.0231,
+      "step": 5546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6915506749118545e-05,
+      "loss": 1.0099,
+      "step": 5547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6914329044260596e-05,
+      "loss": 0.9552,
+      "step": 5548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691315115562768e-05,
+      "loss": 1.0339,
+      "step": 5549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69119730832511e-05,
+      "loss": 0.9419,
+      "step": 5550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691079482716217e-05,
+      "loss": 0.9369,
+      "step": 5551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6909616387392204e-05,
+      "loss": 0.9608,
+      "step": 5552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908437763972524e-05,
+      "loss": 1.0065,
+      "step": 5553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6907258956934468e-05,
+      "loss": 1.0263,
+      "step": 5554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690607996630935e-05,
+      "loss": 1.1011,
+      "step": 5555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6904900792128517e-05,
+      "loss": 1.0427,
+      "step": 5556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6903721434423306e-05,
+      "loss": 1.0266,
+      "step": 5557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690254189322506e-05,
+      "loss": 0.987,
+      "step": 5558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690136216856514e-05,
+      "loss": 1.0131,
+      "step": 5559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900182260474895e-05,
+      "loss": 0.5112,
+      "step": 5560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899002168985683e-05,
+      "loss": 0.945,
+      "step": 5561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6897821894128877e-05,
+      "loss": 0.9314,
+      "step": 5562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689664143593584e-05,
+      "loss": 1.0763,
+      "step": 5563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6895460794437953e-05,
+      "loss": 1.0613,
+      "step": 5564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689427996966659e-05,
+      "loss": 0.4664,
+      "step": 5565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689309896165314e-05,
+      "loss": 1.0236,
+      "step": 5566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6891917770428995e-05,
+      "loss": 0.9559,
+      "step": 5567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890736396025548e-05,
+      "loss": 1.0076,
+      "step": 5568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688955483847419e-05,
+      "loss": 0.9633,
+      "step": 5569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888373097806337e-05,
+      "loss": 1.0355,
+      "step": 5570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887191174053397e-05,
+      "loss": 1.0476,
+      "step": 5571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688600906724678e-05,
+      "loss": 0.9693,
+      "step": 5572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884826777417907e-05,
+      "loss": 0.8631,
+      "step": 5573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68836443045982e-05,
+      "loss": 0.9224,
+      "step": 5574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6882461648819092e-05,
+      "loss": 1.0298,
+      "step": 5575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688127881011201e-05,
+      "loss": 0.9653,
+      "step": 5576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68800957885084e-05,
+      "loss": 0.4159,
+      "step": 5577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6878912584039698e-05,
+      "loss": 0.9124,
+      "step": 5578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687772919673736e-05,
+      "loss": 1.0417,
+      "step": 5579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6876545626632833e-05,
+      "loss": 0.9944,
+      "step": 5580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687536187375758e-05,
+      "loss": 1.0003,
+      "step": 5581
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6874177938143056e-05,
+      "loss": 0.4367,
+      "step": 5582
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6872993819820737e-05,
+      "loss": 0.956,
+      "step": 5583
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6871809518822088e-05,
+      "loss": 0.9971,
+      "step": 5584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870625035178595e-05,
+      "loss": 0.9582,
+      "step": 5585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6869440368921735e-05,
+      "loss": 0.9597,
+      "step": 5586
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868255520082993e-05,
+      "loss": 0.9253,
+      "step": 5587
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686707048869386e-05,
+      "loss": 0.9712,
+      "step": 5588
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6865885274785846e-05,
+      "loss": 0.9153,
+      "step": 5589
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6864699878390438e-05,
+      "loss": 0.9534,
+      "step": 5590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6863514299539145e-05,
+      "loss": 0.9092,
+      "step": 5591
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862328538263484e-05,
+      "loss": 0.9458,
+      "step": 5592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686114259459497e-05,
+      "loss": 1.0198,
+      "step": 5593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685995646856512e-05,
+      "loss": 0.8097,
+      "step": 5594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6858770160205457e-05,
+      "loss": 0.9595,
+      "step": 5595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6857583669547523e-05,
+      "loss": 0.9288,
+      "step": 5596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856396996622844e-05,
+      "loss": 1.1147,
+      "step": 5597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6855210141462964e-05,
+      "loss": 0.9966,
+      "step": 5598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685402310409943e-05,
+      "loss": 0.9289,
+      "step": 5599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6852835884563786e-05,
+      "loss": 0.9951,
+      "step": 5600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6851648482887592e-05,
+      "loss": 0.9731,
+      "step": 5601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685046089910241e-05,
+      "loss": 1.0097,
+      "step": 5602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6849273133239798e-05,
+      "loss": 0.9937,
+      "step": 5603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848085185331332e-05,
+      "loss": 0.9952,
+      "step": 5604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684689705540858e-05,
+      "loss": 0.9677,
+      "step": 5605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845708743503124e-05,
+      "loss": 0.8794,
+      "step": 5606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6844520249646548e-05,
+      "loss": 0.8214,
+      "step": 5607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6843331573870446e-05,
+      "loss": 1.0163,
+      "step": 5608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6842142716206402e-05,
+      "loss": 1.0476,
+      "step": 5609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684095367668602e-05,
+      "loss": 0.9716,
+      "step": 5610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68397644553409e-05,
+      "loss": 0.9935,
+      "step": 5611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6838575052202654e-05,
+      "loss": 0.9081,
+      "step": 5612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6837385467302895e-05,
+      "loss": 1.0328,
+      "step": 5613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683619570067324e-05,
+      "loss": 0.9502,
+      "step": 5614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835005752345307e-05,
+      "loss": 0.9973,
+      "step": 5615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683381562235073e-05,
+      "loss": 0.9859,
+      "step": 5616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683262531072114e-05,
+      "loss": 0.8335,
+      "step": 5617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6831434817488173e-05,
+      "loss": 0.929,
+      "step": 5618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683024414268347e-05,
+      "loss": 1.1021,
+      "step": 5619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6829053286338677e-05,
+      "loss": 0.9391,
+      "step": 5620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6827862248485448e-05,
+      "loss": 0.8827,
+      "step": 5621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682667102915544e-05,
+      "loss": 1.0537,
+      "step": 5622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682547962838031e-05,
+      "loss": 1.0695,
+      "step": 5623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6824288046191727e-05,
+      "loss": 0.4477,
+      "step": 5624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6823096282621366e-05,
+      "loss": 1.0113,
+      "step": 5625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6821904337700892e-05,
+      "loss": 1.0451,
+      "step": 5626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6820712211461998e-05,
+      "loss": 0.9588,
+      "step": 5627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819519903936362e-05,
+      "loss": 0.9204,
+      "step": 5628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818327415155674e-05,
+      "loss": 1.0228,
+      "step": 5629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681713474515163e-05,
+      "loss": 0.9856,
+      "step": 5630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6815941893955935e-05,
+      "loss": 0.9598,
+      "step": 5631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6814748861600282e-05,
+      "loss": 0.9827,
+      "step": 5632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813555648116392e-05,
+      "loss": 1.0248,
+      "step": 5633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6812362253535973e-05,
+      "loss": 0.9617,
+      "step": 5634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681116867789075e-05,
+      "loss": 0.9684,
+      "step": 5635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680997492121244e-05,
+      "loss": 0.9744,
+      "step": 5636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680878098353277e-05,
+      "loss": 0.9429,
+      "step": 5637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807586864883485e-05,
+      "loss": 0.4115,
+      "step": 5638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680639256529631e-05,
+      "loss": 0.9678,
+      "step": 5639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805198084803e-05,
+      "loss": 0.9857,
+      "step": 5640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6804003423435293e-05,
+      "loss": 0.9355,
+      "step": 5641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6802808581224948e-05,
+      "loss": 1.0109,
+      "step": 5642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680161355820372e-05,
+      "loss": 1.0796,
+      "step": 5643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680041835440337e-05,
+      "loss": 0.8954,
+      "step": 5644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679922296985567e-05,
+      "loss": 0.9758,
+      "step": 5645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798027404592387e-05,
+      "loss": 1.026,
+      "step": 5646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6796831658645298e-05,
+      "loss": 0.8771,
+      "step": 5647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679563573204619e-05,
+      "loss": 0.9789,
+      "step": 5648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794439624826843e-05,
+      "loss": 0.492,
+      "step": 5649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6793243337019053e-05,
+      "loss": 0.9579,
+      "step": 5650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679204686865461e-05,
+      "loss": 0.9249,
+      "step": 5651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6790850219765323e-05,
+      "loss": 0.9224,
+      "step": 5652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678965339038299e-05,
+      "loss": 0.9555,
+      "step": 5653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678845638053942e-05,
+      "loss": 0.9962,
+      "step": 5654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678725919026644e-05,
+      "loss": 0.882,
+      "step": 5655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678606181959586e-05,
+      "loss": 0.9483,
+      "step": 5656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678486426855951e-05,
+      "loss": 0.9886,
+      "step": 5657
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783666537189214e-05,
+      "loss": 0.9301,
+      "step": 5658
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678246862551681e-05,
+      "loss": 0.9298,
+      "step": 5659
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6781270533574137e-05,
+      "loss": 0.9535,
+      "step": 5660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6780072261393035e-05,
+      "loss": 0.985,
+      "step": 5661
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6778873809005358e-05,
+      "loss": 1.0201,
+      "step": 5662
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677767517644296e-05,
+      "loss": 1.0443,
+      "step": 5663
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677647636373769e-05,
+      "loss": 0.9484,
+      "step": 5664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775277370921422e-05,
+      "loss": 0.9231,
+      "step": 5665
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677407819802602e-05,
+      "loss": 1.0154,
+      "step": 5666
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6772878845083355e-05,
+      "loss": 0.8882,
+      "step": 5667
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6771679312125305e-05,
+      "loss": 1.0086,
+      "step": 5668
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770479599183755e-05,
+      "loss": 0.9782,
+      "step": 5669
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6769279706290588e-05,
+      "loss": 0.9627,
+      "step": 5670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67680796334777e-05,
+      "loss": 0.9513,
+      "step": 5671
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6766879380776983e-05,
+      "loss": 0.9995,
+      "step": 5672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6765678948220338e-05,
+      "loss": 0.9503,
+      "step": 5673
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676447833583968e-05,
+      "loss": 1.0195,
+      "step": 5674
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676327754366691e-05,
+      "loss": 0.968,
+      "step": 5675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676207657173395e-05,
+      "loss": 0.9057,
+      "step": 5676
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760875420072714e-05,
+      "loss": 0.3771,
+      "step": 5677
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675967408871513e-05,
+      "loss": 1.0049,
+      "step": 5678
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6758472577693132e-05,
+      "loss": 1.0128,
+      "step": 5679
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6757270887038653e-05,
+      "loss": 0.9612,
+      "step": 5680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756069016783628e-05,
+      "loss": 0.9625,
+      "step": 5681
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6754866966960007e-05,
+      "loss": 0.9759,
+      "step": 5682
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6753664737599736e-05,
+      "loss": 0.9789,
+      "step": 5683
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675246232873477e-05,
+      "loss": 1.0544,
+      "step": 5684
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6751259740397067e-05,
+      "loss": 0.8261,
+      "step": 5685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750056972618593e-05,
+      "loss": 0.9619,
+      "step": 5686
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6748854025431313e-05,
+      "loss": 0.9505,
+      "step": 5687
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6747650898867202e-05,
+      "loss": 0.9303,
+      "step": 5688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674644759295823e-05,
+      "loss": 0.9412,
+      "step": 5689
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674524410773639e-05,
+      "loss": 0.9513,
+      "step": 5690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674404044323367e-05,
+      "loss": 1.0245,
+      "step": 5691
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6742836599482053e-05,
+      "loss": 0.9597,
+      "step": 5692
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674163257651354e-05,
+      "loss": 0.9667,
+      "step": 5693
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6740428374360133e-05,
+      "loss": 0.9369,
+      "step": 5694
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673922399305384e-05,
+      "loss": 0.9448,
+      "step": 5695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673801943262667e-05,
+      "loss": 0.4531,
+      "step": 5696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736814693110635e-05,
+      "loss": 0.945,
+      "step": 5697
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6735609774537763e-05,
+      "loss": 1.0303,
+      "step": 5698
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673440467694008e-05,
+      "loss": 1.0491,
+      "step": 5699
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6733199400349606e-05,
+      "loss": 1.008,
+      "step": 5700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673199394479838e-05,
+      "loss": 0.9559,
+      "step": 5701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673078831031845e-05,
+      "loss": 0.9363,
+      "step": 5702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672958249694185e-05,
+      "loss": 0.9088,
+      "step": 5703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6728376504700636e-05,
+      "loss": 1.0136,
+      "step": 5704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672717033362686e-05,
+      "loss": 1.012,
+      "step": 5705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725963983752577e-05,
+      "loss": 0.9951,
+      "step": 5706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672475745510985e-05,
+      "loss": 0.9943,
+      "step": 5707
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6723550747730755e-05,
+      "loss": 1.0317,
+      "step": 5708
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722343861647357e-05,
+      "loss": 0.9246,
+      "step": 5709
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672113679689174e-05,
+      "loss": 0.989,
+      "step": 5710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6719929553495985e-05,
+      "loss": 0.9452,
+      "step": 5711
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671872213149217e-05,
+      "loss": 0.9886,
+      "step": 5712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6717514530912403e-05,
+      "loss": 0.8751,
+      "step": 5713
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6716306751788766e-05,
+      "loss": 0.934,
+      "step": 5714
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6715098794153365e-05,
+      "loss": 0.9998,
+      "step": 5715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6713890658038313e-05,
+      "loss": 0.9548,
+      "step": 5716
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712682343475713e-05,
+      "loss": 0.9168,
+      "step": 5717
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671147385049768e-05,
+      "loss": 0.907,
+      "step": 5718
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671026517913634e-05,
+      "loss": 0.9466,
+      "step": 5719
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6709056329423817e-05,
+      "loss": 0.9618,
+      "step": 5720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707847301392237e-05,
+      "loss": 1.061,
+      "step": 5721
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706638095073733e-05,
+      "loss": 0.9473,
+      "step": 5722
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6705428710500456e-05,
+      "loss": 0.9212,
+      "step": 5723
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6704219147704535e-05,
+      "loss": 0.9646,
+      "step": 5724
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670300940671813e-05,
+      "loss": 0.9754,
+      "step": 5725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670179948757339e-05,
+      "loss": 0.9122,
+      "step": 5726
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700589390302476e-05,
+      "loss": 0.9645,
+      "step": 5727
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6699379114937547e-05,
+      "loss": 0.9605,
+      "step": 5728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669816866151077e-05,
+      "loss": 0.9959,
+      "step": 5729
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6696958030054323e-05,
+      "loss": 0.9427,
+      "step": 5730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669574722060038e-05,
+      "loss": 0.9699,
+      "step": 5731
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6694536233181122e-05,
+      "loss": 0.8878,
+      "step": 5732
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693325067828738e-05,
+      "loss": 1.09,
+      "step": 5733
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669211372457542e-05,
+      "loss": 0.9353,
+      "step": 5734
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669090220345336e-05,
+      "loss": 0.9708,
+      "step": 5735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668969050449476e-05,
+      "loss": 0.9687,
+      "step": 5736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668847862773183e-05,
+      "loss": 0.9747,
+      "step": 5737
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6687266573196773e-05,
+      "loss": 1.118,
+      "step": 5738
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668605434092181e-05,
+      "loss": 1.0184,
+      "step": 5739
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668484193093916e-05,
+      "loss": 0.9691,
+      "step": 5740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683629343281044e-05,
+      "loss": 0.9526,
+      "step": 5741
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6682416577979693e-05,
+      "loss": 0.9876,
+      "step": 5742
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6681203635067344e-05,
+      "loss": 0.9469,
+      "step": 5743
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667999051457623e-05,
+      "loss": 0.9417,
+      "step": 5744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6678777216538603e-05,
+      "loss": 1.0005,
+      "step": 5745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66775637409867e-05,
+      "loss": 0.9832,
+      "step": 5746
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676350087952782e-05,
+      "loss": 0.9061,
+      "step": 5747
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66751362574691e-05,
+      "loss": 1.0373,
+      "step": 5748
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673922249567924e-05,
+      "loss": 0.9697,
+      "step": 5749
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667270806428152e-05,
+      "loss": 0.9214,
+      "step": 5750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667149370164215e-05,
+      "loss": 1.0329,
+      "step": 5751
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66702791616821e-05,
+      "loss": 0.9814,
+      "step": 5752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669064444433645e-05,
+      "loss": 1.0095,
+      "step": 5753
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6667849549929077e-05,
+      "loss": 1.1707,
+      "step": 5754
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6666634478200684e-05,
+      "loss": 0.8538,
+      "step": 5755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666541922928076e-05,
+      "loss": 0.9746,
+      "step": 5756
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6664203803201605e-05,
+      "loss": 0.9814,
+      "step": 5757
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666298819999552e-05,
+      "loss": 0.9623,
+      "step": 5758
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6661772419694827e-05,
+      "loss": 0.9986,
+      "step": 5759
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6660556462331823e-05,
+      "loss": 0.8952,
+      "step": 5760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6659340327938842e-05,
+      "loss": 0.9345,
+      "step": 5761
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66581240165482e-05,
+      "loss": 0.9084,
+      "step": 5762
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6656907528192225e-05,
+      "loss": 0.8711,
+      "step": 5763
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6655690862903248e-05,
+      "loss": 0.954,
+      "step": 5764
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665447402071361e-05,
+      "loss": 0.9249,
+      "step": 5765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6653257001655652e-05,
+      "loss": 0.9739,
+      "step": 5766
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6652039805761727e-05,
+      "loss": 0.9077,
+      "step": 5767
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6650822433064178e-05,
+      "loss": 0.9359,
+      "step": 5768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649604883595367e-05,
+      "loss": 0.967,
+      "step": 5769
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664838715738765e-05,
+      "loss": 1.0085,
+      "step": 5770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6647169254473394e-05,
+      "loss": 0.9155,
+      "step": 5771
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6645951174884974e-05,
+      "loss": 1.0346,
+      "step": 5772
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664473291865476e-05,
+      "loss": 0.9524,
+      "step": 5773
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643514485815136e-05,
+      "loss": 1.0185,
+      "step": 5774
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6642295876398483e-05,
+      "loss": 0.9468,
+      "step": 5775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641077090437196e-05,
+      "loss": 0.9114,
+      "step": 5776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6639858127963657e-05,
+      "loss": 0.9909,
+      "step": 5777
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663863898901028e-05,
+      "loss": 1.0447,
+      "step": 5778
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6637419673609456e-05,
+      "loss": 0.9726,
+      "step": 5779
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66362001817936e-05,
+      "loss": 0.9526,
+      "step": 5780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634980513595122e-05,
+      "loss": 0.916,
+      "step": 5781
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663376066904644e-05,
+      "loss": 0.932,
+      "step": 5782
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6632540648179976e-05,
+      "loss": 0.9283,
+      "step": 5783
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631320451028154e-05,
+      "loss": 0.8648,
+      "step": 5784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663010007762341e-05,
+      "loss": 0.9976,
+      "step": 5785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662887952799818e-05,
+      "loss": 0.9632,
+      "step": 5786
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66276588021849e-05,
+      "loss": 1.0181,
+      "step": 5787
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662643790021602e-05,
+      "loss": 0.9119,
+      "step": 5788
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662521682212399e-05,
+      "loss": 1.0019,
+      "step": 5789
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662399556794126e-05,
+      "loss": 0.9136,
+      "step": 5790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6622774137700298e-05,
+      "loss": 0.9609,
+      "step": 5791
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662155253143356e-05,
+      "loss": 0.417,
+      "step": 5792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6620330749173523e-05,
+      "loss": 0.9188,
+      "step": 5793
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661910879095265e-05,
+      "loss": 0.967,
+      "step": 5794
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6617886656803426e-05,
+      "loss": 1.0292,
+      "step": 5795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6616664346758332e-05,
+      "loss": 0.9725,
+      "step": 5796
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615441860849863e-05,
+      "loss": 0.9943,
+      "step": 5797
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66142191991105e-05,
+      "loss": 1.0014,
+      "step": 5798
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6612996361572745e-05,
+      "loss": 0.9327,
+      "step": 5799
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6611773348269098e-05,
+      "loss": 1.0049,
+      "step": 5800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610550159232067e-05,
+      "loss": 1.0104,
+      "step": 5801
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6609326794494168e-05,
+      "loss": 0.9714,
+      "step": 5802
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6608103254087905e-05,
+      "loss": 1.017,
+      "step": 5803
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660687953804581e-05,
+      "loss": 0.9729,
+      "step": 5804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66056556464004e-05,
+      "loss": 0.9273,
+      "step": 5805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604431579184207e-05,
+      "loss": 0.4186,
+      "step": 5806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6603207336429767e-05,
+      "loss": 1.0102,
+      "step": 5807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6601982918169615e-05,
+      "loss": 0.9723,
+      "step": 5808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600758324436296e-05,
+      "loss": 0.9039,
+      "step": 5809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599533555262365e-05,
+      "loss": 0.9562,
+      "step": 5810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659830861068037e-05,
+      "loss": 0.932,
+      "step": 5811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6597083490722864e-05,
+      "loss": 0.9883,
+      "step": 5812
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595858195422414e-05,
+      "loss": 0.9811,
+      "step": 5813
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6594632724811585e-05,
+      "loss": 0.9654,
+      "step": 5814
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659340707892295e-05,
+      "loss": 0.9666,
+      "step": 5815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659218125778909e-05,
+      "loss": 0.9641,
+      "step": 5816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659095526144258e-05,
+      "loss": 1.0468,
+      "step": 5817
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589729089916004e-05,
+      "loss": 0.9432,
+      "step": 5818
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6588502743241952e-05,
+      "loss": 0.9338,
+      "step": 5819
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587276221453028e-05,
+      "loss": 0.9179,
+      "step": 5820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6586049524581823e-05,
+      "loss": 0.9659,
+      "step": 5821
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6584822652660944e-05,
+      "loss": 0.4245,
+      "step": 5822
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6583595605722996e-05,
+      "loss": 0.964,
+      "step": 5823
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65823683838006e-05,
+      "loss": 0.9732,
+      "step": 5824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6581140986926373e-05,
+      "loss": 0.9168,
+      "step": 5825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657991341513293e-05,
+      "loss": 0.9429,
+      "step": 5826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6578685668452903e-05,
+      "loss": 0.8748,
+      "step": 5827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657745774691893e-05,
+      "loss": 0.9827,
+      "step": 5828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6576229650563636e-05,
+      "loss": 0.967,
+      "step": 5829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657500137941967e-05,
+      "loss": 0.9646,
+      "step": 5830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573772933519677e-05,
+      "loss": 0.9411,
+      "step": 5831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6572544312896307e-05,
+      "loss": 1.0021,
+      "step": 5832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657131551758222e-05,
+      "loss": 0.9469,
+      "step": 5833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6570086547610067e-05,
+      "loss": 0.9855,
+      "step": 5834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656885740301252e-05,
+      "loss": 0.9183,
+      "step": 5835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567628083822245e-05,
+      "loss": 0.959,
+      "step": 5836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6566398590071915e-05,
+      "loss": 1.0212,
+      "step": 5837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656516892179421e-05,
+      "loss": 0.9531,
+      "step": 5838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6563939079021817e-05,
+      "loss": 0.9124,
+      "step": 5839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6562709061787416e-05,
+      "loss": 0.4572,
+      "step": 5840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6561478870123706e-05,
+      "loss": 0.9442,
+      "step": 5841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656024850406338e-05,
+      "loss": 0.9586,
+      "step": 5842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655901796363914e-05,
+      "loss": 0.9644,
+      "step": 5843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6557787248883698e-05,
+      "loss": 0.9846,
+      "step": 5844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6556556359829757e-05,
+      "loss": 0.9947,
+      "step": 5845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6555325296510035e-05,
+      "loss": 0.9671,
+      "step": 5846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655409405895726e-05,
+      "loss": 0.9276,
+      "step": 5847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655286264720414e-05,
+      "loss": 0.8932,
+      "step": 5848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551631061283424e-05,
+      "loss": 1.0186,
+      "step": 5849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550399301227834e-05,
+      "loss": 0.9459,
+      "step": 5850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654916736707011e-05,
+      "loss": 0.9407,
+      "step": 5851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6547935258843e-05,
+      "loss": 0.9688,
+      "step": 5852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654670297657925e-05,
+      "loss": 1.089,
+      "step": 5853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545470520311608e-05,
+      "loss": 0.9537,
+      "step": 5854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654423789007284e-05,
+      "loss": 0.9397,
+      "step": 5855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65430050858957e-05,
+      "loss": 0.9327,
+      "step": 5856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654177210781296e-05,
+      "loss": 0.9685,
+      "step": 5857
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6540538955857385e-05,
+      "loss": 1.0072,
+      "step": 5858
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.653930563006176e-05,
+      "loss": 0.9632,
+      "step": 5859
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6538072130458853e-05,
+      "loss": 0.9283,
+      "step": 5860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653683845708146e-05,
+      "loss": 0.9634,
+      "step": 5861
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535604609962366e-05,
+      "loss": 0.939,
+      "step": 5862
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6534370589134366e-05,
+      "loss": 1.0535,
+      "step": 5863
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653313639463026e-05,
+      "loss": 1.0049,
+      "step": 5864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653190202648285e-05,
+      "loss": 0.9771,
+      "step": 5865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530667484724942e-05,
+      "loss": 0.9675,
+      "step": 5866
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6529432769389353e-05,
+      "loss": 1.0447,
+      "step": 5867
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6528197880508896e-05,
+      "loss": 0.8708,
+      "step": 5868
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6526962818116397e-05,
+      "loss": 0.9634,
+      "step": 5869
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652572758224468e-05,
+      "loss": 0.9585,
+      "step": 5870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652449217292658e-05,
+      "loss": 0.9819,
+      "step": 5871
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523256590194924e-05,
+      "loss": 0.862,
+      "step": 5872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6522020834082562e-05,
+      "loss": 0.9961,
+      "step": 5873
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6520784904622337e-05,
+      "loss": 0.9908,
+      "step": 5874
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651954880184709e-05,
+      "loss": 0.8863,
+      "step": 5875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6518312525789688e-05,
+      "loss": 0.9966,
+      "step": 5876
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651707607648298e-05,
+      "loss": 1.0103,
+      "step": 5877
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6515839453959836e-05,
+      "loss": 0.9727,
+      "step": 5878
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651460265825312e-05,
+      "loss": 0.9221,
+      "step": 5879
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513365689395706e-05,
+      "loss": 1.0578,
+      "step": 5880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651212854742047e-05,
+      "loss": 0.9319,
+      "step": 5881
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6510891232360297e-05,
+      "loss": 0.9602,
+      "step": 5882
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6509653744248068e-05,
+      "loss": 1.0021,
+      "step": 5883
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650841608311668e-05,
+      "loss": 0.9749,
+      "step": 5884
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6507178248999026e-05,
+      "loss": 0.9899,
+      "step": 5885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6505940241928002e-05,
+      "loss": 0.9549,
+      "step": 5886
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650470206193652e-05,
+      "loss": 0.4621,
+      "step": 5887
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6503463709057488e-05,
+      "loss": 1.016,
+      "step": 5888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502225183323816e-05,
+      "loss": 0.8396,
+      "step": 5889
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6500986484768423e-05,
+      "loss": 0.9791,
+      "step": 5890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6499747613424238e-05,
+      "loss": 1.0075,
+      "step": 5891
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6498508569324183e-05,
+      "loss": 0.9259,
+      "step": 5892
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649726935250119e-05,
+      "loss": 0.9927,
+      "step": 5893
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496029962988202e-05,
+      "loss": 1.0194,
+      "step": 5894
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649479040081815e-05,
+      "loss": 1.0395,
+      "step": 5895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6493550666023996e-05,
+      "loss": 0.9803,
+      "step": 5896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6492310758638678e-05,
+      "loss": 1.0046,
+      "step": 5897
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6491070678695154e-05,
+      "loss": 0.4698,
+      "step": 5898
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6489830426226387e-05,
+      "loss": 0.9798,
+      "step": 5899
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488590001265338e-05,
+      "loss": 0.9332,
+      "step": 5900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648734940384498e-05,
+      "loss": 0.9288,
+      "step": 5901
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648610863399828e-05,
+      "loss": 0.8997,
+      "step": 5902
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648486769175822e-05,
+      "loss": 0.9681,
+      "step": 5903
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6483626577157785e-05,
+      "loss": 1.0219,
+      "step": 5904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648238529022996e-05,
+      "loss": 0.9597,
+      "step": 5905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648114383100774e-05,
+      "loss": 1.0275,
+      "step": 5906
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6479902199524116e-05,
+      "loss": 0.994,
+      "step": 5907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478660395812092e-05,
+      "loss": 0.9292,
+      "step": 5908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477418419904676e-05,
+      "loss": 0.9264,
+      "step": 5909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476176271834876e-05,
+      "loss": 1.0496,
+      "step": 5910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474933951635706e-05,
+      "loss": 0.9398,
+      "step": 5911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647369145934019e-05,
+      "loss": 0.9762,
+      "step": 5912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6472448794981346e-05,
+      "loss": 0.9415,
+      "step": 5913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6471205958592204e-05,
+      "loss": 0.9166,
+      "step": 5914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64699629502058e-05,
+      "loss": 0.9537,
+      "step": 5915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6468719769855173e-05,
+      "loss": 1.0089,
+      "step": 5916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6467476417573357e-05,
+      "loss": 0.9882,
+      "step": 5917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646623289339341e-05,
+      "loss": 0.8891,
+      "step": 5918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6464989197348372e-05,
+      "loss": 0.9351,
+      "step": 5919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646374532947131e-05,
+      "loss": 0.9399,
+      "step": 5920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6462501289795277e-05,
+      "loss": 0.9747,
+      "step": 5921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646125707835334e-05,
+      "loss": 1.0032,
+      "step": 5922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6460012695178572e-05,
+      "loss": 0.5455,
+      "step": 5923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458768140304043e-05,
+      "loss": 0.9589,
+      "step": 5924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457523413762834e-05,
+      "loss": 0.9032,
+      "step": 5925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456278515588023e-05,
+      "loss": 0.9926,
+      "step": 5926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645503344581271e-05,
+      "loss": 1.0079,
+      "step": 5927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6453788204469974e-05,
+      "loss": 0.9457,
+      "step": 5928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6452542791592923e-05,
+      "loss": 1.0094,
+      "step": 5929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6451297207214655e-05,
+      "loss": 0.9663,
+      "step": 5930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450051451368272e-05,
+      "loss": 0.8682,
+      "step": 5931
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644880552408689e-05,
+      "loss": 0.9106,
+      "step": 5932
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6447559425403622e-05,
+      "loss": 1.0086,
+      "step": 5933
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644631315535159e-05,
+      "loss": 0.9939,
+      "step": 5934
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6445066713963912e-05,
+      "loss": 0.8796,
+      "step": 5935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443820101273724e-05,
+      "loss": 1.0348,
+      "step": 5936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442573317314158e-05,
+      "loss": 0.5126,
+      "step": 5937
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441326362118354e-05,
+      "loss": 0.9989,
+      "step": 5938
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440079235719445e-05,
+      "loss": 0.9819,
+      "step": 5939
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438831938150593e-05,
+      "loss": 1.0053,
+      "step": 5940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437584469444937e-05,
+      "loss": 1.0125,
+      "step": 5941
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6436336829635638e-05,
+      "loss": 1.0045,
+      "step": 5942
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6435089018755862e-05,
+      "loss": 0.9666,
+      "step": 5943
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6433841036838765e-05,
+      "loss": 0.9454,
+      "step": 5944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643259288391752e-05,
+      "loss": 0.9367,
+      "step": 5945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6431344560025306e-05,
+      "loss": 1.0414,
+      "step": 5946
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64300960651953e-05,
+      "loss": 0.9851,
+      "step": 5947
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642884739946068e-05,
+      "loss": 0.9506,
+      "step": 5948
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427598562854643e-05,
+      "loss": 0.9833,
+      "step": 5949
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6426349555410372e-05,
+      "loss": 1.0211,
+      "step": 5950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6425100377161077e-05,
+      "loss": 0.4547,
+      "step": 5951
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642385102813995e-05,
+      "loss": 0.987,
+      "step": 5952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422601508380193e-05,
+      "loss": 1.0151,
+      "step": 5953
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6421351817915025e-05,
+      "loss": 0.9217,
+      "step": 5954
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6420101956777665e-05,
+      "loss": 1.0379,
+      "step": 5955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6418851925001326e-05,
+      "loss": 1.0032,
+      "step": 5956
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6417601722619233e-05,
+      "loss": 0.9346,
+      "step": 5957
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6416351349664613e-05,
+      "loss": 0.9691,
+      "step": 5958
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6415100806170704e-05,
+      "loss": 1.0046,
+      "step": 5959
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413850092170746e-05,
+      "loss": 0.9918,
+      "step": 5960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641259920769797e-05,
+      "loss": 0.9858,
+      "step": 5961
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641134815278564e-05,
+      "loss": 0.899,
+      "step": 5962
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6410096927466992e-05,
+      "loss": 0.912,
+      "step": 5963
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640884553177529e-05,
+      "loss": 0.9967,
+      "step": 5964
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6407593965743798e-05,
+      "loss": 1.0444,
+      "step": 5965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640634222940577e-05,
+      "loss": 0.918,
+      "step": 5966
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405090322794484e-05,
+      "loss": 0.442,
+      "step": 5967
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6403838245943217e-05,
+      "loss": 0.9356,
+      "step": 5968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640258599888524e-05,
+      "loss": 0.9768,
+      "step": 5969
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6401333581653837e-05,
+      "loss": 0.9049,
+      "step": 5970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64000809942823e-05,
+      "loss": 1.0403,
+      "step": 5971
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6398828236803923e-05,
+      "loss": 0.9984,
+      "step": 5972
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397575309252e-05,
+      "loss": 0.9834,
+      "step": 5973
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6396322211659826e-05,
+      "loss": 0.9657,
+      "step": 5974
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6395068944060718e-05,
+      "loss": 0.9375,
+      "step": 5975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639381550648798e-05,
+      "loss": 1.059,
+      "step": 5976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6392561898974925e-05,
+      "loss": 0.9494,
+      "step": 5977
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639130812155488e-05,
+      "loss": 0.9423,
+      "step": 5978
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6390054174261164e-05,
+      "loss": 1.07,
+      "step": 5979
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388800057127105e-05,
+      "loss": 0.9749,
+      "step": 5980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6387545770186037e-05,
+      "loss": 0.9391,
+      "step": 5981
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63862913134713e-05,
+      "loss": 0.9139,
+      "step": 5982
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638503668701623e-05,
+      "loss": 0.92,
+      "step": 5983
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383781890854182e-05,
+      "loss": 0.9907,
+      "step": 5984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382526925018497e-05,
+      "loss": 0.985,
+      "step": 5985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638127178954254e-05,
+      "loss": 0.9568,
+      "step": 5986
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6380016484459667e-05,
+      "loss": 0.9725,
+      "step": 5987
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637876100980324e-05,
+      "loss": 0.925,
+      "step": 5988
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637750536560663e-05,
+      "loss": 0.9939,
+      "step": 5989
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6376249551903214e-05,
+      "loss": 1.0974,
+      "step": 5990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374993568726367e-05,
+      "loss": 0.9474,
+      "step": 5991
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6373737416109472e-05,
+      "loss": 1.0074,
+      "step": 5992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372481094085913e-05,
+      "loss": 0.9536,
+      "step": 5993
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371224602689088e-05,
+      "loss": 0.4517,
+      "step": 5994
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6369967941952388e-05,
+      "loss": 1.0375,
+      "step": 5995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6368711111909215e-05,
+      "loss": 1.0221,
+      "step": 5996
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367454112592976e-05,
+      "loss": 0.9979,
+      "step": 5997
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636619694403708e-05,
+      "loss": 0.9575,
+      "step": 5998
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364939606274938e-05,
+      "loss": 0.8809,
+      "step": 5999
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636368209933997e-05,
+      "loss": 0.9686,
+      "step": 6000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63624244232656e-05,
+      "loss": 0.937,
+      "step": 6001
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6361166578085254e-05,
+      "loss": 0.8996,
+      "step": 6002
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6359908563832368e-05,
+      "loss": 0.9639,
+      "step": 6003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358650380540376e-05,
+      "loss": 0.9709,
+      "step": 6004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357392028242723e-05,
+      "loss": 0.9472,
+      "step": 6005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6356133506972845e-05,
+      "loss": 0.8608,
+      "step": 6006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354874816764198e-05,
+      "loss": 0.9353,
+      "step": 6007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635361595765024e-05,
+      "loss": 0.9651,
+      "step": 6008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6352356929664422e-05,
+      "loss": 1.0423,
+      "step": 6009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6351097732840212e-05,
+      "loss": 0.9711,
+      "step": 6010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634983836721108e-05,
+      "loss": 0.8546,
+      "step": 6011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63485788328105e-05,
+      "loss": 0.9339,
+      "step": 6012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634731912967194e-05,
+      "loss": 0.9344,
+      "step": 6013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6346059257828888e-05,
+      "loss": 0.9966,
+      "step": 6014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634479921731483e-05,
+      "loss": 0.9978,
+      "step": 6015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6343539008163257e-05,
+      "loss": 0.904,
+      "step": 6016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6342278630407657e-05,
+      "loss": 1.0328,
+      "step": 6017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634101808408154e-05,
+      "loss": 0.9698,
+      "step": 6018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63397573692184e-05,
+      "loss": 1.0163,
+      "step": 6019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633849648585175e-05,
+      "loss": 0.954,
+      "step": 6020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6337235434015107e-05,
+      "loss": 0.9815,
+      "step": 6021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633597421374198e-05,
+      "loss": 0.8882,
+      "step": 6022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6334712825065897e-05,
+      "loss": 0.9644,
+      "step": 6023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6333451268020382e-05,
+      "loss": 0.9452,
+      "step": 6024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633218954263896e-05,
+      "loss": 0.9465,
+      "step": 6025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633092764895518e-05,
+      "loss": 0.9194,
+      "step": 6026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632966558700257e-05,
+      "loss": 0.915,
+      "step": 6027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6328403356814678e-05,
+      "loss": 0.9274,
+      "step": 6028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6327140958425055e-05,
+      "loss": 0.976,
+      "step": 6029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6325878391867252e-05,
+      "loss": 1.0601,
+      "step": 6030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632461565717482e-05,
+      "loss": 1.0874,
+      "step": 6031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632335275438133e-05,
+      "loss": 0.9924,
+      "step": 6032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632208968352035e-05,
+      "loss": 0.9145,
+      "step": 6033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6320826444625445e-05,
+      "loss": 0.9353,
+      "step": 6034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631956303773019e-05,
+      "loss": 1.0074,
+      "step": 6035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6318299462868167e-05,
+      "loss": 1.0576,
+      "step": 6036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631703572007296e-05,
+      "loss": 0.9159,
+      "step": 6037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631577180937816e-05,
+      "loss": 0.9319,
+      "step": 6038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6314507730817357e-05,
+      "loss": 0.9369,
+      "step": 6039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6313243484424152e-05,
+      "loss": 0.9634,
+      "step": 6040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631197907023215e-05,
+      "loss": 0.9758,
+      "step": 6041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6310714488274942e-05,
+      "loss": 0.913,
+      "step": 6042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630944973858616e-05,
+      "loss": 0.9864,
+      "step": 6043
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6308184821199405e-05,
+      "loss": 0.9675,
+      "step": 6044
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6306919736148305e-05,
+      "loss": 1.0184,
+      "step": 6045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6305654483466484e-05,
+      "loss": 0.9817,
+      "step": 6046
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304389063187563e-05,
+      "loss": 1.0303,
+      "step": 6047
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6303123475345182e-05,
+      "loss": 0.9641,
+      "step": 6048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6301857719972977e-05,
+      "loss": 0.998,
+      "step": 6049
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6300591797104597e-05,
+      "loss": 1.0205,
+      "step": 6050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629932570677368e-05,
+      "loss": 0.9425,
+      "step": 6051
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6298059449013877e-05,
+      "loss": 0.9985,
+      "step": 6052
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6296793023858852e-05,
+      "loss": 0.9987,
+      "step": 6053
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6295526431342256e-05,
+      "loss": 0.9786,
+      "step": 6054
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6294259671497758e-05,
+      "loss": 1.0043,
+      "step": 6055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6292992744359028e-05,
+      "loss": 0.8915,
+      "step": 6056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6291725649959738e-05,
+      "loss": 0.9861,
+      "step": 6057
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629045838833357e-05,
+      "loss": 1.0097,
+      "step": 6058
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6289190959514195e-05,
+      "loss": 0.9182,
+      "step": 6059
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628792336353531e-05,
+      "loss": 1.0316,
+      "step": 6060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6286655600430602e-05,
+      "loss": 0.972,
+      "step": 6061
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6285387670233772e-05,
+      "loss": 1.0412,
+      "step": 6062
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6284119572978518e-05,
+      "loss": 0.9362,
+      "step": 6063
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6282851308698537e-05,
+      "loss": 0.9441,
+      "step": 6064
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6281582877427546e-05,
+      "loss": 1.0264,
+      "step": 6065
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6280314279199256e-05,
+      "loss": 0.9561,
+      "step": 6066
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279045514047386e-05,
+      "loss": 0.9791,
+      "step": 6067
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627777658200566e-05,
+      "loss": 0.9523,
+      "step": 6068
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6276507483107798e-05,
+      "loss": 0.9741,
+      "step": 6069
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627523821738754e-05,
+      "loss": 0.9224,
+      "step": 6070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273968784878614e-05,
+      "loss": 1.049,
+      "step": 6071
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6272699185614763e-05,
+      "loss": 0.9941,
+      "step": 6072
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627142941962973e-05,
+      "loss": 0.9964,
+      "step": 6073
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627015948695727e-05,
+      "loss": 0.9497,
+      "step": 6074
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6268889387631133e-05,
+      "loss": 0.907,
+      "step": 6075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6267619121685074e-05,
+      "loss": 0.472,
+      "step": 6076
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6266348689152855e-05,
+      "loss": 0.9412,
+      "step": 6077
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6265078090068243e-05,
+      "loss": 0.9576,
+      "step": 6078
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263807324465018e-05,
+      "loss": 0.9471,
+      "step": 6079
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626253639237694e-05,
+      "loss": 0.9612,
+      "step": 6080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261265293837804e-05,
+      "loss": 0.9637,
+      "step": 6081
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259994028881388e-05,
+      "loss": 1.0318,
+      "step": 6082
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258722597541474e-05,
+      "loss": 0.9506,
+      "step": 6083
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6257450999851865e-05,
+      "loss": 0.9669,
+      "step": 6084
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6256179235846357e-05,
+      "loss": 0.9496,
+      "step": 6085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625490730555875e-05,
+      "loss": 0.9862,
+      "step": 6086
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253635209022844e-05,
+      "loss": 0.903,
+      "step": 6087
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6252362946272466e-05,
+      "loss": 1.0374,
+      "step": 6088
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6251090517341412e-05,
+      "loss": 1.0368,
+      "step": 6089
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6249817922263518e-05,
+      "loss": 0.9123,
+      "step": 6090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62485451610726e-05,
+      "loss": 1.032,
+      "step": 6091
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6247272233802486e-05,
+      "loss": 0.877,
+      "step": 6092
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6245999140487015e-05,
+      "loss": 1.0277,
+      "step": 6093
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624472588116002e-05,
+      "loss": 1.0147,
+      "step": 6094
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624345245585534e-05,
+      "loss": 0.9666,
+      "step": 6095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242178864606826e-05,
+      "loss": 0.8584,
+      "step": 6096
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6240905107448328e-05,
+      "loss": 0.9359,
+      "step": 6097
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62396311844137e-05,
+      "loss": 0.8356,
+      "step": 6098
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62383570955368e-05,
+      "loss": 0.9651,
+      "step": 6099
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6237082840851495e-05,
+      "loss": 0.9475,
+      "step": 6100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6235808420391654e-05,
+      "loss": 0.9412,
+      "step": 6101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623453383419114e-05,
+      "loss": 0.9676,
+      "step": 6102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623325908228384e-05,
+      "loss": 0.9773,
+      "step": 6103
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6231984164703635e-05,
+      "loss": 0.8983,
+      "step": 6104
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623070908148441e-05,
+      "loss": 0.9348,
+      "step": 6105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6229433832660046e-05,
+      "loss": 1.0529,
+      "step": 6106
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6228158418264453e-05,
+      "loss": 0.9611,
+      "step": 6107
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6226882838331516e-05,
+      "loss": 1.0051,
+      "step": 6108
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6225607092895152e-05,
+      "loss": 1.0029,
+      "step": 6109
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622433118198926e-05,
+      "loss": 0.976,
+      "step": 6110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622305510564775e-05,
+      "loss": 1.007,
+      "step": 6111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6221778863904547e-05,
+      "loss": 0.9972,
+      "step": 6112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6220502456793566e-05,
+      "loss": 0.9695,
+      "step": 6113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219225884348734e-05,
+      "loss": 0.947,
+      "step": 6114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6217949146603978e-05,
+      "loss": 0.9411,
+      "step": 6115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6216672243593242e-05,
+      "loss": 0.881,
+      "step": 6116
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6215395175350454e-05,
+      "loss": 0.8787,
+      "step": 6117
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621411794190956e-05,
+      "loss": 0.951,
+      "step": 6118
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621284054330451e-05,
+      "loss": 1.0435,
+      "step": 6119
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621156297956925e-05,
+      "loss": 0.9401,
+      "step": 6120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6210285250737746e-05,
+      "loss": 0.9573,
+      "step": 6121
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620900735684395e-05,
+      "loss": 0.9294,
+      "step": 6122
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6207729297921833e-05,
+      "loss": 0.9064,
+      "step": 6123
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206451074005357e-05,
+      "loss": 0.8981,
+      "step": 6124
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6205172685128504e-05,
+      "loss": 0.9945,
+      "step": 6125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6203894131325248e-05,
+      "loss": 0.9444,
+      "step": 6126
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6202615412629567e-05,
+      "loss": 0.9133,
+      "step": 6127
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6201336529075456e-05,
+      "loss": 0.9737,
+      "step": 6128
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62000574806969e-05,
+      "loss": 0.9361,
+      "step": 6129
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61987782675279e-05,
+      "loss": 0.9028,
+      "step": 6130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619749888960245e-05,
+      "loss": 1.0074,
+      "step": 6131
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619621934695456e-05,
+      "loss": 0.8592,
+      "step": 6132
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194939639618238e-05,
+      "loss": 0.8322,
+      "step": 6133
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6193659767627493e-05,
+      "loss": 0.9607,
+      "step": 6134
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6192379731016346e-05,
+      "loss": 1.0459,
+      "step": 6135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619109952981882e-05,
+      "loss": 0.9461,
+      "step": 6136
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189819164068934e-05,
+      "loss": 0.9213,
+      "step": 6137
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618853863380073e-05,
+      "loss": 1.1049,
+      "step": 6138
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187257939048233e-05,
+      "loss": 0.9315,
+      "step": 6139
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618597707984549e-05,
+      "loss": 1.0038,
+      "step": 6140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184696056226542e-05,
+      "loss": 1.0278,
+      "step": 6141
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6183414868225434e-05,
+      "loss": 0.9885,
+      "step": 6142
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182133515876217e-05,
+      "loss": 1.0257,
+      "step": 6143
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6180851999212956e-05,
+      "loss": 0.9944,
+      "step": 6144
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6179570318269707e-05,
+      "loss": 0.9887,
+      "step": 6145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617828847308054e-05,
+      "loss": 0.9426,
+      "step": 6146
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177006463679517e-05,
+      "loss": 0.9119,
+      "step": 6147
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617572429010072e-05,
+      "loss": 0.9884,
+      "step": 6148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6174441952378222e-05,
+      "loss": 0.9637,
+      "step": 6149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173159450546112e-05,
+      "loss": 0.9303,
+      "step": 6150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171876784638475e-05,
+      "loss": 0.9533,
+      "step": 6151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61705939546894e-05,
+      "loss": 1.0055,
+      "step": 6152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6169310960732984e-05,
+      "loss": 0.9885,
+      "step": 6153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6168027802803332e-05,
+      "loss": 1.0976,
+      "step": 6154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6166744480934543e-05,
+      "loss": 0.9372,
+      "step": 6155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165460995160728e-05,
+      "loss": 0.9867,
+      "step": 6156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6164177345516002e-05,
+      "loss": 0.9597,
+      "step": 6157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6162893532034484e-05,
+      "loss": 0.4726,
+      "step": 6158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161609554750293e-05,
+      "loss": 0.9436,
+      "step": 6159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616032541369756e-05,
+      "loss": 0.9685,
+      "step": 6160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615904110891041e-05,
+      "loss": 1.0044,
+      "step": 6161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6157756640422983e-05,
+      "loss": 0.9379,
+      "step": 6162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6156472008269414e-05,
+      "loss": 1.0016,
+      "step": 6163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155187212483855e-05,
+      "loss": 0.9785,
+      "step": 6164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6153902253100448e-05,
+      "loss": 0.9925,
+      "step": 6165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6152617130153346e-05,
+      "loss": 1.0436,
+      "step": 6166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6151331843676712e-05,
+      "loss": 0.849,
+      "step": 6167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61500463937047e-05,
+      "loss": 0.9504,
+      "step": 6168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148760780271478e-05,
+      "loss": 0.9243,
+      "step": 6169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614747500341122e-05,
+      "loss": 0.3808,
+      "step": 6170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146189063158095e-05,
+      "loss": 0.9572,
+      "step": 6171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6144902959546286e-05,
+      "loss": 0.9613,
+      "step": 6172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614361669260997e-05,
+      "loss": 0.3888,
+      "step": 6173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6142330262383345e-05,
+      "loss": 0.974,
+      "step": 6174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6141043668900595e-05,
+      "loss": 0.9947,
+      "step": 6175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6139756912195917e-05,
+      "loss": 0.8574,
+      "step": 6176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138469992303516e-05,
+      "loss": 1.0083,
+      "step": 6177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613718290925759e-05,
+      "loss": 0.9986,
+      "step": 6178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6135895663092352e-05,
+      "loss": 0.9466,
+      "step": 6179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134608253842013e-05,
+      "loss": 1.0177,
+      "step": 6180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6133320681540797e-05,
+      "loss": 1.0048,
+      "step": 6181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613203294622292e-05,
+      "loss": 0.4423,
+      "step": 6182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613074504792261e-05,
+      "loss": 0.9448,
+      "step": 6183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6129456986674103e-05,
+      "loss": 0.4718,
+      "step": 6184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6128168762511624e-05,
+      "loss": 0.9332,
+      "step": 6185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612688037546942e-05,
+      "loss": 0.9959,
+      "step": 6186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6125591825581735e-05,
+      "loss": 1.0283,
+      "step": 6187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6124303112882816e-05,
+      "loss": 0.9883,
+      "step": 6188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6123014237406912e-05,
+      "loss": 0.941,
+      "step": 6189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612172519918828e-05,
+      "loss": 0.948,
+      "step": 6190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612043599826119e-05,
+      "loss": 0.9831,
+      "step": 6191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6119146634659902e-05,
+      "loss": 0.9278,
+      "step": 6192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117857108418685e-05,
+      "loss": 0.9652,
+      "step": 6193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6116567419571808e-05,
+      "loss": 0.908,
+      "step": 6194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611527756815356e-05,
+      "loss": 1.0273,
+      "step": 6195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113987554198214e-05,
+      "loss": 0.9829,
+      "step": 6196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6112697377740066e-05,
+      "loss": 0.9808,
+      "step": 6197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61114070388134e-05,
+      "loss": 0.9092,
+      "step": 6198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110116537452512e-05,
+      "loss": 0.9122,
+      "step": 6199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6108825873691712e-05,
+      "loss": 1.0145,
+      "step": 6200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107535047565295e-05,
+      "loss": 0.941,
+      "step": 6201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6106244059107567e-05,
+      "loss": 0.417,
+      "step": 6202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6104952908352852e-05,
+      "loss": 1.0155,
+      "step": 6203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103661595335457e-05,
+      "loss": 1.0359,
+      "step": 6204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610237012008971e-05,
+      "loss": 0.9455,
+      "step": 6205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6101078482649938e-05,
+      "loss": 0.9873,
+      "step": 6206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609978668305046e-05,
+      "loss": 1.0678,
+      "step": 6207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6098494721325623e-05,
+      "loss": 1.023,
+      "step": 6208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609720259750976e-05,
+      "loss": 0.8474,
+      "step": 6209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095910311637218e-05,
+      "loss": 0.9746,
+      "step": 6210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6094617863742342e-05,
+      "loss": 1.0231,
+      "step": 6211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6093325253859483e-05,
+      "loss": 0.9564,
+      "step": 6212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6092032482023e-05,
+      "loss": 0.9514,
+      "step": 6213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609073954826725e-05,
+      "loss": 0.9131,
+      "step": 6214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60894464526266e-05,
+      "loss": 0.9946,
+      "step": 6215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088153195135413e-05,
+      "loss": 0.9398,
+      "step": 6216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086859775828073e-05,
+      "loss": 0.9465,
+      "step": 6217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608556619473895e-05,
+      "loss": 1.0174,
+      "step": 6218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6084272451902426e-05,
+      "loss": 0.8983,
+      "step": 6219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6082978547352892e-05,
+      "loss": 0.9544,
+      "step": 6220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6081684481124735e-05,
+      "loss": 0.9711,
+      "step": 6221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608039025325235e-05,
+      "loss": 0.9528,
+      "step": 6222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6079095863770138e-05,
+      "loss": 0.9525,
+      "step": 6223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60778013127125e-05,
+      "loss": 0.9782,
+      "step": 6224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076506600113845e-05,
+      "loss": 0.9938,
+      "step": 6225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6075211726008583e-05,
+      "loss": 0.9878,
+      "step": 6226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6073916690431134e-05,
+      "loss": 1.0014,
+      "step": 6227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6072621493415916e-05,
+      "loss": 0.8544,
+      "step": 6228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071326134997354e-05,
+      "loss": 1.0124,
+      "step": 6229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070030615209878e-05,
+      "loss": 0.8731,
+      "step": 6230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606873493408792e-05,
+      "loss": 0.9864,
+      "step": 6231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6067439091665917e-05,
+      "loss": 1.0234,
+      "step": 6232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6066143087978313e-05,
+      "loss": 0.9236,
+      "step": 6233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6064846923059557e-05,
+      "loss": 0.994,
+      "step": 6234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6063550596944094e-05,
+      "loss": 0.9584,
+      "step": 6235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6062254109666383e-05,
+      "loss": 0.9511,
+      "step": 6236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606095746126088e-05,
+      "loss": 1.0683,
+      "step": 6237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605966065176205e-05,
+      "loss": 1.0358,
+      "step": 6238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605836368120436e-05,
+      "loss": 1.0582,
+      "step": 6239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057066549622284e-05,
+      "loss": 1.0028,
+      "step": 6240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055769257050296e-05,
+      "loss": 0.9075,
+      "step": 6241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605447180352288e-05,
+      "loss": 0.9312,
+      "step": 6242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053174189074515e-05,
+      "loss": 0.9597,
+      "step": 6243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6051876413739696e-05,
+      "loss": 1.0196,
+      "step": 6244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6050578477552915e-05,
+      "loss": 0.9432,
+      "step": 6245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6049280380548665e-05,
+      "loss": 0.9814,
+      "step": 6246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6047982122761454e-05,
+      "loss": 1.0058,
+      "step": 6247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046683704225784e-05,
+      "loss": 0.9658,
+      "step": 6248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6045385124976167e-05,
+      "loss": 0.9972,
+      "step": 6249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6044086385047117e-05,
+      "loss": 0.9125,
+      "step": 6250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604278748447316e-05,
+      "loss": 1.0107,
+      "step": 6251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6041488423288804e-05,
+      "loss": 0.9386,
+      "step": 6252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6040189201528592e-05,
+      "loss": 1.0353,
+      "step": 6253
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038889819227047e-05,
+      "loss": 0.9879,
+      "step": 6254
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603759027641871e-05,
+      "loss": 0.9699,
+      "step": 6255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6036290573138115e-05,
+      "loss": 0.9677,
+      "step": 6256
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6034990709419815e-05,
+      "loss": 0.9287,
+      "step": 6257
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603369068529835e-05,
+      "loss": 1.0501,
+      "step": 6258
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6032390500808283e-05,
+      "loss": 0.9183,
+      "step": 6259
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6031090155984162e-05,
+      "loss": 0.998,
+      "step": 6260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6029789650860554e-05,
+      "loss": 0.9245,
+      "step": 6261
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6028488985472024e-05,
+      "loss": 1.0393,
+      "step": 6262
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027188159853144e-05,
+      "loss": 0.952,
+      "step": 6263
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6025887174038484e-05,
+      "loss": 0.8802,
+      "step": 6264
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602458602806263e-05,
+      "loss": 0.9437,
+      "step": 6265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6023284721960153e-05,
+      "loss": 0.9623,
+      "step": 6266
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602198325576565e-05,
+      "loss": 0.8369,
+      "step": 6267
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602068162951371e-05,
+      "loss": 0.8954,
+      "step": 6268
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6019379843238928e-05,
+      "loss": 1.0052,
+      "step": 6269
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6018077896975906e-05,
+      "loss": 0.9093,
+      "step": 6270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6016775790759243e-05,
+      "loss": 1.0079,
+      "step": 6271
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015473524623556e-05,
+      "loss": 0.9773,
+      "step": 6272
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601417109860345e-05,
+      "loss": 0.8409,
+      "step": 6273
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012868512733546e-05,
+      "loss": 0.8947,
+      "step": 6274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6011565767048468e-05,
+      "loss": 1.0817,
+      "step": 6275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010262861582833e-05,
+      "loss": 1.0248,
+      "step": 6276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6008959796371276e-05,
+      "loss": 0.9635,
+      "step": 6277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6007656571448432e-05,
+      "loss": 0.9667,
+      "step": 6278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6006353186848937e-05,
+      "loss": 0.9853,
+      "step": 6279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600504964260743e-05,
+      "loss": 0.9753,
+      "step": 6280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600374593875857e-05,
+      "loss": 0.9936,
+      "step": 6281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6002442075336997e-05,
+      "loss": 0.9825,
+      "step": 6282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600113805237737e-05,
+      "loss": 0.9337,
+      "step": 6283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5999833869914345e-05,
+      "loss": 0.9632,
+      "step": 6284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5998529527982587e-05,
+      "loss": 1.0221,
+      "step": 6285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599722502661677e-05,
+      "loss": 0.9368,
+      "step": 6286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995920365851557e-05,
+      "loss": 1.0091,
+      "step": 6287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599461554572163e-05,
+      "loss": 0.9953,
+      "step": 6288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599331056626167e-05,
+      "loss": 0.9836,
+      "step": 6289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5992005427506354e-05,
+      "loss": 0.9289,
+      "step": 6290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599070012949038e-05,
+      "loss": 0.9921,
+      "step": 6291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989394672248444e-05,
+      "loss": 0.9549,
+      "step": 6292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5988089055815233e-05,
+      "loss": 0.9512,
+      "step": 6293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5986783280225456e-05,
+      "loss": 0.9057,
+      "step": 6294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985477345513818e-05,
+      "loss": 1.0569,
+      "step": 6295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984171251715026e-05,
+      "loss": 0.9928,
+      "step": 6296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982864998863797e-05,
+      "loss": 0.8994,
+      "step": 6297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598155858699485e-05,
+      "loss": 0.9439,
+      "step": 6298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980252016142905e-05,
+      "loss": 0.9644,
+      "step": 6299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5978945286342697e-05,
+      "loss": 0.9368,
+      "step": 6300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5977638397628946e-05,
+      "loss": 0.9187,
+      "step": 6301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5976331350036396e-05,
+      "loss": 0.915,
+      "step": 6302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5975024143599784e-05,
+      "loss": 1.0537,
+      "step": 6303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5973716778353855e-05,
+      "loss": 1.0399,
+      "step": 6304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5972409254333356e-05,
+      "loss": 1.0284,
+      "step": 6305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597110157157304e-05,
+      "loss": 0.8444,
+      "step": 6306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5969793730107667e-05,
+      "loss": 0.9703,
+      "step": 6307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596848572997199e-05,
+      "loss": 0.9317,
+      "step": 6308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5967177571200783e-05,
+      "loss": 0.9446,
+      "step": 6309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596586925382881e-05,
+      "loss": 0.9938,
+      "step": 6310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964560777890844e-05,
+      "loss": 1.0619,
+      "step": 6311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5963252143421667e-05,
+      "loss": 0.9225,
+      "step": 6312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5961943350456058e-05,
+      "loss": 1.0061,
+      "step": 6313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5960634399028803e-05,
+      "loss": 0.9403,
+      "step": 6314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5959325289174693e-05,
+      "loss": 0.4768,
+      "step": 6315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5958016020928526e-05,
+      "loss": 0.9224,
+      "step": 6316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5956706594325098e-05,
+      "loss": 0.9196,
+      "step": 6317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595539700939921e-05,
+      "loss": 0.9792,
+      "step": 6318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595408726618567e-05,
+      "loss": 1.0311,
+      "step": 6319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952777364719293e-05,
+      "loss": 0.9834,
+      "step": 6320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595146730503489e-05,
+      "loss": 0.9204,
+      "step": 6321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5950157087167284e-05,
+      "loss": 0.9959,
+      "step": 6322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5948846711151296e-05,
+      "loss": 0.9882,
+      "step": 6323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5947536177021762e-05,
+      "loss": 0.9677,
+      "step": 6324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594622548481351e-05,
+      "loss": 1.0133,
+      "step": 6325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594491463456137e-05,
+      "loss": 0.9161,
+      "step": 6326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594360362630019e-05,
+      "loss": 0.9111,
+      "step": 6327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942292460064813e-05,
+      "loss": 0.9366,
+      "step": 6328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5940981135890095e-05,
+      "loss": 0.9546,
+      "step": 6329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5939669653810882e-05,
+      "loss": 0.9692,
+      "step": 6330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593835801386203e-05,
+      "loss": 0.9866,
+      "step": 6331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937046216078402e-05,
+      "loss": 0.9674,
+      "step": 6332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935734260494873e-05,
+      "loss": 0.9916,
+      "step": 6333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59344221471463e-05,
+      "loss": 1.0091,
+      "step": 6334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5933109876067573e-05,
+      "loss": 0.9594,
+      "step": 6335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5931797447293553e-05,
+      "loss": 1.0012,
+      "step": 6336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930484860859136e-05,
+      "loss": 1.0017,
+      "step": 6337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5929172116799203e-05,
+      "loss": 0.944,
+      "step": 6338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927859215148647e-05,
+      "loss": 0.9731,
+      "step": 6339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5926546155942367e-05,
+      "loss": 0.884,
+      "step": 6340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5925232939215254e-05,
+      "loss": 0.8663,
+      "step": 6341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5923919565002218e-05,
+      "loss": 1.0247,
+      "step": 6342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5922606033338165e-05,
+      "loss": 0.9055,
+      "step": 6343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592129234425801e-05,
+      "loss": 0.8932,
+      "step": 6344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919978497796665e-05,
+      "loss": 0.9561,
+      "step": 6345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5918664493989055e-05,
+      "loss": 0.9963,
+      "step": 6346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5917350332870102e-05,
+      "loss": 0.9451,
+      "step": 6347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5916036014474736e-05,
+      "loss": 0.9989,
+      "step": 6348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591472153883789e-05,
+      "loss": 0.8649,
+      "step": 6349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5913406905994496e-05,
+      "loss": 0.9542,
+      "step": 6350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5912092115979506e-05,
+      "loss": 0.9829,
+      "step": 6351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5910777168827857e-05,
+      "loss": 1.0054,
+      "step": 6352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5909462064574502e-05,
+      "loss": 1.0536,
+      "step": 6353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5908146803254395e-05,
+      "loss": 1.0051,
+      "step": 6354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906831384902494e-05,
+      "loss": 0.9411,
+      "step": 6355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590551580955376e-05,
+      "loss": 1.0227,
+      "step": 6356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590420007724316e-05,
+      "loss": 0.9173,
+      "step": 6357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5902884188005664e-05,
+      "loss": 0.9955,
+      "step": 6358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5901568141876253e-05,
+      "loss": 0.9946,
+      "step": 6359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5900251938889896e-05,
+      "loss": 0.8982,
+      "step": 6360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5898935579081586e-05,
+      "loss": 0.9864,
+      "step": 6361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5897619062486302e-05,
+      "loss": 0.944,
+      "step": 6362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589630238913904e-05,
+      "loss": 0.9429,
+      "step": 6363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589498555907479e-05,
+      "loss": 0.9225,
+      "step": 6364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893668572328564e-05,
+      "loss": 0.9596,
+      "step": 6365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5892351428935357e-05,
+      "loss": 0.4597,
+      "step": 6366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5891034128930175e-05,
+      "loss": 0.9879,
+      "step": 6367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588971667234804e-05,
+      "loss": 0.9588,
+      "step": 6368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888399059223957e-05,
+      "loss": 0.9471,
+      "step": 6369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5887081289592955e-05,
+      "loss": 0.9108,
+      "step": 6370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885763363490056e-05,
+      "loss": 1.0111,
+      "step": 6371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884445280950288e-05,
+      "loss": 0.9979,
+      "step": 6372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5883127042008685e-05,
+      "loss": 0.9857,
+      "step": 6373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5881808646700287e-05,
+      "loss": 0.9139,
+      "step": 6374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880490095060127e-05,
+      "loss": 0.9353,
+      "step": 6375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5879171387123262e-05,
+      "loss": 0.9648,
+      "step": 6376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877852522924733e-05,
+      "loss": 0.974,
+      "step": 6377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58765335024996e-05,
+      "loss": 0.9213,
+      "step": 6378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875214325882914e-05,
+      "loss": 1.0495,
+      "step": 6379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5873894993109743e-05,
+      "loss": 0.9464,
+      "step": 6380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5872575504215152e-05,
+      "loss": 0.9935,
+      "step": 6381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5871255859234215e-05,
+      "loss": 0.9777,
+      "step": 6382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5869936058202e-05,
+      "loss": 0.8895,
+      "step": 6383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5868616101153586e-05,
+      "loss": 1.0592,
+      "step": 6384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586729598812406e-05,
+      "loss": 0.9554,
+      "step": 6385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586597571914851e-05,
+      "loss": 1.0418,
+      "step": 6386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5864655294262025e-05,
+      "loss": 0.8983,
+      "step": 6387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863334713499698e-05,
+      "loss": 0.9328,
+      "step": 6388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586201397689663e-05,
+      "loss": 0.9737,
+      "step": 6389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586069308448793e-05,
+      "loss": 0.9526,
+      "step": 6390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5859372036308697e-05,
+      "loss": 1.0533,
+      "step": 6391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585805083239405e-05,
+      "loss": 0.9518,
+      "step": 6392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856729472779103e-05,
+      "loss": 0.8546,
+      "step": 6393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5855407957498973e-05,
+      "loss": 0.9316,
+      "step": 6394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585408628658879e-05,
+      "loss": 1.0093,
+      "step": 6395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5852764460083675e-05,
+      "loss": 1.0008,
+      "step": 6396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5851442478018765e-05,
+      "loss": 0.9278,
+      "step": 6397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5850120340429203e-05,
+      "loss": 0.8451,
+      "step": 6398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584879804735012e-05,
+      "loss": 1.0902,
+      "step": 6399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5847475598816666e-05,
+      "loss": 0.9739,
+      "step": 6400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5846152994863988e-05,
+      "loss": 0.9661,
+      "step": 6401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584483023552724e-05,
+      "loss": 0.9732,
+      "step": 6402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584350732084158e-05,
+      "loss": 0.919,
+      "step": 6403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5842184250842172e-05,
+      "loss": 0.9838,
+      "step": 6404
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584086102556418e-05,
+      "loss": 0.976,
+      "step": 6405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5839537645042766e-05,
+      "loss": 0.9455,
+      "step": 6406
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5838214109313117e-05,
+      "loss": 1.0339,
+      "step": 6407
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5836890418410403e-05,
+      "loss": 1.0331,
+      "step": 6408
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835566572369803e-05,
+      "loss": 1.0077,
+      "step": 6409
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5834242571226514e-05,
+      "loss": 0.9507,
+      "step": 6410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583291841501572e-05,
+      "loss": 0.9103,
+      "step": 6411
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5831594103772615e-05,
+      "loss": 0.9141,
+      "step": 6412
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5830269637532404e-05,
+      "loss": 0.9491,
+      "step": 6413
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582894501633028e-05,
+      "loss": 0.8861,
+      "step": 6414
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5827620240201453e-05,
+      "loss": 0.9741,
+      "step": 6415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5826295309181138e-05,
+      "loss": 0.9874,
+      "step": 6416
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582497022330455e-05,
+      "loss": 0.981,
+      "step": 6417
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5823644982606905e-05,
+      "loss": 0.8748,
+      "step": 6418
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822319587123426e-05,
+      "loss": 1.0071,
+      "step": 6419
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5820994036889342e-05,
+      "loss": 0.9527,
+      "step": 6420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581966833193989e-05,
+      "loss": 1.0153,
+      "step": 6421
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5818342472310293e-05,
+      "loss": 1.0166,
+      "step": 6422
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5817016458035802e-05,
+      "loss": 1.0071,
+      "step": 6423
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581569028915166e-05,
+      "loss": 0.9441,
+      "step": 6424
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581436396569311e-05,
+      "loss": 0.946,
+      "step": 6425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5813037487695405e-05,
+      "loss": 0.892,
+      "step": 6426
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5811710855193802e-05,
+      "loss": 0.9753,
+      "step": 6427
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5810384068223564e-05,
+      "loss": 0.9855,
+      "step": 6428
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580905712681995e-05,
+      "loss": 0.9904,
+      "step": 6429
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5807730031018235e-05,
+      "loss": 0.921,
+      "step": 6430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580640278085369e-05,
+      "loss": 0.9031,
+      "step": 6431
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580507537636159e-05,
+      "loss": 0.953,
+      "step": 6432
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5803747817577215e-05,
+      "loss": 0.9314,
+      "step": 6433
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5802420104535854e-05,
+      "loss": 0.9536,
+      "step": 6434
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5801092237272792e-05,
+      "loss": 1.0298,
+      "step": 6435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5799764215823326e-05,
+      "loss": 0.9604,
+      "step": 6436
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5798436040222748e-05,
+      "loss": 0.9276,
+      "step": 6437
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579710771050636e-05,
+      "loss": 0.9875,
+      "step": 6438
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5795779226709478e-05,
+      "loss": 0.9895,
+      "step": 6439
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5794450588867397e-05,
+      "loss": 0.9119,
+      "step": 6440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5793121797015442e-05,
+      "loss": 0.978,
+      "step": 6441
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579179285118892e-05,
+      "loss": 0.4443,
+      "step": 6442
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5790463751423164e-05,
+      "loss": 0.9638,
+      "step": 6443
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789134497753495e-05,
+      "loss": 0.4525,
+      "step": 6444
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787805090215238e-05,
+      "loss": 0.8953,
+      "step": 6445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5786475528843737e-05,
+      "loss": 0.9172,
+      "step": 6446
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5785145813674322e-05,
+      "loss": 0.9322,
+      "step": 6447
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783815944742342e-05,
+      "loss": 0.9944,
+      "step": 6448
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578248592208314e-05,
+      "loss": 1.0197,
+      "step": 6449
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578115574573206e-05,
+      "loss": 0.9462,
+      "step": 6450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779825415724468e-05,
+      "loss": 0.9552,
+      "step": 6451
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5778494932095714e-05,
+      "loss": 0.9205,
+      "step": 6452
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577716429488117e-05,
+      "loss": 0.9127,
+      "step": 6453
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5775833504116194e-05,
+      "loss": 0.949,
+      "step": 6454
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5774502559836162e-05,
+      "loss": 0.9626,
+      "step": 6455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5773171462076443e-05,
+      "loss": 0.9553,
+      "step": 6456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5771840210872423e-05,
+      "loss": 0.9975,
+      "step": 6457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5770508806259484e-05,
+      "loss": 1.0315,
+      "step": 6458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576917724827301e-05,
+      "loss": 0.9953,
+      "step": 6459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5767845536948393e-05,
+      "loss": 1.0831,
+      "step": 6460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576651367232103e-05,
+      "loss": 0.9576,
+      "step": 6461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576518165442632e-05,
+      "loss": 0.9656,
+      "step": 6462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5763849483299663e-05,
+      "loss": 0.9902,
+      "step": 6463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576251715897647e-05,
+      "loss": 0.9047,
+      "step": 6464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761184681492156e-05,
+      "loss": 0.9686,
+      "step": 6465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5759852050882133e-05,
+      "loss": 1.032,
+      "step": 6466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575851926718182e-05,
+      "loss": 1.0335,
+      "step": 6467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757186330426635e-05,
+      "loss": 0.9888,
+      "step": 6468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5755853240652022e-05,
+      "loss": 0.9644,
+      "step": 6469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57545199978934e-05,
+      "loss": 0.9927,
+      "step": 6470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5753186602186207e-05,
+      "loss": 0.8878,
+      "step": 6471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5751853053565888e-05,
+      "loss": 0.9809,
+      "step": 6472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5750519352067882e-05,
+      "loss": 1.0126,
+      "step": 6473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5749185497727643e-05,
+      "loss": 1.0151,
+      "step": 6474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747851490580614e-05,
+      "loss": 0.9096,
+      "step": 6475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5746517330662258e-05,
+      "loss": 0.9419,
+      "step": 6476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574518301800804e-05,
+      "loss": 0.9754,
+      "step": 6477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5743848552653414e-05,
+      "loss": 0.9408,
+      "step": 6478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5742513934633853e-05,
+      "loss": 0.9628,
+      "step": 6479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5741179163984833e-05,
+      "loss": 0.9269,
+      "step": 6480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573984424074183e-05,
+      "loss": 0.9196,
+      "step": 6481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5738509164940317e-05,
+      "loss": 0.9699,
+      "step": 6482
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5737173936615787e-05,
+      "loss": 0.9793,
+      "step": 6483
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5735838555803728e-05,
+      "loss": 1.0056,
+      "step": 6484
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573450302253963e-05,
+      "loss": 0.9574,
+      "step": 6485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573316733685899e-05,
+      "loss": 0.8594,
+      "step": 6486
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5731831498797308e-05,
+      "loss": 0.911,
+      "step": 6487
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5730495508390094e-05,
+      "loss": 0.9984,
+      "step": 6488
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5729159365672858e-05,
+      "loss": 0.9407,
+      "step": 6489
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5727823070681103e-05,
+      "loss": 0.9815,
+      "step": 6490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5726486623450355e-05,
+      "loss": 0.4987,
+      "step": 6491
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572515002401613e-05,
+      "loss": 0.4754,
+      "step": 6492
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572381327241396e-05,
+      "loss": 0.9065,
+      "step": 6493
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5722476368679373e-05,
+      "loss": 0.9021,
+      "step": 6494
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5721139312847893e-05,
+      "loss": 1.0114,
+      "step": 6495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571980210495507e-05,
+      "loss": 0.9034,
+      "step": 6496
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5718464745036435e-05,
+      "loss": 1.0129,
+      "step": 6497
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571712723312754e-05,
+      "loss": 0.9869,
+      "step": 6498
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571578956926393e-05,
+      "loss": 1.0119,
+      "step": 6499
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571445175348117e-05,
+      "loss": 0.9175,
+      "step": 6500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5713113785814802e-05,
+      "loss": 0.9937,
+      "step": 6501
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711775666300395e-05,
+      "loss": 0.9665,
+      "step": 6502
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5710437394973516e-05,
+      "loss": 0.9992,
+      "step": 6503
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570909897186973e-05,
+      "loss": 0.97,
+      "step": 6504
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5707760397024616e-05,
+      "loss": 0.9912,
+      "step": 6505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570642167047375e-05,
+      "loss": 0.9675,
+      "step": 6506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570508279225271e-05,
+      "loss": 0.9106,
+      "step": 6507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570374376239709e-05,
+      "loss": 0.4886,
+      "step": 6508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570240458094247e-05,
+      "loss": 0.4811,
+      "step": 6509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701065247924455e-05,
+      "loss": 1.0017,
+      "step": 6510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569972576337863e-05,
+      "loss": 0.8787,
+      "step": 6511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5698386127340613e-05,
+      "loss": 0.9096,
+      "step": 6512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5697046339845995e-05,
+      "loss": 0.9934,
+      "step": 6513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5695706400930394e-05,
+      "loss": 1.0667,
+      "step": 6514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569436631062942e-05,
+      "loss": 0.9275,
+      "step": 6515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693026068978692e-05,
+      "loss": 0.9658,
+      "step": 6516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5691685676013834e-05,
+      "loss": 1.0367,
+      "step": 6517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5690345131770474e-05,
+      "loss": 0.9405,
+      "step": 6518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5689004436284238e-05,
+      "loss": 1.0492,
+      "step": 6519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5687663589590762e-05,
+      "loss": 0.9616,
+      "step": 6520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568632259172568e-05,
+      "loss": 0.9553,
+      "step": 6521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568498144272464e-05,
+      "loss": 0.8952,
+      "step": 6522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568364014262329e-05,
+      "loss": 1.0014,
+      "step": 6523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682298691457273e-05,
+      "loss": 0.9521,
+      "step": 6524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5680957089262246e-05,
+      "loss": 1.0057,
+      "step": 6525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5679615336073867e-05,
+      "loss": 0.9143,
+      "step": 6526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56782734319278e-05,
+      "loss": 0.9308,
+      "step": 6527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676931376859714e-05,
+      "loss": 0.5452,
+      "step": 6528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567558917090527e-05,
+      "loss": 0.9678,
+      "step": 6529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5674246814100153e-05,
+      "loss": 1.0469,
+      "step": 6530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5672904306480035e-05,
+      "loss": 0.9215,
+      "step": 6531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56715616480806e-05,
+      "loss": 0.9527,
+      "step": 6532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567021883893753e-05,
+      "loss": 1.0055,
+      "step": 6533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566887587908652e-05,
+      "loss": 1.0103,
+      "step": 6534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5667532768563266e-05,
+      "loss": 0.9811,
+      "step": 6535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5666189507403465e-05,
+      "loss": 1.0095,
+      "step": 6536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566484609564282e-05,
+      "loss": 0.9575,
+      "step": 6537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566350253331703e-05,
+      "loss": 0.9141,
+      "step": 6538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5662158820461814e-05,
+      "loss": 0.9057,
+      "step": 6539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566081495711288e-05,
+      "loss": 0.9963,
+      "step": 6540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659470943305956e-05,
+      "loss": 1.0141,
+      "step": 6541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5658126779076754e-05,
+      "loss": 0.9296,
+      "step": 6542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5656782464461002e-05,
+      "loss": 0.9659,
+      "step": 6543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5655437999494436e-05,
+      "loss": 1.0957,
+      "step": 6544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5654093384212787e-05,
+      "loss": 0.9459,
+      "step": 6545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652748618651795e-05,
+      "loss": 0.8735,
+      "step": 6546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56514037028472e-05,
+      "loss": 1.0121,
+      "step": 6547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5650058636834745e-05,
+      "loss": 0.9554,
+      "step": 6548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564871342065019e-05,
+      "loss": 1.0293,
+      "step": 6549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564736805432928e-05,
+      "loss": 0.9616,
+      "step": 6550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5646022537907778e-05,
+      "loss": 0.9882,
+      "step": 6551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5644676871421445e-05,
+      "loss": 0.999,
+      "step": 6552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564333105490605e-05,
+      "loss": 0.4833,
+      "step": 6553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641985088397357e-05,
+      "loss": 0.9441,
+      "step": 6554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5640638971931145e-05,
+      "loss": 0.9535,
+      "step": 6555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5639292705543194e-05,
+      "loss": 1.017,
+      "step": 6556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563794628926928e-05,
+      "loss": 0.9689,
+      "step": 6557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5636599723145196e-05,
+      "loss": 0.9237,
+      "step": 6558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5635253007206726e-05,
+      "loss": 1.0218,
+      "step": 6559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5633906141489674e-05,
+      "loss": 0.9752,
+      "step": 6560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5632559126029823e-05,
+      "loss": 0.9352,
+      "step": 6561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5631211960862985e-05,
+      "loss": 0.9121,
+      "step": 6562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562986464602497e-05,
+      "loss": 0.9492,
+      "step": 6563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562851718155158e-05,
+      "loss": 0.9747,
+      "step": 6564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5627169567478627e-05,
+      "loss": 0.9904,
+      "step": 6565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5625821803841937e-05,
+      "loss": 0.9932,
+      "step": 6566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5624473890677328e-05,
+      "loss": 0.9241,
+      "step": 6567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562312582802063e-05,
+      "loss": 0.939,
+      "step": 6568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562177761590767e-05,
+      "loss": 0.9191,
+      "step": 6569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5620429254374277e-05,
+      "loss": 1.0057,
+      "step": 6570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5619080743456298e-05,
+      "loss": 1.063,
+      "step": 6571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5617732083189568e-05,
+      "loss": 0.9737,
+      "step": 6572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5616383273609933e-05,
+      "loss": 0.9441,
+      "step": 6573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5615034314753247e-05,
+      "loss": 0.9844,
+      "step": 6574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5613685206655363e-05,
+      "loss": 0.9207,
+      "step": 6575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5612335949352137e-05,
+      "loss": 0.9408,
+      "step": 6576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5610986542879433e-05,
+      "loss": 0.9446,
+      "step": 6577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5609636987273115e-05,
+      "loss": 0.973,
+      "step": 6578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560828728256905e-05,
+      "loss": 0.9692,
+      "step": 6579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5606937428803114e-05,
+      "loss": 0.8646,
+      "step": 6580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560558742601119e-05,
+      "loss": 0.9619,
+      "step": 6581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560423727422915e-05,
+      "loss": 0.4484,
+      "step": 6582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5602886973492884e-05,
+      "loss": 0.9279,
+      "step": 6583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560153652383828e-05,
+      "loss": 1.0122,
+      "step": 6584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5600185925301237e-05,
+      "loss": 0.9861,
+      "step": 6585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5598835177917644e-05,
+      "loss": 0.9497,
+      "step": 6586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559748428172341e-05,
+      "loss": 1.0259,
+      "step": 6587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5596133236754435e-05,
+      "loss": 0.9797,
+      "step": 6588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559478204304663e-05,
+      "loss": 0.9039,
+      "step": 6589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559343070063591e-05,
+      "loss": 0.9275,
+      "step": 6590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559207920955819e-05,
+      "loss": 0.8914,
+      "step": 6591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590727569849388e-05,
+      "loss": 0.9285,
+      "step": 6592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5589375781545434e-05,
+      "loss": 0.9306,
+      "step": 6593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5588023844682255e-05,
+      "loss": 0.9209,
+      "step": 6594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586671759295786e-05,
+      "loss": 0.9447,
+      "step": 6595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5585319525421964e-05,
+      "loss": 0.9882,
+      "step": 6596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5583967143096725e-05,
+      "loss": 0.9788,
+      "step": 6597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582614612356012e-05,
+      "loss": 1.0248,
+      "step": 6598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5581261933235788e-05,
+      "loss": 0.4951,
+      "step": 6599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557990910577199e-05,
+      "loss": 0.9347,
+      "step": 6600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5578556130000584e-05,
+      "loss": 0.9791,
+      "step": 6601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5577203005957527e-05,
+      "loss": 0.8908,
+      "step": 6602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557584973367878e-05,
+      "loss": 0.9348,
+      "step": 6603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557449631320032e-05,
+      "loss": 1.0043,
+      "step": 6604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5573142744558115e-05,
+      "loss": 0.8545,
+      "step": 6605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5571789027788134e-05,
+      "loss": 0.8802,
+      "step": 6606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5570435162926372e-05,
+      "loss": 0.9106,
+      "step": 6607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5569081150008804e-05,
+      "loss": 0.407,
+      "step": 6608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556772698907142e-05,
+      "loss": 1.0111,
+      "step": 6609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556637268015021e-05,
+      "loss": 0.9961,
+      "step": 6610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5565018223281168e-05,
+      "loss": 1.0031,
+      "step": 6611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55636636185003e-05,
+      "loss": 0.9558,
+      "step": 6612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5562308865843608e-05,
+      "loss": 0.9262,
+      "step": 6613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55609539653471e-05,
+      "loss": 0.9234,
+      "step": 6614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5559598917046787e-05,
+      "loss": 1.0905,
+      "step": 6615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5558243720978684e-05,
+      "loss": 0.9994,
+      "step": 6616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5556888377178812e-05,
+      "loss": 0.9423,
+      "step": 6617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555553288568319e-05,
+      "loss": 1.0335,
+      "step": 6618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5554177246527853e-05,
+      "loss": 0.943,
+      "step": 6619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5552821459748825e-05,
+      "loss": 0.9739,
+      "step": 6620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5551465525382146e-05,
+      "loss": 1.0225,
+      "step": 6621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5550109443463856e-05,
+      "loss": 1.0672,
+      "step": 6622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554875321402999e-05,
+      "loss": 0.9222,
+      "step": 6623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554739683711661e-05,
+      "loss": 0.9427,
+      "step": 6624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554604031275975e-05,
+      "loss": 0.9471,
+      "step": 6625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5544683640995473e-05,
+      "loss": 0.8781,
+      "step": 6626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5543326821859838e-05,
+      "loss": 0.47,
+      "step": 6627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554196985538891e-05,
+      "loss": 0.9265,
+      "step": 6628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554061274161875e-05,
+      "loss": 0.9577,
+      "step": 6629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539255480585433e-05,
+      "loss": 0.9877,
+      "step": 6630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5537898072325033e-05,
+      "loss": 0.9099,
+      "step": 6631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5536540516873625e-05,
+      "loss": 1.0339,
+      "step": 6632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553518281426729e-05,
+      "loss": 1.0688,
+      "step": 6633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5533824964542122e-05,
+      "loss": 0.8568,
+      "step": 6634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5532466967734208e-05,
+      "loss": 0.9159,
+      "step": 6635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553110882387964e-05,
+      "loss": 0.8682,
+      "step": 6636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5529750533014515e-05,
+      "loss": 0.9245,
+      "step": 6637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552839209517494e-05,
+      "loss": 0.8717,
+      "step": 6638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5527033510397008e-05,
+      "loss": 0.9514,
+      "step": 6639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5525674778716845e-05,
+      "loss": 0.9554,
+      "step": 6640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524315900170557e-05,
+      "loss": 0.9073,
+      "step": 6641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552295687479426e-05,
+      "loss": 0.9247,
+      "step": 6642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5521597702624078e-05,
+      "loss": 0.9758,
+      "step": 6643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5520238383696135e-05,
+      "loss": 1.0083,
+      "step": 6644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5518878918046558e-05,
+      "loss": 0.9626,
+      "step": 6645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5517519305711487e-05,
+      "loss": 0.9979,
+      "step": 6646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5516159546727055e-05,
+      "loss": 0.9186,
+      "step": 6647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55147996411294e-05,
+      "loss": 0.9609,
+      "step": 6648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551343958895467e-05,
+      "loss": 0.9182,
+      "step": 6649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512079390239008e-05,
+      "loss": 0.9478,
+      "step": 6650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510719045018574e-05,
+      "loss": 1.0794,
+      "step": 6651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5509358553329524e-05,
+      "loss": 1.0311,
+      "step": 6652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5507997915208017e-05,
+      "loss": 0.9811,
+      "step": 6653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550663713069021e-05,
+      "loss": 0.949,
+      "step": 6654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550527619981228e-05,
+      "loss": 0.9435,
+      "step": 6655
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5503915122610398e-05,
+      "loss": 1.0973,
+      "step": 6656
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5502553899120738e-05,
+      "loss": 0.8644,
+      "step": 6657
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5501192529379476e-05,
+      "loss": 0.9782,
+      "step": 6658
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5499831013422804e-05,
+      "loss": 0.9885,
+      "step": 6659
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5498469351286904e-05,
+      "loss": 1.002,
+      "step": 6660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5497107543007967e-05,
+      "loss": 1.0181,
+      "step": 6661
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5495745588622193e-05,
+      "loss": 0.9331,
+      "step": 6662
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5494383488165774e-05,
+      "loss": 0.9806,
+      "step": 6663
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549302124167492e-05,
+      "loss": 0.9056,
+      "step": 6664
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5491658849185832e-05,
+      "loss": 0.9441,
+      "step": 6665
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490296310734727e-05,
+      "loss": 0.9485,
+      "step": 6666
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5488933626357815e-05,
+      "loss": 0.9316,
+      "step": 6667
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5487570796091317e-05,
+      "loss": 1.016,
+      "step": 6668
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5486207819971454e-05,
+      "loss": 0.4451,
+      "step": 6669
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5484844698034453e-05,
+      "loss": 0.9054,
+      "step": 6670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5483481430316545e-05,
+      "loss": 0.9652,
+      "step": 6671
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548211801685396e-05,
+      "loss": 0.9944,
+      "step": 6672
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5480754457682947e-05,
+      "loss": 0.9982,
+      "step": 6673
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479390752839735e-05,
+      "loss": 0.9207,
+      "step": 6674
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547802690236057e-05,
+      "loss": 1.016,
+      "step": 6675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5476662906281716e-05,
+      "loss": 0.9801,
+      "step": 6676
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547529876463941e-05,
+      "loss": 0.904,
+      "step": 6677
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5473934477469922e-05,
+      "loss": 0.961,
+      "step": 6678
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472570044809503e-05,
+      "loss": 0.9019,
+      "step": 6679
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547120546669442e-05,
+      "loss": 1.035,
+      "step": 6680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546984074316095e-05,
+      "loss": 1.0242,
+      "step": 6681
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546847587424536e-05,
+      "loss": 0.9707,
+      "step": 6682
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5467110859983922e-05,
+      "loss": 0.8878,
+      "step": 6683
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5465745700412926e-05,
+      "loss": 0.934,
+      "step": 6684
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546438039556865e-05,
+      "loss": 0.9324,
+      "step": 6685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463014945487384e-05,
+      "loss": 0.9637,
+      "step": 6686
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546164935020542e-05,
+      "loss": 0.9834,
+      "step": 6687
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5460283609759054e-05,
+      "loss": 0.9654,
+      "step": 6688
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545891772418459e-05,
+      "loss": 0.9416,
+      "step": 6689
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545755169351832e-05,
+      "loss": 1.0046,
+      "step": 6690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5456185517796565e-05,
+      "loss": 0.9361,
+      "step": 6691
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454819197055627e-05,
+      "loss": 0.9395,
+      "step": 6692
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5453452731331827e-05,
+      "loss": 0.95,
+      "step": 6693
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545208612066148e-05,
+      "loss": 1.0332,
+      "step": 6694
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5450719365080914e-05,
+      "loss": 0.4123,
+      "step": 6695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544935246462645e-05,
+      "loss": 0.9754,
+      "step": 6696
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544798541933442e-05,
+      "loss": 0.9856,
+      "step": 6697
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544661822924116e-05,
+      "loss": 0.9918,
+      "step": 6698
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544525089438301e-05,
+      "loss": 0.9358,
+      "step": 6699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544388341479631e-05,
+      "loss": 0.9293,
+      "step": 6700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5442515790517406e-05,
+      "loss": 0.9321,
+      "step": 6701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544114802158265e-05,
+      "loss": 1.0175,
+      "step": 6702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439780108028393e-05,
+      "loss": 0.9362,
+      "step": 6703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5438412049890993e-05,
+      "loss": 0.9925,
+      "step": 6704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5437043847206808e-05,
+      "loss": 0.969,
+      "step": 6705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5435675500012212e-05,
+      "loss": 0.9422,
+      "step": 6706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543430700834357e-05,
+      "loss": 0.9699,
+      "step": 6707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543293837223725e-05,
+      "loss": 0.9895,
+      "step": 6708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543156959172964e-05,
+      "loss": 1.0587,
+      "step": 6709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5430200666857106e-05,
+      "loss": 1.0105,
+      "step": 6710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5428831597656045e-05,
+      "loss": 0.902,
+      "step": 6711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542746238416284e-05,
+      "loss": 0.9895,
+      "step": 6712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542609302641388e-05,
+      "loss": 1.0552,
+      "step": 6713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542472352444557e-05,
+      "loss": 0.9311,
+      "step": 6714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54233538782943e-05,
+      "loss": 0.8979,
+      "step": 6715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542198408799648e-05,
+      "loss": 0.9877,
+      "step": 6716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5420614153588514e-05,
+      "loss": 0.981,
+      "step": 6717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5419244075106815e-05,
+      "loss": 0.4616,
+      "step": 6718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54178738525878e-05,
+      "loss": 0.9259,
+      "step": 6719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5416503486067887e-05,
+      "loss": 1.0673,
+      "step": 6720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415132975583494e-05,
+      "loss": 0.9672,
+      "step": 6721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5413762321171052e-05,
+      "loss": 0.9056,
+      "step": 6722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541239152286699e-05,
+      "loss": 0.8895,
+      "step": 6723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411020580707743e-05,
+      "loss": 0.9667,
+      "step": 6724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540964949472975e-05,
+      "loss": 1.0218,
+      "step": 6725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408278264969452e-05,
+      "loss": 0.9654,
+      "step": 6726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5406906891463294e-05,
+      "loss": 0.9813,
+      "step": 6727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5405535374247724e-05,
+      "loss": 0.9758,
+      "step": 6728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5404163713359198e-05,
+      "loss": 0.897,
+      "step": 6729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5402791908834175e-05,
+      "loss": 1.0454,
+      "step": 6730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5401419960709116e-05,
+      "loss": 0.9815,
+      "step": 6731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540004786902048e-05,
+      "loss": 1.1499,
+      "step": 6732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539867563380474e-05,
+      "loss": 0.9642,
+      "step": 6733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5397303255098366e-05,
+      "loss": 1.0418,
+      "step": 6734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5395930732937834e-05,
+      "loss": 1.0114,
+      "step": 6735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539455806735963e-05,
+      "loss": 0.9259,
+      "step": 6736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5393185258400234e-05,
+      "loss": 0.9133,
+      "step": 6737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539181230609613e-05,
+      "loss": 0.9796,
+      "step": 6738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390439210483816e-05,
+      "loss": 0.9866,
+      "step": 6739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538906597159978e-05,
+      "loss": 1.0276,
+      "step": 6740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538769258948053e-05,
+      "loss": 0.8795,
+      "step": 6741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386319064162563e-05,
+      "loss": 1.0763,
+      "step": 6742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5384945395682387e-05,
+      "loss": 0.9624,
+      "step": 6743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5383571584076512e-05,
+      "loss": 0.9868,
+      "step": 6744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5382197629381453e-05,
+      "loss": 0.9722,
+      "step": 6745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5380823531633727e-05,
+      "loss": 0.9679,
+      "step": 6746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537944929086986e-05,
+      "loss": 1.0036,
+      "step": 6747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5378074907126374e-05,
+      "loss": 0.9846,
+      "step": 6748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5376700380439795e-05,
+      "loss": 0.9833,
+      "step": 6749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5375325710846668e-05,
+      "loss": 0.9365,
+      "step": 6750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373950898383515e-05,
+      "loss": 0.9768,
+      "step": 6751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537257594308689e-05,
+      "loss": 1.0194,
+      "step": 6752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371200844993332e-05,
+      "loss": 0.9662,
+      "step": 6753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536982560413939e-05,
+      "loss": 0.9769,
+      "step": 6754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368450220561616e-05,
+      "loss": 1.0711,
+      "step": 6755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5367074694296565e-05,
+      "loss": 1.0459,
+      "step": 6756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5365699025380795e-05,
+      "loss": 0.8381,
+      "step": 6757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536432321385088e-05,
+      "loss": 0.9852,
+      "step": 6758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536294725974338e-05,
+      "loss": 1.0071,
+      "step": 6759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361571163094865e-05,
+      "loss": 1.0164,
+      "step": 6760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5360194923941912e-05,
+      "loss": 0.8791,
+      "step": 6761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53588185423211e-05,
+      "loss": 1.0525,
+      "step": 6762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5357442018269008e-05,
+      "loss": 0.86,
+      "step": 6763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535606535182223e-05,
+      "loss": 1.0465,
+      "step": 6764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535468854301735e-05,
+      "loss": 1.0473,
+      "step": 6765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353311591890968e-05,
+      "loss": 0.9793,
+      "step": 6766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5351934498479676e-05,
+      "loss": 1.0181,
+      "step": 6767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350557262820077e-05,
+      "loss": 0.9182,
+      "step": 6768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349179884948776e-05,
+      "loss": 0.9722,
+      "step": 6769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5347802364902384e-05,
+      "loss": 0.9005,
+      "step": 6770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534642470271751e-05,
+      "loss": 0.9434,
+      "step": 6771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345046898430783e-05,
+      "loss": 0.9509,
+      "step": 6772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5343668952078805e-05,
+      "loss": 0.985,
+      "step": 6773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5342290863698215e-05,
+      "loss": 0.9791,
+      "step": 6774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5340912633325627e-05,
+      "loss": 0.9474,
+      "step": 6775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533953426099769e-05,
+      "loss": 0.4495,
+      "step": 6776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5338155746751028e-05,
+      "loss": 0.9707,
+      "step": 6777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533677709062228e-05,
+      "loss": 0.8505,
+      "step": 6778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5335398292648094e-05,
+      "loss": 0.9342,
+      "step": 6779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5334019352865115e-05,
+      "loss": 0.9654,
+      "step": 6780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533264027130999e-05,
+      "loss": 1.0305,
+      "step": 6781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5331261048019385e-05,
+      "loss": 0.4138,
+      "step": 6782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5329881683029942e-05,
+      "loss": 0.9256,
+      "step": 6783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532850217637834e-05,
+      "loss": 1.0122,
+      "step": 6784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532712252810123e-05,
+      "loss": 1.0422,
+      "step": 6785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5325742738235288e-05,
+      "loss": 0.9486,
+      "step": 6786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5324362806817186e-05,
+      "loss": 0.9679,
+      "step": 6787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5322982733883602e-05,
+      "loss": 0.9308,
+      "step": 6788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532160251947122e-05,
+      "loss": 0.9378,
+      "step": 6789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532022216361672e-05,
+      "loss": 0.9161,
+      "step": 6790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5318841666356786e-05,
+      "loss": 0.9384,
+      "step": 6791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531746102772812e-05,
+      "loss": 0.9899,
+      "step": 6792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5316080247767412e-05,
+      "loss": 0.9655,
+      "step": 6793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531469932651136e-05,
+      "loss": 1.0088,
+      "step": 6794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313318263996673e-05,
+      "loss": 0.9439,
+      "step": 6795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531193706026005e-05,
+      "loss": 0.4357,
+      "step": 6796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5310555715338212e-05,
+      "loss": 0.8945,
+      "step": 6797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309174229267866e-05,
+      "loss": 0.9859,
+      "step": 6798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5307792602085734e-05,
+      "loss": 0.9585,
+      "step": 6799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5306410833828534e-05,
+      "loss": 1.0226,
+      "step": 6800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5305028924532996e-05,
+      "loss": 0.9731,
+      "step": 6801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530364687423585e-05,
+      "loss": 0.4221,
+      "step": 6802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5302264682973828e-05,
+      "loss": 0.9593,
+      "step": 6803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5300882350783664e-05,
+      "loss": 0.9294,
+      "step": 6804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52994998777021e-05,
+      "loss": 1.0405,
+      "step": 6805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298117263765885e-05,
+      "loss": 1.0203,
+      "step": 6806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5296734509011763e-05,
+      "loss": 0.9701,
+      "step": 6807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529535161347649e-05,
+      "loss": 0.9477,
+      "step": 6808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5293968577196814e-05,
+      "loss": 1.0569,
+      "step": 6809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52925854002095e-05,
+      "loss": 0.8828,
+      "step": 6810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529120208255131e-05,
+      "loss": 1.0393,
+      "step": 6811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528981862425902e-05,
+      "loss": 0.9325,
+      "step": 6812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5288435025369385e-05,
+      "loss": 0.8868,
+      "step": 6813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287051285919187e-05,
+      "loss": 0.9472,
+      "step": 6814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5285667405945207e-05,
+      "loss": 1.0176,
+      "step": 6815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284283385484222e-05,
+      "loss": 1.0515,
+      "step": 6816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528289922457302e-05,
+      "loss": 0.8948,
+      "step": 6817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5281514923248393e-05,
+      "loss": 0.9695,
+      "step": 6818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528013048154713e-05,
+      "loss": 0.9378,
+      "step": 6819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5278745899506028e-05,
+      "loss": 0.9893,
+      "step": 6820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5277361177161895e-05,
+      "loss": 0.9581,
+      "step": 6821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275976314551518e-05,
+      "loss": 0.8736,
+      "step": 6822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5274591311711723e-05,
+      "loss": 0.9692,
+      "step": 6823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5273206168679315e-05,
+      "loss": 0.9618,
+      "step": 6824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5271820885491113e-05,
+      "loss": 1.0062,
+      "step": 6825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527043546218393e-05,
+      "loss": 0.9816,
+      "step": 6826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5269049898794585e-05,
+      "loss": 1.0345,
+      "step": 6827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5267664195359917e-05,
+      "loss": 1.028,
+      "step": 6828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526627835191675e-05,
+      "loss": 0.9572,
+      "step": 6829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264892368501923e-05,
+      "loss": 0.9729,
+      "step": 6830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5263506245152267e-05,
+      "loss": 0.9329,
+      "step": 6831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5262119981904627e-05,
+      "loss": 1.0335,
+      "step": 6832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526073357879585e-05,
+      "loss": 0.9395,
+      "step": 6833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525934703586278e-05,
+      "loss": 0.885,
+      "step": 6834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525796035314227e-05,
+      "loss": 0.9228,
+      "step": 6835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256573530671181e-05,
+      "loss": 0.9541,
+      "step": 6836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5255186568486374e-05,
+      "loss": 0.9364,
+      "step": 6837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5253799466624708e-05,
+      "loss": 1.0304,
+      "step": 6838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5252412225123052e-05,
+      "loss": 0.9895,
+      "step": 6839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251024844018277e-05,
+      "loss": 0.4622,
+      "step": 6840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524963732334726e-05,
+      "loss": 0.9707,
+      "step": 6841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248249663146882e-05,
+      "loss": 1.0144,
+      "step": 6842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246861863454017e-05,
+      "loss": 0.9183,
+      "step": 6843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5245473924305558e-05,
+      "loss": 0.9235,
+      "step": 6844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5244085845738394e-05,
+      "loss": 0.9759,
+      "step": 6845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5242697627789413e-05,
+      "loss": 0.9368,
+      "step": 6846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241309270495524e-05,
+      "loss": 0.9881,
+      "step": 6847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5239920773893615e-05,
+      "loss": 0.87,
+      "step": 6848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5238532138020601e-05,
+      "loss": 0.9289,
+      "step": 6849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523714336291338e-05,
+      "loss": 0.959,
+      "step": 6850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5235754448608874e-05,
+      "loss": 0.9767,
+      "step": 6851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5234365395143992e-05,
+      "loss": 0.9654,
+      "step": 6852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523297620255566e-05,
+      "loss": 0.9688,
+      "step": 6853
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5231586870880794e-05,
+      "loss": 0.8557,
+      "step": 6854
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230197400156325e-05,
+      "loss": 0.9623,
+      "step": 6855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5228807790419184e-05,
+      "loss": 0.453,
+      "step": 6856
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5227418041706302e-05,
+      "loss": 0.9817,
+      "step": 6857
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5226028154054617e-05,
+      "loss": 1.0514,
+      "step": 6858
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224638127501075e-05,
+      "loss": 0.9593,
+      "step": 6859
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5223247962082619e-05,
+      "loss": 0.8681,
+      "step": 6860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5221857657836197e-05,
+      "loss": 0.876,
+      "step": 6861
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5220467214798762e-05,
+      "loss": 0.8583,
+      "step": 6862
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521907663300727e-05,
+      "loss": 0.9812,
+      "step": 6863
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217685912498681e-05,
+      "loss": 0.965,
+      "step": 6864
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216295053309963e-05,
+      "loss": 0.98,
+      "step": 6865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.521490405547808e-05,
+      "loss": 0.903,
+      "step": 6866
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5213512919040001e-05,
+      "loss": 0.9123,
+      "step": 6867
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212121644032703e-05,
+      "loss": 0.9112,
+      "step": 6868
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210730230493163e-05,
+      "loss": 0.9485,
+      "step": 6869
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520933867845837e-05,
+      "loss": 0.9555,
+      "step": 6870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5207946987965299e-05,
+      "loss": 0.9381,
+      "step": 6871
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206555159050949e-05,
+      "loss": 0.9868,
+      "step": 6872
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205163191752306e-05,
+      "loss": 1.0113,
+      "step": 6873
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5203771086106373e-05,
+      "loss": 0.9237,
+      "step": 6874
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5202378842150145e-05,
+      "loss": 1.0285,
+      "step": 6875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520098645992063e-05,
+      "loss": 0.9344,
+      "step": 6876
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519959393945484e-05,
+      "loss": 0.9773,
+      "step": 6877
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198201280789778e-05,
+      "loss": 1.0129,
+      "step": 6878
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519680848396246e-05,
+      "loss": 0.8828,
+      "step": 6879
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195415549009913e-05,
+      "loss": 0.9298,
+      "step": 6880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519402247596915e-05,
+      "loss": 0.9305,
+      "step": 6881
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5192629264877203e-05,
+      "loss": 0.8997,
+      "step": 6882
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5191235915771101e-05,
+      "loss": 1.0278,
+      "step": 6883
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189842428687875e-05,
+      "loss": 1.0204,
+      "step": 6884
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5188448803664569e-05,
+      "loss": 0.9321,
+      "step": 6885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187055040738217e-05,
+      "loss": 0.8503,
+      "step": 6886
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5185661139945866e-05,
+      "loss": 0.9546,
+      "step": 6887
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184267101324566e-05,
+      "loss": 1.0092,
+      "step": 6888
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182872924911367e-05,
+      "loss": 0.9921,
+      "step": 6889
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5181478610743323e-05,
+      "loss": 0.9976,
+      "step": 6890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51800841588575e-05,
+      "loss": 0.8556,
+      "step": 6891
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178689569290951e-05,
+      "loss": 1.0503,
+      "step": 6892
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5177294842080752e-05,
+      "loss": 0.9677,
+      "step": 6893
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5175899977263963e-05,
+      "loss": 1.0554,
+      "step": 6894
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174504974877672e-05,
+      "loss": 0.9976,
+      "step": 6895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173109834958945e-05,
+      "loss": 0.9771,
+      "step": 6896
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171714557544866e-05,
+      "loss": 0.9267,
+      "step": 6897
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5170319142672523e-05,
+      "loss": 0.4779,
+      "step": 6898
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168923590378999e-05,
+      "loss": 0.9211,
+      "step": 6899
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516752790070139e-05,
+      "loss": 0.9169,
+      "step": 6900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166132073676796e-05,
+      "loss": 0.9104,
+      "step": 6901
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164736109342308e-05,
+      "loss": 0.9685,
+      "step": 6902
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5163340007735032e-05,
+      "loss": 0.997,
+      "step": 6903
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5161943768892076e-05,
+      "loss": 0.9323,
+      "step": 6904
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516054739285055e-05,
+      "loss": 0.9613,
+      "step": 6905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515915087964757e-05,
+      "loss": 0.9931,
+      "step": 6906
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157754229320252e-05,
+      "loss": 0.9794,
+      "step": 6907
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156357441905718e-05,
+      "loss": 0.9202,
+      "step": 6908
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5154960517441093e-05,
+      "loss": 0.8843,
+      "step": 6909
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5153563455963501e-05,
+      "loss": 0.9453,
+      "step": 6910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5152166257510078e-05,
+      "loss": 0.9443,
+      "step": 6911
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150768922117964e-05,
+      "loss": 0.9117,
+      "step": 6912
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5149371449824292e-05,
+      "loss": 1.0246,
+      "step": 6913
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514797384066621e-05,
+      "loss": 0.9558,
+      "step": 6914
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146576094680861e-05,
+      "loss": 1.0021,
+      "step": 6915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51451782119054e-05,
+      "loss": 0.9422,
+      "step": 6916
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5143780192376975e-05,
+      "loss": 0.9534,
+      "step": 6917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514238203613275e-05,
+      "loss": 0.8807,
+      "step": 6918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5140983743209882e-05,
+      "loss": 0.9811,
+      "step": 6919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513958531364554e-05,
+      "loss": 0.9229,
+      "step": 6920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5138186747476888e-05,
+      "loss": 0.8781,
+      "step": 6921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136788044741101e-05,
+      "loss": 0.9839,
+      "step": 6922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5135389205475352e-05,
+      "loss": 0.9652,
+      "step": 6923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133990229716828e-05,
+      "loss": 0.9548,
+      "step": 6924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5132591117502706e-05,
+      "loss": 0.9456,
+      "step": 6925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5131191868870172e-05,
+      "loss": 0.4661,
+      "step": 6926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5129792483856421e-05,
+      "loss": 0.9628,
+      "step": 6927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5128392962498641e-05,
+      "loss": 0.4331,
+      "step": 6928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126993304834037e-05,
+      "loss": 1.0047,
+      "step": 6929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125593510899808e-05,
+      "loss": 0.8892,
+      "step": 6930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512419358073315e-05,
+      "loss": 0.8932,
+      "step": 6931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5122793514371289e-05,
+      "loss": 0.9858,
+      "step": 6932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512139331185142e-05,
+      "loss": 0.9217,
+      "step": 6933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119992973210767e-05,
+      "loss": 0.943,
+      "step": 6934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5118592498486548e-05,
+      "loss": 0.9372,
+      "step": 6935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5117191887715987e-05,
+      "loss": 1.0333,
+      "step": 6936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5115791140936313e-05,
+      "loss": 0.974,
+      "step": 6937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114390258184748e-05,
+      "loss": 0.9442,
+      "step": 6938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5112989239498536e-05,
+      "loss": 0.9573,
+      "step": 6939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111588084914907e-05,
+      "loss": 0.9568,
+      "step": 6940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5110186794471105e-05,
+      "loss": 1.0185,
+      "step": 6941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5108785368204376e-05,
+      "loss": 0.9602,
+      "step": 6942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107383806151963e-05,
+      "loss": 0.9622,
+      "step": 6943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5105982108351125e-05,
+      "loss": 0.9942,
+      "step": 6944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5104580274839116e-05,
+      "loss": 0.9099,
+      "step": 6945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5103178305653185e-05,
+      "loss": 0.8559,
+      "step": 6946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101776200830612e-05,
+      "loss": 0.9985,
+      "step": 6947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100373960408653e-05,
+      "loss": 0.9577,
+      "step": 6948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098971584424573e-05,
+      "loss": 0.9728,
+      "step": 6949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097569072915658e-05,
+      "loss": 0.9229,
+      "step": 6950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5096166425919176e-05,
+      "loss": 1.0042,
+      "step": 6951
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5094763643472413e-05,
+      "loss": 0.9552,
+      "step": 6952
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093360725612649e-05,
+      "loss": 0.8912,
+      "step": 6953
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5091957672377176e-05,
+      "loss": 0.9855,
+      "step": 6954
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5090554483803281e-05,
+      "loss": 1.0043,
+      "step": 6955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5089151159928264e-05,
+      "loss": 0.8948,
+      "step": 6956
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508774770078942e-05,
+      "loss": 1.0142,
+      "step": 6957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086344106424054e-05,
+      "loss": 1.0282,
+      "step": 6958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5084940376869471e-05,
+      "loss": 0.8731,
+      "step": 6959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508353651216298e-05,
+      "loss": 1.0101,
+      "step": 6960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5082132512341897e-05,
+      "loss": 0.9627,
+      "step": 6961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508072837744353e-05,
+      "loss": 0.892,
+      "step": 6962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507932410750521e-05,
+      "loss": 0.9537,
+      "step": 6963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077919702564255e-05,
+      "loss": 0.9044,
+      "step": 6964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5076515162657995e-05,
+      "loss": 1.012,
+      "step": 6965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507511048782376e-05,
+      "loss": 0.9768,
+      "step": 6966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5073705678098882e-05,
+      "loss": 1.0488,
+      "step": 6967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072300733520706e-05,
+      "loss": 0.9714,
+      "step": 6968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5070895654126565e-05,
+      "loss": 0.9853,
+      "step": 6969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5069490439953812e-05,
+      "loss": 0.9213,
+      "step": 6970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068085091039792e-05,
+      "loss": 0.9334,
+      "step": 6971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5066679607421857e-05,
+      "loss": 0.9393,
+      "step": 6972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506527398913737e-05,
+      "loss": 0.9212,
+      "step": 6973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506386823622368e-05,
+      "loss": 0.8485,
+      "step": 6974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5062462348718158e-05,
+      "loss": 0.9084,
+      "step": 6975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506105632665817e-05,
+      "loss": 0.9739,
+      "step": 6976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505965017008108e-05,
+      "loss": 0.9489,
+      "step": 6977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5058243879024272e-05,
+      "loss": 0.959,
+      "step": 6978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056837453525119e-05,
+      "loss": 0.9209,
+      "step": 6979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055430893621e-05,
+      "loss": 0.9886,
+      "step": 6980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5054024199349302e-05,
+      "loss": 0.9481,
+      "step": 6981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052617370747414e-05,
+      "loss": 0.9769,
+      "step": 6982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051210407852725e-05,
+      "loss": 0.4862,
+      "step": 6983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5049803310702635e-05,
+      "loss": 1.0009,
+      "step": 6984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504839607933454e-05,
+      "loss": 0.9764,
+      "step": 6985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504698871378584e-05,
+      "loss": 0.9533,
+      "step": 6986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5045581214093946e-05,
+      "loss": 0.9601,
+      "step": 6987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044173580296267e-05,
+      "loss": 1.157,
+      "step": 6988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042765812430213e-05,
+      "loss": 0.9295,
+      "step": 6989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5041357910533205e-05,
+      "loss": 0.9577,
+      "step": 6990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5039949874642661e-05,
+      "loss": 0.957,
+      "step": 6991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038541704796004e-05,
+      "loss": 0.9443,
+      "step": 6992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5037133401030665e-05,
+      "loss": 1.0107,
+      "step": 6993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035724963384076e-05,
+      "loss": 0.9837,
+      "step": 6994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034316391893662e-05,
+      "loss": 0.9826,
+      "step": 6995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5032907686596874e-05,
+      "loss": 0.897,
+      "step": 6996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031498847531144e-05,
+      "loss": 0.8283,
+      "step": 6997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5030089874733923e-05,
+      "loss": 1.0009,
+      "step": 6998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5028680768242657e-05,
+      "loss": 0.9762,
+      "step": 6999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50272715280948e-05,
+      "loss": 0.9993,
+      "step": 7000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5025862154327806e-05,
+      "loss": 0.9886,
+      "step": 7001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5024452646979139e-05,
+      "loss": 0.4371,
+      "step": 7002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023043006086258e-05,
+      "loss": 0.8995,
+      "step": 7003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502163323168663e-05,
+      "loss": 0.9928,
+      "step": 7004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5020223323817726e-05,
+      "loss": 0.9767,
+      "step": 7005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501881328251702e-05,
+      "loss": 0.9299,
+      "step": 7006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017403107821991e-05,
+      "loss": 0.9905,
+      "step": 7007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015992799770114e-05,
+      "loss": 0.9146,
+      "step": 7008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5014582358398882e-05,
+      "loss": 0.8168,
+      "step": 7009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013171783745774e-05,
+      "loss": 0.8988,
+      "step": 7010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5011761075848283e-05,
+      "loss": 0.9514,
+      "step": 7011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5010350234743913e-05,
+      "loss": 0.969,
+      "step": 7012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008939260470153e-05,
+      "loss": 0.9192,
+      "step": 7013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5007528153064508e-05,
+      "loss": 0.8699,
+      "step": 7014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5006116912564487e-05,
+      "loss": 0.938,
+      "step": 7015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500470553900759e-05,
+      "loss": 0.9026,
+      "step": 7016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003294032431341e-05,
+      "loss": 1.0388,
+      "step": 7017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001882392873251e-05,
+      "loss": 1.0123,
+      "step": 7018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000470620370835e-05,
+      "loss": 1.0712,
+      "step": 7019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999058714961626e-05,
+      "loss": 0.9327,
+      "step": 7020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4997646676683144e-05,
+      "loss": 1.0013,
+      "step": 7021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4996234505572923e-05,
+      "loss": 0.9649,
+      "step": 7022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4994822201668488e-05,
+      "loss": 0.9996,
+      "step": 7023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993409765007389e-05,
+      "loss": 0.9492,
+      "step": 7024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991997195627162e-05,
+      "loss": 1.0416,
+      "step": 7025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499058449356535e-05,
+      "loss": 0.92,
+      "step": 7026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989171658859502e-05,
+      "loss": 0.9568,
+      "step": 7027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4987758691547169e-05,
+      "loss": 0.9296,
+      "step": 7028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986345591665906e-05,
+      "loss": 0.8994,
+      "step": 7029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984932359253272e-05,
+      "loss": 0.4835,
+      "step": 7030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4983518994346831e-05,
+      "loss": 1.0007,
+      "step": 7031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4982105496984148e-05,
+      "loss": 0.9674,
+      "step": 7032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.498069186720279e-05,
+      "loss": 0.9415,
+      "step": 7033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4979278105040328e-05,
+      "loss": 1.0058,
+      "step": 7034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4977864210534341e-05,
+      "loss": 0.9185,
+      "step": 7035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976450183722411e-05,
+      "loss": 0.9439,
+      "step": 7036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4975036024642118e-05,
+      "loss": 0.9088,
+      "step": 7037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497362173333105e-05,
+      "loss": 0.99,
+      "step": 7038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972207309826796e-05,
+      "loss": 1.0339,
+      "step": 7039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4970792754166949e-05,
+      "loss": 0.8818,
+      "step": 7040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496937806638911e-05,
+      "loss": 0.939,
+      "step": 7041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4967963246530877e-05,
+      "loss": 0.9634,
+      "step": 7042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4966548294629856e-05,
+      "loss": 0.9399,
+      "step": 7043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965133210723653e-05,
+      "loss": 0.9072,
+      "step": 7044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496371799484988e-05,
+      "loss": 0.9697,
+      "step": 7045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496230264704615e-05,
+      "loss": 0.9897,
+      "step": 7046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4960887167350082e-05,
+      "loss": 0.9659,
+      "step": 7047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4959471555799303e-05,
+      "loss": 1.0132,
+      "step": 7048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4958055812431433e-05,
+      "loss": 0.8836,
+      "step": 7049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4956639937284103e-05,
+      "loss": 0.9672,
+      "step": 7050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955223930394945e-05,
+      "loss": 0.9147,
+      "step": 7051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495380779180159e-05,
+      "loss": 0.4599,
+      "step": 7052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952391521541685e-05,
+      "loss": 1.0659,
+      "step": 7053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950975119652872e-05,
+      "loss": 0.9622,
+      "step": 7054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4949558586172788e-05,
+      "loss": 0.4193,
+      "step": 7055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4948141921139095e-05,
+      "loss": 1.0103,
+      "step": 7056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946725124589438e-05,
+      "loss": 0.9266,
+      "step": 7057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4945308196561479e-05,
+      "loss": 1.052,
+      "step": 7058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943891137092877e-05,
+      "loss": 0.9173,
+      "step": 7059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4942473946221292e-05,
+      "loss": 0.9355,
+      "step": 7060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4941056623984396e-05,
+      "loss": 0.9631,
+      "step": 7061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493963917041986e-05,
+      "loss": 0.8729,
+      "step": 7062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4938221585565353e-05,
+      "loss": 0.9349,
+      "step": 7063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4936803869458558e-05,
+      "loss": 0.9235,
+      "step": 7064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935386022137155e-05,
+      "loss": 0.9336,
+      "step": 7065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4933968043638826e-05,
+      "loss": 0.922,
+      "step": 7066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4932549934001266e-05,
+      "loss": 0.9731,
+      "step": 7067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931131693262159e-05,
+      "loss": 0.9379,
+      "step": 7068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4929713321459204e-05,
+      "loss": 0.9433,
+      "step": 7069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928294818630099e-05,
+      "loss": 0.9042,
+      "step": 7070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4926876184812544e-05,
+      "loss": 0.9443,
+      "step": 7071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925457420044252e-05,
+      "loss": 1.0418,
+      "step": 7072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924038524362924e-05,
+      "loss": 0.9798,
+      "step": 7073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922619497806276e-05,
+      "loss": 0.952,
+      "step": 7074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4921200340412025e-05,
+      "loss": 0.9392,
+      "step": 7075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919781052217888e-05,
+      "loss": 0.896,
+      "step": 7076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491836163326159e-05,
+      "loss": 0.9484,
+      "step": 7077
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4916942083580858e-05,
+      "loss": 0.9878,
+      "step": 7078
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4915522403213422e-05,
+      "loss": 0.8489,
+      "step": 7079
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4914102592197012e-05,
+      "loss": 0.9833,
+      "step": 7080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4912682650569372e-05,
+      "loss": 1.041,
+      "step": 7081
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4911262578368233e-05,
+      "loss": 0.9376,
+      "step": 7082
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909842375631345e-05,
+      "loss": 0.9592,
+      "step": 7083
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908422042396455e-05,
+      "loss": 1.0072,
+      "step": 7084
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4907001578701316e-05,
+      "loss": 1.0314,
+      "step": 7085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4905580984583675e-05,
+      "loss": 0.945,
+      "step": 7086
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4904160260081298e-05,
+      "loss": 1.0264,
+      "step": 7087
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4902739405231938e-05,
+      "loss": 0.9562,
+      "step": 7088
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4901318420073364e-05,
+      "loss": 0.9449,
+      "step": 7089
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489989730464335e-05,
+      "loss": 1.0182,
+      "step": 7090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4898476058979655e-05,
+      "loss": 0.9436,
+      "step": 7091
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897054683120065e-05,
+      "loss": 0.9559,
+      "step": 7092
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4895633177102353e-05,
+      "loss": 0.9965,
+      "step": 7093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894211540964303e-05,
+      "loss": 1.0296,
+      "step": 7094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48927897747437e-05,
+      "loss": 0.9264,
+      "step": 7095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4891367878478333e-05,
+      "loss": 0.9635,
+      "step": 7096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4889945852205995e-05,
+      "loss": 1.0217,
+      "step": 7097
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888523695964482e-05,
+      "loss": 0.4164,
+      "step": 7098
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4887101409791588e-05,
+      "loss": 0.9331,
+      "step": 7099
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885678993725124e-05,
+      "loss": 0.9242,
+      "step": 7100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4884256447802889e-05,
+      "loss": 0.9384,
+      "step": 7101
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48828337720627e-05,
+      "loss": 0.9133,
+      "step": 7102
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881410966542364e-05,
+      "loss": 0.9466,
+      "step": 7103
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4879988031279699e-05,
+      "loss": 0.9276,
+      "step": 7104
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4878564966312525e-05,
+      "loss": 0.9885,
+      "step": 7105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487714177167867e-05,
+      "loss": 0.9867,
+      "step": 7106
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4875718447415952e-05,
+      "loss": 0.9194,
+      "step": 7107
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4874294993562211e-05,
+      "loss": 0.8859,
+      "step": 7108
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4872871410155271e-05,
+      "loss": 0.9794,
+      "step": 7109
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4871447697232978e-05,
+      "loss": 0.9886,
+      "step": 7110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870023854833168e-05,
+      "loss": 1.0062,
+      "step": 7111
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4868599882993686e-05,
+      "loss": 0.9373,
+      "step": 7112
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867175781752381e-05,
+      "loss": 0.9119,
+      "step": 7113
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4865751551147098e-05,
+      "loss": 0.9472,
+      "step": 7114
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4864327191215702e-05,
+      "loss": 0.9684,
+      "step": 7115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862902701996043e-05,
+      "loss": 0.8867,
+      "step": 7116
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486147808352598e-05,
+      "loss": 0.8925,
+      "step": 7117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860053335843385e-05,
+      "loss": 0.9895,
+      "step": 7118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4858628458986124e-05,
+      "loss": 0.9274,
+      "step": 7119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4857203452992065e-05,
+      "loss": 0.9666,
+      "step": 7120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485577831789909e-05,
+      "loss": 0.9573,
+      "step": 7121
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485435305374507e-05,
+      "loss": 0.9819,
+      "step": 7122
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852927660567889e-05,
+      "loss": 0.8991,
+      "step": 7123
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4851502138405436e-05,
+      "loss": 1.0491,
+      "step": 7124
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4850076487295594e-05,
+      "loss": 0.9619,
+      "step": 7125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848650707276262e-05,
+      "loss": 0.9514,
+      "step": 7126
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4847224798385329e-05,
+      "loss": 0.8977,
+      "step": 7127
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4845798760660702e-05,
+      "loss": 0.9682,
+      "step": 7128
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844372594140271e-05,
+      "loss": 0.4596,
+      "step": 7129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484294629886195e-05,
+      "loss": 0.9667,
+      "step": 7130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841519874863655e-05,
+      "loss": 1.0003,
+      "step": 7131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4840093322183284e-05,
+      "loss": 0.8765,
+      "step": 7132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4838666640858766e-05,
+      "loss": 0.9481,
+      "step": 7133
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483723983092801e-05,
+      "loss": 0.9384,
+      "step": 7134
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835812892428945e-05,
+      "loss": 0.9368,
+      "step": 7135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4834385825399498e-05,
+      "loss": 1.0019,
+      "step": 7136
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832958629877596e-05,
+      "loss": 0.9922,
+      "step": 7137
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4831531305901172e-05,
+      "loss": 0.9179,
+      "step": 7138
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4830103853508169e-05,
+      "loss": 0.9052,
+      "step": 7139
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4828676272736517e-05,
+      "loss": 0.9666,
+      "step": 7140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827248563624165e-05,
+      "loss": 0.9145,
+      "step": 7141
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825820726209062e-05,
+      "loss": 0.4422,
+      "step": 7142
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4824392760529153e-05,
+      "loss": 0.9529,
+      "step": 7143
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4822964666622399e-05,
+      "loss": 0.9495,
+      "step": 7144
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4821536444526747e-05,
+      "loss": 1.06,
+      "step": 7145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4820108094280169e-05,
+      "loss": 0.8977,
+      "step": 7146
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818679615920618e-05,
+      "loss": 1.0572,
+      "step": 7147
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4817251009486065e-05,
+      "loss": 0.9793,
+      "step": 7148
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4815822275014486e-05,
+      "loss": 0.9884,
+      "step": 7149
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4814393412543849e-05,
+      "loss": 0.8573,
+      "step": 7150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4812964422112136e-05,
+      "loss": 0.8843,
+      "step": 7151
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4811535303757326e-05,
+      "loss": 0.9544,
+      "step": 7152
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4810106057517397e-05,
+      "loss": 1.035,
+      "step": 7153
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4808676683430348e-05,
+      "loss": 0.9799,
+      "step": 7154
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4807247181534163e-05,
+      "loss": 1.0074,
+      "step": 7155
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4805817551866839e-05,
+      "loss": 0.917,
+      "step": 7156
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804387794466375e-05,
+      "loss": 0.9348,
+      "step": 7157
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802957909370768e-05,
+      "loss": 0.9323,
+      "step": 7158
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4801527896618025e-05,
+      "loss": 0.9763,
+      "step": 7159
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800097756246154e-05,
+      "loss": 0.9176,
+      "step": 7160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4798667488293168e-05,
+      "loss": 0.9424,
+      "step": 7161
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797237092797083e-05,
+      "loss": 0.9486,
+      "step": 7162
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4795806569795912e-05,
+      "loss": 0.9875,
+      "step": 7163
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4794375919327679e-05,
+      "loss": 1.0012,
+      "step": 7164
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4792945141430413e-05,
+      "loss": 1.0193,
+      "step": 7165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791514236142136e-05,
+      "loss": 0.9425,
+      "step": 7166
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4790083203500886e-05,
+      "loss": 0.8775,
+      "step": 7167
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4788652043544693e-05,
+      "loss": 0.8634,
+      "step": 7168
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.47872207563116e-05,
+      "loss": 0.9219,
+      "step": 7169
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785789341839644e-05,
+      "loss": 0.9455,
+      "step": 7170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4784357800166874e-05,
+      "loss": 1.0182,
+      "step": 7171
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4782926131331339e-05,
+      "loss": 0.9216,
+      "step": 7172
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4781494335371092e-05,
+      "loss": 0.9575,
+      "step": 7173
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780062412324184e-05,
+      "loss": 0.9745,
+      "step": 7174
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778630362228676e-05,
+      "loss": 1.028,
+      "step": 7175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477719818512263e-05,
+      "loss": 1.0189,
+      "step": 7176
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4775765881044115e-05,
+      "loss": 0.9669,
+      "step": 7177
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4774333450031196e-05,
+      "loss": 0.974,
+      "step": 7178
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4772900892121946e-05,
+      "loss": 0.921,
+      "step": 7179
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4771468207354443e-05,
+      "loss": 0.898,
+      "step": 7180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770035395766761e-05,
+      "loss": 0.9712,
+      "step": 7181
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476860245739699e-05,
+      "loss": 0.9387,
+      "step": 7182
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4767169392283211e-05,
+      "loss": 0.9728,
+      "step": 7183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4765736200463513e-05,
+      "loss": 0.8903,
+      "step": 7184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764302881975993e-05,
+      "loss": 1.0205,
+      "step": 7185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762869436858743e-05,
+      "loss": 0.9935,
+      "step": 7186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476143586514986e-05,
+      "loss": 0.9682,
+      "step": 7187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476000216688745e-05,
+      "loss": 0.994,
+      "step": 7188
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475856834210962e-05,
+      "loss": 0.9091,
+      "step": 7189
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475713439085448e-05,
+      "loss": 0.8181,
+      "step": 7190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755700313160139e-05,
+      "loss": 0.9349,
+      "step": 7191
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4754266109064717e-05,
+      "loss": 1.0083,
+      "step": 7192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752831778606331e-05,
+      "loss": 1.0188,
+      "step": 7193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4751397321823102e-05,
+      "loss": 0.9459,
+      "step": 7194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474996273875316e-05,
+      "loss": 0.909,
+      "step": 7195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4748528029434637e-05,
+      "loss": 1.0351,
+      "step": 7196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4747093193905658e-05,
+      "loss": 0.9869,
+      "step": 7197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4745658232204366e-05,
+      "loss": 0.8986,
+      "step": 7198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4744223144368897e-05,
+      "loss": 0.9808,
+      "step": 7199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4742787930437395e-05,
+      "loss": 0.9153,
+      "step": 7200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4741352590448009e-05,
+      "loss": 0.9256,
+      "step": 7201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739917124438883e-05,
+      "loss": 0.8654,
+      "step": 7202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4738481532448174e-05,
+      "loss": 0.9574,
+      "step": 7203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4737045814514042e-05,
+      "loss": 1.0041,
+      "step": 7204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4735609970674638e-05,
+      "loss": 0.9872,
+      "step": 7205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734174000968129e-05,
+      "loss": 0.9349,
+      "step": 7206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732737905432686e-05,
+      "loss": 0.8919,
+      "step": 7207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473130168410647e-05,
+      "loss": 0.9228,
+      "step": 7208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4729865337027663e-05,
+      "loss": 0.9478,
+      "step": 7209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728428864234434e-05,
+      "loss": 0.9178,
+      "step": 7210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472699226576497e-05,
+      "loss": 0.9161,
+      "step": 7211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4725555541657446e-05,
+      "loss": 0.9533,
+      "step": 7212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4724118691950055e-05,
+      "loss": 0.9269,
+      "step": 7213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722681716680987e-05,
+      "loss": 0.9646,
+      "step": 7214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4721244615888428e-05,
+      "loss": 0.4151,
+      "step": 7215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4719807389610583e-05,
+      "loss": 1.0001,
+      "step": 7216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718370037885648e-05,
+      "loss": 0.9193,
+      "step": 7217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4716932560751822e-05,
+      "loss": 1.0289,
+      "step": 7218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471549495824732e-05,
+      "loss": 0.9397,
+      "step": 7219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4714057230410345e-05,
+      "loss": 0.9502,
+      "step": 7220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4712619377279116e-05,
+      "loss": 0.9378,
+      "step": 7221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4711181398891846e-05,
+      "loss": 0.927,
+      "step": 7222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709743295286751e-05,
+      "loss": 0.8919,
+      "step": 7223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4708305066502062e-05,
+      "loss": 0.9537,
+      "step": 7224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4706866712576e-05,
+      "loss": 0.9908,
+      "step": 7225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4705428233546797e-05,
+      "loss": 0.9635,
+      "step": 7226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703989629452689e-05,
+      "loss": 0.4991,
+      "step": 7227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4702550900331903e-05,
+      "loss": 0.9402,
+      "step": 7228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701112046222688e-05,
+      "loss": 1.01,
+      "step": 7229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4699673067163284e-05,
+      "loss": 0.9767,
+      "step": 7230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4698233963191937e-05,
+      "loss": 0.8901,
+      "step": 7231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4696794734346897e-05,
+      "loss": 0.9933,
+      "step": 7232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4695355380666419e-05,
+      "loss": 0.8831,
+      "step": 7233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4693915902188756e-05,
+      "loss": 0.9686,
+      "step": 7234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4692476298952169e-05,
+      "loss": 0.9501,
+      "step": 7235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469103657099492e-05,
+      "loss": 0.967,
+      "step": 7236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689596718355281e-05,
+      "loss": 1.0098,
+      "step": 7237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688156741071513e-05,
+      "loss": 0.9847,
+      "step": 7238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4686716639181895e-05,
+      "loss": 0.9324,
+      "step": 7239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4685276412724702e-05,
+      "loss": 0.9834,
+      "step": 7240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4683836061738212e-05,
+      "loss": 0.9248,
+      "step": 7241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468239558626071e-05,
+      "loss": 0.9401,
+      "step": 7242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680954986330483e-05,
+      "loss": 0.931,
+      "step": 7243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4679514261985812e-05,
+      "loss": 0.9479,
+      "step": 7244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4678073413265004e-05,
+      "loss": 0.9593,
+      "step": 7245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676632440206342e-05,
+      "loss": 0.9858,
+      "step": 7246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4675191342848132e-05,
+      "loss": 0.9262,
+      "step": 7247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4673750121228678e-05,
+      "loss": 0.9883,
+      "step": 7248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4672308775386281e-05,
+      "loss": 0.9731,
+      "step": 7249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670867305359257e-05,
+      "loss": 0.933,
+      "step": 7250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4669425711185912e-05,
+      "loss": 0.875,
+      "step": 7251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667983992904565e-05,
+      "loss": 1.0377,
+      "step": 7252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4666542150553533e-05,
+      "loss": 0.9679,
+      "step": 7253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4665100184171142e-05,
+      "loss": 1.0111,
+      "step": 7254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663658093795716e-05,
+      "loss": 1.0159,
+      "step": 7255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4662215879465587e-05,
+      "loss": 1.1273,
+      "step": 7256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4660773541219079e-05,
+      "loss": 0.9548,
+      "step": 7257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465933107909454e-05,
+      "loss": 0.9587,
+      "step": 7258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4657888493130294e-05,
+      "loss": 1.0031,
+      "step": 7259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4656445783364697e-05,
+      "loss": 0.9391,
+      "step": 7260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655002949836091e-05,
+      "loss": 1.0282,
+      "step": 7261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653559992582819e-05,
+      "loss": 0.9062,
+      "step": 7262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4652116911643242e-05,
+      "loss": 1.0215,
+      "step": 7263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4650673707055707e-05,
+      "loss": 0.9602,
+      "step": 7264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649230378858575e-05,
+      "loss": 0.9346,
+      "step": 7265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647786927090213e-05,
+      "loss": 0.959,
+      "step": 7266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4646343351788982e-05,
+      "loss": 0.8454,
+      "step": 7267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464489965299325e-05,
+      "loss": 0.9831,
+      "step": 7268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4643455830741392e-05,
+      "loss": 0.9904,
+      "step": 7269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464201188507178e-05,
+      "loss": 0.9566,
+      "step": 7270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640567816022793e-05,
+      "loss": 0.4889,
+      "step": 7271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4639123623632813e-05,
+      "loss": 0.9913,
+      "step": 7272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637679307940229e-05,
+      "loss": 1.02,
+      "step": 7273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636234868983425e-05,
+      "loss": 0.9923,
+      "step": 7274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463479030680079e-05,
+      "loss": 0.9606,
+      "step": 7275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4633345621430728e-05,
+      "loss": 0.8795,
+      "step": 7276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4631900812911626e-05,
+      "loss": 0.4802,
+      "step": 7277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630455881281891e-05,
+      "loss": 0.8774,
+      "step": 7278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4629010826579928e-05,
+      "loss": 0.9112,
+      "step": 7279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4627565648844144e-05,
+      "loss": 0.9547,
+      "step": 7280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462612034811295e-05,
+      "loss": 1.0242,
+      "step": 7281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462467492442476e-05,
+      "loss": 0.8969,
+      "step": 7282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623229377817994e-05,
+      "loss": 1.0377,
+      "step": 7283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4621783708331071e-05,
+      "loss": 0.9498,
+      "step": 7284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4620337916002414e-05,
+      "loss": 0.9308,
+      "step": 7285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4618892000870455e-05,
+      "loss": 0.923,
+      "step": 7286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4617445962973623e-05,
+      "loss": 0.8903,
+      "step": 7287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4615999802350344e-05,
+      "loss": 1.0011,
+      "step": 7288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461455351903907e-05,
+      "loss": 0.449,
+      "step": 7289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461310711307823e-05,
+      "loss": 1.0152,
+      "step": 7290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611660584506271e-05,
+      "loss": 0.971,
+      "step": 7291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4610213933361644e-05,
+      "loss": 0.9977,
+      "step": 7292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4608767159682793e-05,
+      "loss": 0.9963,
+      "step": 7293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607320263508173e-05,
+      "loss": 0.8998,
+      "step": 7294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605873244876246e-05,
+      "loss": 1.0111,
+      "step": 7295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4604426103825466e-05,
+      "loss": 1.0081,
+      "step": 7296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4602978840394302e-05,
+      "loss": 1.0018,
+      "step": 7297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601531454621216e-05,
+      "loss": 1.0545,
+      "step": 7298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600083946544678e-05,
+      "loss": 0.9258,
+      "step": 7299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4598636316203164e-05,
+      "loss": 0.8632,
+      "step": 7300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4597188563635148e-05,
+      "loss": 0.8558,
+      "step": 7301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595740688879112e-05,
+      "loss": 0.9513,
+      "step": 7302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4594292691973536e-05,
+      "loss": 0.8974,
+      "step": 7303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592844572956908e-05,
+      "loss": 0.9394,
+      "step": 7304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4591396331867718e-05,
+      "loss": 0.4773,
+      "step": 7305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4589947968744454e-05,
+      "loss": 0.9453,
+      "step": 7306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588499483625618e-05,
+      "loss": 0.9744,
+      "step": 7307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4587050876549702e-05,
+      "loss": 0.9564,
+      "step": 7308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4585602147555218e-05,
+      "loss": 0.9658,
+      "step": 7309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4584153296680666e-05,
+      "loss": 0.9725,
+      "step": 7310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4582704323964552e-05,
+      "loss": 0.9066,
+      "step": 7311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4581255229445392e-05,
+      "loss": 1.0064,
+      "step": 7312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4579806013161701e-05,
+      "loss": 0.9655,
+      "step": 7313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578356675151994e-05,
+      "loss": 0.9236,
+      "step": 7314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45769072154548e-05,
+      "loss": 0.9219,
+      "step": 7315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4575457634108636e-05,
+      "loss": 0.974,
+      "step": 7316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574007931152037e-05,
+      "loss": 0.9713,
+      "step": 7317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4572558106623528e-05,
+      "loss": 0.9503,
+      "step": 7318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571108160561648e-05,
+      "loss": 1.0028,
+      "step": 7319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569658093004935e-05,
+      "loss": 0.9785,
+      "step": 7320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568207903991927e-05,
+      "loss": 0.9172,
+      "step": 7321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566757593561174e-05,
+      "loss": 0.9048,
+      "step": 7322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4565307161751216e-05,
+      "loss": 0.9105,
+      "step": 7323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4563856608600605e-05,
+      "loss": 0.9342,
+      "step": 7324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4562405934147906e-05,
+      "loss": 0.9847,
+      "step": 7325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4560955138431662e-05,
+      "loss": 0.9357,
+      "step": 7326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4559504221490441e-05,
+      "loss": 0.9249,
+      "step": 7327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558053183362808e-05,
+      "loss": 0.8847,
+      "step": 7328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4556602024087323e-05,
+      "loss": 0.9589,
+      "step": 7329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555150743702562e-05,
+      "loss": 0.9338,
+      "step": 7330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553699342247098e-05,
+      "loss": 0.9916,
+      "step": 7331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4552247819759504e-05,
+      "loss": 0.8663,
+      "step": 7332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4550796176278363e-05,
+      "loss": 0.8669,
+      "step": 7333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549344411842257e-05,
+      "loss": 1.0113,
+      "step": 7334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4547892526489773e-05,
+      "loss": 0.4359,
+      "step": 7335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45464405202595e-05,
+      "loss": 0.9736,
+      "step": 7336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4544988393190029e-05,
+      "loss": 1.0348,
+      "step": 7337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543536145319958e-05,
+      "loss": 0.9283,
+      "step": 7338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4542083776687887e-05,
+      "loss": 0.9256,
+      "step": 7339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540631287332416e-05,
+      "loss": 0.9198,
+      "step": 7340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4539178677292152e-05,
+      "loss": 1.0063,
+      "step": 7341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4537725946605698e-05,
+      "loss": 0.9682,
+      "step": 7342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536273095311677e-05,
+      "loss": 0.9095,
+      "step": 7343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45348201234487e-05,
+      "loss": 0.9796,
+      "step": 7344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533367031055377e-05,
+      "loss": 0.8967,
+      "step": 7345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4531913818170338e-05,
+      "loss": 0.9838,
+      "step": 7346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4530460484832203e-05,
+      "loss": 0.9745,
+      "step": 7347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4529007031079607e-05,
+      "loss": 0.9651,
+      "step": 7348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4527553456951176e-05,
+      "loss": 0.9502,
+      "step": 7349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526099762485544e-05,
+      "loss": 0.8797,
+      "step": 7350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4524645947721349e-05,
+      "loss": 0.9247,
+      "step": 7351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523192012697234e-05,
+      "loss": 0.9485,
+      "step": 7352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4521737957451838e-05,
+      "loss": 0.9433,
+      "step": 7353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4520283782023816e-05,
+      "loss": 0.9098,
+      "step": 7354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4518829486451811e-05,
+      "loss": 1.0033,
+      "step": 7355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4517375070774483e-05,
+      "loss": 0.9413,
+      "step": 7356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451592053503048e-05,
+      "loss": 0.9208,
+      "step": 7357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514465879258469e-05,
+      "loss": 0.9649,
+      "step": 7358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451301110349711e-05,
+      "loss": 0.9411,
+      "step": 7359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451155620778507e-05,
+      "loss": 0.9165,
+      "step": 7360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451010119216102e-05,
+      "loss": 0.9145,
+      "step": 7361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450864605666363e-05,
+      "loss": 0.8975,
+      "step": 7362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450719080133158e-05,
+      "loss": 1.0125,
+      "step": 7363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505735426203545e-05,
+      "loss": 1.0124,
+      "step": 7364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4504279931318206e-05,
+      "loss": 0.9144,
+      "step": 7365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4502824316714252e-05,
+      "loss": 1.0096,
+      "step": 7366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4501368582430374e-05,
+      "loss": 0.8966,
+      "step": 7367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4499912728505255e-05,
+      "loss": 0.9436,
+      "step": 7368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4498456754977596e-05,
+      "loss": 0.9833,
+      "step": 7369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497000661886096e-05,
+      "loss": 0.8457,
+      "step": 7370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495544449269452e-05,
+      "loss": 0.9419,
+      "step": 7371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4494088117166374e-05,
+      "loss": 0.9336,
+      "step": 7372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4492631665615563e-05,
+      "loss": 0.8785,
+      "step": 7373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491175094655738e-05,
+      "loss": 0.9412,
+      "step": 7374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4489718404325608e-05,
+      "loss": 0.9834,
+      "step": 7375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488261594663887e-05,
+      "loss": 0.4615,
+      "step": 7376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44868046657093e-05,
+      "loss": 1.0189,
+      "step": 7377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448534761750057e-05,
+      "loss": 1.0049,
+      "step": 7378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483890450076423e-05,
+      "loss": 0.8448,
+      "step": 7379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4482433163475591e-05,
+      "loss": 0.9284,
+      "step": 7380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4480975757736802e-05,
+      "loss": 0.9891,
+      "step": 7381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479518232898797e-05,
+      "loss": 0.9116,
+      "step": 7382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4478060589000311e-05,
+      "loss": 0.9561,
+      "step": 7383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4476602826080088e-05,
+      "loss": 0.9896,
+      "step": 7384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447514494417688e-05,
+      "loss": 0.9156,
+      "step": 7385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4473686943329426e-05,
+      "loss": 0.8853,
+      "step": 7386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4472228823576484e-05,
+      "loss": 0.9624,
+      "step": 7387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470770584956808e-05,
+      "loss": 0.8946,
+      "step": 7388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469312227509153e-05,
+      "loss": 0.9546,
+      "step": 7389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446785375127229e-05,
+      "loss": 0.9068,
+      "step": 7390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446639515628497e-05,
+      "loss": 1.0272,
+      "step": 7391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464936442585973e-05,
+      "loss": 1.0294,
+      "step": 7392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4463477610214065e-05,
+      "loss": 0.9378,
+      "step": 7393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462018659208017e-05,
+      "loss": 0.9832,
+      "step": 7394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4460559589606609e-05,
+      "loss": 0.9216,
+      "step": 7395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4459100401448621e-05,
+      "loss": 0.9568,
+      "step": 7396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457641094772838e-05,
+      "loss": 1.0039,
+      "step": 7397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445618166961805e-05,
+      "loss": 0.894,
+      "step": 7398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454722126023038e-05,
+      "loss": 0.9554,
+      "step": 7399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44532624640266e-05,
+      "loss": 0.9491,
+      "step": 7400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4451802683667532e-05,
+      "loss": 0.9093,
+      "step": 7401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450342784984632e-05,
+      "loss": 0.488,
+      "step": 7402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4448882768016707e-05,
+      "loss": 0.9505,
+      "step": 7403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4447422632802557e-05,
+      "loss": 0.9454,
+      "step": 7404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445962379380993e-05,
+      "loss": 0.9378,
+      "step": 7405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4444502007790829e-05,
+      "loss": 0.9113,
+      "step": 7406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4443041518070873e-05,
+      "loss": 0.9362,
+      "step": 7407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441580910259951e-05,
+      "loss": 1.0232,
+      "step": 7408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444012018439688e-05,
+      "loss": 0.4875,
+      "step": 7409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443865934052049e-05,
+      "loss": 0.9196,
+      "step": 7410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4437198378669598e-05,
+      "loss": 0.9264,
+      "step": 7411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4435737298883044e-05,
+      "loss": 0.9592,
+      "step": 7412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4434276101199657e-05,
+      "loss": 0.9192,
+      "step": 7413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4432814785658278e-05,
+      "loss": 0.9252,
+      "step": 7414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431353352297744e-05,
+      "loss": 1.0155,
+      "step": 7415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4429891801156898e-05,
+      "loss": 0.91,
+      "step": 7416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4428430132274587e-05,
+      "loss": 0.842,
+      "step": 7417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442696834568966e-05,
+      "loss": 0.9833,
+      "step": 7418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4425506441440976e-05,
+      "loss": 0.9809,
+      "step": 7419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424044419567377e-05,
+      "loss": 0.4771,
+      "step": 7420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4422582280107738e-05,
+      "loss": 0.9721,
+      "step": 7421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4421120023100907e-05,
+      "loss": 0.9719,
+      "step": 7422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419657648585758e-05,
+      "loss": 0.9213,
+      "step": 7423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4418195156601155e-05,
+      "loss": 0.8911,
+      "step": 7424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4416732547185972e-05,
+      "loss": 0.9909,
+      "step": 7425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4415269820379083e-05,
+      "loss": 0.9284,
+      "step": 7426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4413806976219364e-05,
+      "loss": 0.9544,
+      "step": 7427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4412344014745698e-05,
+      "loss": 1.0154,
+      "step": 7428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410880935996967e-05,
+      "loss": 0.9284,
+      "step": 7429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409417740012055e-05,
+      "loss": 0.9294,
+      "step": 7430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440795442682986e-05,
+      "loss": 0.9737,
+      "step": 7431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4406490996489269e-05,
+      "loss": 0.9063,
+      "step": 7432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4405027449029178e-05,
+      "loss": 0.9828,
+      "step": 7433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4403563784488493e-05,
+      "loss": 0.9424,
+      "step": 7434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402100002906108e-05,
+      "loss": 0.9228,
+      "step": 7435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400636104320933e-05,
+      "loss": 0.9325,
+      "step": 7436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4399172088771878e-05,
+      "loss": 0.44,
+      "step": 7437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4397707956297851e-05,
+      "loss": 0.976,
+      "step": 7438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439624370693777e-05,
+      "loss": 0.9105,
+      "step": 7439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4394779340730554e-05,
+      "loss": 0.9553,
+      "step": 7440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4393314857715122e-05,
+      "loss": 0.9215,
+      "step": 7441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4391850257930396e-05,
+      "loss": 0.9502,
+      "step": 7442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4390385541415309e-05,
+      "loss": 0.4197,
+      "step": 7443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4388920708208787e-05,
+      "loss": 1.0099,
+      "step": 7444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4387455758349766e-05,
+      "loss": 0.9517,
+      "step": 7445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385990691877182e-05,
+      "loss": 0.9154,
+      "step": 7446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4384525508829977e-05,
+      "loss": 0.9048,
+      "step": 7447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4383060209247086e-05,
+      "loss": 0.9835,
+      "step": 7448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4381594793167464e-05,
+      "loss": 0.9282,
+      "step": 7449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4380129260630057e-05,
+      "loss": 0.9811,
+      "step": 7450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378663611673815e-05,
+      "loss": 0.9655,
+      "step": 7451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4377197846337698e-05,
+      "loss": 0.9361,
+      "step": 7452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375731964660657e-05,
+      "loss": 0.8978,
+      "step": 7453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374265966681662e-05,
+      "loss": 0.9774,
+      "step": 7454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4372799852439672e-05,
+      "loss": 0.9374,
+      "step": 7455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371333621973655e-05,
+      "loss": 0.9193,
+      "step": 7456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369867275322587e-05,
+      "loss": 0.9598,
+      "step": 7457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4368400812525434e-05,
+      "loss": 0.9391,
+      "step": 7458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4366934233621176e-05,
+      "loss": 0.9554,
+      "step": 7459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4365467538648796e-05,
+      "loss": 0.9641,
+      "step": 7460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364000727647272e-05,
+      "loss": 0.4515,
+      "step": 7461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4362533800655597e-05,
+      "loss": 0.9564,
+      "step": 7462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361066757712755e-05,
+      "loss": 0.8308,
+      "step": 7463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359599598857736e-05,
+      "loss": 0.9988,
+      "step": 7464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435813232412954e-05,
+      "loss": 1.0088,
+      "step": 7465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4356664933567164e-05,
+      "loss": 0.9548,
+      "step": 7466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4355197427209611e-05,
+      "loss": 0.9719,
+      "step": 7467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4353729805095886e-05,
+      "loss": 0.935,
+      "step": 7468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4352262067264993e-05,
+      "loss": 0.9214,
+      "step": 7469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4350794213755947e-05,
+      "loss": 0.4819,
+      "step": 7470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349326244607758e-05,
+      "loss": 0.9272,
+      "step": 7471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347858159859444e-05,
+      "loss": 0.9303,
+      "step": 7472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434638995955003e-05,
+      "loss": 1.0113,
+      "step": 7473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434492164371853e-05,
+      "loss": 0.9828,
+      "step": 7474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343453212403976e-05,
+      "loss": 0.9631,
+      "step": 7475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4341984665645399e-05,
+      "loss": 0.948,
+      "step": 7476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4340516003481825e-05,
+      "loss": 0.9054,
+      "step": 7477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4339047225952296e-05,
+      "loss": 0.9402,
+      "step": 7478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4337578333095847e-05,
+      "loss": 1.0066,
+      "step": 7479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433610932495152e-05,
+      "loss": 0.8917,
+      "step": 7480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433464020155836e-05,
+      "loss": 0.9345,
+      "step": 7481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333170962955411e-05,
+      "loss": 0.9764,
+      "step": 7482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433170160918173e-05,
+      "loss": 0.9986,
+      "step": 7483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330232140276365e-05,
+      "loss": 0.9887,
+      "step": 7484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328762556278377e-05,
+      "loss": 0.8637,
+      "step": 7485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4327292857226825e-05,
+      "loss": 0.9302,
+      "step": 7486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432582304316077e-05,
+      "loss": 0.9971,
+      "step": 7487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324353114119278e-05,
+      "loss": 0.9862,
+      "step": 7488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432288307014142e-05,
+      "loss": 0.9479,
+      "step": 7489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321412911266265e-05,
+      "loss": 0.9371,
+      "step": 7490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4319942637532895e-05,
+      "loss": 1.0342,
+      "step": 7491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431847224898038e-05,
+      "loss": 0.9256,
+      "step": 7492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4317001745647808e-05,
+      "loss": 1.0038,
+      "step": 7493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4315531127574258e-05,
+      "loss": 0.9707,
+      "step": 7494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4314060394798817e-05,
+      "loss": 0.8603,
+      "step": 7495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431258954736058e-05,
+      "loss": 0.9862,
+      "step": 7496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4311118585298638e-05,
+      "loss": 1.0055,
+      "step": 7497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4309647508652088e-05,
+      "loss": 0.9835,
+      "step": 7498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4308176317460031e-05,
+      "loss": 0.9081,
+      "step": 7499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4306705011761566e-05,
+      "loss": 0.9177,
+      "step": 7500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.43052335915958e-05,
+      "loss": 1.0522,
+      "step": 7501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4303762057001844e-05,
+      "loss": 0.9845,
+      "step": 7502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302290408018806e-05,
+      "loss": 0.9778,
+      "step": 7503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4300818644685805e-05,
+      "loss": 0.9547,
+      "step": 7504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4299346767041956e-05,
+      "loss": 1.0427,
+      "step": 7505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297874775126379e-05,
+      "loss": 0.9185,
+      "step": 7506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.42964026689782e-05,
+      "loss": 0.8832,
+      "step": 7507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294930448636541e-05,
+      "loss": 1.0056,
+      "step": 7508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4293458114140542e-05,
+      "loss": 1.0097,
+      "step": 7509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4291985665529327e-05,
+      "loss": 0.9681,
+      "step": 7510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4290513102842036e-05,
+      "loss": 0.4599,
+      "step": 7511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4289040426117806e-05,
+      "loss": 0.9244,
+      "step": 7512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287567635395778e-05,
+      "loss": 0.9923,
+      "step": 7513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4286094730715104e-05,
+      "loss": 0.915,
+      "step": 7514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4284621712114923e-05,
+      "loss": 1.0171,
+      "step": 7515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4283148579634394e-05,
+      "loss": 0.9968,
+      "step": 7516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281675333312666e-05,
+      "loss": 0.9328,
+      "step": 7517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280201973188896e-05,
+      "loss": 0.9671,
+      "step": 7518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4278728499302247e-05,
+      "loss": 0.9651,
+      "step": 7519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277254911691882e-05,
+      "loss": 0.9249,
+      "step": 7520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4275781210396966e-05,
+      "loss": 0.9713,
+      "step": 7521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427430739545667e-05,
+      "loss": 0.973,
+      "step": 7522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4272833466910163e-05,
+      "loss": 1.07,
+      "step": 7523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4271359424796625e-05,
+      "loss": 1.0458,
+      "step": 7524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426988526915523e-05,
+      "loss": 0.9392,
+      "step": 7525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4268411000025163e-05,
+      "loss": 0.9227,
+      "step": 7526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4266936617445604e-05,
+      "loss": 0.9674,
+      "step": 7527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265462121455745e-05,
+      "loss": 1.0022,
+      "step": 7528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263987512094775e-05,
+      "loss": 0.8832,
+      "step": 7529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4262512789401885e-05,
+      "loss": 1.0031,
+      "step": 7530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261037953416273e-05,
+      "loss": 0.956,
+      "step": 7531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425956300417714e-05,
+      "loss": 0.8602,
+      "step": 7532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4258087941723688e-05,
+      "loss": 0.8691,
+      "step": 7533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425661276609512e-05,
+      "loss": 0.9983,
+      "step": 7534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255137477330647e-05,
+      "loss": 0.9928,
+      "step": 7535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4253662075469477e-05,
+      "loss": 0.835,
+      "step": 7536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4252186560550832e-05,
+      "loss": 0.8994,
+      "step": 7537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250710932613922e-05,
+      "loss": 0.9714,
+      "step": 7538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4249235191697972e-05,
+      "loss": 0.907,
+      "step": 7539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4247759337842202e-05,
+      "loss": 0.9808,
+      "step": 7540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424628337108584e-05,
+      "loss": 0.9754,
+      "step": 7541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244807291468115e-05,
+      "loss": 0.932,
+      "step": 7542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243331099028263e-05,
+      "loss": 0.9463,
+      "step": 7543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4241854793805515e-05,
+      "loss": 0.9576,
+      "step": 7544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4240378375839112e-05,
+      "loss": 0.9367,
+      "step": 7545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4238901845168294e-05,
+      "loss": 1.0193,
+      "step": 7546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4237425201832307e-05,
+      "loss": 0.9424,
+      "step": 7547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4235948445870395e-05,
+      "loss": 0.9884,
+      "step": 7548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4234471577321814e-05,
+      "loss": 0.9319,
+      "step": 7549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4232994596225815e-05,
+      "loss": 0.9593,
+      "step": 7550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231517502621654e-05,
+      "loss": 0.9567,
+      "step": 7551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4230040296548588e-05,
+      "loss": 0.8854,
+      "step": 7552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4228562978045885e-05,
+      "loss": 0.998,
+      "step": 7553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4227085547152806e-05,
+      "loss": 0.8989,
+      "step": 7554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4225608003908622e-05,
+      "loss": 0.8875,
+      "step": 7555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4224130348352605e-05,
+      "loss": 1.0155,
+      "step": 7556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4222652580524023e-05,
+      "loss": 0.8686,
+      "step": 7557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221174700462164e-05,
+      "loss": 0.9431,
+      "step": 7558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4219696708206296e-05,
+      "loss": 0.9852,
+      "step": 7559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4218218603795713e-05,
+      "loss": 0.917,
+      "step": 7560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4216740387269696e-05,
+      "loss": 0.8948,
+      "step": 7561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4215262058667537e-05,
+      "loss": 0.9404,
+      "step": 7562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4213783618028525e-05,
+      "loss": 0.9219,
+      "step": 7563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4212305065391958e-05,
+      "loss": 0.9371,
+      "step": 7564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4210826400797133e-05,
+      "loss": 0.9159,
+      "step": 7565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4209347624283352e-05,
+      "loss": 0.9185,
+      "step": 7566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4207868735889916e-05,
+      "loss": 0.9211,
+      "step": 7567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420638973565614e-05,
+      "loss": 0.9221,
+      "step": 7568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204910623621327e-05,
+      "loss": 0.9322,
+      "step": 7569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4203431399824788e-05,
+      "loss": 0.9033,
+      "step": 7570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4201952064305848e-05,
+      "loss": 0.9563,
+      "step": 7571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200472617103818e-05,
+      "loss": 0.9149,
+      "step": 7572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198993058258023e-05,
+      "loss": 0.9672,
+      "step": 7573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419751338780779e-05,
+      "loss": 0.9445,
+      "step": 7574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4196033605792446e-05,
+      "loss": 1.0266,
+      "step": 7575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194553712251323e-05,
+      "loss": 0.9497,
+      "step": 7576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193073707223748e-05,
+      "loss": 0.9419,
+      "step": 7577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4191593590749064e-05,
+      "loss": 0.9727,
+      "step": 7578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4190113362866611e-05,
+      "loss": 0.9382,
+      "step": 7579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4188633023615728e-05,
+      "loss": 0.9533,
+      "step": 7580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4187152573035767e-05,
+      "loss": 0.881,
+      "step": 7581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4185672011166072e-05,
+      "loss": 0.8959,
+      "step": 7582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4184191338045992e-05,
+      "loss": 0.9597,
+      "step": 7583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4182710553714887e-05,
+      "loss": 0.9836,
+      "step": 7584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181229658212112e-05,
+      "loss": 0.8598,
+      "step": 7585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417974865157703e-05,
+      "loss": 0.9193,
+      "step": 7586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4178267533849002e-05,
+      "loss": 1.0178,
+      "step": 7587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176786305067394e-05,
+      "loss": 0.957,
+      "step": 7588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417530496527158e-05,
+      "loss": 1.0055,
+      "step": 7589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4173823514500924e-05,
+      "loss": 0.9074,
+      "step": 7590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4172341952794806e-05,
+      "loss": 0.9167,
+      "step": 7591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4170860280192606e-05,
+      "loss": 0.9381,
+      "step": 7592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169378496733703e-05,
+      "loss": 1.0227,
+      "step": 7593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4167896602457482e-05,
+      "loss": 0.9827,
+      "step": 7594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166414597403329e-05,
+      "loss": 0.9086,
+      "step": 7595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164932481610633e-05,
+      "loss": 0.983,
+      "step": 7596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163450255118791e-05,
+      "loss": 0.9793,
+      "step": 7597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4161967917967194e-05,
+      "loss": 0.9097,
+      "step": 7598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4160485470195245e-05,
+      "loss": 0.9278,
+      "step": 7599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159002911842345e-05,
+      "loss": 0.9883,
+      "step": 7600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157520242947893e-05,
+      "loss": 0.9765,
+      "step": 7601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4156037463551307e-05,
+      "loss": 0.9795,
+      "step": 7602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4154554573691987e-05,
+      "loss": 1.0166,
+      "step": 7603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4153071573409354e-05,
+      "loss": 0.9598,
+      "step": 7604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4151588462742823e-05,
+      "loss": 0.9566,
+      "step": 7605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4150105241731811e-05,
+      "loss": 0.989,
+      "step": 7606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4148621910415738e-05,
+      "loss": 0.9988,
+      "step": 7607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147138468834038e-05,
+      "loss": 0.9827,
+      "step": 7608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414565491702613e-05,
+      "loss": 0.9269,
+      "step": 7609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4144171255031454e-05,
+      "loss": 0.9346,
+      "step": 7610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4142687482889435e-05,
+      "loss": 0.9484,
+      "step": 7611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4141203600639516e-05,
+      "loss": 0.9857,
+      "step": 7612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139719608321133e-05,
+      "loss": 0.8906,
+      "step": 7613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413823550597373e-05,
+      "loss": 0.9806,
+      "step": 7614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4136751293636759e-05,
+      "loss": 0.93,
+      "step": 7615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135266971349657e-05,
+      "loss": 1.0555,
+      "step": 7616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4133782539151886e-05,
+      "loss": 0.9282,
+      "step": 7617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4132297997082898e-05,
+      "loss": 0.9729,
+      "step": 7618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4130813345182142e-05,
+      "loss": 0.9414,
+      "step": 7619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412932858348909e-05,
+      "loss": 0.9727,
+      "step": 7620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4127843712043202e-05,
+      "loss": 0.9462,
+      "step": 7621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4126358730883936e-05,
+      "loss": 0.9656,
+      "step": 7622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4124873640050774e-05,
+      "loss": 1.0296,
+      "step": 7623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123388439583179e-05,
+      "loss": 1.068,
+      "step": 7624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4121903129520627e-05,
+      "loss": 0.9179,
+      "step": 7625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41204177099026e-05,
+      "loss": 1.0216,
+      "step": 7626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4118932180768576e-05,
+      "loss": 0.9112,
+      "step": 7627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411744654215804e-05,
+      "loss": 0.9748,
+      "step": 7628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115960794110475e-05,
+      "loss": 0.9143,
+      "step": 7629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4114474936665375e-05,
+      "loss": 0.8977,
+      "step": 7630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4112988969862227e-05,
+      "loss": 0.9052,
+      "step": 7631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4111502893740532e-05,
+      "loss": 0.8474,
+      "step": 7632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4110016708339787e-05,
+      "loss": 0.9286,
+      "step": 7633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410853041369949e-05,
+      "loss": 0.9782,
+      "step": 7634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410704400985915e-05,
+      "loss": 0.9454,
+      "step": 7635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4105557496858268e-05,
+      "loss": 0.9968,
+      "step": 7636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4104070874736355e-05,
+      "loss": 0.9443,
+      "step": 7637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4102584143532929e-05,
+      "loss": 0.9498,
+      "step": 7638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4101097303287504e-05,
+      "loss": 0.9402,
+      "step": 7639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4099610354039593e-05,
+      "loss": 0.8422,
+      "step": 7640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4098123295828721e-05,
+      "loss": 0.9487,
+      "step": 7641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4096636128694412e-05,
+      "loss": 0.8857,
+      "step": 7642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095148852676195e-05,
+      "loss": 1.0108,
+      "step": 7643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093661467813597e-05,
+      "loss": 0.9457,
+      "step": 7644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4092173974146155e-05,
+      "loss": 0.9317,
+      "step": 7645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4090686371713403e-05,
+      "loss": 0.9762,
+      "step": 7646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4089198660554875e-05,
+      "loss": 0.9929,
+      "step": 7647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.408771084071012e-05,
+      "loss": 0.9071,
+      "step": 7648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408622291221868e-05,
+      "loss": 0.9641,
+      "step": 7649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4084734875120098e-05,
+      "loss": 0.8964,
+      "step": 7650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4083246729453932e-05,
+      "loss": 1.0249,
+      "step": 7651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408175847525973e-05,
+      "loss": 0.9767,
+      "step": 7652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080270112577051e-05,
+      "loss": 0.9442,
+      "step": 7653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407878164144545e-05,
+      "loss": 0.8452,
+      "step": 7654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407729306190449e-05,
+      "loss": 0.9052,
+      "step": 7655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4075804373993743e-05,
+      "loss": 1.0031,
+      "step": 7656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074315577752768e-05,
+      "loss": 0.929,
+      "step": 7657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072826673221136e-05,
+      "loss": 0.9107,
+      "step": 7658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4071337660438426e-05,
+      "loss": 0.4769,
+      "step": 7659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4069848539444208e-05,
+      "loss": 1.003,
+      "step": 7660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068359310278066e-05,
+      "loss": 0.9453,
+      "step": 7661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406686997297958e-05,
+      "loss": 0.9128,
+      "step": 7662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4065380527588333e-05,
+      "loss": 0.883,
+      "step": 7663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406389097414392e-05,
+      "loss": 0.9003,
+      "step": 7664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4062401312685921e-05,
+      "loss": 0.4501,
+      "step": 7665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4060911543253935e-05,
+      "loss": 0.9129,
+      "step": 7666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405942166588756e-05,
+      "loss": 0.4785,
+      "step": 7667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405793168062639e-05,
+      "loss": 0.9497,
+      "step": 7668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056441587510037e-05,
+      "loss": 0.9578,
+      "step": 7669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4054951386578095e-05,
+      "loss": 0.9587,
+      "step": 7670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4053461077870178e-05,
+      "loss": 0.9101,
+      "step": 7671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4051970661425894e-05,
+      "loss": 0.4383,
+      "step": 7672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4050480137284856e-05,
+      "loss": 0.8907,
+      "step": 7673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4048989505486685e-05,
+      "loss": 0.9827,
+      "step": 7674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047498766070997e-05,
+      "loss": 0.9439,
+      "step": 7675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4046007919077414e-05,
+      "loss": 0.9581,
+      "step": 7676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4044516964545563e-05,
+      "loss": 0.9602,
+      "step": 7677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4043025902515067e-05,
+      "loss": 0.9985,
+      "step": 7678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4041534733025559e-05,
+      "loss": 0.9382,
+      "step": 7679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040043456116678e-05,
+      "loss": 0.9271,
+      "step": 7680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4038552071828054e-05,
+      "loss": 0.9387,
+      "step": 7681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4037060580199331e-05,
+      "loss": 0.4494,
+      "step": 7682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4035568981270145e-05,
+      "loss": 0.9172,
+      "step": 7683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4034077275080145e-05,
+      "loss": 0.9114,
+      "step": 7684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4032585461668981e-05,
+      "loss": 0.9466,
+      "step": 7685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4031093541076298e-05,
+      "loss": 0.9768,
+      "step": 7686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4029601513341756e-05,
+      "loss": 0.933,
+      "step": 7687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4028109378505006e-05,
+      "loss": 0.9816,
+      "step": 7688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402661713660571e-05,
+      "loss": 1.0457,
+      "step": 7689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4025124787683527e-05,
+      "loss": 0.8345,
+      "step": 7690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4023632331778128e-05,
+      "loss": 1.015,
+      "step": 7691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022139768929173e-05,
+      "loss": 0.9262,
+      "step": 7692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402064709917634e-05,
+      "loss": 0.922,
+      "step": 7693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4019154322559296e-05,
+      "loss": 0.9809,
+      "step": 7694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4017661439117724e-05,
+      "loss": 0.9547,
+      "step": 7695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016168448891296e-05,
+      "loss": 0.9827,
+      "step": 7696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40146753519197e-05,
+      "loss": 0.9172,
+      "step": 7697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4013182148242618e-05,
+      "loss": 0.9589,
+      "step": 7698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4011688837899737e-05,
+      "loss": 0.9826,
+      "step": 7699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401019542093075e-05,
+      "loss": 1.0025,
+      "step": 7700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008701897375349e-05,
+      "loss": 0.9058,
+      "step": 7701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4007208267273227e-05,
+      "loss": 0.9814,
+      "step": 7702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4005714530664087e-05,
+      "loss": 0.9362,
+      "step": 7703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004220687587632e-05,
+      "loss": 0.9816,
+      "step": 7704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4002726738083563e-05,
+      "loss": 0.8913,
+      "step": 7705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001232682191591e-05,
+      "loss": 1.0213,
+      "step": 7706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399973851995142e-05,
+      "loss": 1.0175,
+      "step": 7707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3998244251402768e-05,
+      "loss": 0.9706,
+      "step": 7708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996749876585351e-05,
+      "loss": 0.9018,
+      "step": 7709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3995255395538889e-05,
+      "loss": 0.9881,
+      "step": 7710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3993760808303101e-05,
+      "loss": 0.894,
+      "step": 7711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3992266114917709e-05,
+      "loss": 0.9169,
+      "step": 7712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3990771315422446e-05,
+      "loss": 0.9205,
+      "step": 7713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989276409857038e-05,
+      "loss": 0.9852,
+      "step": 7714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398778139826122e-05,
+      "loss": 0.9517,
+      "step": 7715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986286280674728e-05,
+      "loss": 0.9678,
+      "step": 7716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984791057137298e-05,
+      "loss": 0.9983,
+      "step": 7717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3983295727688673e-05,
+      "loss": 0.975,
+      "step": 7718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3981800292368598e-05,
+      "loss": 1.0274,
+      "step": 7719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3980304751216816e-05,
+      "loss": 0.9197,
+      "step": 7720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3978809104273082e-05,
+      "loss": 0.9211,
+      "step": 7721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3977313351577144e-05,
+      "loss": 0.9732,
+      "step": 7722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3975817493168764e-05,
+      "loss": 0.9537,
+      "step": 7723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974321529087694e-05,
+      "loss": 0.9801,
+      "step": 7724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972825459373694e-05,
+      "loss": 0.9084,
+      "step": 7725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3971329284066533e-05,
+      "loss": 0.9862,
+      "step": 7726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969833003205976e-05,
+      "loss": 1.0465,
+      "step": 7727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3968336616831791e-05,
+      "loss": 0.9498,
+      "step": 7728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396684012498375e-05,
+      "loss": 0.9658,
+      "step": 7729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3965343527701629e-05,
+      "loss": 0.9397,
+      "step": 7730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3963846825025208e-05,
+      "loss": 0.9172,
+      "step": 7731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962350016994265e-05,
+      "loss": 0.9306,
+      "step": 7732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3960853103648581e-05,
+      "loss": 0.9464,
+      "step": 7733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3959356085027949e-05,
+      "loss": 1.0083,
+      "step": 7734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3957858961172152e-05,
+      "loss": 0.9625,
+      "step": 7735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3956361732120987e-05,
+      "loss": 0.9304,
+      "step": 7736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954864397914241e-05,
+      "loss": 0.9751,
+      "step": 7737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395336695859172e-05,
+      "loss": 0.9238,
+      "step": 7738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395186941419322e-05,
+      "loss": 0.9287,
+      "step": 7739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950371764758543e-05,
+      "loss": 0.913,
+      "step": 7740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3948874010327495e-05,
+      "loss": 0.917,
+      "step": 7741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947376150939891e-05,
+      "loss": 0.9331,
+      "step": 7742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394587818663553e-05,
+      "loss": 0.9405,
+      "step": 7743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3944380117454239e-05,
+      "loss": 1.0558,
+      "step": 7744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3942881943435828e-05,
+      "loss": 0.8859,
+      "step": 7745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3941383664620118e-05,
+      "loss": 0.9752,
+      "step": 7746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939885281046932e-05,
+      "loss": 0.876,
+      "step": 7747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938386792756096e-05,
+      "loss": 0.9209,
+      "step": 7748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3936888199787433e-05,
+      "loss": 0.9335,
+      "step": 7749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935389502180783e-05,
+      "loss": 0.9105,
+      "step": 7750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.393389069997597e-05,
+      "loss": 1.0195,
+      "step": 7751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.393239179321284e-05,
+      "loss": 0.9826,
+      "step": 7752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3930892781931225e-05,
+      "loss": 0.9006,
+      "step": 7753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929393666170969e-05,
+      "loss": 0.9879,
+      "step": 7754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3927894445971916e-05,
+      "loss": 0.909,
+      "step": 7755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3926395121373917e-05,
+      "loss": 0.9129,
+      "step": 7756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392489569241682e-05,
+      "loss": 1.019,
+      "step": 7757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392339615914048e-05,
+      "loss": 0.9407,
+      "step": 7758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921896521584746e-05,
+      "loss": 0.9829,
+      "step": 7759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3920396779789487e-05,
+      "loss": 0.8891,
+      "step": 7760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3918896933794553e-05,
+      "loss": 1.006,
+      "step": 7761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3917396983639818e-05,
+      "loss": 1.0146,
+      "step": 7762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915896929365147e-05,
+      "loss": 0.4412,
+      "step": 7763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3914396771010403e-05,
+      "loss": 0.9935,
+      "step": 7764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3912896508615468e-05,
+      "loss": 1.0285,
+      "step": 7765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3911396142220211e-05,
+      "loss": 0.9136,
+      "step": 7766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909895671864511e-05,
+      "loss": 1.0209,
+      "step": 7767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908395097588252e-05,
+      "loss": 1.0419,
+      "step": 7768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390689441943131e-05,
+      "loss": 0.9356,
+      "step": 7769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3905393637433582e-05,
+      "loss": 0.96,
+      "step": 7770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903892751634949e-05,
+      "loss": 0.9485,
+      "step": 7771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3902391762075307e-05,
+      "loss": 0.9556,
+      "step": 7772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3900890668794547e-05,
+      "loss": 0.9425,
+      "step": 7773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899389471832567e-05,
+      "loss": 0.93,
+      "step": 7774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3897888171229272e-05,
+      "loss": 0.9236,
+      "step": 7775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3896386767024561e-05,
+      "loss": 0.8495,
+      "step": 7776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3894885259258337e-05,
+      "loss": 0.9597,
+      "step": 7777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3893383647970513e-05,
+      "loss": 0.4569,
+      "step": 7778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3891881933200999e-05,
+      "loss": 0.8408,
+      "step": 7779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3890380114989705e-05,
+      "loss": 0.9203,
+      "step": 7780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888878193376556e-05,
+      "loss": 0.9724,
+      "step": 7781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388737616840146e-05,
+      "loss": 1.0159,
+      "step": 7782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388587404010435e-05,
+      "loss": 0.9084,
+      "step": 7783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3884371808525145e-05,
+      "loss": 1.0117,
+      "step": 7784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3882869473703773e-05,
+      "loss": 0.8826,
+      "step": 7785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881367035680164e-05,
+      "loss": 0.9942,
+      "step": 7786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3879864494494252e-05,
+      "loss": 0.8784,
+      "step": 7787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3878361850185977e-05,
+      "loss": 0.8955,
+      "step": 7788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876859102795272e-05,
+      "loss": 0.983,
+      "step": 7789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3875356252362075e-05,
+      "loss": 0.9698,
+      "step": 7790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3873853298926339e-05,
+      "loss": 0.9174,
+      "step": 7791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3872350242528007e-05,
+      "loss": 1.0649,
+      "step": 7792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387084708320703e-05,
+      "loss": 0.8574,
+      "step": 7793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3869343821003356e-05,
+      "loss": 0.9644,
+      "step": 7794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386784045595694e-05,
+      "loss": 0.9384,
+      "step": 7795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3866336988107745e-05,
+      "loss": 0.5078,
+      "step": 7796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386483341749573e-05,
+      "loss": 0.9596,
+      "step": 7797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863329744160851e-05,
+      "loss": 0.4718,
+      "step": 7798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3861825968143085e-05,
+      "loss": 0.9432,
+      "step": 7799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3860322089482394e-05,
+      "loss": 0.9372,
+      "step": 7800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3858818108218752e-05,
+      "loss": 0.9113,
+      "step": 7801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385731402439213e-05,
+      "loss": 1.0095,
+      "step": 7802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855809838042507e-05,
+      "loss": 0.9211,
+      "step": 7803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3854305549209864e-05,
+      "loss": 0.9039,
+      "step": 7804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852801157934179e-05,
+      "loss": 0.9817,
+      "step": 7805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3851296664255443e-05,
+      "loss": 0.956,
+      "step": 7806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3849792068213639e-05,
+      "loss": 1.0104,
+      "step": 7807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3848287369848754e-05,
+      "loss": 0.9447,
+      "step": 7808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3846782569200792e-05,
+      "loss": 0.9107,
+      "step": 7809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845277666309738e-05,
+      "loss": 0.9073,
+      "step": 7810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38437726612156e-05,
+      "loss": 1.0341,
+      "step": 7811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3842267553958373e-05,
+      "loss": 0.8799,
+      "step": 7812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384076234457806e-05,
+      "loss": 0.9871,
+      "step": 7813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3839257033114672e-05,
+      "loss": 0.9867,
+      "step": 7814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383775161960822e-05,
+      "loss": 0.8849,
+      "step": 7815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383624610409871e-05,
+      "loss": 0.967,
+      "step": 7816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3834740486626162e-05,
+      "loss": 0.9342,
+      "step": 7817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833234767230588e-05,
+      "loss": 1.0394,
+      "step": 7818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3831728945952016e-05,
+      "loss": 0.9726,
+      "step": 7819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3830223022830462e-05,
+      "loss": 0.9267,
+      "step": 7820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3828716997905953e-05,
+      "loss": 0.9307,
+      "step": 7821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3827210871218524e-05,
+      "loss": 0.8509,
+      "step": 7822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3825704642808197e-05,
+      "loss": 0.8743,
+      "step": 7823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3824198312715011e-05,
+      "loss": 1.0656,
+      "step": 7824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3822691880979005e-05,
+      "loss": 0.9699,
+      "step": 7825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821185347640209e-05,
+      "loss": 0.903,
+      "step": 7826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819678712738673e-05,
+      "loss": 0.8828,
+      "step": 7827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3818171976314442e-05,
+      "loss": 0.9131,
+      "step": 7828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3816665138407559e-05,
+      "loss": 0.9745,
+      "step": 7829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815158199058077e-05,
+      "loss": 1.0051,
+      "step": 7830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3813651158306044e-05,
+      "loss": 0.9196,
+      "step": 7831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3812144016191517e-05,
+      "loss": 0.9277,
+      "step": 7832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810636772754563e-05,
+      "loss": 1.0182,
+      "step": 7833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809129428035229e-05,
+      "loss": 0.4407,
+      "step": 7834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380762198207359e-05,
+      "loss": 0.9329,
+      "step": 7835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3806114434909704e-05,
+      "loss": 0.9462,
+      "step": 7836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3804606786583644e-05,
+      "loss": 1.0159,
+      "step": 7837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803099037135484e-05,
+      "loss": 0.9038,
+      "step": 7838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3801591186605292e-05,
+      "loss": 0.9258,
+      "step": 7839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380008323503315e-05,
+      "loss": 0.8814,
+      "step": 7840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3798575182459134e-05,
+      "loss": 1.0183,
+      "step": 7841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379706702892333e-05,
+      "loss": 0.9637,
+      "step": 7842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3795558774465818e-05,
+      "loss": 0.8666,
+      "step": 7843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379405041912669e-05,
+      "loss": 0.9707,
+      "step": 7844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3792541962946038e-05,
+      "loss": 0.9081,
+      "step": 7845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379103340596395e-05,
+      "loss": 0.9463,
+      "step": 7846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3789524748220524e-05,
+      "loss": 0.9547,
+      "step": 7847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3788015989755858e-05,
+      "loss": 0.8984,
+      "step": 7848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3786507130610051e-05,
+      "loss": 0.9772,
+      "step": 7849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784998170823212e-05,
+      "loss": 0.9409,
+      "step": 7850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3783489110435443e-05,
+      "loss": 0.9953,
+      "step": 7851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3781979949486856e-05,
+      "loss": 0.9119,
+      "step": 7852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780470688017562e-05,
+      "loss": 0.9352,
+      "step": 7853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778961326067672e-05,
+      "loss": 0.8669,
+      "step": 7854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777451863677306e-05,
+      "loss": 0.9286,
+      "step": 7855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3775942300886587e-05,
+      "loss": 0.9416,
+      "step": 7856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774432637735633e-05,
+      "loss": 0.9378,
+      "step": 7857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3772922874264572e-05,
+      "loss": 0.9483,
+      "step": 7858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3771413010513528e-05,
+      "loss": 0.9809,
+      "step": 7859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3769903046522632e-05,
+      "loss": 0.8931,
+      "step": 7860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376839298233202e-05,
+      "loss": 0.8013,
+      "step": 7861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766882817981824e-05,
+      "loss": 0.898,
+      "step": 7862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376537255351219e-05,
+      "loss": 0.9057,
+      "step": 7863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3763862188963254e-05,
+      "loss": 1.0106,
+      "step": 7864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762351724375153e-05,
+      "loss": 0.957,
+      "step": 7865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3760841159788045e-05,
+      "loss": 0.9827,
+      "step": 7866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3759330495242074e-05,
+      "loss": 0.9686,
+      "step": 7867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3757819730777392e-05,
+      "loss": 0.9084,
+      "step": 7868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756308866434153e-05,
+      "loss": 0.9316,
+      "step": 7869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3754797902252515e-05,
+      "loss": 0.9892,
+      "step": 7870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3753286838272637e-05,
+      "loss": 0.9767,
+      "step": 7871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375177567453468e-05,
+      "loss": 0.9123,
+      "step": 7872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375026441107881e-05,
+      "loss": 0.8749,
+      "step": 7873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748753047945198e-05,
+      "loss": 0.9765,
+      "step": 7874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374724158517401e-05,
+      "loss": 0.9176,
+      "step": 7875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3745730022805422e-05,
+      "loss": 0.9437,
+      "step": 7876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3744218360879606e-05,
+      "loss": 0.9824,
+      "step": 7877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742706599436744e-05,
+      "loss": 0.9612,
+      "step": 7878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3741194738517012e-05,
+      "loss": 0.9374,
+      "step": 7879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3739682778160599e-05,
+      "loss": 1.0301,
+      "step": 7880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3738170718407689e-05,
+      "loss": 0.4561,
+      "step": 7881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736658559298473e-05,
+      "loss": 0.9963,
+      "step": 7882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3735146300873137e-05,
+      "loss": 0.9859,
+      "step": 7883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3733633943171878e-05,
+      "loss": 0.9838,
+      "step": 7884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732121486234892e-05,
+      "loss": 0.9828,
+      "step": 7885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3730608930102382e-05,
+      "loss": 0.9747,
+      "step": 7886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3729096274814549e-05,
+      "loss": 0.9444,
+      "step": 7887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3727583520411592e-05,
+      "loss": 0.8699,
+      "step": 7888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3726070666933724e-05,
+      "loss": 0.9637,
+      "step": 7889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724557714421152e-05,
+      "loss": 1.0202,
+      "step": 7890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3723044662914091e-05,
+      "loss": 0.9517,
+      "step": 7891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3721531512452757e-05,
+      "loss": 0.9802,
+      "step": 7892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720018263077361e-05,
+      "loss": 1.0025,
+      "step": 7893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3718504914828135e-05,
+      "loss": 1.0322,
+      "step": 7894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716991467745291e-05,
+      "loss": 1.002,
+      "step": 7895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3715477921869058e-05,
+      "loss": 0.8865,
+      "step": 7896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3713964277239667e-05,
+      "loss": 0.969,
+      "step": 7897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712450533897348e-05,
+      "loss": 0.9007,
+      "step": 7898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371093669188233e-05,
+      "loss": 0.9293,
+      "step": 7899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3709422751234857e-05,
+      "loss": 0.9798,
+      "step": 7900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3707908711995163e-05,
+      "loss": 0.9209,
+      "step": 7901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370639457420349e-05,
+      "loss": 0.9483,
+      "step": 7902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3704880337900081e-05,
+      "loss": 1.0288,
+      "step": 7903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3703366003125188e-05,
+      "loss": 0.9479,
+      "step": 7904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3701851569919054e-05,
+      "loss": 0.9198,
+      "step": 7905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700337038321932e-05,
+      "loss": 1.0091,
+      "step": 7906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369882240837408e-05,
+      "loss": 0.9103,
+      "step": 7907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369730768011575e-05,
+      "loss": 0.9718,
+      "step": 7908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3695792853587204e-05,
+      "loss": 1.0045,
+      "step": 7909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694277928828707e-05,
+      "loss": 0.9598,
+      "step": 7910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369276290588052e-05,
+      "loss": 0.9294,
+      "step": 7911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3691247784782918e-05,
+      "loss": 0.9123,
+      "step": 7912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3689732565576162e-05,
+      "loss": 0.9458,
+      "step": 7913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3688217248300526e-05,
+      "loss": 0.9286,
+      "step": 7914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686701832996292e-05,
+      "loss": 0.8729,
+      "step": 7915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3685186319703733e-05,
+      "loss": 0.8235,
+      "step": 7916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3683670708463128e-05,
+      "loss": 0.9384,
+      "step": 7917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3682154999314765e-05,
+      "loss": 0.9543,
+      "step": 7918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3680639192298928e-05,
+      "loss": 0.4866,
+      "step": 7919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679123287455901e-05,
+      "loss": 0.9514,
+      "step": 7920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677607284825984e-05,
+      "loss": 1.004,
+      "step": 7921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676091184449462e-05,
+      "loss": 0.9331,
+      "step": 7922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367457498636664e-05,
+      "loss": 1.011,
+      "step": 7923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3673058690617807e-05,
+      "loss": 0.8741,
+      "step": 7924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367154229724327e-05,
+      "loss": 0.8349,
+      "step": 7925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3670025806283332e-05,
+      "loss": 0.9637,
+      "step": 7926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3668509217778299e-05,
+      "loss": 0.9264,
+      "step": 7927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3666992531768482e-05,
+      "loss": 0.8616,
+      "step": 7928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3665475748294191e-05,
+      "loss": 0.8727,
+      "step": 7929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3663958867395741e-05,
+      "loss": 0.9794,
+      "step": 7930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3662441889113451e-05,
+      "loss": 0.982,
+      "step": 7931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3660924813487633e-05,
+      "loss": 0.9932,
+      "step": 7932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365940764055862e-05,
+      "loss": 0.9082,
+      "step": 7933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3657890370366731e-05,
+      "loss": 0.9393,
+      "step": 7934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365637300295229e-05,
+      "loss": 0.9092,
+      "step": 7935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3654855538355634e-05,
+      "loss": 0.8873,
+      "step": 7936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3653337976617089e-05,
+      "loss": 1.0202,
+      "step": 7937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3651820317776995e-05,
+      "loss": 0.9843,
+      "step": 7938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365030256187569e-05,
+      "loss": 0.9538,
+      "step": 7939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648784708953509e-05,
+      "loss": 0.8555,
+      "step": 7940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3647266759050799e-05,
+      "loss": 0.979,
+      "step": 7941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3645748712207903e-05,
+      "loss": 0.971,
+      "step": 7942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364423056846517e-05,
+      "loss": 0.9579,
+      "step": 7943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3642712327862952e-05,
+      "loss": 0.9975,
+      "step": 7944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641193990441597e-05,
+      "loss": 0.9437,
+      "step": 7945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363967555624147e-05,
+      "loss": 0.9699,
+      "step": 7946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363815702530292e-05,
+      "loss": 0.8826,
+      "step": 7947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3636638397666314e-05,
+      "loss": 0.9444,
+      "step": 7948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363511967337201e-05,
+      "loss": 1.0426,
+      "step": 7949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3633600852460376e-05,
+      "loss": 1.0218,
+      "step": 7950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632081934971787e-05,
+      "loss": 0.9415,
+      "step": 7951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3630562920946603e-05,
+      "loss": 0.9049,
+      "step": 7952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3629043810425204e-05,
+      "loss": 0.9903,
+      "step": 7953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3627524603447968e-05,
+      "loss": 0.9807,
+      "step": 7954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3626005300055267e-05,
+      "loss": 0.9861,
+      "step": 7955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3624485900287488e-05,
+      "loss": 0.9988,
+      "step": 7956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3622966404185017e-05,
+      "loss": 0.8944,
+      "step": 7957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362144681178823e-05,
+      "loss": 0.9343,
+      "step": 7958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619927123137529e-05,
+      "loss": 0.9273,
+      "step": 7959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3618407338273296e-05,
+      "loss": 0.9947,
+      "step": 7960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3616887457235928e-05,
+      "loss": 0.9261,
+      "step": 7961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615367480065825e-05,
+      "loss": 0.8397,
+      "step": 7962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361384740680338e-05,
+      "loss": 0.958,
+      "step": 7963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3612327237489003e-05,
+      "loss": 0.9404,
+      "step": 7964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3610806972163092e-05,
+      "loss": 0.891,
+      "step": 7965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609286610866055e-05,
+      "loss": 0.9681,
+      "step": 7966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36077661536383e-05,
+      "loss": 0.9541,
+      "step": 7967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3606245600520242e-05,
+      "loss": 0.9667,
+      "step": 7968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3604724951552298e-05,
+      "loss": 0.8548,
+      "step": 7969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3603204206774877e-05,
+      "loss": 0.9465,
+      "step": 7970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601683366228407e-05,
+      "loss": 0.9547,
+      "step": 7971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3600162429953307e-05,
+      "loss": 0.9121,
+      "step": 7972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359864139799e-05,
+      "loss": 1.0136,
+      "step": 7973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597120270378912e-05,
+      "loss": 0.9548,
+      "step": 7974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359559904716048e-05,
+      "loss": 0.8848,
+      "step": 7975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594077728375129e-05,
+      "loss": 0.9889,
+      "step": 7976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3592556314063299e-05,
+      "loss": 0.8349,
+      "step": 7977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591034804265424e-05,
+      "loss": 0.9669,
+      "step": 7978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3589513199021944e-05,
+      "loss": 0.9211,
+      "step": 7979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587991498373305e-05,
+      "loss": 1.0317,
+      "step": 7980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3586469702359948e-05,
+      "loss": 0.418,
+      "step": 7981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3584947811022324e-05,
+      "loss": 0.87,
+      "step": 7982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3583425824400882e-05,
+      "loss": 0.9879,
+      "step": 7983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3581903742536074e-05,
+      "loss": 0.9602,
+      "step": 7984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3580381565468353e-05,
+      "loss": 0.8779,
+      "step": 7985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578859293238182e-05,
+      "loss": 0.9541,
+      "step": 7986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3577336925886017e-05,
+      "loss": 0.9086,
+      "step": 7987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3575814463452323e-05,
+      "loss": 0.8921,
+      "step": 7988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574291905977564e-05,
+      "loss": 0.9416,
+      "step": 7989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357276925350221e-05,
+      "loss": 0.9207,
+      "step": 7990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3571246506066726e-05,
+      "loss": 0.861,
+      "step": 7991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3569723663711591e-05,
+      "loss": 0.8856,
+      "step": 7992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3568200726477279e-05,
+      "loss": 0.8782,
+      "step": 7993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3566677694404265e-05,
+      "loss": 0.9558,
+      "step": 7994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565154567533031e-05,
+      "loss": 0.9303,
+      "step": 7995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3563631345904064e-05,
+      "loss": 0.9742,
+      "step": 7996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3562108029557842e-05,
+      "loss": 0.9497,
+      "step": 7997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356058461853486e-05,
+      "loss": 0.8862,
+      "step": 7998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3559061112875602e-05,
+      "loss": 0.9987,
+      "step": 7999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3557537512620567e-05,
+      "loss": 0.9794,
+      "step": 8000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3556013817810249e-05,
+      "loss": 1.0094,
+      "step": 8001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554490028485143e-05,
+      "loss": 1.0031,
+      "step": 8002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355296614468575e-05,
+      "loss": 0.8543,
+      "step": 8003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355144216645258e-05,
+      "loss": 0.966,
+      "step": 8004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3549918093826127e-05,
+      "loss": 1.0034,
+      "step": 8005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354839392684691e-05,
+      "loss": 0.9943,
+      "step": 8006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3546869665555433e-05,
+      "loss": 0.938,
+      "step": 8007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3545345309992212e-05,
+      "loss": 0.897,
+      "step": 8008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543820860197762e-05,
+      "loss": 0.8716,
+      "step": 8009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35422963162126e-05,
+      "loss": 0.9704,
+      "step": 8010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3540771678077249e-05,
+      "loss": 0.9168,
+      "step": 8011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3539246945832229e-05,
+      "loss": 0.97,
+      "step": 8012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3537722119518071e-05,
+      "loss": 0.8533,
+      "step": 8013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536197199175297e-05,
+      "loss": 1.0086,
+      "step": 8014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3534672184844437e-05,
+      "loss": 0.999,
+      "step": 8015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3533147076566033e-05,
+      "loss": 0.9602,
+      "step": 8016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3531621874380614e-05,
+      "loss": 0.9631,
+      "step": 8017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3530096578328716e-05,
+      "loss": 0.9251,
+      "step": 8018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3528571188450885e-05,
+      "loss": 0.9134,
+      "step": 8019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3527045704787661e-05,
+      "loss": 0.975,
+      "step": 8020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3525520127379587e-05,
+      "loss": 0.9541,
+      "step": 8021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352399445626722e-05,
+      "loss": 1.0019,
+      "step": 8022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3522468691491102e-05,
+      "loss": 0.9571,
+      "step": 8023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3520942833091792e-05,
+      "loss": 1.0532,
+      "step": 8024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3519416881109838e-05,
+      "loss": 0.8779,
+      "step": 8025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3517890835585808e-05,
+      "loss": 0.9052,
+      "step": 8026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3516364696560256e-05,
+      "loss": 0.9351,
+      "step": 8027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514838464073745e-05,
+      "loss": 0.9293,
+      "step": 8028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513312138166843e-05,
+      "loss": 0.9448,
+      "step": 8029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3511785718880117e-05,
+      "loss": 1.0154,
+      "step": 8030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510259206254137e-05,
+      "loss": 0.9697,
+      "step": 8031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508732600329478e-05,
+      "loss": 1.0208,
+      "step": 8032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350720590114671e-05,
+      "loss": 0.9986,
+      "step": 8033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3505679108746419e-05,
+      "loss": 0.9365,
+      "step": 8034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350415222316918e-05,
+      "loss": 0.9396,
+      "step": 8035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3502625244455576e-05,
+      "loss": 0.8815,
+      "step": 8036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3501098172646194e-05,
+      "loss": 0.8718,
+      "step": 8037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3499571007781618e-05,
+      "loss": 0.9029,
+      "step": 8038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498043749902445e-05,
+      "loss": 0.9153,
+      "step": 8039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3496516399049267e-05,
+      "loss": 0.9829,
+      "step": 8040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3494988955262671e-05,
+      "loss": 0.9768,
+      "step": 8041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3493461418583263e-05,
+      "loss": 0.9775,
+      "step": 8042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491933789051639e-05,
+      "loss": 0.8646,
+      "step": 8043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3490406066708404e-05,
+      "loss": 0.4262,
+      "step": 8044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488878251594165e-05,
+      "loss": 0.9146,
+      "step": 8045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3487350343749521e-05,
+      "loss": 0.882,
+      "step": 8046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3485822343215091e-05,
+      "loss": 1.0172,
+      "step": 8047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3484294250031484e-05,
+      "loss": 0.9523,
+      "step": 8048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3482766064239317e-05,
+      "loss": 0.8322,
+      "step": 8049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3481237785879202e-05,
+      "loss": 0.8732,
+      "step": 8050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479709414991763e-05,
+      "loss": 0.9669,
+      "step": 8051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478180951617626e-05,
+      "loss": 0.9685,
+      "step": 8052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.347665239579741e-05,
+      "loss": 0.9013,
+      "step": 8053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475123747571744e-05,
+      "loss": 0.9553,
+      "step": 8054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473595006981256e-05,
+      "loss": 0.9293,
+      "step": 8055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.347206617406658e-05,
+      "loss": 0.9271,
+      "step": 8056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3470537248868352e-05,
+      "loss": 0.9594,
+      "step": 8057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469008231427207e-05,
+      "loss": 0.9288,
+      "step": 8058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3467479121783783e-05,
+      "loss": 0.9693,
+      "step": 8059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465949919978728e-05,
+      "loss": 0.9272,
+      "step": 8060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3464420626052684e-05,
+      "loss": 0.8405,
+      "step": 8061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3462891240046291e-05,
+      "loss": 1.0059,
+      "step": 8062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3461361762000209e-05,
+      "loss": 0.9819,
+      "step": 8063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345983219195508e-05,
+      "loss": 0.9188,
+      "step": 8064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3458302529951568e-05,
+      "loss": 0.9296,
+      "step": 8065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3456772776030322e-05,
+      "loss": 0.9159,
+      "step": 8066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3455242930232002e-05,
+      "loss": 0.941,
+      "step": 8067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3453712992597273e-05,
+      "loss": 0.9583,
+      "step": 8068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3452182963166792e-05,
+      "loss": 0.9322,
+      "step": 8069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3450652841981238e-05,
+      "loss": 0.9867,
+      "step": 8070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3449122629081273e-05,
+      "loss": 1.0155,
+      "step": 8071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447592324507561e-05,
+      "loss": 0.9875,
+      "step": 8072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3446061928300788e-05,
+      "loss": 0.9798,
+      "step": 8073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3444531440501622e-05,
+      "loss": 0.9135,
+      "step": 8074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443000861150743e-05,
+      "loss": 0.9366,
+      "step": 8075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344147019028884e-05,
+      "loss": 1.0687,
+      "step": 8076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3439939427956583e-05,
+      "loss": 1.0097,
+      "step": 8077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438408574194668e-05,
+      "loss": 0.9088,
+      "step": 8078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3436877629043776e-05,
+      "loss": 0.9277,
+      "step": 8079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435346592544604e-05,
+      "loss": 0.9758,
+      "step": 8080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3433815464737846e-05,
+      "loss": 1.0528,
+      "step": 8081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3432284245664193e-05,
+      "loss": 0.9613,
+      "step": 8082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430752935364346e-05,
+      "loss": 0.9361,
+      "step": 8083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3429221533879004e-05,
+      "loss": 0.9552,
+      "step": 8084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3427690041248866e-05,
+      "loss": 1.0023,
+      "step": 8085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3426158457514647e-05,
+      "loss": 0.9207,
+      "step": 8086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3424626782717046e-05,
+      "loss": 0.8862,
+      "step": 8087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342309501689678e-05,
+      "loss": 0.9812,
+      "step": 8088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3421563160094556e-05,
+      "loss": 0.9405,
+      "step": 8089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3420031212351087e-05,
+      "loss": 0.9706,
+      "step": 8090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34184991737071e-05,
+      "loss": 1.0178,
+      "step": 8091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416967044203309e-05,
+      "loss": 0.9134,
+      "step": 8092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3415434823880432e-05,
+      "loss": 0.9576,
+      "step": 8093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413902512779202e-05,
+      "loss": 0.9125,
+      "step": 8094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341237011094034e-05,
+      "loss": 0.9699,
+      "step": 8095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341083761840458e-05,
+      "loss": 0.9621,
+      "step": 8096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340930503521265e-05,
+      "loss": 0.3984,
+      "step": 8097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3407772361405285e-05,
+      "loss": 0.9122,
+      "step": 8098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3406239597023225e-05,
+      "loss": 0.8955,
+      "step": 8099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3404706742107205e-05,
+      "loss": 0.8762,
+      "step": 8100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340317379669797e-05,
+      "loss": 0.9308,
+      "step": 8101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3401640760836262e-05,
+      "loss": 0.9117,
+      "step": 8102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400107634562826e-05,
+      "loss": 0.9871,
+      "step": 8103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3398574417918416e-05,
+      "loss": 0.9945,
+      "step": 8104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3397041110943775e-05,
+      "loss": 0.8726,
+      "step": 8105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3395507713679665e-05,
+      "loss": 1.032,
+      "step": 8106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393974226166838e-05,
+      "loss": 0.8577,
+      "step": 8107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3392440648446047e-05,
+      "loss": 0.8982,
+      "step": 8108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3390906980558063e-05,
+      "loss": 0.9017,
+      "step": 8109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389373222543642e-05,
+      "loss": 1.0361,
+      "step": 8110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3387839374443551e-05,
+      "loss": 0.5301,
+      "step": 8111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3386305436298563e-05,
+      "loss": 0.939,
+      "step": 8112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.338477140814944e-05,
+      "loss": 0.9715,
+      "step": 8113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3383237290036962e-05,
+      "loss": 0.8941,
+      "step": 8114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3381703082001897e-05,
+      "loss": 0.9507,
+      "step": 8115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3380168784085028e-05,
+      "loss": 0.9269,
+      "step": 8116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3378634396327135e-05,
+      "loss": 0.4561,
+      "step": 8117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3377099918768996e-05,
+      "loss": 0.9904,
+      "step": 8118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33755653514514e-05,
+      "loss": 0.9045,
+      "step": 8119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3374030694415132e-05,
+      "loss": 1.0223,
+      "step": 8120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3372495947700979e-05,
+      "loss": 1.0127,
+      "step": 8121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3370961111349738e-05,
+      "loss": 0.9113,
+      "step": 8122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33694261854022e-05,
+      "loss": 1.0053,
+      "step": 8123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3367891169899166e-05,
+      "loss": 0.9476,
+      "step": 8124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3366356064881427e-05,
+      "loss": 0.9461,
+      "step": 8125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336482087038979e-05,
+      "loss": 0.891,
+      "step": 8126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3363285586465054e-05,
+      "loss": 1.0086,
+      "step": 8127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361750213148033e-05,
+      "loss": 0.9297,
+      "step": 8128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3360214750479527e-05,
+      "loss": 0.9184,
+      "step": 8129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3358679198500352e-05,
+      "loss": 0.9354,
+      "step": 8130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335714355725132e-05,
+      "loss": 0.9698,
+      "step": 8131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335560782677325e-05,
+      "loss": 0.8578,
+      "step": 8132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335407200710695e-05,
+      "loss": 0.9685,
+      "step": 8133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3352536098293249e-05,
+      "loss": 1.0055,
+      "step": 8134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335100010037297e-05,
+      "loss": 1.0494,
+      "step": 8135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349464013386935e-05,
+      "loss": 0.92,
+      "step": 8136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334792783737597e-05,
+      "loss": 1.0008,
+      "step": 8137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334639157238091e-05,
+      "loss": 0.9191,
+      "step": 8138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3344855218442579e-05,
+      "loss": 0.8998,
+      "step": 8139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334331877560182e-05,
+      "loss": 0.8639,
+      "step": 8140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3341782243899467e-05,
+      "loss": 0.9556,
+      "step": 8141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3340245623376355e-05,
+      "loss": 0.9782,
+      "step": 8142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3338708914073333e-05,
+      "loss": 0.9162,
+      "step": 8143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337172116031239e-05,
+      "loss": 1.0381,
+      "step": 8144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333563522929092e-05,
+      "loss": 0.9576,
+      "step": 8145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334098253893231e-05,
+      "loss": 0.98,
+      "step": 8146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3332561189879015e-05,
+      "loss": 1.0731,
+      "step": 8147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333102403728913e-05,
+      "loss": 0.9558,
+      "step": 8148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3329486796164428e-05,
+      "loss": 0.9911,
+      "step": 8149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3327949466545773e-05,
+      "loss": 0.9943,
+      "step": 8150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332641204847402e-05,
+      "loss": 0.962,
+      "step": 8151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3324874541990034e-05,
+      "loss": 1.035,
+      "step": 8152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323336947134681e-05,
+      "loss": 0.9794,
+      "step": 8153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332179926394883e-05,
+      "loss": 0.8676,
+      "step": 8154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320261492473345e-05,
+      "loss": 0.9163,
+      "step": 8155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33187236327491e-05,
+      "loss": 0.8973,
+      "step": 8156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317185684816974e-05,
+      "loss": 1.0362,
+      "step": 8157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331564764871784e-05,
+      "loss": 0.9574,
+      "step": 8158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331410952449258e-05,
+      "loss": 0.854,
+      "step": 8159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3312571312182073e-05,
+      "loss": 0.9354,
+      "step": 8160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3311033011827203e-05,
+      "loss": 0.9788,
+      "step": 8161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309494623468857e-05,
+      "loss": 0.8453,
+      "step": 8162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3307956147147924e-05,
+      "loss": 0.9291,
+      "step": 8163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3306417582905296e-05,
+      "loss": 0.9831,
+      "step": 8164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3304878930781863e-05,
+      "loss": 0.9951,
+      "step": 8165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3303340190818525e-05,
+      "loss": 0.9292,
+      "step": 8166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3301801363056175e-05,
+      "loss": 1.0639,
+      "step": 8167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300262447535717e-05,
+      "loss": 0.9318,
+      "step": 8168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298723444298052e-05,
+      "loss": 1.0269,
+      "step": 8169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3297184353384085e-05,
+      "loss": 0.9694,
+      "step": 8170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295645174834726e-05,
+      "loss": 0.9325,
+      "step": 8171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3294105908690882e-05,
+      "loss": 0.9737,
+      "step": 8172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292566554993462e-05,
+      "loss": 0.9114,
+      "step": 8173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3291027113783383e-05,
+      "loss": 0.8712,
+      "step": 8174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3289487585101564e-05,
+      "loss": 0.9135,
+      "step": 8175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287947968988923e-05,
+      "loss": 0.9113,
+      "step": 8176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3286408265486381e-05,
+      "loss": 0.9373,
+      "step": 8177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3284868474634857e-05,
+      "loss": 0.9738,
+      "step": 8178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3283328596475284e-05,
+      "loss": 0.9683,
+      "step": 8179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281788631048585e-05,
+      "loss": 1.0239,
+      "step": 8180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328024857839569e-05,
+      "loss": 0.9259,
+      "step": 8181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327870843855754e-05,
+      "loss": 0.8764,
+      "step": 8182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327716821157506e-05,
+      "loss": 0.8794,
+      "step": 8183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3275627897489193e-05,
+      "loss": 0.9267,
+      "step": 8184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327408749634088e-05,
+      "loss": 0.8797,
+      "step": 8185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272547008171055e-05,
+      "loss": 0.4511,
+      "step": 8186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3271006433020675e-05,
+      "loss": 0.9134,
+      "step": 8187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3269465770930674e-05,
+      "loss": 0.4918,
+      "step": 8188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326792502194201e-05,
+      "loss": 0.987,
+      "step": 8189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326638418609563e-05,
+      "loss": 0.8892,
+      "step": 8190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264843263432491e-05,
+      "loss": 0.9139,
+      "step": 8191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3263302253993544e-05,
+      "loss": 0.8791,
+      "step": 8192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261761157819752e-05,
+      "loss": 0.9474,
+      "step": 8193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3260219974952074e-05,
+      "loss": 0.8984,
+      "step": 8194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3258678705431472e-05,
+      "loss": 0.8771,
+      "step": 8195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3257137349298912e-05,
+      "loss": 0.9436,
+      "step": 8196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255595906595359e-05,
+      "loss": 0.906,
+      "step": 8197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3254054377361787e-05,
+      "loss": 0.9817,
+      "step": 8198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3252512761639165e-05,
+      "loss": 1.0376,
+      "step": 8199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3250971059468472e-05,
+      "loss": 0.9873,
+      "step": 8200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3249429270890677e-05,
+      "loss": 0.9038,
+      "step": 8201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247887395946766e-05,
+      "loss": 0.9305,
+      "step": 8202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246345434677719e-05,
+      "loss": 1.0316,
+      "step": 8203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3244803387124512e-05,
+      "loss": 0.9638,
+      "step": 8204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3243261253328141e-05,
+      "loss": 0.9232,
+      "step": 8205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3241719033329587e-05,
+      "loss": 1.0166,
+      "step": 8206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3240176727169849e-05,
+      "loss": 0.9394,
+      "step": 8207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323863433488991e-05,
+      "loss": 0.9599,
+      "step": 8208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3237091856530768e-05,
+      "loss": 0.9898,
+      "step": 8209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3235549292133425e-05,
+      "loss": 1.0405,
+      "step": 8210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3234006641738877e-05,
+      "loss": 0.9395,
+      "step": 8211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3232463905388125e-05,
+      "loss": 0.9921,
+      "step": 8212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230921083122175e-05,
+      "loss": 0.941,
+      "step": 8213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.322937817498203e-05,
+      "loss": 0.8685,
+      "step": 8214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3227835181008704e-05,
+      "loss": 0.9103,
+      "step": 8215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3226292101243206e-05,
+      "loss": 0.935,
+      "step": 8216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224748935726545e-05,
+      "loss": 1.0095,
+      "step": 8217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3223205684499745e-05,
+      "loss": 0.7993,
+      "step": 8218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3221662347603816e-05,
+      "loss": 0.9427,
+      "step": 8219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3220118925079783e-05,
+      "loss": 0.9087,
+      "step": 8220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218575416968666e-05,
+      "loss": 1.0,
+      "step": 8221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3217031823311488e-05,
+      "loss": 0.9241,
+      "step": 8222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3215488144149285e-05,
+      "loss": 0.9307,
+      "step": 8223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3213944379523076e-05,
+      "loss": 0.9022,
+      "step": 8224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3212400529473893e-05,
+      "loss": 0.9374,
+      "step": 8225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3210856594042775e-05,
+      "loss": 0.865,
+      "step": 8226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209312573270754e-05,
+      "loss": 0.8843,
+      "step": 8227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3207768467198874e-05,
+      "loss": 0.9459,
+      "step": 8228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3206224275868171e-05,
+      "loss": 0.918,
+      "step": 8229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3204679999319688e-05,
+      "loss": 0.97,
+      "step": 8230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203135637594471e-05,
+      "loss": 0.9738,
+      "step": 8231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3201591190733564e-05,
+      "loss": 0.9193,
+      "step": 8232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3200046658778023e-05,
+      "loss": 0.994,
+      "step": 8233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3198502041768897e-05,
+      "loss": 0.967,
+      "step": 8234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3196957339747236e-05,
+      "loss": 0.8991,
+      "step": 8235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3195412552754103e-05,
+      "loss": 0.913,
+      "step": 8236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193867680830554e-05,
+      "loss": 0.8783,
+      "step": 8237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3192322724017652e-05,
+      "loss": 0.9311,
+      "step": 8238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3190777682356452e-05,
+      "loss": 0.9891,
+      "step": 8239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318923255588803e-05,
+      "loss": 1.0018,
+      "step": 8240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187687344653448e-05,
+      "loss": 1.0662,
+      "step": 8241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3186142048693777e-05,
+      "loss": 0.8366,
+      "step": 8242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318459666805009e-05,
+      "loss": 1.0173,
+      "step": 8243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3183051202763459e-05,
+      "loss": 0.9593,
+      "step": 8244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181505652874962e-05,
+      "loss": 0.9605,
+      "step": 8245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179960018425682e-05,
+      "loss": 0.9383,
+      "step": 8246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3178414299456695e-05,
+      "loss": 0.95,
+      "step": 8247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3176868496009084e-05,
+      "loss": 0.8779,
+      "step": 8248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317532260812394e-05,
+      "loss": 0.9109,
+      "step": 8249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3173776635842343e-05,
+      "loss": 0.9505,
+      "step": 8250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317223057920539e-05,
+      "loss": 0.9374,
+      "step": 8251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3170684438254173e-05,
+      "loss": 0.9835,
+      "step": 8252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169138213029782e-05,
+      "loss": 0.9129,
+      "step": 8253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167591903573321e-05,
+      "loss": 1.0319,
+      "step": 8254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.316604550992588e-05,
+      "loss": 0.9035,
+      "step": 8255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164499032128571e-05,
+      "loss": 0.9857,
+      "step": 8256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3162952470222488e-05,
+      "loss": 0.9854,
+      "step": 8257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3161405824248738e-05,
+      "loss": 0.9661,
+      "step": 8258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3159859094248437e-05,
+      "loss": 0.8092,
+      "step": 8259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315831228026269e-05,
+      "loss": 0.9059,
+      "step": 8260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3156765382332605e-05,
+      "loss": 1.0368,
+      "step": 8261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3155218400499307e-05,
+      "loss": 0.9272,
+      "step": 8262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3153671334803906e-05,
+      "loss": 0.9761,
+      "step": 8263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315212418528752e-05,
+      "loss": 0.9768,
+      "step": 8264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150576951991277e-05,
+      "loss": 0.9955,
+      "step": 8265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3149029634956294e-05,
+      "loss": 0.9662,
+      "step": 8266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3147482234223704e-05,
+      "loss": 0.9411,
+      "step": 8267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3145934749834624e-05,
+      "loss": 0.9459,
+      "step": 8268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144387181830195e-05,
+      "loss": 0.9285,
+      "step": 8269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3142839530251548e-05,
+      "loss": 0.9154,
+      "step": 8270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3141291795139812e-05,
+      "loss": 0.9207,
+      "step": 8271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3139743976536133e-05,
+      "loss": 1.0123,
+      "step": 8272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3138196074481641e-05,
+      "loss": 0.9751,
+      "step": 8273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3136648089017481e-05,
+      "loss": 0.903,
+      "step": 8274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31351000201848e-05,
+      "loss": 1.0147,
+      "step": 8275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3133551868024736e-05,
+      "loss": 0.9629,
+      "step": 8276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3132003632578448e-05,
+      "loss": 0.4779,
+      "step": 8277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313045531388708e-05,
+      "loss": 0.9504,
+      "step": 8278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312890691199178e-05,
+      "loss": 0.9952,
+      "step": 8279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3127358426933712e-05,
+      "loss": 0.9512,
+      "step": 8280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125809858754028e-05,
+      "loss": 1.0142,
+      "step": 8281
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3124261207493887e-05,
+      "loss": 0.8892,
+      "step": 8282
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3122712473194455e-05,
+      "loss": 0.9946,
+      "step": 8283
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3121163655896889e-05,
+      "loss": 0.8942,
+      "step": 8284
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3119614755642358e-05,
+      "loss": 0.9928,
+      "step": 8285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3118065772472028e-05,
+      "loss": 0.9292,
+      "step": 8286
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116516706427071e-05,
+      "loss": 0.9808,
+      "step": 8287
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3114967557548664e-05,
+      "loss": 1.0354,
+      "step": 8288
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113418325877974e-05,
+      "loss": 0.9338,
+      "step": 8289
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3111869011456179e-05,
+      "loss": 0.901,
+      "step": 8290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3110319614324464e-05,
+      "loss": 0.8928,
+      "step": 8291
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108770134524001e-05,
+      "loss": 0.9605,
+      "step": 8292
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3107220572095982e-05,
+      "loss": 0.9786,
+      "step": 8293
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3105670927081588e-05,
+      "loss": 0.9601,
+      "step": 8294
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3104121199522009e-05,
+      "loss": 0.8708,
+      "step": 8295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3102571389458434e-05,
+      "loss": 0.9443,
+      "step": 8296
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101021496932051e-05,
+      "loss": 0.9132,
+      "step": 8297
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309947152198406e-05,
+      "loss": 0.9413,
+      "step": 8298
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3097921464655659e-05,
+      "loss": 0.9467,
+      "step": 8299
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309637132498804e-05,
+      "loss": 0.9259,
+      "step": 8300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3094821103022408e-05,
+      "loss": 1.0332,
+      "step": 8301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3093270798799966e-05,
+      "loss": 1.0378,
+      "step": 8302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309172041236192e-05,
+      "loss": 0.955,
+      "step": 8303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3090169943749475e-05,
+      "loss": 0.9074,
+      "step": 8304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088619393003844e-05,
+      "loss": 1.0293,
+      "step": 8305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3087068760166237e-05,
+      "loss": 0.9644,
+      "step": 8306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3085518045277865e-05,
+      "loss": 0.9156,
+      "step": 8307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3083967248379952e-05,
+      "loss": 0.9482,
+      "step": 8308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3082416369513711e-05,
+      "loss": 0.8833,
+      "step": 8309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3080865408720358e-05,
+      "loss": 0.9013,
+      "step": 8310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3079314366041128e-05,
+      "loss": 0.9511,
+      "step": 8311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3077763241517236e-05,
+      "loss": 0.9075,
+      "step": 8312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076212035189912e-05,
+      "loss": 0.8853,
+      "step": 8313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3074660747100386e-05,
+      "loss": 1.0022,
+      "step": 8314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073109377289888e-05,
+      "loss": 0.9124,
+      "step": 8315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3071557925799652e-05,
+      "loss": 1.0231,
+      "step": 8316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3070006392670917e-05,
+      "loss": 0.9683,
+      "step": 8317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3068454777944916e-05,
+      "loss": 0.914,
+      "step": 8318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3066903081662894e-05,
+      "loss": 0.9822,
+      "step": 8319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3065351303866086e-05,
+      "loss": 0.8535,
+      "step": 8320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3063799444595748e-05,
+      "loss": 0.9746,
+      "step": 8321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3062247503893116e-05,
+      "loss": 0.9203,
+      "step": 8322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3060695481799442e-05,
+      "loss": 0.952,
+      "step": 8323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3059143378355981e-05,
+      "loss": 0.8807,
+      "step": 8324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305759119360398e-05,
+      "loss": 0.9809,
+      "step": 8325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30560389275847e-05,
+      "loss": 0.4235,
+      "step": 8326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3054486580339394e-05,
+      "loss": 0.9033,
+      "step": 8327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305293415190932e-05,
+      "loss": 0.8838,
+      "step": 8328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3051381642335748e-05,
+      "loss": 0.9504,
+      "step": 8329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3049829051659934e-05,
+      "loss": 0.991,
+      "step": 8330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3048276379923148e-05,
+      "loss": 0.8948,
+      "step": 8331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3046723627166658e-05,
+      "loss": 1.0689,
+      "step": 8332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045170793431731e-05,
+      "loss": 0.874,
+      "step": 8333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043617878759645e-05,
+      "loss": 0.9023,
+      "step": 8334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3042064883191672e-05,
+      "loss": 0.9896,
+      "step": 8335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3040511806769087e-05,
+      "loss": 0.9158,
+      "step": 8336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3038958649533174e-05,
+      "loss": 0.9348,
+      "step": 8337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037405411525208e-05,
+      "loss": 0.9942,
+      "step": 8338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3035852092786475e-05,
+      "loss": 1.0244,
+      "step": 8339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3034298693358263e-05,
+      "loss": 0.9928,
+      "step": 8340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3032745213281854e-05,
+      "loss": 1.0122,
+      "step": 8341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3031191652598545e-05,
+      "loss": 1.0383,
+      "step": 8342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3029638011349618e-05,
+      "loss": 0.9705,
+      "step": 8343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302808428957638e-05,
+      "loss": 0.9333,
+      "step": 8344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026530487320115e-05,
+      "loss": 0.9927,
+      "step": 8345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024976604622125e-05,
+      "loss": 0.9996,
+      "step": 8346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3023422641523718e-05,
+      "loss": 0.9841,
+      "step": 8347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3021868598066187e-05,
+      "loss": 0.9706,
+      "step": 8348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3020314474290838e-05,
+      "loss": 0.9485,
+      "step": 8349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018760270238984e-05,
+      "loss": 0.9669,
+      "step": 8350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017205985951926e-05,
+      "loss": 0.9478,
+      "step": 8351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3015651621470978e-05,
+      "loss": 0.9403,
+      "step": 8352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3014097176837457e-05,
+      "loss": 0.9409,
+      "step": 8353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012542652092674e-05,
+      "loss": 0.952,
+      "step": 8354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301098804727795e-05,
+      "loss": 0.9141,
+      "step": 8355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30094333624346e-05,
+      "loss": 0.9567,
+      "step": 8356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300787859760395e-05,
+      "loss": 0.9061,
+      "step": 8357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3006323752827321e-05,
+      "loss": 1.0088,
+      "step": 8358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3004768828146039e-05,
+      "loss": 1.0317,
+      "step": 8359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003213823601438e-05,
+      "loss": 0.9553,
+      "step": 8360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3001658739234841e-05,
+      "loss": 0.9867,
+      "step": 8361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3000103575087581e-05,
+      "loss": 0.8431,
+      "step": 8362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2998548331200998e-05,
+      "loss": 0.8991,
+      "step": 8363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299699300761642e-05,
+      "loss": 0.9542,
+      "step": 8364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2995437604375196e-05,
+      "loss": 0.8987,
+      "step": 8365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2993882121518659e-05,
+      "loss": 0.9336,
+      "step": 8366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2992326559088155e-05,
+      "loss": 0.9207,
+      "step": 8367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2990770917125031e-05,
+      "loss": 0.8762,
+      "step": 8368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2989215195670628e-05,
+      "loss": 0.9819,
+      "step": 8369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987659394766299e-05,
+      "loss": 1.0373,
+      "step": 8370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2986103514453398e-05,
+      "loss": 0.9366,
+      "step": 8371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2984547554773274e-05,
+      "loss": 0.8383,
+      "step": 8372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2982991515767287e-05,
+      "loss": 0.9445,
+      "step": 8373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2981435397476792e-05,
+      "loss": 0.9682,
+      "step": 8374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2979879199943145e-05,
+      "loss": 0.9387,
+      "step": 8375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978322923207717e-05,
+      "loss": 0.8578,
+      "step": 8376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2976766567311868e-05,
+      "loss": 0.8638,
+      "step": 8377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2975210132296963e-05,
+      "loss": 0.9296,
+      "step": 8378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297365361820437e-05,
+      "loss": 0.9856,
+      "step": 8379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972097025075456e-05,
+      "loss": 0.9883,
+      "step": 8380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2970540352951602e-05,
+      "loss": 1.0094,
+      "step": 8381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2968983601874176e-05,
+      "loss": 0.8843,
+      "step": 8382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2967426771884557e-05,
+      "loss": 0.9573,
+      "step": 8383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965869863024125e-05,
+      "loss": 0.872,
+      "step": 8384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2964312875334258e-05,
+      "loss": 0.9381,
+      "step": 8385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2962755808856341e-05,
+      "loss": 0.9936,
+      "step": 8386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2961198663631759e-05,
+      "loss": 0.8964,
+      "step": 8387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959641439701897e-05,
+      "loss": 0.4352,
+      "step": 8388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2958084137108147e-05,
+      "loss": 0.9189,
+      "step": 8389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956526755891899e-05,
+      "loss": 0.9561,
+      "step": 8390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954969296094547e-05,
+      "loss": 0.9024,
+      "step": 8391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2953411757757484e-05,
+      "loss": 0.8978,
+      "step": 8392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.295185414092211e-05,
+      "loss": 0.9134,
+      "step": 8393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2950296445629827e-05,
+      "loss": 0.4289,
+      "step": 8394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294873867192203e-05,
+      "loss": 0.997,
+      "step": 8395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294718081984013e-05,
+      "loss": 0.9838,
+      "step": 8396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294562288942553e-05,
+      "loss": 1.0228,
+      "step": 8397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2944064880719634e-05,
+      "loss": 1.0357,
+      "step": 8398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2942506793763858e-05,
+      "loss": 0.9351,
+      "step": 8399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2940948628599614e-05,
+      "loss": 0.9328,
+      "step": 8400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2939390385268314e-05,
+      "loss": 0.9715,
+      "step": 8401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2937832063811374e-05,
+      "loss": 0.8524,
+      "step": 8402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936273664270211e-05,
+      "loss": 0.9314,
+      "step": 8403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2934715186686249e-05,
+      "loss": 0.9623,
+      "step": 8404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293315663110091e-05,
+      "loss": 0.9893,
+      "step": 8405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2931597997555617e-05,
+      "loss": 1.0277,
+      "step": 8406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2930039286091798e-05,
+      "loss": 0.9167,
+      "step": 8407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292848049675088e-05,
+      "loss": 0.993,
+      "step": 8408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2926921629574296e-05,
+      "loss": 0.9133,
+      "step": 8409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925362684603478e-05,
+      "loss": 0.9768,
+      "step": 8410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292380366187986e-05,
+      "loss": 0.4372,
+      "step": 8411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2922244561444881e-05,
+      "loss": 0.9662,
+      "step": 8412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2920685383339977e-05,
+      "loss": 0.9195,
+      "step": 8413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2919126127606593e-05,
+      "loss": 0.9026,
+      "step": 8414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2917566794286171e-05,
+      "loss": 1.0063,
+      "step": 8415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916007383420152e-05,
+      "loss": 0.9215,
+      "step": 8416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914447895049993e-05,
+      "loss": 0.8891,
+      "step": 8417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2912888329217138e-05,
+      "loss": 0.9605,
+      "step": 8418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2911328685963032e-05,
+      "loss": 0.9537,
+      "step": 8419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2909768965329139e-05,
+      "loss": 0.9057,
+      "step": 8420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908209167356906e-05,
+      "loss": 1.0289,
+      "step": 8421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2906649292087799e-05,
+      "loss": 0.9562,
+      "step": 8422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2905089339563273e-05,
+      "loss": 0.9337,
+      "step": 8423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2903529309824788e-05,
+      "loss": 0.9215,
+      "step": 8424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2901969202913813e-05,
+      "loss": 0.9301,
+      "step": 8425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900409018871807e-05,
+      "loss": 0.9564,
+      "step": 8426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898848757740246e-05,
+      "loss": 0.8563,
+      "step": 8427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2897288419560594e-05,
+      "loss": 0.8516,
+      "step": 8428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895728004374321e-05,
+      "loss": 0.8978,
+      "step": 8429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2894167512222913e-05,
+      "loss": 0.9145,
+      "step": 8430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2892606943147835e-05,
+      "loss": 0.9022,
+      "step": 8431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2891046297190566e-05,
+      "loss": 1.0219,
+      "step": 8432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288948557439259e-05,
+      "loss": 0.8709,
+      "step": 8433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2887924774795382e-05,
+      "loss": 1.0006,
+      "step": 8434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2886363898440437e-05,
+      "loss": 0.9232,
+      "step": 8435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2884802945369237e-05,
+      "loss": 0.9815,
+      "step": 8436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2883241915623267e-05,
+      "loss": 0.9423,
+      "step": 8437
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2881680809244023e-05,
+      "loss": 1.0034,
+      "step": 8438
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288011962627299e-05,
+      "loss": 0.9107,
+      "step": 8439
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.287855836675167e-05,
+      "loss": 1.03,
+      "step": 8440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2876997030721557e-05,
+      "loss": 0.8942,
+      "step": 8441
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875435618224145e-05,
+      "loss": 0.9576,
+      "step": 8442
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.287387412930094e-05,
+      "loss": 0.8184,
+      "step": 8443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872312563993444e-05,
+      "loss": 0.9623,
+      "step": 8444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.287075092234316e-05,
+      "loss": 0.8803,
+      "step": 8445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2869189204391595e-05,
+      "loss": 0.9535,
+      "step": 8446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2867627410180257e-05,
+      "loss": 0.9904,
+      "step": 8447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2866065539750661e-05,
+      "loss": 0.9429,
+      "step": 8448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2864503593144315e-05,
+      "loss": 0.8727,
+      "step": 8449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862941570402731e-05,
+      "loss": 0.9115,
+      "step": 8450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2861379471567434e-05,
+      "loss": 0.9474,
+      "step": 8451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2859817296679936e-05,
+      "loss": 0.9339,
+      "step": 8452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858255045781762e-05,
+      "loss": 0.4429,
+      "step": 8453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2856692718914435e-05,
+      "loss": 0.963,
+      "step": 8454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2855130316119476e-05,
+      "loss": 0.9983,
+      "step": 8455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2853567837438415e-05,
+      "loss": 0.9407,
+      "step": 8456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852005282912779e-05,
+      "loss": 1.0083,
+      "step": 8457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2850442652584098e-05,
+      "loss": 0.935,
+      "step": 8458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848879946493911e-05,
+      "loss": 0.9461,
+      "step": 8459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2847317164683745e-05,
+      "loss": 0.4791,
+      "step": 8460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2845754307195144e-05,
+      "loss": 0.9258,
+      "step": 8461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2844191374069641e-05,
+      "loss": 1.0639,
+      "step": 8462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284262836534878e-05,
+      "loss": 0.9842,
+      "step": 8463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2841065281074105e-05,
+      "loss": 0.9477,
+      "step": 8464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283950212128716e-05,
+      "loss": 0.9406,
+      "step": 8465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283793888602949e-05,
+      "loss": 0.9812,
+      "step": 8466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836375575342648e-05,
+      "loss": 0.9252,
+      "step": 8467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283481218926818e-05,
+      "loss": 1.0073,
+      "step": 8468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2833248727847643e-05,
+      "loss": 0.8704,
+      "step": 8469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2831685191122592e-05,
+      "loss": 0.9213,
+      "step": 8470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2830121579134578e-05,
+      "loss": 0.9086,
+      "step": 8471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282855789192517e-05,
+      "loss": 0.8329,
+      "step": 8472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2826994129535922e-05,
+      "loss": 0.9743,
+      "step": 8473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2825430292008398e-05,
+      "loss": 0.8263,
+      "step": 8474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2823866379384164e-05,
+      "loss": 0.9249,
+      "step": 8475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2822302391704786e-05,
+      "loss": 0.87,
+      "step": 8476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2820738329011837e-05,
+      "loss": 0.9255,
+      "step": 8477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2819174191346882e-05,
+      "loss": 0.9413,
+      "step": 8478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2817609978751497e-05,
+      "loss": 0.9151,
+      "step": 8479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2816045691267258e-05,
+      "loss": 0.9837,
+      "step": 8480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2814481328935737e-05,
+      "loss": 0.9244,
+      "step": 8481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2812916891798521e-05,
+      "loss": 0.9736,
+      "step": 8482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811352379897187e-05,
+      "loss": 0.9304,
+      "step": 8483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2809787793273314e-05,
+      "loss": 0.9013,
+      "step": 8484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808223131968496e-05,
+      "loss": 0.9787,
+      "step": 8485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2806658396024309e-05,
+      "loss": 0.902,
+      "step": 8486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280509358548235e-05,
+      "loss": 0.9277,
+      "step": 8487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2803528700384207e-05,
+      "loss": 0.94,
+      "step": 8488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280196374077147e-05,
+      "loss": 0.9468,
+      "step": 8489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2800398706685742e-05,
+      "loss": 0.8521,
+      "step": 8490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2798833598168612e-05,
+      "loss": 0.8837,
+      "step": 8491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2797268415261681e-05,
+      "loss": 0.9519,
+      "step": 8492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2795703158006551e-05,
+      "loss": 0.856,
+      "step": 8493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2794137826444823e-05,
+      "loss": 0.9433,
+      "step": 8494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2792572420618105e-05,
+      "loss": 0.953,
+      "step": 8495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2791006940568e-05,
+      "loss": 0.9245,
+      "step": 8496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789441386336119e-05,
+      "loss": 0.9107,
+      "step": 8497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2787875757964072e-05,
+      "loss": 0.956,
+      "step": 8498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786310055493469e-05,
+      "loss": 0.9595,
+      "step": 8499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278474427896593e-05,
+      "loss": 0.9201,
+      "step": 8500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783178428423066e-05,
+      "loss": 0.9878,
+      "step": 8501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27816125039065e-05,
+      "loss": 0.9147,
+      "step": 8502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2780046505457852e-05,
+      "loss": 0.8904,
+      "step": 8503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277848043311874e-05,
+      "loss": 0.9085,
+      "step": 8504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2776914286930793e-05,
+      "loss": 0.959,
+      "step": 8505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775348066935636e-05,
+      "loss": 0.9363,
+      "step": 8506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2773781773174894e-05,
+      "loss": 1.0996,
+      "step": 8507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2772215405690205e-05,
+      "loss": 0.9257,
+      "step": 8508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2770648964523194e-05,
+      "loss": 0.9582,
+      "step": 8509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769082449715498e-05,
+      "loss": 0.9264,
+      "step": 8510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767515861308753e-05,
+      "loss": 0.9954,
+      "step": 8511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765949199344596e-05,
+      "loss": 0.982,
+      "step": 8512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2764382463864671e-05,
+      "loss": 0.9404,
+      "step": 8513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762815654910617e-05,
+      "loss": 0.9897,
+      "step": 8514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2761248772524077e-05,
+      "loss": 0.9843,
+      "step": 8515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2759681816746699e-05,
+      "loss": 0.9432,
+      "step": 8516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.275811478762013e-05,
+      "loss": 0.9037,
+      "step": 8517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2756547685186021e-05,
+      "loss": 0.9939,
+      "step": 8518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2754980509486021e-05,
+      "loss": 0.9426,
+      "step": 8519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2753413260561786e-05,
+      "loss": 1.0202,
+      "step": 8520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2751845938454972e-05,
+      "loss": 0.9743,
+      "step": 8521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750278543207234e-05,
+      "loss": 0.9488,
+      "step": 8522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748711074860235e-05,
+      "loss": 0.9313,
+      "step": 8523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2747143533455637e-05,
+      "loss": 0.8475,
+      "step": 8524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745575919035098e-05,
+      "loss": 0.9443,
+      "step": 8525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2744008231640291e-05,
+      "loss": 0.9723,
+      "step": 8526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742440471312874e-05,
+      "loss": 0.8953,
+      "step": 8527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2740872638094524e-05,
+      "loss": 0.4838,
+      "step": 8528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273930473202691e-05,
+      "loss": 0.9178,
+      "step": 8529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2737736753151703e-05,
+      "loss": 0.9801,
+      "step": 8530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273616870151058e-05,
+      "loss": 0.9142,
+      "step": 8531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273460057714522e-05,
+      "loss": 0.9744,
+      "step": 8532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2733032380097301e-05,
+      "loss": 0.8841,
+      "step": 8533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2731464110408498e-05,
+      "loss": 1.007,
+      "step": 8534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2729895768120501e-05,
+      "loss": 1.0197,
+      "step": 8535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2728327353274995e-05,
+      "loss": 0.9762,
+      "step": 8536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2726758865913662e-05,
+      "loss": 0.9197,
+      "step": 8537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272519030607819e-05,
+      "loss": 0.9198,
+      "step": 8538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723621673810277e-05,
+      "loss": 0.9264,
+      "step": 8539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2722052969151607e-05,
+      "loss": 0.9648,
+      "step": 8540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2720484192143875e-05,
+      "loss": 0.9633,
+      "step": 8541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2718915342828786e-05,
+      "loss": 0.9666,
+      "step": 8542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2717346421248031e-05,
+      "loss": 0.982,
+      "step": 8543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2715777427443311e-05,
+      "loss": 0.9502,
+      "step": 8544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2714208361456327e-05,
+      "loss": 0.8369,
+      "step": 8545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2712639223328789e-05,
+      "loss": 0.912,
+      "step": 8546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2711070013102396e-05,
+      "loss": 0.9849,
+      "step": 8547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709500730818857e-05,
+      "loss": 0.9995,
+      "step": 8548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2707931376519886e-05,
+      "loss": 1.016,
+      "step": 8549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270636195024719e-05,
+      "loss": 0.9242,
+      "step": 8550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2704792452042486e-05,
+      "loss": 0.9101,
+      "step": 8551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2703222881947485e-05,
+      "loss": 0.9621,
+      "step": 8552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2701653240003908e-05,
+      "loss": 0.9489,
+      "step": 8553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2700083526253475e-05,
+      "loss": 1.0236,
+      "step": 8554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698513740737908e-05,
+      "loss": 0.4899,
+      "step": 8555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2696943883498921e-05,
+      "loss": 0.9311,
+      "step": 8556
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2695373954578252e-05,
+      "loss": 0.8708,
+      "step": 8557
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693803954017618e-05,
+      "loss": 0.9347,
+      "step": 8558
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2692233881858753e-05,
+      "loss": 0.8487,
+      "step": 8559
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2690663738143385e-05,
+      "loss": 0.9393,
+      "step": 8560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268909352291325e-05,
+      "loss": 0.8892,
+      "step": 8561
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268752323621008e-05,
+      "loss": 1.0108,
+      "step": 8562
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268595287807561e-05,
+      "loss": 0.9577,
+      "step": 8563
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268438244855158e-05,
+      "loss": 0.4115,
+      "step": 8564
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2682811947679731e-05,
+      "loss": 0.9552,
+      "step": 8565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681241375501805e-05,
+      "loss": 0.9885,
+      "step": 8566
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2679670732059546e-05,
+      "loss": 0.9202,
+      "step": 8567
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.26781000173947e-05,
+      "loss": 0.9442,
+      "step": 8568
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2676529231549013e-05,
+      "loss": 0.8812,
+      "step": 8569
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2674958374564234e-05,
+      "loss": 0.9904,
+      "step": 8570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2673387446482118e-05,
+      "loss": 0.9099,
+      "step": 8571
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2671816447344418e-05,
+      "loss": 0.9684,
+      "step": 8572
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267024537719289e-05,
+      "loss": 0.8791,
+      "step": 8573
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668674236069287e-05,
+      "loss": 0.8762,
+      "step": 8574
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.266710302401537e-05,
+      "loss": 0.9515,
+      "step": 8575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2665531741072902e-05,
+      "loss": 1.0061,
+      "step": 8576
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663960387283645e-05,
+      "loss": 0.9258,
+      "step": 8577
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662388962689365e-05,
+      "loss": 0.9132,
+      "step": 8578
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660817467331823e-05,
+      "loss": 0.9159,
+      "step": 8579
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2659245901252794e-05,
+      "loss": 0.9133,
+      "step": 8580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2657674264494047e-05,
+      "loss": 0.9861,
+      "step": 8581
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265610255709735e-05,
+      "loss": 0.9269,
+      "step": 8582
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654530779104484e-05,
+      "loss": 1.0105,
+      "step": 8583
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265295893055722e-05,
+      "loss": 0.9316,
+      "step": 8584
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2651387011497339e-05,
+      "loss": 0.4713,
+      "step": 8585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264981502196662e-05,
+      "loss": 0.8393,
+      "step": 8586
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264824296200684e-05,
+      "loss": 0.9645,
+      "step": 8587
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2646670831659789e-05,
+      "loss": 0.9236,
+      "step": 8588
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2645098630967251e-05,
+      "loss": 0.9537,
+      "step": 8589
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2643526359971013e-05,
+      "loss": 0.4665,
+      "step": 8590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2641954018712863e-05,
+      "loss": 0.875,
+      "step": 8591
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2640381607234591e-05,
+      "loss": 0.9194,
+      "step": 8592
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263880912557799e-05,
+      "loss": 0.8992,
+      "step": 8593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637236573784862e-05,
+      "loss": 0.981,
+      "step": 8594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2635663951896992e-05,
+      "loss": 0.9453,
+      "step": 8595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634091259956187e-05,
+      "loss": 0.8709,
+      "step": 8596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2632518498004245e-05,
+      "loss": 1.0109,
+      "step": 8597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263094566608297e-05,
+      "loss": 0.9311,
+      "step": 8598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262937276423416e-05,
+      "loss": 0.9724,
+      "step": 8599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627799792499625e-05,
+      "loss": 0.949,
+      "step": 8600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2626226750921175e-05,
+      "loss": 0.8177,
+      "step": 8601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2624653639540617e-05,
+      "loss": 0.9289,
+      "step": 8602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2623080458399763e-05,
+      "loss": 0.8725,
+      "step": 8603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621507207540426e-05,
+      "loss": 0.9371,
+      "step": 8604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2619933887004418e-05,
+      "loss": 0.9465,
+      "step": 8605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2618360496833563e-05,
+      "loss": 0.9576,
+      "step": 8606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261678703706968e-05,
+      "loss": 0.8951,
+      "step": 8607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2615213507754583e-05,
+      "loss": 1.015,
+      "step": 8608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2613639908930097e-05,
+      "loss": 0.4251,
+      "step": 8609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2612066240638046e-05,
+      "loss": 0.9797,
+      "step": 8610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610492502920258e-05,
+      "loss": 1.0041,
+      "step": 8611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608918695818564e-05,
+      "loss": 0.9669,
+      "step": 8612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607344819374788e-05,
+      "loss": 0.8479,
+      "step": 8613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2605770873630764e-05,
+      "loss": 0.9185,
+      "step": 8614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2604196858628328e-05,
+      "loss": 0.8872,
+      "step": 8615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2602622774409315e-05,
+      "loss": 1.0237,
+      "step": 8616
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2601048621015558e-05,
+      "loss": 0.9316,
+      "step": 8617
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.25994743984889e-05,
+      "loss": 0.912,
+      "step": 8618
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2597900106871184e-05,
+      "loss": 0.9829,
+      "step": 8619
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2596325746204246e-05,
+      "loss": 0.9827,
+      "step": 8620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2594751316529939e-05,
+      "loss": 0.8838,
+      "step": 8621
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593176817890105e-05,
+      "loss": 0.9779,
+      "step": 8622
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2591602250326588e-05,
+      "loss": 0.9251,
+      "step": 8623
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2590027613881249e-05,
+      "loss": 0.9076,
+      "step": 8624
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2588452908595934e-05,
+      "loss": 0.9749,
+      "step": 8625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2586878134512492e-05,
+      "loss": 0.92,
+      "step": 8626
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585303291672788e-05,
+      "loss": 0.987,
+      "step": 8627
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583728380118674e-05,
+      "loss": 0.9276,
+      "step": 8628
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2582153399892007e-05,
+      "loss": 0.9875,
+      "step": 8629
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2580578351034657e-05,
+      "loss": 1.0286,
+      "step": 8630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257900323358848e-05,
+      "loss": 0.8747,
+      "step": 8631
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2577428047595343e-05,
+      "loss": 0.9385,
+      "step": 8632
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575852793097113e-05,
+      "loss": 0.9618,
+      "step": 8633
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2574277470135656e-05,
+      "loss": 0.9867,
+      "step": 8634
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572702078752844e-05,
+      "loss": 0.9706,
+      "step": 8635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2571126618990546e-05,
+      "loss": 1.0235,
+      "step": 8636
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2569551090890643e-05,
+      "loss": 0.96,
+      "step": 8637
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2567975494495008e-05,
+      "loss": 0.9785,
+      "step": 8638
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566399829845515e-05,
+      "loss": 0.9195,
+      "step": 8639
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2564824096984044e-05,
+      "loss": 0.931,
+      "step": 8640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256324829595248e-05,
+      "loss": 0.9168,
+      "step": 8641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2561672426792704e-05,
+      "loss": 0.8715,
+      "step": 8642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560096489546601e-05,
+      "loss": 0.934,
+      "step": 8643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2558520484256054e-05,
+      "loss": 0.9662,
+      "step": 8644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255694441096296e-05,
+      "loss": 0.8607,
+      "step": 8645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2555368269709198e-05,
+      "loss": 0.96,
+      "step": 8646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2553792060536667e-05,
+      "loss": 0.8997,
+      "step": 8647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2552215783487264e-05,
+      "loss": 0.9476,
+      "step": 8648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2550639438602876e-05,
+      "loss": 0.9381,
+      "step": 8649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2549063025925407e-05,
+      "loss": 1.0368,
+      "step": 8650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2547486545496755e-05,
+      "loss": 0.8837,
+      "step": 8651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2545909997358816e-05,
+      "loss": 0.8668,
+      "step": 8652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2544333381553502e-05,
+      "loss": 0.8202,
+      "step": 8653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254275669812271e-05,
+      "loss": 0.869,
+      "step": 8654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254117994710835e-05,
+      "loss": 0.8601,
+      "step": 8655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539603128552331e-05,
+      "loss": 0.9622,
+      "step": 8656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2538026242496556e-05,
+      "loss": 0.878,
+      "step": 8657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536449288982945e-05,
+      "loss": 1.0596,
+      "step": 8658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2534872268053408e-05,
+      "loss": 0.9881,
+      "step": 8659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2533295179749865e-05,
+      "loss": 0.9978,
+      "step": 8660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2531718024114228e-05,
+      "loss": 0.9286,
+      "step": 8661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2530140801188412e-05,
+      "loss": 0.8694,
+      "step": 8662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2528563511014351e-05,
+      "loss": 0.9613,
+      "step": 8663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2526986153633956e-05,
+      "loss": 0.9354,
+      "step": 8664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2525408729089153e-05,
+      "loss": 0.4789,
+      "step": 8665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2523831237421873e-05,
+      "loss": 0.9009,
+      "step": 8666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2522253678674041e-05,
+      "loss": 0.9192,
+      "step": 8667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2520676052887588e-05,
+      "loss": 0.8946,
+      "step": 8668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2519098360104443e-05,
+      "loss": 0.993,
+      "step": 8669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2517520600366538e-05,
+      "loss": 1.0381,
+      "step": 8670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2515942773715815e-05,
+      "loss": 0.9286,
+      "step": 8671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2514364880194206e-05,
+      "loss": 0.9712,
+      "step": 8672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2512786919843649e-05,
+      "loss": 0.9707,
+      "step": 8673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511208892706085e-05,
+      "loss": 0.9443,
+      "step": 8674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509630798823457e-05,
+      "loss": 0.932,
+      "step": 8675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2508052638237708e-05,
+      "loss": 1.0006,
+      "step": 8676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2506474410990788e-05,
+      "loss": 0.9155,
+      "step": 8677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2504896117124636e-05,
+      "loss": 0.92,
+      "step": 8678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250331775668121e-05,
+      "loss": 0.9282,
+      "step": 8679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2501739329702453e-05,
+      "loss": 0.8875,
+      "step": 8680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2500160836230326e-05,
+      "loss": 0.8691,
+      "step": 8681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2498582276306779e-05,
+      "loss": 0.4265,
+      "step": 8682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2497003649973767e-05,
+      "loss": 0.9228,
+      "step": 8683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249542495727325e-05,
+      "loss": 0.9334,
+      "step": 8684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2493846198247189e-05,
+      "loss": 0.4158,
+      "step": 8685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2492267372937543e-05,
+      "loss": 1.0473,
+      "step": 8686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490688481386277e-05,
+      "loss": 1.0423,
+      "step": 8687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2489109523635351e-05,
+      "loss": 0.8615,
+      "step": 8688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2487530499726746e-05,
+      "loss": 0.884,
+      "step": 8689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485951409702415e-05,
+      "loss": 0.9804,
+      "step": 8690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2484372253604338e-05,
+      "loss": 1.03,
+      "step": 8691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482793031474483e-05,
+      "loss": 0.8766,
+      "step": 8692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2481213743354823e-05,
+      "loss": 0.9191,
+      "step": 8693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2479634389287336e-05,
+      "loss": 0.8373,
+      "step": 8694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2478054969314002e-05,
+      "loss": 0.929,
+      "step": 8695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2476475483476794e-05,
+      "loss": 0.9055,
+      "step": 8696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2474895931817699e-05,
+      "loss": 0.8717,
+      "step": 8697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2473316314378693e-05,
+      "loss": 0.9237,
+      "step": 8698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2471736631201762e-05,
+      "loss": 0.9741,
+      "step": 8699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2470156882328901e-05,
+      "loss": 1.0471,
+      "step": 8700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2468577067802087e-05,
+      "loss": 0.9446,
+      "step": 8701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2466997187663317e-05,
+      "loss": 0.9441,
+      "step": 8702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2465417241954578e-05,
+      "loss": 0.9145,
+      "step": 8703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2463837230717865e-05,
+      "loss": 0.9219,
+      "step": 8704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246225715399517e-05,
+      "loss": 0.8949,
+      "step": 8705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2460677011828493e-05,
+      "loss": 0.973,
+      "step": 8706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459096804259833e-05,
+      "loss": 0.9583,
+      "step": 8707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2457516531331187e-05,
+      "loss": 0.3658,
+      "step": 8708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245593619308456e-05,
+      "loss": 0.8851,
+      "step": 8709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2454355789561954e-05,
+      "loss": 0.9385,
+      "step": 8710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245277532080537e-05,
+      "loss": 0.8777,
+      "step": 8711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2451194786856824e-05,
+      "loss": 0.8978,
+      "step": 8712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2449614187758319e-05,
+      "loss": 0.9778,
+      "step": 8713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2448033523551866e-05,
+      "loss": 0.9296,
+      "step": 8714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2446452794279478e-05,
+      "loss": 0.4183,
+      "step": 8715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2444871999983168e-05,
+      "loss": 0.8688,
+      "step": 8716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2443291140704951e-05,
+      "loss": 0.9052,
+      "step": 8717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2441710216486851e-05,
+      "loss": 1.0523,
+      "step": 8718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244012922737088e-05,
+      "loss": 0.9853,
+      "step": 8719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438548173399058e-05,
+      "loss": 0.9106,
+      "step": 8720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2436967054613414e-05,
+      "loss": 0.9646,
+      "step": 8721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2435385871055969e-05,
+      "loss": 0.8962,
+      "step": 8722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2433804622768749e-05,
+      "loss": 0.4874,
+      "step": 8723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2432223309793778e-05,
+      "loss": 0.9625,
+      "step": 8724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2430641932173094e-05,
+      "loss": 0.9626,
+      "step": 8725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2429060489948724e-05,
+      "loss": 1.0087,
+      "step": 8726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2427478983162694e-05,
+      "loss": 0.887,
+      "step": 8727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242589741185705e-05,
+      "loss": 0.949,
+      "step": 8728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2424315776073821e-05,
+      "loss": 0.9021,
+      "step": 8729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.242273407585505e-05,
+      "loss": 0.967,
+      "step": 8730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2421152311242774e-05,
+      "loss": 0.9489,
+      "step": 8731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2419570482279035e-05,
+      "loss": 0.9119,
+      "step": 8732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417988589005875e-05,
+      "loss": 0.9768,
+      "step": 8733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241640663146534e-05,
+      "loss": 0.96,
+      "step": 8734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2414824609699476e-05,
+      "loss": 0.9131,
+      "step": 8735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2413242523750332e-05,
+      "loss": 0.9107,
+      "step": 8736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241166037365996e-05,
+      "loss": 0.9256,
+      "step": 8737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410078159470408e-05,
+      "loss": 0.9641,
+      "step": 8738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240849588122373e-05,
+      "loss": 0.9528,
+      "step": 8739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2406913538961984e-05,
+      "loss": 0.9649,
+      "step": 8740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2405331132727225e-05,
+      "loss": 0.9076,
+      "step": 8741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240374866256151e-05,
+      "loss": 0.9157,
+      "step": 8742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2402166128506902e-05,
+      "loss": 0.9008,
+      "step": 8743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400583530605463e-05,
+      "loss": 0.9109,
+      "step": 8744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2399000868899253e-05,
+      "loss": 0.954,
+      "step": 8745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397418143430344e-05,
+      "loss": 1.0003,
+      "step": 8746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2395835354240795e-05,
+      "loss": 0.894,
+      "step": 8747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2394252501372679e-05,
+      "loss": 0.9812,
+      "step": 8748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2392669584868067e-05,
+      "loss": 0.9108,
+      "step": 8749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239108660476903e-05,
+      "loss": 0.9939,
+      "step": 8750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389503561117642e-05,
+      "loss": 0.4474,
+      "step": 8751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387920453955977e-05,
+      "loss": 0.4292,
+      "step": 8752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2386337283326114e-05,
+      "loss": 0.9853,
+      "step": 8753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2384754049270134e-05,
+      "loss": 0.8777,
+      "step": 8754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2383170751830112e-05,
+      "loss": 1.0068,
+      "step": 8755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2381587391048135e-05,
+      "loss": 0.9297,
+      "step": 8756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2380003966966285e-05,
+      "loss": 0.9632,
+      "step": 8757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2378420479626648e-05,
+      "loss": 0.427,
+      "step": 8758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237683692907131e-05,
+      "loss": 0.9259,
+      "step": 8759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237525331534236e-05,
+      "loss": 0.899,
+      "step": 8760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237366963848189e-05,
+      "loss": 0.9297,
+      "step": 8761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2372085898531995e-05,
+      "loss": 0.7923,
+      "step": 8762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2370502095534764e-05,
+      "loss": 0.8425,
+      "step": 8763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2368918229532294e-05,
+      "loss": 0.8068,
+      "step": 8764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2367334300566688e-05,
+      "loss": 1.0049,
+      "step": 8765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2365750308680034e-05,
+      "loss": 0.9315,
+      "step": 8766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236416625391444e-05,
+      "loss": 0.921,
+      "step": 8767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2362582136312008e-05,
+      "loss": 0.9627,
+      "step": 8768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2360997955914842e-05,
+      "loss": 1.0278,
+      "step": 8769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2359413712765045e-05,
+      "loss": 0.9664,
+      "step": 8770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2357829406904726e-05,
+      "loss": 1.036,
+      "step": 8771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356245038375995e-05,
+      "loss": 0.9834,
+      "step": 8772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2354660607220963e-05,
+      "loss": 0.8775,
+      "step": 8773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2353076113481742e-05,
+      "loss": 0.8832,
+      "step": 8774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351491557200444e-05,
+      "loss": 0.9263,
+      "step": 8775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2349906938419183e-05,
+      "loss": 0.8956,
+      "step": 8776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2348322257180082e-05,
+      "loss": 0.4808,
+      "step": 8777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234673751352526e-05,
+      "loss": 0.867,
+      "step": 8778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345152707496833e-05,
+      "loss": 0.9176,
+      "step": 8779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2343567839136928e-05,
+      "loss": 0.8614,
+      "step": 8780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2341982908487661e-05,
+      "loss": 0.8997,
+      "step": 8781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340397915591167e-05,
+      "loss": 0.8729,
+      "step": 8782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2338812860489569e-05,
+      "loss": 0.8965,
+      "step": 8783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2337227743224994e-05,
+      "loss": 0.9898,
+      "step": 8784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335642563839578e-05,
+      "loss": 0.9327,
+      "step": 8785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2334057322375448e-05,
+      "loss": 0.9232,
+      "step": 8786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332472018874741e-05,
+      "loss": 0.9302,
+      "step": 8787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330886653379592e-05,
+      "loss": 0.8873,
+      "step": 8788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2329301225932137e-05,
+      "loss": 0.9477,
+      "step": 8789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2327715736574516e-05,
+      "loss": 0.9453,
+      "step": 8790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232613018534887e-05,
+      "loss": 0.9787,
+      "step": 8791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232454457229734e-05,
+      "loss": 0.9244,
+      "step": 8792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322958897462072e-05,
+      "loss": 0.9258,
+      "step": 8793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2321373160885204e-05,
+      "loss": 1.0147,
+      "step": 8794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2319787362608895e-05,
+      "loss": 0.9951,
+      "step": 8795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2318201502675285e-05,
+      "loss": 0.9886,
+      "step": 8796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2316615581126526e-05,
+      "loss": 0.999,
+      "step": 8797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2315029598004774e-05,
+      "loss": 0.8608,
+      "step": 8798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2313443553352173e-05,
+      "loss": 0.8897,
+      "step": 8799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2311857447210886e-05,
+      "loss": 0.8963,
+      "step": 8800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2310271279623071e-05,
+      "loss": 1.0063,
+      "step": 8801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2308685050630885e-05,
+      "loss": 0.9403,
+      "step": 8802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307098760276486e-05,
+      "loss": 1.0206,
+      "step": 8803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305512408602037e-05,
+      "loss": 0.8878,
+      "step": 8804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2303925995649698e-05,
+      "loss": 0.976,
+      "step": 8805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2302339521461642e-05,
+      "loss": 0.8185,
+      "step": 8806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2300752986080027e-05,
+      "loss": 0.9359,
+      "step": 8807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299166389547028e-05,
+      "loss": 1.005,
+      "step": 8808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2297579731904812e-05,
+      "loss": 0.9792,
+      "step": 8809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229599301319555e-05,
+      "loss": 0.9363,
+      "step": 8810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2294406233461418e-05,
+      "loss": 0.8855,
+      "step": 8811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292819392744585e-05,
+      "loss": 0.918,
+      "step": 8812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2291232491087235e-05,
+      "loss": 1.0189,
+      "step": 8813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2289645528531542e-05,
+      "loss": 1.0016,
+      "step": 8814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2288058505119684e-05,
+      "loss": 0.9158,
+      "step": 8815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286471420893843e-05,
+      "loss": 1.0218,
+      "step": 8816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284884275896205e-05,
+      "loss": 0.8967,
+      "step": 8817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283297070168951e-05,
+      "loss": 0.9354,
+      "step": 8818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228170980375427e-05,
+      "loss": 0.9778,
+      "step": 8819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2280122476694349e-05,
+      "loss": 0.9575,
+      "step": 8820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278535089031377e-05,
+      "loss": 1.0068,
+      "step": 8821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2276947640807544e-05,
+      "loss": 1.0122,
+      "step": 8822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2275360132065042e-05,
+      "loss": 0.4089,
+      "step": 8823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227377256284607e-05,
+      "loss": 0.9743,
+      "step": 8824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2272184933192814e-05,
+      "loss": 0.8578,
+      "step": 8825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2270597243147485e-05,
+      "loss": 0.9645,
+      "step": 8826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226900949275227e-05,
+      "loss": 0.866,
+      "step": 8827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267421682049374e-05,
+      "loss": 0.9154,
+      "step": 8828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2265833811080999e-05,
+      "loss": 0.9559,
+      "step": 8829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2264245879889349e-05,
+      "loss": 0.8697,
+      "step": 8830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2262657888516631e-05,
+      "loss": 0.8673,
+      "step": 8831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2261069837005053e-05,
+      "loss": 0.8919,
+      "step": 8832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2259481725396815e-05,
+      "loss": 0.9301,
+      "step": 8833
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.225789355373414e-05,
+      "loss": 1.042,
+      "step": 8834
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2256305322059228e-05,
+      "loss": 0.8209,
+      "step": 8835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254717030414298e-05,
+      "loss": 0.9093,
+      "step": 8836
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2253128678841568e-05,
+      "loss": 0.9896,
+      "step": 8837
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2251540267383249e-05,
+      "loss": 0.961,
+      "step": 8838
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2249951796081561e-05,
+      "loss": 0.9495,
+      "step": 8839
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2248363264978723e-05,
+      "loss": 0.9482,
+      "step": 8840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2246774674116956e-05,
+      "loss": 0.8547,
+      "step": 8841
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224518602353849e-05,
+      "loss": 0.9084,
+      "step": 8842
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2243597313285538e-05,
+      "loss": 0.9346,
+      "step": 8843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2242008543400337e-05,
+      "loss": 0.9082,
+      "step": 8844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2240419713925107e-05,
+      "loss": 0.9703,
+      "step": 8845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2238830824902078e-05,
+      "loss": 0.9733,
+      "step": 8846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237241876373482e-05,
+      "loss": 0.8847,
+      "step": 8847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2235652868381553e-05,
+      "loss": 0.9375,
+      "step": 8848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2234063800968528e-05,
+      "loss": 0.4524,
+      "step": 8849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2232474674176637e-05,
+      "loss": 0.9193,
+      "step": 8850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2230885488048117e-05,
+      "loss": 0.9408,
+      "step": 8851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222929624262521e-05,
+      "loss": 0.4649,
+      "step": 8852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2227706937950155e-05,
+      "loss": 1.0756,
+      "step": 8853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222611757406519e-05,
+      "loss": 0.9246,
+      "step": 8854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2224528151012569e-05,
+      "loss": 0.8676,
+      "step": 8855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2222938668834525e-05,
+      "loss": 1.0413,
+      "step": 8856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2221349127573314e-05,
+      "loss": 0.966,
+      "step": 8857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219759527271178e-05,
+      "loss": 0.9346,
+      "step": 8858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2218169867970366e-05,
+      "loss": 0.9149,
+      "step": 8859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216580149713137e-05,
+      "loss": 0.9282,
+      "step": 8860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214990372541738e-05,
+      "loss": 0.9062,
+      "step": 8861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2213400536498425e-05,
+      "loss": 0.9716,
+      "step": 8862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2211810641625455e-05,
+      "loss": 0.9302,
+      "step": 8863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2210220687965081e-05,
+      "loss": 0.9441,
+      "step": 8864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208630675559568e-05,
+      "loss": 0.9158,
+      "step": 8865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2207040604451171e-05,
+      "loss": 0.9717,
+      "step": 8866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220545047468216e-05,
+      "loss": 1.0333,
+      "step": 8867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220386028629479e-05,
+      "loss": 0.9799,
+      "step": 8868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2202270039331332e-05,
+      "loss": 0.9283,
+      "step": 8869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200679733834054e-05,
+      "loss": 0.8554,
+      "step": 8870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2199089369845218e-05,
+      "loss": 0.9446,
+      "step": 8871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2197498947407101e-05,
+      "loss": 0.9459,
+      "step": 8872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219590846656197e-05,
+      "loss": 0.9374,
+      "step": 8873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2194317927352102e-05,
+      "loss": 0.8571,
+      "step": 8874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219272732981977e-05,
+      "loss": 0.9092,
+      "step": 8875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2191136674007246e-05,
+      "loss": 1.0255,
+      "step": 8876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2189545959956814e-05,
+      "loss": 0.9293,
+      "step": 8877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187955187710752e-05,
+      "loss": 0.9406,
+      "step": 8878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2186364357311337e-05,
+      "loss": 0.9405,
+      "step": 8879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218477346880086e-05,
+      "loss": 0.8783,
+      "step": 8880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2183182522221597e-05,
+      "loss": 0.9792,
+      "step": 8881
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2181591517615831e-05,
+      "loss": 0.9524,
+      "step": 8882
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218000045502586e-05,
+      "loss": 0.906,
+      "step": 8883
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2178409334493966e-05,
+      "loss": 0.9414,
+      "step": 8884
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176818156062438e-05,
+      "loss": 0.9985,
+      "step": 8885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2175226919773571e-05,
+      "loss": 0.9382,
+      "step": 8886
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2173635625669656e-05,
+      "loss": 0.9537,
+      "step": 8887
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172044273792987e-05,
+      "loss": 0.8899,
+      "step": 8888
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2170452864185864e-05,
+      "loss": 0.9611,
+      "step": 8889
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216886139689058e-05,
+      "loss": 0.923,
+      "step": 8890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2167269871949443e-05,
+      "loss": 0.9202,
+      "step": 8891
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2165678289404743e-05,
+      "loss": 0.9263,
+      "step": 8892
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164086649298792e-05,
+      "loss": 0.8432,
+      "step": 8893
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2162494951673884e-05,
+      "loss": 0.9154,
+      "step": 8894
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2160903196572331e-05,
+      "loss": 0.9489,
+      "step": 8895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159311384036443e-05,
+      "loss": 1.0203,
+      "step": 8896
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2157719514108522e-05,
+      "loss": 0.8999,
+      "step": 8897
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2156127586830881e-05,
+      "loss": 0.9174,
+      "step": 8898
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2154535602245832e-05,
+      "loss": 0.9538,
+      "step": 8899
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2152943560395684e-05,
+      "loss": 0.9917,
+      "step": 8900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151351461322757e-05,
+      "loss": 0.9245,
+      "step": 8901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149759305069368e-05,
+      "loss": 0.8951,
+      "step": 8902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214816709167783e-05,
+      "loss": 1.0443,
+      "step": 8903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146574821190464e-05,
+      "loss": 0.9343,
+      "step": 8904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144982493649591e-05,
+      "loss": 0.9834,
+      "step": 8905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2143390109097533e-05,
+      "loss": 0.893,
+      "step": 8906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2141797667576613e-05,
+      "loss": 0.9284,
+      "step": 8907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2140205169129157e-05,
+      "loss": 0.9539,
+      "step": 8908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138612613797496e-05,
+      "loss": 1.0561,
+      "step": 8909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2137020001623949e-05,
+      "loss": 0.8812,
+      "step": 8910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2135427332650855e-05,
+      "loss": 0.9341,
+      "step": 8911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2133834606920539e-05,
+      "loss": 0.9162,
+      "step": 8912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2132241824475334e-05,
+      "loss": 0.9372,
+      "step": 8913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130648985357583e-05,
+      "loss": 0.9406,
+      "step": 8914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212905608960961e-05,
+      "loss": 0.9382,
+      "step": 8915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212746313727376e-05,
+      "loss": 0.9433,
+      "step": 8916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212587012839237e-05,
+      "loss": 0.9953,
+      "step": 8917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2124277063007776e-05,
+      "loss": 0.8905,
+      "step": 8918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212268394116233e-05,
+      "loss": 0.9269,
+      "step": 8919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2121090762898367e-05,
+      "loss": 0.9181,
+      "step": 8920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2119497528258233e-05,
+      "loss": 0.8812,
+      "step": 8921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117904237284275e-05,
+      "loss": 0.9202,
+      "step": 8922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2116310890018841e-05,
+      "loss": 0.9796,
+      "step": 8923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114717486504282e-05,
+      "loss": 0.9144,
+      "step": 8924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2113124026782948e-05,
+      "loss": 0.8911,
+      "step": 8925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2111530510897187e-05,
+      "loss": 0.9325,
+      "step": 8926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109936938889362e-05,
+      "loss": 0.9953,
+      "step": 8927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2108343310801818e-05,
+      "loss": 0.9468,
+      "step": 8928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210674962667692e-05,
+      "loss": 0.9223,
+      "step": 8929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210515588655702e-05,
+      "loss": 0.9933,
+      "step": 8930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2103562090484484e-05,
+      "loss": 0.9787,
+      "step": 8931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2101968238501669e-05,
+      "loss": 1.0101,
+      "step": 8932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2100374330650941e-05,
+      "loss": 0.9675,
+      "step": 8933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2098780366974658e-05,
+      "loss": 0.8802,
+      "step": 8934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2097186347515196e-05,
+      "loss": 0.4544,
+      "step": 8935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2095592272314907e-05,
+      "loss": 0.9626,
+      "step": 8936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2093998141416175e-05,
+      "loss": 0.929,
+      "step": 8937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2092403954861364e-05,
+      "loss": 0.9515,
+      "step": 8938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2090809712692843e-05,
+      "loss": 1.0093,
+      "step": 8939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089215414952991e-05,
+      "loss": 0.4345,
+      "step": 8940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2087621061684178e-05,
+      "loss": 0.9029,
+      "step": 8941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208602665292878e-05,
+      "loss": 0.9991,
+      "step": 8942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2084432188729178e-05,
+      "loss": 0.9104,
+      "step": 8943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2082837669127748e-05,
+      "loss": 0.9225,
+      "step": 8944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2081243094166872e-05,
+      "loss": 0.994,
+      "step": 8945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079648463888932e-05,
+      "loss": 0.8994,
+      "step": 8946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2078053778336307e-05,
+      "loss": 0.4398,
+      "step": 8947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2076459037551391e-05,
+      "loss": 0.9614,
+      "step": 8948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074864241576563e-05,
+      "loss": 0.906,
+      "step": 8949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2073269390454215e-05,
+      "loss": 0.9657,
+      "step": 8950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2071674484226734e-05,
+      "loss": 0.976,
+      "step": 8951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070079522936511e-05,
+      "loss": 0.9012,
+      "step": 8952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2068484506625936e-05,
+      "loss": 0.927,
+      "step": 8953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066889435337407e-05,
+      "loss": 0.9633,
+      "step": 8954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2065294309113319e-05,
+      "loss": 0.9892,
+      "step": 8955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2063699127996067e-05,
+      "loss": 0.8852,
+      "step": 8956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062103892028048e-05,
+      "loss": 0.9409,
+      "step": 8957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2060508601251663e-05,
+      "loss": 0.9343,
+      "step": 8958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205891325570931e-05,
+      "loss": 0.9252,
+      "step": 8959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2057317855443395e-05,
+      "loss": 0.8868,
+      "step": 8960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2055722400496326e-05,
+      "loss": 0.9706,
+      "step": 8961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2054126890910499e-05,
+      "loss": 0.7608,
+      "step": 8962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2052531326728328e-05,
+      "loss": 0.9338,
+      "step": 8963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2050935707992215e-05,
+      "loss": 0.9709,
+      "step": 8964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2049340034744574e-05,
+      "loss": 1.0276,
+      "step": 8965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047744307027817e-05,
+      "loss": 0.9245,
+      "step": 8966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2046148524884355e-05,
+      "loss": 0.9375,
+      "step": 8967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2044552688356604e-05,
+      "loss": 0.9217,
+      "step": 8968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2042956797486976e-05,
+      "loss": 0.9029,
+      "step": 8969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204136085231789e-05,
+      "loss": 0.8625,
+      "step": 8970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203976485289176e-05,
+      "loss": 0.9599,
+      "step": 8971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2038168799251016e-05,
+      "loss": 0.9254,
+      "step": 8972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036572691438068e-05,
+      "loss": 1.0181,
+      "step": 8973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034976529495347e-05,
+      "loss": 0.9206,
+      "step": 8974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2033380313465273e-05,
+      "loss": 0.9668,
+      "step": 8975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031784043390273e-05,
+      "loss": 0.942,
+      "step": 8976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203018771931277e-05,
+      "loss": 0.9708,
+      "step": 8977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20285913412752e-05,
+      "loss": 1.0161,
+      "step": 8978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2026994909319992e-05,
+      "loss": 0.9288,
+      "step": 8979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2025398423489566e-05,
+      "loss": 1.0098,
+      "step": 8980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2023801883826372e-05,
+      "loss": 0.9821,
+      "step": 8981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202220529037283e-05,
+      "loss": 0.9618,
+      "step": 8982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2020608643171382e-05,
+      "loss": 0.9471,
+      "step": 8983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019011942264465e-05,
+      "loss": 0.9515,
+      "step": 8984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2017415187694512e-05,
+      "loss": 0.8753,
+      "step": 8985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2015818379503975e-05,
+      "loss": 0.9834,
+      "step": 8986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2014221517735286e-05,
+      "loss": 0.9757,
+      "step": 8987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2012624602430884e-05,
+      "loss": 0.9188,
+      "step": 8988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2011027633633221e-05,
+      "loss": 0.8923,
+      "step": 8989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2009430611384742e-05,
+      "loss": 0.8731,
+      "step": 8990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200783353572789e-05,
+      "loss": 0.8492,
+      "step": 8991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006236406705122e-05,
+      "loss": 0.9259,
+      "step": 8992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004639224358874e-05,
+      "loss": 0.8879,
+      "step": 8993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003041988731609e-05,
+      "loss": 0.9038,
+      "step": 8994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001444699865774e-05,
+      "loss": 1.0008,
+      "step": 8995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999847357803824e-05,
+      "loss": 0.9811,
+      "step": 8996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199824996258822e-05,
+      "loss": 1.024,
+      "step": 8997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996652514261409e-05,
+      "loss": 0.9631,
+      "step": 8998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995055012865858e-05,
+      "loss": 0.8044,
+      "step": 8999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993457458444022e-05,
+      "loss": 0.9983,
+      "step": 9000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991859851038362e-05,
+      "loss": 0.9386,
+      "step": 9001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990262190691344e-05,
+      "loss": 0.9987,
+      "step": 9002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198866447744543e-05,
+      "loss": 0.9179,
+      "step": 9003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987066711343085e-05,
+      "loss": 0.4369,
+      "step": 9004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985468892426778e-05,
+      "loss": 0.9564,
+      "step": 9005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198387102073897e-05,
+      "loss": 0.9078,
+      "step": 9006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198227309632214e-05,
+      "loss": 0.8946,
+      "step": 9007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980675119218756e-05,
+      "loss": 0.8748,
+      "step": 9008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979077089471288e-05,
+      "loss": 0.9025,
+      "step": 9009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1977479007122213e-05,
+      "loss": 0.9149,
+      "step": 9010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975880872214e-05,
+      "loss": 0.9534,
+      "step": 9011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974282684789133e-05,
+      "loss": 0.8888,
+      "step": 9012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197268444489009e-05,
+      "loss": 0.8574,
+      "step": 9013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971086152559344e-05,
+      "loss": 0.9055,
+      "step": 9014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196948780783938e-05,
+      "loss": 0.9287,
+      "step": 9015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196788941077268e-05,
+      "loss": 0.8925,
+      "step": 9016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966290961401728e-05,
+      "loss": 0.988,
+      "step": 9017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964692459769003e-05,
+      "loss": 0.976,
+      "step": 9018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963093905917e-05,
+      "loss": 0.9836,
+      "step": 9019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961495299888204e-05,
+      "loss": 0.9772,
+      "step": 9020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1959896641725105e-05,
+      "loss": 0.8834,
+      "step": 9021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958297931470186e-05,
+      "loss": 0.9558,
+      "step": 9022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1956699169165948e-05,
+      "loss": 0.9859,
+      "step": 9023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955100354854877e-05,
+      "loss": 1.0203,
+      "step": 9024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1953501488579476e-05,
+      "loss": 0.8998,
+      "step": 9025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1951902570382236e-05,
+      "loss": 1.0169,
+      "step": 9026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1950303600305653e-05,
+      "loss": 0.8756,
+      "step": 9027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1948704578392229e-05,
+      "loss": 0.9368,
+      "step": 9028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1947105504684463e-05,
+      "loss": 0.9284,
+      "step": 9029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1945506379224856e-05,
+      "loss": 0.9082,
+      "step": 9030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1943907202055912e-05,
+      "loss": 0.9315,
+      "step": 9031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1942307973220133e-05,
+      "loss": 0.9003,
+      "step": 9032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194070869276003e-05,
+      "loss": 0.9949,
+      "step": 9033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1939109360718102e-05,
+      "loss": 0.9169,
+      "step": 9034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937509977136866e-05,
+      "loss": 1.0031,
+      "step": 9035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935910542058824e-05,
+      "loss": 0.9164,
+      "step": 9036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934311055526493e-05,
+      "loss": 0.9206,
+      "step": 9037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932711517582383e-05,
+      "loss": 0.4428,
+      "step": 9038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1931111928269011e-05,
+      "loss": 0.9274,
+      "step": 9039
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929512287628883e-05,
+      "loss": 0.9281,
+      "step": 9040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927912595704529e-05,
+      "loss": 0.8896,
+      "step": 9041
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926312852538456e-05,
+      "loss": 0.9397,
+      "step": 9042
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924713058173188e-05,
+      "loss": 0.9844,
+      "step": 9043
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923113212651245e-05,
+      "loss": 0.9229,
+      "step": 9044
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192151331601515e-05,
+      "loss": 0.9714,
+      "step": 9045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919913368307425e-05,
+      "loss": 0.9158,
+      "step": 9046
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918313369570595e-05,
+      "loss": 0.8924,
+      "step": 9047
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916713319847186e-05,
+      "loss": 0.983,
+      "step": 9048
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1915113219179729e-05,
+      "loss": 0.9479,
+      "step": 9049
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1913513067610747e-05,
+      "loss": 1.0132,
+      "step": 9050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1911912865182776e-05,
+      "loss": 0.9103,
+      "step": 9051
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1910312611938344e-05,
+      "loss": 0.9117,
+      "step": 9052
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1908712307919983e-05,
+      "loss": 0.923,
+      "step": 9053
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1907111953170228e-05,
+      "loss": 0.8894,
+      "step": 9054
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1905511547731616e-05,
+      "loss": 0.9492,
+      "step": 9055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903911091646684e-05,
+      "loss": 0.8951,
+      "step": 9056
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902310584957972e-05,
+      "loss": 0.903,
+      "step": 9057
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1900710027708013e-05,
+      "loss": 0.9504,
+      "step": 9058
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1899109419939356e-05,
+      "loss": 0.9924,
+      "step": 9059
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1897508761694538e-05,
+      "loss": 0.9306,
+      "step": 9060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895908053016103e-05,
+      "loss": 0.9357,
+      "step": 9061
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18943072939466e-05,
+      "loss": 0.9573,
+      "step": 9062
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1892706484528569e-05,
+      "loss": 0.9364,
+      "step": 9063
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1891105624804566e-05,
+      "loss": 0.8666,
+      "step": 9064
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889504714817131e-05,
+      "loss": 0.9973,
+      "step": 9065
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188790375460882e-05,
+      "loss": 0.8448,
+      "step": 9066
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1886302744222184e-05,
+      "loss": 0.9173,
+      "step": 9067
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1884701683699775e-05,
+      "loss": 0.9501,
+      "step": 9068
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188310057308415e-05,
+      "loss": 0.9616,
+      "step": 9069
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188149941241786e-05,
+      "loss": 0.9291,
+      "step": 9070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1879898201743462e-05,
+      "loss": 0.9492,
+      "step": 9071
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1878296941103522e-05,
+      "loss": 0.9115,
+      "step": 9072
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1876695630540593e-05,
+      "loss": 0.8885,
+      "step": 9073
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1875094270097237e-05,
+      "loss": 0.8972,
+      "step": 9074
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1873492859816018e-05,
+      "loss": 0.9967,
+      "step": 9075
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1871891399739497e-05,
+      "loss": 0.8585,
+      "step": 9076
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187028988991024e-05,
+      "loss": 0.9134,
+      "step": 9077
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1868688330370816e-05,
+      "loss": 0.9013,
+      "step": 9078
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1867086721163787e-05,
+      "loss": 0.8361,
+      "step": 9079
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1865485062331729e-05,
+      "loss": 0.9433,
+      "step": 9080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1863883353917206e-05,
+      "loss": 0.8788,
+      "step": 9081
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1862281595962796e-05,
+      "loss": 0.9113,
+      "step": 9082
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1860679788511066e-05,
+      "loss": 0.9179,
+      "step": 9083
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185907793160459e-05,
+      "loss": 0.9133,
+      "step": 9084
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857476025285948e-05,
+      "loss": 0.9404,
+      "step": 9085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1855874069597712e-05,
+      "loss": 0.412,
+      "step": 9086
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1854272064582469e-05,
+      "loss": 0.9487,
+      "step": 9087
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1852670010282787e-05,
+      "loss": 0.9849,
+      "step": 9088
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851067906741252e-05,
+      "loss": 0.9173,
+      "step": 9089
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1849465754000449e-05,
+      "loss": 0.907,
+      "step": 9090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1847863552102959e-05,
+      "loss": 0.9722,
+      "step": 9091
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1846261301091361e-05,
+      "loss": 0.8408,
+      "step": 9092
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1844659001008252e-05,
+      "loss": 1.0155,
+      "step": 9093
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184305665189621e-05,
+      "loss": 0.9987,
+      "step": 9094
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1841454253797828e-05,
+      "loss": 0.9156,
+      "step": 9095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1839851806755696e-05,
+      "loss": 0.9114,
+      "step": 9096
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838249310812404e-05,
+      "loss": 0.8949,
+      "step": 9097
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836646766010546e-05,
+      "loss": 0.9972,
+      "step": 9098
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183504417239271e-05,
+      "loss": 0.9918,
+      "step": 9099
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833441530001502e-05,
+      "loss": 0.9588,
+      "step": 9100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1831838838879508e-05,
+      "loss": 0.9027,
+      "step": 9101
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183023609906933e-05,
+      "loss": 0.9709,
+      "step": 9102
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828633310613569e-05,
+      "loss": 0.9477,
+      "step": 9103
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827030473554822e-05,
+      "loss": 1.0261,
+      "step": 9104
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1825427587935692e-05,
+      "loss": 0.9317,
+      "step": 9105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1823824653798781e-05,
+      "loss": 1.0551,
+      "step": 9106
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.182222167118669e-05,
+      "loss": 0.9036,
+      "step": 9107
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820618640142034e-05,
+      "loss": 0.9909,
+      "step": 9108
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1819015560707414e-05,
+      "loss": 0.8218,
+      "step": 9109
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1817412432925435e-05,
+      "loss": 0.9022,
+      "step": 9110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181580925683871e-05,
+      "loss": 0.9509,
+      "step": 9111
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1814206032489847e-05,
+      "loss": 0.9815,
+      "step": 9112
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1812602759921461e-05,
+      "loss": 0.9183,
+      "step": 9113
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1810999439176164e-05,
+      "loss": 0.9171,
+      "step": 9114
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180939607029657e-05,
+      "loss": 0.9552,
+      "step": 9115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1807792653325298e-05,
+      "loss": 0.951,
+      "step": 9116
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1806189188304957e-05,
+      "loss": 0.957,
+      "step": 9117
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804585675278175e-05,
+      "loss": 0.9126,
+      "step": 9118
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1802982114287565e-05,
+      "loss": 0.9589,
+      "step": 9119
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1801378505375747e-05,
+      "loss": 0.9864,
+      "step": 9120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799774848585348e-05,
+      "loss": 1.0323,
+      "step": 9121
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179817114395899e-05,
+      "loss": 0.9434,
+      "step": 9122
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796567391539296e-05,
+      "loss": 0.9602,
+      "step": 9123
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1794963591368893e-05,
+      "loss": 0.8624,
+      "step": 9124
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793359743490404e-05,
+      "loss": 0.9086,
+      "step": 9125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791755847946465e-05,
+      "loss": 0.9005,
+      "step": 9126
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1790151904779703e-05,
+      "loss": 0.9015,
+      "step": 9127
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1788547914032744e-05,
+      "loss": 0.9467,
+      "step": 9128
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178694387574823e-05,
+      "loss": 0.8968,
+      "step": 9129
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785339789968782e-05,
+      "loss": 0.9779,
+      "step": 9130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783735656737042e-05,
+      "loss": 0.9211,
+      "step": 9131
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178213147609565e-05,
+      "loss": 0.9134,
+      "step": 9132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1780527248087237e-05,
+      "loss": 0.9573,
+      "step": 9133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1778922972754443e-05,
+      "loss": 0.8136,
+      "step": 9134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1777318650139906e-05,
+      "loss": 0.9615,
+      "step": 9135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1775714280286272e-05,
+      "loss": 0.9735,
+      "step": 9136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177410986323618e-05,
+      "loss": 0.9311,
+      "step": 9137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1772505399032273e-05,
+      "loss": 0.4392,
+      "step": 9138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770900887717198e-05,
+      "loss": 0.8776,
+      "step": 9139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1769296329333601e-05,
+      "loss": 0.9608,
+      "step": 9140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1767691723924126e-05,
+      "loss": 0.9584,
+      "step": 9141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766087071531424e-05,
+      "loss": 0.891,
+      "step": 9142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764482372198145e-05,
+      "loss": 0.9052,
+      "step": 9143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1762877625966939e-05,
+      "loss": 0.9438,
+      "step": 9144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1761272832880464e-05,
+      "loss": 0.9815,
+      "step": 9145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759667992981362e-05,
+      "loss": 1.0394,
+      "step": 9146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1758063106312298e-05,
+      "loss": 0.9532,
+      "step": 9147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1756458172915925e-05,
+      "loss": 0.8933,
+      "step": 9148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1754853192834897e-05,
+      "loss": 0.974,
+      "step": 9149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175324816611188e-05,
+      "loss": 0.9423,
+      "step": 9150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751643092789524e-05,
+      "loss": 0.8488,
+      "step": 9151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1750037972910499e-05,
+      "loss": 0.9932,
+      "step": 9152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174843280651746e-05,
+      "loss": 0.9895,
+      "step": 9153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746827593653076e-05,
+      "loss": 0.9209,
+      "step": 9154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1745222334360009e-05,
+      "loss": 0.9535,
+      "step": 9155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1743617028680927e-05,
+      "loss": 0.8778,
+      "step": 9156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1742011676658496e-05,
+      "loss": 0.4651,
+      "step": 9157
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1740406278335384e-05,
+      "loss": 0.9566,
+      "step": 9158
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1738800833754258e-05,
+      "loss": 0.949,
+      "step": 9159
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1737195342957792e-05,
+      "loss": 0.9166,
+      "step": 9160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1735589805988661e-05,
+      "loss": 0.9367,
+      "step": 9161
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1733984222889533e-05,
+      "loss": 0.9548,
+      "step": 9162
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1732378593703087e-05,
+      "loss": 0.8967,
+      "step": 9163
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730772918471994e-05,
+      "loss": 0.8861,
+      "step": 9164
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1729167197238936e-05,
+      "loss": 0.9641,
+      "step": 9165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1727561430046585e-05,
+      "loss": 0.45,
+      "step": 9166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1725955616937626e-05,
+      "loss": 0.9826,
+      "step": 9167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172434975795474e-05,
+      "loss": 0.9756,
+      "step": 9168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1722743853140603e-05,
+      "loss": 0.8292,
+      "step": 9169
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721137902537907e-05,
+      "loss": 0.4164,
+      "step": 9170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1719531906189327e-05,
+      "loss": 0.9621,
+      "step": 9171
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1717925864137553e-05,
+      "loss": 0.4691,
+      "step": 9172
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1716319776425275e-05,
+      "loss": 0.9189,
+      "step": 9173
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1714713643095175e-05,
+      "loss": 0.9391,
+      "step": 9174
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1713107464189946e-05,
+      "loss": 0.8992,
+      "step": 9175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1711501239752277e-05,
+      "loss": 0.902,
+      "step": 9176
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709894969824857e-05,
+      "loss": 0.9678,
+      "step": 9177
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1708288654450384e-05,
+      "loss": 0.9325,
+      "step": 9178
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1706682293671552e-05,
+      "loss": 0.919,
+      "step": 9179
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170507588753105e-05,
+      "loss": 0.9246,
+      "step": 9180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170346943607158e-05,
+      "loss": 0.895,
+      "step": 9181
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1701862939335835e-05,
+      "loss": 0.9386,
+      "step": 9182
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170025639736652e-05,
+      "loss": 1.0131,
+      "step": 9183
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1698649810206332e-05,
+      "loss": 0.9332,
+      "step": 9184
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169704317789797e-05,
+      "loss": 0.8867,
+      "step": 9185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169543650048414e-05,
+      "loss": 0.9448,
+      "step": 9186
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693829778007544e-05,
+      "loss": 0.9957,
+      "step": 9187
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1692223010510889e-05,
+      "loss": 0.9432,
+      "step": 9188
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1690616198036876e-05,
+      "loss": 0.9098,
+      "step": 9189
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1689009340628215e-05,
+      "loss": 0.8113,
+      "step": 9190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1687402438327617e-05,
+      "loss": 0.9189,
+      "step": 9191
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1685795491177788e-05,
+      "loss": 0.9477,
+      "step": 9192
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168418849922144e-05,
+      "loss": 0.4991,
+      "step": 9193
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682581462501287e-05,
+      "loss": 0.9694,
+      "step": 9194
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1680974381060037e-05,
+      "loss": 0.9143,
+      "step": 9195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1679367254940412e-05,
+      "loss": 0.9908,
+      "step": 9196
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1677760084185123e-05,
+      "loss": 0.9227,
+      "step": 9197
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1676152868836885e-05,
+      "loss": 0.9098,
+      "step": 9198
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167454560893842e-05,
+      "loss": 0.9126,
+      "step": 9199
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1672938304532441e-05,
+      "loss": 0.978,
+      "step": 9200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1671330955661677e-05,
+      "loss": 0.9045,
+      "step": 9201
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1669723562368844e-05,
+      "loss": 0.9565,
+      "step": 9202
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1668116124696665e-05,
+      "loss": 0.9108,
+      "step": 9203
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1666508642687866e-05,
+      "loss": 0.9698,
+      "step": 9204
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1664901116385168e-05,
+      "loss": 0.9342,
+      "step": 9205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1663293545831302e-05,
+      "loss": 0.8981,
+      "step": 9206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661685931068991e-05,
+      "loss": 0.9467,
+      "step": 9207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1660078272140966e-05,
+      "loss": 0.9231,
+      "step": 9208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1658470569089961e-05,
+      "loss": 0.9897,
+      "step": 9209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1656862821958698e-05,
+      "loss": 0.897,
+      "step": 9210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1655255030789916e-05,
+      "loss": 0.9057,
+      "step": 9211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1653647195626345e-05,
+      "loss": 0.9313,
+      "step": 9212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652039316510716e-05,
+      "loss": 0.9441,
+      "step": 9213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1650431393485775e-05,
+      "loss": 0.9398,
+      "step": 9214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1648823426594251e-05,
+      "loss": 0.9205,
+      "step": 9215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1647215415878881e-05,
+      "loss": 0.9712,
+      "step": 9216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1645607361382409e-05,
+      "loss": 1.0085,
+      "step": 9217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1643999263147571e-05,
+      "loss": 0.9789,
+      "step": 9218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164239112121711e-05,
+      "loss": 0.9846,
+      "step": 9219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1640782935633772e-05,
+      "loss": 0.9413,
+      "step": 9220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1639174706440295e-05,
+      "loss": 0.8707,
+      "step": 9221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1637566433679428e-05,
+      "loss": 0.9498,
+      "step": 9222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635958117393913e-05,
+      "loss": 0.9208,
+      "step": 9223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1634349757626503e-05,
+      "loss": 0.934,
+      "step": 9224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1632741354419939e-05,
+      "loss": 0.437,
+      "step": 9225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1631132907816976e-05,
+      "loss": 0.9409,
+      "step": 9226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1629524417860365e-05,
+      "loss": 0.9358,
+      "step": 9227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627915884592855e-05,
+      "loss": 0.9033,
+      "step": 9228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1626307308057199e-05,
+      "loss": 0.9241,
+      "step": 9229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1624698688296154e-05,
+      "loss": 0.9232,
+      "step": 9230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162309002535247e-05,
+      "loss": 0.9603,
+      "step": 9231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162148131926891e-05,
+      "loss": 0.9503,
+      "step": 9232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1619872570088231e-05,
+      "loss": 0.8967,
+      "step": 9233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618263777853185e-05,
+      "loss": 0.8971,
+      "step": 9234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1616654942606539e-05,
+      "loss": 0.9844,
+      "step": 9235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1615046064391047e-05,
+      "loss": 0.6904,
+      "step": 9236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1613437143249476e-05,
+      "loss": 0.9444,
+      "step": 9237
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1611828179224592e-05,
+      "loss": 0.8935,
+      "step": 9238
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610219172359151e-05,
+      "loss": 0.965,
+      "step": 9239
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1608610122695928e-05,
+      "loss": 1.0564,
+      "step": 9240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160700103027768e-05,
+      "loss": 0.9294,
+      "step": 9241
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1605391895147186e-05,
+      "loss": 0.9356,
+      "step": 9242
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1603782717347205e-05,
+      "loss": 0.9092,
+      "step": 9243
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160217349692051e-05,
+      "loss": 0.96,
+      "step": 9244
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1600564233909877e-05,
+      "loss": 0.9967,
+      "step": 9245
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1598954928358074e-05,
+      "loss": 0.9482,
+      "step": 9246
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1597345580307874e-05,
+      "loss": 0.8994,
+      "step": 9247
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1595736189802053e-05,
+      "loss": 0.9356,
+      "step": 9248
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1594126756883387e-05,
+      "loss": 0.9391,
+      "step": 9249
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592517281594651e-05,
+      "loss": 0.9983,
+      "step": 9250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1590907763978628e-05,
+      "loss": 0.9158,
+      "step": 9251
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158929820407809e-05,
+      "loss": 0.8553,
+      "step": 9252
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1587688601935822e-05,
+      "loss": 0.9812,
+      "step": 9253
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586078957594605e-05,
+      "loss": 0.9755,
+      "step": 9254
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1584469271097219e-05,
+      "loss": 0.9766,
+      "step": 9255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1582859542486452e-05,
+      "loss": 0.97,
+      "step": 9256
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581249771805082e-05,
+      "loss": 0.9093,
+      "step": 9257
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1579639959095906e-05,
+      "loss": 1.0196,
+      "step": 9258
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15780301044017e-05,
+      "loss": 0.9932,
+      "step": 9259
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576420207765254e-05,
+      "loss": 0.9689,
+      "step": 9260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1574810269229362e-05,
+      "loss": 0.8976,
+      "step": 9261
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.157320028883681e-05,
+      "loss": 0.9493,
+      "step": 9262
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571590266630392e-05,
+      "loss": 0.9513,
+      "step": 9263
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569980202652902e-05,
+      "loss": 0.9951,
+      "step": 9264
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1568370096947128e-05,
+      "loss": 0.4575,
+      "step": 9265
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1566759949555867e-05,
+      "loss": 0.4291,
+      "step": 9266
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1565149760521921e-05,
+      "loss": 0.9119,
+      "step": 9267
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1563539529888078e-05,
+      "loss": 0.8965,
+      "step": 9268
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1561929257697143e-05,
+      "loss": 0.9234,
+      "step": 9269
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560318943991908e-05,
+      "loss": 0.9505,
+      "step": 9270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1558708588815179e-05,
+      "loss": 0.937,
+      "step": 9271
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1557098192209758e-05,
+      "loss": 0.9128,
+      "step": 9272
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555487754218442e-05,
+      "loss": 0.8969,
+      "step": 9273
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1553877274884041e-05,
+      "loss": 0.9802,
+      "step": 9274
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552266754249354e-05,
+      "loss": 0.885,
+      "step": 9275
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550656192357195e-05,
+      "loss": 0.9156,
+      "step": 9276
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549045589250363e-05,
+      "loss": 0.9451,
+      "step": 9277
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547434944971662e-05,
+      "loss": 1.0164,
+      "step": 9278
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1545824259563915e-05,
+      "loss": 0.9786,
+      "step": 9279
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544213533069922e-05,
+      "loss": 0.9718,
+      "step": 9280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1542602765532498e-05,
+      "loss": 0.9067,
+      "step": 9281
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1540991956994453e-05,
+      "loss": 0.9081,
+      "step": 9282
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539381107498604e-05,
+      "loss": 0.8678,
+      "step": 9283
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153777021708776e-05,
+      "loss": 0.9904,
+      "step": 9284
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1536159285804742e-05,
+      "loss": 0.9713,
+      "step": 9285
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1534548313692364e-05,
+      "loss": 0.8676,
+      "step": 9286
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1532937300793445e-05,
+      "loss": 0.9846,
+      "step": 9287
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1531326247150802e-05,
+      "loss": 0.8711,
+      "step": 9288
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152971515280726e-05,
+      "loss": 0.9583,
+      "step": 9289
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528104017805632e-05,
+      "loss": 1.0578,
+      "step": 9290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1526492842188746e-05,
+      "loss": 0.9439,
+      "step": 9291
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1524881625999425e-05,
+      "loss": 0.9623,
+      "step": 9292
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152327036928049e-05,
+      "loss": 0.8817,
+      "step": 9293
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1521659072074768e-05,
+      "loss": 0.852,
+      "step": 9294
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152004773442509e-05,
+      "loss": 0.9564,
+      "step": 9295
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1518436356374272e-05,
+      "loss": 0.9566,
+      "step": 9296
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1516824937965156e-05,
+      "loss": 0.4441,
+      "step": 9297
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1515213479240566e-05,
+      "loss": 0.8976,
+      "step": 9298
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1513601980243327e-05,
+      "loss": 1.0502,
+      "step": 9299
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1511990441016281e-05,
+      "loss": 0.9252,
+      "step": 9300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510378861602253e-05,
+      "loss": 0.9023,
+      "step": 9301
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1508767242044078e-05,
+      "loss": 0.9845,
+      "step": 9302
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15071555823846e-05,
+      "loss": 0.9036,
+      "step": 9303
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150554388266664e-05,
+      "loss": 0.8309,
+      "step": 9304
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150393214293305e-05,
+      "loss": 0.8983,
+      "step": 9305
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1502320363226655e-05,
+      "loss": 0.886,
+      "step": 9306
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1500708543590306e-05,
+      "loss": 0.9305,
+      "step": 9307
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1499096684066834e-05,
+      "loss": 1.0108,
+      "step": 9308
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1497484784699084e-05,
+      "loss": 0.921,
+      "step": 9309
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1495872845529902e-05,
+      "loss": 0.968,
+      "step": 9310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1494260866602126e-05,
+      "loss": 0.9264,
+      "step": 9311
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1492648847958606e-05,
+      "loss": 0.8954,
+      "step": 9312
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149103678964218e-05,
+      "loss": 0.8814,
+      "step": 9313
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1489424691695698e-05,
+      "loss": 0.9777,
+      "step": 9314
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487812554162014e-05,
+      "loss": 0.9497,
+      "step": 9315
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148620037708397e-05,
+      "loss": 0.9684,
+      "step": 9316
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1484588160504414e-05,
+      "loss": 0.3928,
+      "step": 9317
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1482975904466204e-05,
+      "loss": 0.9422,
+      "step": 9318
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481363609012188e-05,
+      "loss": 0.9553,
+      "step": 9319
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1479751274185218e-05,
+      "loss": 0.9942,
+      "step": 9320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1478138900028148e-05,
+      "loss": 0.9563,
+      "step": 9321
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1476526486583837e-05,
+      "loss": 0.958,
+      "step": 9322
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1474914033895137e-05,
+      "loss": 0.961,
+      "step": 9323
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473301542004904e-05,
+      "loss": 0.9423,
+      "step": 9324
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1471689010956002e-05,
+      "loss": 0.9416,
+      "step": 9325
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470076440791287e-05,
+      "loss": 0.8885,
+      "step": 9326
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1468463831553618e-05,
+      "loss": 0.4064,
+      "step": 9327
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1466851183285859e-05,
+      "loss": 1.0321,
+      "step": 9328
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1465238496030868e-05,
+      "loss": 0.9117,
+      "step": 9329
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1463625769831512e-05,
+      "loss": 0.962,
+      "step": 9330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462013004730657e-05,
+      "loss": 0.996,
+      "step": 9331
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1460400200771162e-05,
+      "loss": 1.0167,
+      "step": 9332
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1458787357995902e-05,
+      "loss": 0.9906,
+      "step": 9333
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1457174476447739e-05,
+      "loss": 0.8409,
+      "step": 9334
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145556155616954e-05,
+      "loss": 0.9161,
+      "step": 9335
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1453948597204181e-05,
+      "loss": 0.9853,
+      "step": 9336
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452335599594526e-05,
+      "loss": 0.8996,
+      "step": 9337
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145072256338345e-05,
+      "loss": 0.9042,
+      "step": 9338
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1449109488613827e-05,
+      "loss": 0.9389,
+      "step": 9339
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447496375328526e-05,
+      "loss": 0.9363,
+      "step": 9340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1445883223570426e-05,
+      "loss": 0.9547,
+      "step": 9341
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1444270033382401e-05,
+      "loss": 0.9928,
+      "step": 9342
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442656804807328e-05,
+      "loss": 0.9843,
+      "step": 9343
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1441043537888085e-05,
+      "loss": 0.9129,
+      "step": 9344
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143943023266755e-05,
+      "loss": 0.8664,
+      "step": 9345
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1437816889188606e-05,
+      "loss": 0.9832,
+      "step": 9346
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1436203507494131e-05,
+      "loss": 0.9195,
+      "step": 9347
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1434590087627005e-05,
+      "loss": 0.4247,
+      "step": 9348
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432976629630113e-05,
+      "loss": 0.9516,
+      "step": 9349
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1431363133546336e-05,
+      "loss": 0.9115,
+      "step": 9350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429749599418566e-05,
+      "loss": 0.8579,
+      "step": 9351
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1428136027289684e-05,
+      "loss": 0.9571,
+      "step": 9352
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1426522417202577e-05,
+      "loss": 0.9253,
+      "step": 9353
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1424908769200137e-05,
+      "loss": 0.969,
+      "step": 9354
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423295083325242e-05,
+      "loss": 0.9131,
+      "step": 9355
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1421681359620793e-05,
+      "loss": 0.8915,
+      "step": 9356
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1420067598129679e-05,
+      "loss": 0.9617,
+      "step": 9357
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1418453798894786e-05,
+      "loss": 0.8596,
+      "step": 9358
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1416839961959015e-05,
+      "loss": 0.9498,
+      "step": 9359
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1415226087365251e-05,
+      "loss": 0.9238,
+      "step": 9360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1413612175156396e-05,
+      "loss": 0.9299,
+      "step": 9361
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141199822537535e-05,
+      "loss": 0.9572,
+      "step": 9362
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410384238064996e-05,
+      "loss": 0.9195,
+      "step": 9363
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1408770213268244e-05,
+      "loss": 0.9411,
+      "step": 9364
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1407156151027986e-05,
+      "loss": 1.0433,
+      "step": 9365
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1405542051387127e-05,
+      "loss": 0.9558,
+      "step": 9366
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1403927914388563e-05,
+      "loss": 0.9883,
+      "step": 9367
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402313740075202e-05,
+      "loss": 0.872,
+      "step": 9368
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140069952848994e-05,
+      "loss": 1.0691,
+      "step": 9369
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1399085279675688e-05,
+      "loss": 0.939,
+      "step": 9370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397470993675343e-05,
+      "loss": 0.8811,
+      "step": 9371
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1395856670531815e-05,
+      "loss": 0.8664,
+      "step": 9372
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1394242310288015e-05,
+      "loss": 0.9412,
+      "step": 9373
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1392627912986845e-05,
+      "loss": 0.4196,
+      "step": 9374
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1391013478671216e-05,
+      "loss": 0.8992,
+      "step": 9375
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1389399007384034e-05,
+      "loss": 0.929,
+      "step": 9376
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387784499168218e-05,
+      "loss": 0.9456,
+      "step": 9377
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138616995406667e-05,
+      "loss": 0.8552,
+      "step": 9378
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138455537212231e-05,
+      "loss": 0.9305,
+      "step": 9379
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1382940753378051e-05,
+      "loss": 0.9669,
+      "step": 9380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1381326097876805e-05,
+      "loss": 0.918,
+      "step": 9381
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137971140566149e-05,
+      "loss": 0.9009,
+      "step": 9382
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1378096676775023e-05,
+      "loss": 0.9751,
+      "step": 9383
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1376481911260315e-05,
+      "loss": 0.8519,
+      "step": 9384
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1374867109160295e-05,
+      "loss": 0.9438,
+      "step": 9385
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1373252270517877e-05,
+      "loss": 0.9297,
+      "step": 9386
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1371637395375978e-05,
+      "loss": 0.9346,
+      "step": 9387
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137002248377753e-05,
+      "loss": 0.9604,
+      "step": 9388
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1368407535765443e-05,
+      "loss": 0.811,
+      "step": 9389
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1366792551382649e-05,
+      "loss": 0.9303,
+      "step": 9390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365177530672069e-05,
+      "loss": 0.9243,
+      "step": 9391
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136356247367663e-05,
+      "loss": 0.8484,
+      "step": 9392
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136194738043926e-05,
+      "loss": 0.9239,
+      "step": 9393
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1360332251002882e-05,
+      "loss": 0.9438,
+      "step": 9394
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1358717085410427e-05,
+      "loss": 0.9184,
+      "step": 9395
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1357101883704821e-05,
+      "loss": 0.8956,
+      "step": 9396
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1355486645929e-05,
+      "loss": 0.8892,
+      "step": 9397
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1353871372125891e-05,
+      "loss": 0.9655,
+      "step": 9398
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1352256062338428e-05,
+      "loss": 0.4089,
+      "step": 9399
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350640716609544e-05,
+      "loss": 0.8902,
+      "step": 9400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1349025334982172e-05,
+      "loss": 0.9522,
+      "step": 9401
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1347409917499243e-05,
+      "loss": 0.8661,
+      "step": 9402
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13457944642037e-05,
+      "loss": 0.9345,
+      "step": 9403
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134417897513848e-05,
+      "loss": 0.9173,
+      "step": 9404
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1342563450346516e-05,
+      "loss": 0.8654,
+      "step": 9405
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134094788987075e-05,
+      "loss": 0.9103,
+      "step": 9406
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339332293754119e-05,
+      "loss": 0.9343,
+      "step": 9407
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1337716662039564e-05,
+      "loss": 0.9851,
+      "step": 9408
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1336100994770031e-05,
+      "loss": 0.9797,
+      "step": 9409
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1334485291988458e-05,
+      "loss": 0.937,
+      "step": 9410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133286955373779e-05,
+      "loss": 0.8713,
+      "step": 9411
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331253780060972e-05,
+      "loss": 0.9994,
+      "step": 9412
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132963797100095e-05,
+      "loss": 0.9067,
+      "step": 9413
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1328022126600666e-05,
+      "loss": 0.9334,
+      "step": 9414
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1326406246903072e-05,
+      "loss": 0.9568,
+      "step": 9415
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1324790331951117e-05,
+      "loss": 0.8312,
+      "step": 9416
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323174381787747e-05,
+      "loss": 0.9016,
+      "step": 9417
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1321558396455913e-05,
+      "loss": 0.9541,
+      "step": 9418
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1319942375998567e-05,
+      "loss": 0.9805,
+      "step": 9419
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1318326320458655e-05,
+      "loss": 0.9856,
+      "step": 9420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131671022987914e-05,
+      "loss": 0.837,
+      "step": 9421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1315094104302973e-05,
+      "loss": 0.9506,
+      "step": 9422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1313477943773102e-05,
+      "loss": 0.8299,
+      "step": 9423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1311861748332488e-05,
+      "loss": 0.9215,
+      "step": 9424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1310245518024087e-05,
+      "loss": 0.9389,
+      "step": 9425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130862925289086e-05,
+      "loss": 0.8571,
+      "step": 9426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130701295297576e-05,
+      "loss": 1.1023,
+      "step": 9427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1305396618321747e-05,
+      "loss": 0.9519,
+      "step": 9428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1303780248971784e-05,
+      "loss": 0.9156,
+      "step": 9429
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1302163844968832e-05,
+      "loss": 0.9212,
+      "step": 9430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1300547406355854e-05,
+      "loss": 0.9917,
+      "step": 9431
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1298930933175805e-05,
+      "loss": 0.8784,
+      "step": 9432
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1297314425471657e-05,
+      "loss": 0.9809,
+      "step": 9433
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1295697883286379e-05,
+      "loss": 0.9003,
+      "step": 9434
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294081306662928e-05,
+      "loss": 0.9099,
+      "step": 9435
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1292464695644272e-05,
+      "loss": 0.9646,
+      "step": 9436
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290848050273379e-05,
+      "loss": 0.8526,
+      "step": 9437
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1289231370593222e-05,
+      "loss": 0.4216,
+      "step": 9438
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1287614656646768e-05,
+      "loss": 0.9385,
+      "step": 9439
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1285997908476988e-05,
+      "loss": 0.9035,
+      "step": 9440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.128438112612685e-05,
+      "loss": 0.9464,
+      "step": 9441
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.128276430963933e-05,
+      "loss": 0.9961,
+      "step": 9442
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1281147459057398e-05,
+      "loss": 0.9891,
+      "step": 9443
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127953057442403e-05,
+      "loss": 0.4565,
+      "step": 9444
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1277913655782203e-05,
+      "loss": 0.9637,
+      "step": 9445
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1276296703174887e-05,
+      "loss": 0.9149,
+      "step": 9446
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1274679716645064e-05,
+      "loss": 0.9025,
+      "step": 9447
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127306269623571e-05,
+      "loss": 0.8797,
+      "step": 9448
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1271445641989805e-05,
+      "loss": 0.9542,
+      "step": 9449
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1269828553950327e-05,
+      "loss": 0.9659,
+      "step": 9450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1268211432160252e-05,
+      "loss": 0.915,
+      "step": 9451
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126659427666257e-05,
+      "loss": 0.9435,
+      "step": 9452
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126497708750026e-05,
+      "loss": 0.9897,
+      "step": 9453
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1263359864716299e-05,
+      "loss": 0.396,
+      "step": 9454
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1261742608353677e-05,
+      "loss": 0.9725,
+      "step": 9455
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126012531845538e-05,
+      "loss": 0.9162,
+      "step": 9456
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1258507995064387e-05,
+      "loss": 0.9267,
+      "step": 9457
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1256890638223695e-05,
+      "loss": 0.938,
+      "step": 9458
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125527324797628e-05,
+      "loss": 0.8632,
+      "step": 9459
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1253655824365137e-05,
+      "loss": 0.9265,
+      "step": 9460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1252038367433254e-05,
+      "loss": 0.9345,
+      "step": 9461
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125042087722362e-05,
+      "loss": 0.9866,
+      "step": 9462
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1248803353779233e-05,
+      "loss": 0.9397,
+      "step": 9463
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1247185797143071e-05,
+      "loss": 0.9682,
+      "step": 9464
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1245568207358142e-05,
+      "loss": 0.8583,
+      "step": 9465
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1243950584467429e-05,
+      "loss": 0.897,
+      "step": 9466
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1242332928513929e-05,
+      "loss": 0.9266,
+      "step": 9467
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124071523954064e-05,
+      "loss": 0.9519,
+      "step": 9468
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1239097517590559e-05,
+      "loss": 0.913,
+      "step": 9469
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1237479762706678e-05,
+      "loss": 0.8535,
+      "step": 9470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1235861974932e-05,
+      "loss": 0.9039,
+      "step": 9471
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1234244154309522e-05,
+      "loss": 0.9223,
+      "step": 9472
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232626300882243e-05,
+      "loss": 0.9276,
+      "step": 9473
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1231008414693167e-05,
+      "loss": 0.9047,
+      "step": 9474
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1229390495785291e-05,
+      "loss": 0.9083,
+      "step": 9475
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1227772544201622e-05,
+      "loss": 0.8597,
+      "step": 9476
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1226154559985159e-05,
+      "loss": 0.9272,
+      "step": 9477
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122453654317891e-05,
+      "loss": 0.9264,
+      "step": 9478
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1222918493825876e-05,
+      "loss": 1.0044,
+      "step": 9479
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1221300411969065e-05,
+      "loss": 0.9087,
+      "step": 9480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1219682297651487e-05,
+      "loss": 0.9383,
+      "step": 9481
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1218064150916143e-05,
+      "loss": 1.0351,
+      "step": 9482
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1216445971806047e-05,
+      "loss": 0.9235,
+      "step": 9483
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121482776036421e-05,
+      "loss": 0.883,
+      "step": 9484
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121320951663363e-05,
+      "loss": 0.9469,
+      "step": 9485
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1211591240657334e-05,
+      "loss": 0.4081,
+      "step": 9486
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209972932478322e-05,
+      "loss": 0.9428,
+      "step": 9487
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1208354592139614e-05,
+      "loss": 0.9247,
+      "step": 9488
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1206736219684223e-05,
+      "loss": 0.9833,
+      "step": 9489
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1205117815155156e-05,
+      "loss": 0.9993,
+      "step": 9490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1203499378595434e-05,
+      "loss": 0.8885,
+      "step": 9491
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1201880910048079e-05,
+      "loss": 0.9189,
+      "step": 9492
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1200262409556098e-05,
+      "loss": 0.8549,
+      "step": 9493
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1198643877162513e-05,
+      "loss": 1.01,
+      "step": 9494
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197025312910344e-05,
+      "loss": 0.8852,
+      "step": 9495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1195406716842607e-05,
+      "loss": 0.9649,
+      "step": 9496
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1193788089002329e-05,
+      "loss": 1.0303,
+      "step": 9497
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1192169429432524e-05,
+      "loss": 0.8742,
+      "step": 9498
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1190550738176217e-05,
+      "loss": 0.8953,
+      "step": 9499
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188932015276432e-05,
+      "loss": 0.9154,
+      "step": 9500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1187313260776194e-05,
+      "loss": 0.9823,
+      "step": 9501
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1185694474718523e-05,
+      "loss": 0.8755,
+      "step": 9502
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1184075657146445e-05,
+      "loss": 0.9137,
+      "step": 9503
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1182456808102993e-05,
+      "loss": 0.9784,
+      "step": 9504
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180837927631188e-05,
+      "loss": 0.9013,
+      "step": 9505
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1179219015774059e-05,
+      "loss": 0.9353,
+      "step": 9506
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1177600072574635e-05,
+      "loss": 0.9872,
+      "step": 9507
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1175981098075943e-05,
+      "loss": 0.9281,
+      "step": 9508
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117436209232102e-05,
+      "loss": 0.8663,
+      "step": 9509
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1172743055352897e-05,
+      "loss": 0.9665,
+      "step": 9510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1171123987214598e-05,
+      "loss": 0.9002,
+      "step": 9511
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169504887949163e-05,
+      "loss": 0.4323,
+      "step": 9512
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1167885757599623e-05,
+      "loss": 0.9462,
+      "step": 9513
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1166266596209013e-05,
+      "loss": 0.4612,
+      "step": 9514
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116464740382037e-05,
+      "loss": 0.8858,
+      "step": 9515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163028180476729e-05,
+      "loss": 1.0089,
+      "step": 9516
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1161408926221129e-05,
+      "loss": 0.8935,
+      "step": 9517
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1159789641096603e-05,
+      "loss": 0.8448,
+      "step": 9518
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1158170325146194e-05,
+      "loss": 0.9207,
+      "step": 9519
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115655097841294e-05,
+      "loss": 0.9695,
+      "step": 9520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1154931600939883e-05,
+      "loss": 0.9963,
+      "step": 9521
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1153312192770062e-05,
+      "loss": 0.9251,
+      "step": 9522
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1151692753946523e-05,
+      "loss": 0.9432,
+      "step": 9523
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1150073284512304e-05,
+      "loss": 0.8635,
+      "step": 9524
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1148453784510452e-05,
+      "loss": 0.9299,
+      "step": 9525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146834253984008e-05,
+      "loss": 0.8681,
+      "step": 9526
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1145214692976016e-05,
+      "loss": 0.8012,
+      "step": 9527
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143595101529532e-05,
+      "loss": 0.9364,
+      "step": 9528
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1141975479687592e-05,
+      "loss": 0.8717,
+      "step": 9529
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114035582749325e-05,
+      "loss": 0.8494,
+      "step": 9530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138736144989553e-05,
+      "loss": 0.8991,
+      "step": 9531
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137116432219548e-05,
+      "loss": 0.9817,
+      "step": 9532
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1135496689226289e-05,
+      "loss": 0.9098,
+      "step": 9533
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1133876916052822e-05,
+      "loss": 0.9413,
+      "step": 9534
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1132257112742204e-05,
+      "loss": 1.0076,
+      "step": 9535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130637279337484e-05,
+      "loss": 0.9142,
+      "step": 9536
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1129017415881718e-05,
+      "loss": 0.9836,
+      "step": 9537
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1127397522417955e-05,
+      "loss": 0.9409,
+      "step": 9538
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1125777598989255e-05,
+      "loss": 0.9541,
+      "step": 9539
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124157645638675e-05,
+      "loss": 0.9707,
+      "step": 9540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122537662409267e-05,
+      "loss": 0.929,
+      "step": 9541
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1120917649344089e-05,
+      "loss": 0.8825,
+      "step": 9542
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119297606486204e-05,
+      "loss": 0.9043,
+      "step": 9543
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1117677533878662e-05,
+      "loss": 0.9706,
+      "step": 9544
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111605743156453e-05,
+      "loss": 0.8283,
+      "step": 9545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1114437299586864e-05,
+      "loss": 0.987,
+      "step": 9546
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111281713798873e-05,
+      "loss": 0.9115,
+      "step": 9547
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1111196946813188e-05,
+      "loss": 0.9325,
+      "step": 9548
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109576726103297e-05,
+      "loss": 0.8339,
+      "step": 9549
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1107956475902125e-05,
+      "loss": 0.927,
+      "step": 9550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106336196252735e-05,
+      "loss": 0.9167,
+      "step": 9551
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1104715887198194e-05,
+      "loss": 0.8931,
+      "step": 9552
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103095548781566e-05,
+      "loss": 0.9427,
+      "step": 9553
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.110147518104592e-05,
+      "loss": 0.9076,
+      "step": 9554
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1099854784034319e-05,
+      "loss": 0.9603,
+      "step": 9555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1098234357789832e-05,
+      "loss": 0.9239,
+      "step": 9556
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096613902355532e-05,
+      "loss": 0.8749,
+      "step": 9557
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1094993417774489e-05,
+      "loss": 0.9681,
+      "step": 9558
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1093372904089772e-05,
+      "loss": 0.7883,
+      "step": 9559
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1091752361344448e-05,
+      "loss": 0.8555,
+      "step": 9560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1090131789581595e-05,
+      "loss": 0.8585,
+      "step": 9561
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1088511188844284e-05,
+      "loss": 0.8961,
+      "step": 9562
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108689055917559e-05,
+      "loss": 0.9572,
+      "step": 9563
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085269900618587e-05,
+      "loss": 0.9414,
+      "step": 9564
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1083649213216346e-05,
+      "loss": 0.379,
+      "step": 9565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108202849701195e-05,
+      "loss": 0.8655,
+      "step": 9566
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108040775204847e-05,
+      "loss": 0.8878,
+      "step": 9567
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1078786978368985e-05,
+      "loss": 0.9341,
+      "step": 9568
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1077166176016579e-05,
+      "loss": 0.9029,
+      "step": 9569
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107554534503432e-05,
+      "loss": 0.9823,
+      "step": 9570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1073924485465299e-05,
+      "loss": 0.9187,
+      "step": 9571
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1072303597352592e-05,
+      "loss": 0.9137,
+      "step": 9572
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1070682680739275e-05,
+      "loss": 0.4476,
+      "step": 9573
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1069061735668439e-05,
+      "loss": 0.8335,
+      "step": 9574
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1067440762183164e-05,
+      "loss": 0.9093,
+      "step": 9575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106581976032653e-05,
+      "loss": 1.0193,
+      "step": 9576
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1064198730141626e-05,
+      "loss": 0.9599,
+      "step": 9577
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1062577671671533e-05,
+      "loss": 0.9574,
+      "step": 9578
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106095658495934e-05,
+      "loss": 0.8345,
+      "step": 9579
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1059335470048134e-05,
+      "loss": 0.8833,
+      "step": 9580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1057714326981e-05,
+      "loss": 0.9213,
+      "step": 9581
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1056093155801026e-05,
+      "loss": 0.9209,
+      "step": 9582
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1054471956551303e-05,
+      "loss": 0.9502,
+      "step": 9583
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1052850729274922e-05,
+      "loss": 0.8797,
+      "step": 9584
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1051229474014968e-05,
+      "loss": 0.92,
+      "step": 9585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1049608190814536e-05,
+      "loss": 0.8876,
+      "step": 9586
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1047986879716718e-05,
+      "loss": 0.9452,
+      "step": 9587
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1046365540764605e-05,
+      "loss": 0.8969,
+      "step": 9588
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1044744174001292e-05,
+      "loss": 0.9731,
+      "step": 9589
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1043122779469872e-05,
+      "loss": 0.9352,
+      "step": 9590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1041501357213435e-05,
+      "loss": 0.857,
+      "step": 9591
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103987990727509e-05,
+      "loss": 0.4572,
+      "step": 9592
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103825842969792e-05,
+      "loss": 0.941,
+      "step": 9593
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1036636924525026e-05,
+      "loss": 0.9582,
+      "step": 9594
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1035015391799509e-05,
+      "loss": 0.8738,
+      "step": 9595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1033393831564461e-05,
+      "loss": 0.9652,
+      "step": 9596
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1031772243862988e-05,
+      "loss": 0.9028,
+      "step": 9597
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1030150628738187e-05,
+      "loss": 1.0034,
+      "step": 9598
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1028528986233157e-05,
+      "loss": 0.9856,
+      "step": 9599
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1026907316391004e-05,
+      "loss": 0.8531,
+      "step": 9600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1025285619254827e-05,
+      "loss": 0.823,
+      "step": 9601
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1023663894867728e-05,
+      "loss": 0.8812,
+      "step": 9602
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1022042143272813e-05,
+      "loss": 1.0298,
+      "step": 9603
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1020420364513182e-05,
+      "loss": 0.854,
+      "step": 9604
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018798558631947e-05,
+      "loss": 0.9869,
+      "step": 9605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1017176725672207e-05,
+      "loss": 0.9039,
+      "step": 9606
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1015554865677073e-05,
+      "loss": 0.901,
+      "step": 9607
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1013932978689654e-05,
+      "loss": 0.9438,
+      "step": 9608
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1012311064753048e-05,
+      "loss": 0.908,
+      "step": 9609
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1010689123910374e-05,
+      "loss": 0.9278,
+      "step": 9610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1009067156204736e-05,
+      "loss": 0.9957,
+      "step": 9611
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1007445161679247e-05,
+      "loss": 0.902,
+      "step": 9612
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005823140377014e-05,
+      "loss": 0.8754,
+      "step": 9613
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1004201092341152e-05,
+      "loss": 0.9117,
+      "step": 9614
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100257901761477e-05,
+      "loss": 0.9677,
+      "step": 9615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000956916240985e-05,
+      "loss": 0.975,
+      "step": 9616
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0999334788262907e-05,
+      "loss": 0.9757,
+      "step": 9617
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0997712633723653e-05,
+      "loss": 0.9176,
+      "step": 9618
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0996090452666336e-05,
+      "loss": 0.8723,
+      "step": 9619
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0994468245134071e-05,
+      "loss": 0.9288,
+      "step": 9620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0992846011169978e-05,
+      "loss": 0.9108,
+      "step": 9621
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099122375081717e-05,
+      "loss": 0.398,
+      "step": 9622
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0989601464118771e-05,
+      "loss": 0.4492,
+      "step": 9623
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0987979151117894e-05,
+      "loss": 0.9867,
+      "step": 9624
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0986356811857654e-05,
+      "loss": 0.993,
+      "step": 9625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098473444638118e-05,
+      "loss": 0.9549,
+      "step": 9626
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0983112054731589e-05,
+      "loss": 0.8869,
+      "step": 9627
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0981489636952005e-05,
+      "loss": 0.9912,
+      "step": 9628
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979867193085546e-05,
+      "loss": 0.9198,
+      "step": 9629
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0978244723175332e-05,
+      "loss": 0.8899,
+      "step": 9630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0976622227264497e-05,
+      "loss": 0.9061,
+      "step": 9631
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974999705396155e-05,
+      "loss": 0.968,
+      "step": 9632
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0973377157613434e-05,
+      "loss": 0.9664,
+      "step": 9633
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0971754583959462e-05,
+      "loss": 0.8748,
+      "step": 9634
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0970131984477362e-05,
+      "loss": 0.9588,
+      "step": 9635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968509359210262e-05,
+      "loss": 0.9983,
+      "step": 9636
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966886708201288e-05,
+      "loss": 0.965,
+      "step": 9637
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0965264031493567e-05,
+      "loss": 0.8518,
+      "step": 9638
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963641329130237e-05,
+      "loss": 0.9207,
+      "step": 9639
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962018601154416e-05,
+      "loss": 0.9374,
+      "step": 9640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0960395847609244e-05,
+      "loss": 0.9162,
+      "step": 9641
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0958773068537844e-05,
+      "loss": 0.8895,
+      "step": 9642
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0957150263983351e-05,
+      "loss": 0.8342,
+      "step": 9643
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0955527433988897e-05,
+      "loss": 0.8872,
+      "step": 9644
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0953904578597616e-05,
+      "loss": 0.9339,
+      "step": 9645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0952281697852642e-05,
+      "loss": 0.8885,
+      "step": 9646
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0950658791797106e-05,
+      "loss": 1.0475,
+      "step": 9647
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0949035860474146e-05,
+      "loss": 0.9579,
+      "step": 9648
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0947412903926897e-05,
+      "loss": 0.9584,
+      "step": 9649
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0945789922198493e-05,
+      "loss": 0.9315,
+      "step": 9650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944166915332075e-05,
+      "loss": 0.966,
+      "step": 9651
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094254388337078e-05,
+      "loss": 0.8231,
+      "step": 9652
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0940920826357742e-05,
+      "loss": 0.9271,
+      "step": 9653
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0939297744336105e-05,
+      "loss": 0.8914,
+      "step": 9654
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0937674637349005e-05,
+      "loss": 0.9726,
+      "step": 9655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0936051505439584e-05,
+      "loss": 0.4184,
+      "step": 9656
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934428348650987e-05,
+      "loss": 0.8956,
+      "step": 9657
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0932805167026347e-05,
+      "loss": 0.9241,
+      "step": 9658
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0931181960608815e-05,
+      "loss": 0.9535,
+      "step": 9659
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092955872944153e-05,
+      "loss": 0.8744,
+      "step": 9660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0927935473567632e-05,
+      "loss": 0.942,
+      "step": 9661
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0926312193030269e-05,
+      "loss": 0.9834,
+      "step": 9662
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0924688887872588e-05,
+      "loss": 0.954,
+      "step": 9663
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092306555813773e-05,
+      "loss": 0.9045,
+      "step": 9664
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0921442203868848e-05,
+      "loss": 0.9637,
+      "step": 9665
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0919818825109081e-05,
+      "loss": 0.9159,
+      "step": 9666
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918195421901583e-05,
+      "loss": 0.927,
+      "step": 9667
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0916571994289497e-05,
+      "loss": 0.917,
+      "step": 9668
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0914948542315976e-05,
+      "loss": 0.9082,
+      "step": 9669
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913325066024164e-05,
+      "loss": 0.8979,
+      "step": 9670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0911701565457217e-05,
+      "loss": 0.9312,
+      "step": 9671
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910078040658287e-05,
+      "loss": 0.9635,
+      "step": 9672
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090845449167052e-05,
+      "loss": 0.9733,
+      "step": 9673
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906830918537066e-05,
+      "loss": 0.9433,
+      "step": 9674
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0905207321301087e-05,
+      "loss": 0.9243,
+      "step": 9675
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090358370000573e-05,
+      "loss": 0.9106,
+      "step": 9676
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090196005469415e-05,
+      "loss": 0.9392,
+      "step": 9677
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0900336385409503e-05,
+      "loss": 0.9251,
+      "step": 9678
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0898712692194939e-05,
+      "loss": 0.8911,
+      "step": 9679
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897088975093622e-05,
+      "loss": 0.9584,
+      "step": 9680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895465234148703e-05,
+      "loss": 0.8698,
+      "step": 9681
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0893841469403341e-05,
+      "loss": 0.9534,
+      "step": 9682
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0892217680900696e-05,
+      "loss": 0.9526,
+      "step": 9683
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.089059386868392e-05,
+      "loss": 0.9111,
+      "step": 9684
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0888970032796177e-05,
+      "loss": 1.0081,
+      "step": 9685
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088734617328063e-05,
+      "loss": 0.7995,
+      "step": 9686
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088572229018043e-05,
+      "loss": 0.9467,
+      "step": 9687
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0884098383538745e-05,
+      "loss": 0.8098,
+      "step": 9688
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0882474453398734e-05,
+      "loss": 0.9271,
+      "step": 9689
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0880850499803564e-05,
+      "loss": 0.9307,
+      "step": 9690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0879226522796389e-05,
+      "loss": 0.9888,
+      "step": 9691
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0877602522420379e-05,
+      "loss": 0.9556,
+      "step": 9692
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0875978498718695e-05,
+      "loss": 1.0239,
+      "step": 9693
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0874354451734503e-05,
+      "loss": 0.9466,
+      "step": 9694
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0872730381510972e-05,
+      "loss": 0.9026,
+      "step": 9695
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0871106288091262e-05,
+      "loss": 0.9513,
+      "step": 9696
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086948217151854e-05,
+      "loss": 0.8095,
+      "step": 9697
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0867858031835975e-05,
+      "loss": 0.9245,
+      "step": 9698
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0866233869086737e-05,
+      "loss": 0.8942,
+      "step": 9699
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0864609683313992e-05,
+      "loss": 0.8616,
+      "step": 9700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086298547456091e-05,
+      "loss": 0.9169,
+      "step": 9701
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861361242870656e-05,
+      "loss": 0.4621,
+      "step": 9702
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0859736988286405e-05,
+      "loss": 0.9744,
+      "step": 9703
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0858112710851327e-05,
+      "loss": 0.8502,
+      "step": 9704
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0856488410608595e-05,
+      "loss": 0.9868,
+      "step": 9705
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854864087601377e-05,
+      "loss": 0.9212,
+      "step": 9706
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085323974187285e-05,
+      "loss": 0.8501,
+      "step": 9707
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0851615373466185e-05,
+      "loss": 1.0178,
+      "step": 9708
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0849990982424554e-05,
+      "loss": 0.8667,
+      "step": 9709
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0848366568791133e-05,
+      "loss": 0.9698,
+      "step": 9710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08467421326091e-05,
+      "loss": 0.8855,
+      "step": 9711
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0845117673921628e-05,
+      "loss": 0.8396,
+      "step": 9712
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0843493192771893e-05,
+      "loss": 0.9592,
+      "step": 9713
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0841868689203072e-05,
+      "loss": 0.9475,
+      "step": 9714
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.084024416325834e-05,
+      "loss": 0.9326,
+      "step": 9715
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.083861961498088e-05,
+      "loss": 0.899,
+      "step": 9716
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836995044413872e-05,
+      "loss": 0.9514,
+      "step": 9717
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0835370451600487e-05,
+      "loss": 0.8981,
+      "step": 9718
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833745836583914e-05,
+      "loss": 0.9235,
+      "step": 9719
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0832121199407324e-05,
+      "loss": 0.9536,
+      "step": 9720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830496540113904e-05,
+      "loss": 0.4783,
+      "step": 9721
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828871858746837e-05,
+      "loss": 0.9344,
+      "step": 9722
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827247155349299e-05,
+      "loss": 0.8924,
+      "step": 9723
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082562242996448e-05,
+      "loss": 0.981,
+      "step": 9724
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0823997682635557e-05,
+      "loss": 0.8826,
+      "step": 9725
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0822372913405719e-05,
+      "loss": 0.915,
+      "step": 9726
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820748122318145e-05,
+      "loss": 0.877,
+      "step": 9727
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0819123309416024e-05,
+      "loss": 0.9551,
+      "step": 9728
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817498474742545e-05,
+      "loss": 0.9051,
+      "step": 9729
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0815873618340889e-05,
+      "loss": 0.942,
+      "step": 9730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0814248740254241e-05,
+      "loss": 0.9056,
+      "step": 9731
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0812623840525793e-05,
+      "loss": 0.9743,
+      "step": 9732
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0810998919198731e-05,
+      "loss": 0.9108,
+      "step": 9733
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0809373976316245e-05,
+      "loss": 0.8506,
+      "step": 9734
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0807749011921522e-05,
+      "loss": 0.9333,
+      "step": 9735
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0806124026057754e-05,
+      "loss": 0.8737,
+      "step": 9736
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080449901876813e-05,
+      "loss": 0.8645,
+      "step": 9737
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0802873990095839e-05,
+      "loss": 0.9015,
+      "step": 9738
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0801248940084074e-05,
+      "loss": 0.9176,
+      "step": 9739
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0799623868776031e-05,
+      "loss": 0.9374,
+      "step": 9740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0797998776214896e-05,
+      "loss": 0.9162,
+      "step": 9741
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796373662443866e-05,
+      "loss": 0.9575,
+      "step": 9742
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.079474852750613e-05,
+      "loss": 0.9512,
+      "step": 9743
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.079312337144489e-05,
+      "loss": 0.952,
+      "step": 9744
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791498194303339e-05,
+      "loss": 1.016,
+      "step": 9745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789872996124664e-05,
+      "loss": 0.919,
+      "step": 9746
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078824777695207e-05,
+      "loss": 0.9237,
+      "step": 9747
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0786622536828753e-05,
+      "loss": 0.906,
+      "step": 9748
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0784997275797901e-05,
+      "loss": 0.9306,
+      "step": 9749
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783371993902722e-05,
+      "loss": 0.9427,
+      "step": 9750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078174669118641e-05,
+      "loss": 0.9693,
+      "step": 9751
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0780121367692162e-05,
+      "loss": 0.9614,
+      "step": 9752
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0778496023463183e-05,
+      "loss": 0.4433,
+      "step": 9753
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0776870658542665e-05,
+      "loss": 0.8764,
+      "step": 9754
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0775245272973814e-05,
+      "loss": 0.9245,
+      "step": 9755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0773619866799826e-05,
+      "loss": 0.4316,
+      "step": 9756
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0771994440063906e-05,
+      "loss": 0.8969,
+      "step": 9757
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0770368992809259e-05,
+      "loss": 0.9762,
+      "step": 9758
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0768743525079082e-05,
+      "loss": 0.9544,
+      "step": 9759
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0767118036916582e-05,
+      "loss": 0.9014,
+      "step": 9760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076549252836496e-05,
+      "loss": 0.9423,
+      "step": 9761
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0763866999467416e-05,
+      "loss": 0.9352,
+      "step": 9762
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0762241450267165e-05,
+      "loss": 1.0003,
+      "step": 9763
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0760615880807403e-05,
+      "loss": 0.8603,
+      "step": 9764
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0758990291131344e-05,
+      "loss": 0.8319,
+      "step": 9765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.075736468128219e-05,
+      "loss": 1.027,
+      "step": 9766
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755739051303145e-05,
+      "loss": 0.9507,
+      "step": 9767
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754113401237418e-05,
+      "loss": 0.9676,
+      "step": 9768
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.075248773112822e-05,
+      "loss": 0.8776,
+      "step": 9769
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750862041018756e-05,
+      "loss": 0.9098,
+      "step": 9770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074923633095224e-05,
+      "loss": 0.8437,
+      "step": 9771
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0747610600971874e-05,
+      "loss": 0.9969,
+      "step": 9772
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0745984851120877e-05,
+      "loss": 0.874,
+      "step": 9773
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0744359081442451e-05,
+      "loss": 0.9032,
+      "step": 9774
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074273329197981e-05,
+      "loss": 0.9391,
+      "step": 9775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0741107482776168e-05,
+      "loss": 0.9563,
+      "step": 9776
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0739481653874735e-05,
+      "loss": 0.9239,
+      "step": 9777
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0737855805318724e-05,
+      "loss": 0.833,
+      "step": 9778
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073622993715135e-05,
+      "loss": 0.9409,
+      "step": 9779
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0734604049415822e-05,
+      "loss": 0.9663,
+      "step": 9780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0732978142155361e-05,
+      "loss": 0.9153,
+      "step": 9781
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0731352215413178e-05,
+      "loss": 0.9387,
+      "step": 9782
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0729726269232486e-05,
+      "loss": 0.9494,
+      "step": 9783
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0728100303656505e-05,
+      "loss": 0.9008,
+      "step": 9784
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0726474318728447e-05,
+      "loss": 0.8963,
+      "step": 9785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0724848314491531e-05,
+      "loss": 0.8932,
+      "step": 9786
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072322229098898e-05,
+      "loss": 0.9776,
+      "step": 9787
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0721596248264001e-05,
+      "loss": 0.9491,
+      "step": 9788
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071997018635982e-05,
+      "loss": 0.9786,
+      "step": 9789
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0718344105319654e-05,
+      "loss": 0.9108,
+      "step": 9790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716718005186724e-05,
+      "loss": 1.0033,
+      "step": 9791
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0715091886004242e-05,
+      "loss": 0.9638,
+      "step": 9792
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0713465747815439e-05,
+      "loss": 1.0639,
+      "step": 9793
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0711839590663531e-05,
+      "loss": 0.9777,
+      "step": 9794
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071021341459174e-05,
+      "loss": 1.0122,
+      "step": 9795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708587219643287e-05,
+      "loss": 0.9828,
+      "step": 9796
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0706961005861396e-05,
+      "loss": 0.9181,
+      "step": 9797
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0705334773289285e-05,
+      "loss": 0.9171,
+      "step": 9798
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0703708521970183e-05,
+      "loss": 0.4775,
+      "step": 9799
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0702082251947314e-05,
+      "loss": 0.8923,
+      "step": 9800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.07004559632639e-05,
+      "loss": 0.4117,
+      "step": 9801
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0698829655963168e-05,
+      "loss": 0.9129,
+      "step": 9802
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0697203330088334e-05,
+      "loss": 0.9209,
+      "step": 9803
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0695576985682641e-05,
+      "loss": 0.9614,
+      "step": 9804
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0693950622789304e-05,
+      "loss": 0.9801,
+      "step": 9805
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0692324241451551e-05,
+      "loss": 0.9816,
+      "step": 9806
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0690697841712609e-05,
+      "loss": 0.932,
+      "step": 9807
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0689071423615708e-05,
+      "loss": 0.9572,
+      "step": 9808
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0687444987204077e-05,
+      "loss": 0.8897,
+      "step": 9809
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685818532520945e-05,
+      "loss": 0.9576,
+      "step": 9810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0684192059609535e-05,
+      "loss": 0.953,
+      "step": 9811
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682565568513084e-05,
+      "loss": 0.9334,
+      "step": 9812
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0680939059274818e-05,
+      "loss": 0.9206,
+      "step": 9813
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679312531937972e-05,
+      "loss": 1.0317,
+      "step": 9814
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0677685986545774e-05,
+      "loss": 0.9567,
+      "step": 9815
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676059423141454e-05,
+      "loss": 0.9448,
+      "step": 9816
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.067443284176825e-05,
+      "loss": 0.963,
+      "step": 9817
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0672806242469392e-05,
+      "loss": 0.9494,
+      "step": 9818
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671179625288109e-05,
+      "loss": 0.9192,
+      "step": 9819
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0669552990267642e-05,
+      "loss": 0.9733,
+      "step": 9820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0667926337451217e-05,
+      "loss": 0.9743,
+      "step": 9821
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0666299666882074e-05,
+      "loss": 0.914,
+      "step": 9822
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066467297860345e-05,
+      "loss": 0.9619,
+      "step": 9823
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0663046272658573e-05,
+      "loss": 0.8896,
+      "step": 9824
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0661419549090687e-05,
+      "loss": 0.9147,
+      "step": 9825
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0659792807943021e-05,
+      "loss": 0.8836,
+      "step": 9826
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0658166049258818e-05,
+      "loss": 0.8909,
+      "step": 9827
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0656539273081315e-05,
+      "loss": 0.9559,
+      "step": 9828
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0654912479453743e-05,
+      "loss": 1.0007,
+      "step": 9829
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0653285668419351e-05,
+      "loss": 0.8477,
+      "step": 9830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0651658840021368e-05,
+      "loss": 0.9189,
+      "step": 9831
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0650031994303038e-05,
+      "loss": 0.9855,
+      "step": 9832
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.06484051313076e-05,
+      "loss": 0.933,
+      "step": 9833
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0646778251078293e-05,
+      "loss": 0.8888,
+      "step": 9834
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0645151353658363e-05,
+      "loss": 0.9785,
+      "step": 9835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0643524439091046e-05,
+      "loss": 0.9715,
+      "step": 9836
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064189750741958e-05,
+      "loss": 0.9826,
+      "step": 9837
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640270558687216e-05,
+      "loss": 0.9935,
+      "step": 9838
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063864359293719e-05,
+      "loss": 0.9241,
+      "step": 9839
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0637016610212744e-05,
+      "loss": 0.9502,
+      "step": 9840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063538961055713e-05,
+      "loss": 1.0026,
+      "step": 9841
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0633762594013582e-05,
+      "loss": 0.9719,
+      "step": 9842
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0632135560625348e-05,
+      "loss": 0.906,
+      "step": 9843
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0630508510435674e-05,
+      "loss": 0.8544,
+      "step": 9844
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06288814434878e-05,
+      "loss": 0.4058,
+      "step": 9845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0627254359824981e-05,
+      "loss": 0.4283,
+      "step": 9846
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625627259490454e-05,
+      "loss": 0.9231,
+      "step": 9847
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0624000142527471e-05,
+      "loss": 0.9075,
+      "step": 9848
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0622373008979277e-05,
+      "loss": 0.9107,
+      "step": 9849
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0620745858889116e-05,
+      "loss": 0.9512,
+      "step": 9850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061911869230024e-05,
+      "loss": 0.9515,
+      "step": 9851
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06174915092559e-05,
+      "loss": 0.8847,
+      "step": 9852
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0615864309799337e-05,
+      "loss": 0.8662,
+      "step": 9853
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0614237093973806e-05,
+      "loss": 0.8676,
+      "step": 9854
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0612609861822556e-05,
+      "loss": 0.9073,
+      "step": 9855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0610982613388834e-05,
+      "loss": 1.0082,
+      "step": 9856
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0609355348715891e-05,
+      "loss": 0.969,
+      "step": 9857
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0607728067846978e-05,
+      "loss": 0.9523,
+      "step": 9858
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0606100770825349e-05,
+      "loss": 0.916,
+      "step": 9859
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0604473457694252e-05,
+      "loss": 1.0116,
+      "step": 9860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0602846128496945e-05,
+      "loss": 0.8554,
+      "step": 9861
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0601218783276673e-05,
+      "loss": 0.9118,
+      "step": 9862
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0599591422076692e-05,
+      "loss": 0.8574,
+      "step": 9863
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597964044940257e-05,
+      "loss": 0.9379,
+      "step": 9864
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059633665191062e-05,
+      "loss": 0.9001,
+      "step": 9865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0594709243031035e-05,
+      "loss": 0.9376,
+      "step": 9866
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0593081818344761e-05,
+      "loss": 0.9246,
+      "step": 9867
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0591454377895045e-05,
+      "loss": 0.9107,
+      "step": 9868
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058982692172515e-05,
+      "loss": 0.8608,
+      "step": 9869
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058819944987833e-05,
+      "loss": 0.912,
+      "step": 9870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586571962397839e-05,
+      "loss": 0.8778,
+      "step": 9871
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0584944459326936e-05,
+      "loss": 0.951,
+      "step": 9872
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0583316940708875e-05,
+      "loss": 0.9741,
+      "step": 9873
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0581689406586916e-05,
+      "loss": 0.897,
+      "step": 9874
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058006185700432e-05,
+      "loss": 0.8947,
+      "step": 9875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057843429200434e-05,
+      "loss": 0.9136,
+      "step": 9876
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0576806711630238e-05,
+      "loss": 0.8976,
+      "step": 9877
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0575179115925272e-05,
+      "loss": 0.9077,
+      "step": 9878
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573551504932702e-05,
+      "loss": 0.9062,
+      "step": 9879
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571923878695788e-05,
+      "loss": 0.9423,
+      "step": 9880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0570296237257791e-05,
+      "loss": 0.8362,
+      "step": 9881
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0568668580661972e-05,
+      "loss": 0.8514,
+      "step": 9882
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.056704090895159e-05,
+      "loss": 0.9269,
+      "step": 9883
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.056541322216991e-05,
+      "loss": 0.9548,
+      "step": 9884
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0563785520360193e-05,
+      "loss": 0.9831,
+      "step": 9885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562157803565696e-05,
+      "loss": 0.903,
+      "step": 9886
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0560530071829689e-05,
+      "loss": 0.863,
+      "step": 9887
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0558902325195436e-05,
+      "loss": 0.9759,
+      "step": 9888
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0557274563706195e-05,
+      "loss": 0.9843,
+      "step": 9889
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0555646787405233e-05,
+      "loss": 0.9494,
+      "step": 9890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0554018996335813e-05,
+      "loss": 0.9298,
+      "step": 9891
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05523911905412e-05,
+      "loss": 0.9942,
+      "step": 9892
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0550763370064662e-05,
+      "loss": 1.024,
+      "step": 9893
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0549135534949461e-05,
+      "loss": 0.9787,
+      "step": 9894
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0547507685238868e-05,
+      "loss": 0.8896,
+      "step": 9895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0545879820976143e-05,
+      "loss": 0.8943,
+      "step": 9896
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544251942204557e-05,
+      "loss": 0.9459,
+      "step": 9897
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0542624048967376e-05,
+      "loss": 0.938,
+      "step": 9898
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0540996141307866e-05,
+      "loss": 0.4267,
+      "step": 9899
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05393682192693e-05,
+      "loss": 0.8722,
+      "step": 9900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0537740282894937e-05,
+      "loss": 0.9099,
+      "step": 9901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536112332228057e-05,
+      "loss": 0.8728,
+      "step": 9902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0534484367311923e-05,
+      "loss": 0.8444,
+      "step": 9903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05328563881898e-05,
+      "loss": 0.4949,
+      "step": 9904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0531228394904966e-05,
+      "loss": 0.8931,
+      "step": 9905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0529600387500692e-05,
+      "loss": 0.9038,
+      "step": 9906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527972366020238e-05,
+      "loss": 0.9535,
+      "step": 9907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0526344330506886e-05,
+      "loss": 0.9391,
+      "step": 9908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05247162810039e-05,
+      "loss": 0.8745,
+      "step": 9909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523088217554554e-05,
+      "loss": 0.8708,
+      "step": 9910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0521460140202125e-05,
+      "loss": 0.9701,
+      "step": 9911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519832048989875e-05,
+      "loss": 1.0367,
+      "step": 9912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0518203943961088e-05,
+      "loss": 0.9603,
+      "step": 9913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0516575825159028e-05,
+      "loss": 0.9347,
+      "step": 9914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514947692626976e-05,
+      "loss": 0.969,
+      "step": 9915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05133195464082e-05,
+      "loss": 0.8974,
+      "step": 9916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0511691386545977e-05,
+      "loss": 1.0089,
+      "step": 9917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0510063213083584e-05,
+      "loss": 0.9219,
+      "step": 9918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508435026064291e-05,
+      "loss": 0.9019,
+      "step": 9919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0506806825531377e-05,
+      "loss": 0.9898,
+      "step": 9920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0505178611528115e-05,
+      "loss": 0.9748,
+      "step": 9921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.050355038409778e-05,
+      "loss": 0.8966,
+      "step": 9922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0501922143283657e-05,
+      "loss": 0.8569,
+      "step": 9923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0500293889129014e-05,
+      "loss": 0.9336,
+      "step": 9924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0498665621677129e-05,
+      "loss": 0.9534,
+      "step": 9925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497037340971286e-05,
+      "loss": 0.9266,
+      "step": 9926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0495409047054753e-05,
+      "loss": 0.8677,
+      "step": 9927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0493780739970813e-05,
+      "loss": 0.9094,
+      "step": 9928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0492152419762748e-05,
+      "loss": 0.961,
+      "step": 9929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0490524086473832e-05,
+      "loss": 0.8779,
+      "step": 9930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488895740147349e-05,
+      "loss": 1.0006,
+      "step": 9931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048726738082657e-05,
+      "loss": 0.9331,
+      "step": 9932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485639008554784e-05,
+      "loss": 0.8295,
+      "step": 9933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0484010623375272e-05,
+      "loss": 0.9486,
+      "step": 9934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482382225331305e-05,
+      "loss": 0.8246,
+      "step": 9935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0480753814466171e-05,
+      "loss": 0.9187,
+      "step": 9936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0479125390823151e-05,
+      "loss": 0.9601,
+      "step": 9937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047749695444552e-05,
+      "loss": 0.9244,
+      "step": 9938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047586850537657e-05,
+      "loss": 0.961,
+      "step": 9939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0474240043659578e-05,
+      "loss": 0.8727,
+      "step": 9940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0472611569337827e-05,
+      "loss": 0.4969,
+      "step": 9941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04709830824546e-05,
+      "loss": 1.0033,
+      "step": 9942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046935458305318e-05,
+      "loss": 0.8633,
+      "step": 9943
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467726071176854e-05,
+      "loss": 0.9415,
+      "step": 9944
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04660975468689e-05,
+      "loss": 0.9235,
+      "step": 9945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464469010172605e-05,
+      "loss": 0.8612,
+      "step": 9946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462840461131259e-05,
+      "loss": 0.9912,
+      "step": 9947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0461211899788137e-05,
+      "loss": 0.902,
+      "step": 9948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0459583326186532e-05,
+      "loss": 0.949,
+      "step": 9949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0457954740369727e-05,
+      "loss": 0.9476,
+      "step": 9950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456326142381005e-05,
+      "loss": 0.466,
+      "step": 9951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045469753226366e-05,
+      "loss": 0.9095,
+      "step": 9952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0453068910060971e-05,
+      "loss": 0.9144,
+      "step": 9953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0451440275816229e-05,
+      "loss": 0.8679,
+      "step": 9954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449811629572721e-05,
+      "loss": 0.9637,
+      "step": 9955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0448182971373733e-05,
+      "loss": 0.9478,
+      "step": 9956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044655430126255e-05,
+      "loss": 0.8899,
+      "step": 9957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0444925619282468e-05,
+      "loss": 0.9298,
+      "step": 9958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044329692547677e-05,
+      "loss": 0.8463,
+      "step": 9959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0441668219888745e-05,
+      "loss": 0.9345,
+      "step": 9960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0440039502561683e-05,
+      "loss": 0.8785,
+      "step": 9961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0438410773538875e-05,
+      "loss": 0.9143,
+      "step": 9962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0436782032863608e-05,
+      "loss": 0.8594,
+      "step": 9963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0435153280579174e-05,
+      "loss": 0.9783,
+      "step": 9964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433524516728863e-05,
+      "loss": 0.9211,
+      "step": 9965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0431895741355967e-05,
+      "loss": 0.9821,
+      "step": 9966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0430266954503774e-05,
+      "loss": 0.9852,
+      "step": 9967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0428638156215579e-05,
+      "loss": 0.9172,
+      "step": 9968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427009346534665e-05,
+      "loss": 0.915,
+      "step": 9969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0425380525504337e-05,
+      "loss": 0.9142,
+      "step": 9970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0423751693167877e-05,
+      "loss": 0.9058,
+      "step": 9971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0422122849568581e-05,
+      "loss": 0.9859,
+      "step": 9972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0420493994749746e-05,
+      "loss": 0.9622,
+      "step": 9973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0418865128754653e-05,
+      "loss": 0.8952,
+      "step": 9974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0417236251626608e-05,
+      "loss": 0.8844,
+      "step": 9975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0415607363408901e-05,
+      "loss": 0.8386,
+      "step": 9976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413978464144823e-05,
+      "loss": 0.9669,
+      "step": 9977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0412349553877673e-05,
+      "loss": 0.9625,
+      "step": 9978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041072063265074e-05,
+      "loss": 0.968,
+      "step": 9979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0409091700507322e-05,
+      "loss": 0.8638,
+      "step": 9980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0407462757490713e-05,
+      "loss": 0.8284,
+      "step": 9981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040583380364421e-05,
+      "loss": 0.931,
+      "step": 9982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040420483901111e-05,
+      "loss": 0.9323,
+      "step": 9983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402575863634702e-05,
+      "loss": 0.9107,
+      "step": 9984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0400946877558293e-05,
+      "loss": 1.0124,
+      "step": 9985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0399317880825171e-05,
+      "loss": 0.9804,
+      "step": 9986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0397688873478635e-05,
+      "loss": 0.8828,
+      "step": 9987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0396059855561986e-05,
+      "loss": 0.8739,
+      "step": 9988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0394430827118519e-05,
+      "loss": 0.8711,
+      "step": 9989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0392801788191526e-05,
+      "loss": 0.964,
+      "step": 9990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391172738824315e-05,
+      "loss": 0.9219,
+      "step": 9991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0389543679060175e-05,
+      "loss": 0.9162,
+      "step": 9992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0387914608942414e-05,
+      "loss": 0.8838,
+      "step": 9993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0386285528514323e-05,
+      "loss": 0.8942,
+      "step": 9994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0384656437819203e-05,
+      "loss": 0.9436,
+      "step": 9995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0383027336900356e-05,
+      "loss": 0.9619,
+      "step": 9996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381398225801079e-05,
+      "loss": 0.8635,
+      "step": 9997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037976910456467e-05,
+      "loss": 0.9222,
+      "step": 9998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0378139973234436e-05,
+      "loss": 0.9225,
+      "step": 9999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037651083185367e-05,
+      "loss": 0.96,
+      "step": 10000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0374881680465681e-05,
+      "loss": 0.9742,
+      "step": 10001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037325251911376e-05,
+      "loss": 0.8977,
+      "step": 10002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371623347841217e-05,
+      "loss": 0.9136,
+      "step": 10003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0369994166691348e-05,
+      "loss": 0.984,
+      "step": 10004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0368364975707457e-05,
+      "loss": 0.8805,
+      "step": 10005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0366735774932846e-05,
+      "loss": 0.9133,
+      "step": 10006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.036510656441082e-05,
+      "loss": 0.8881,
+      "step": 10007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363477344184671e-05,
+      "loss": 0.8929,
+      "step": 10008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0361848114297714e-05,
+      "loss": 0.9182,
+      "step": 10009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0360218874793245e-05,
+      "loss": 0.9554,
+      "step": 10010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0358589625714572e-05,
+      "loss": 0.8973,
+      "step": 10011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0356960367104997e-05,
+      "loss": 1.0035,
+      "step": 10012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355331099007822e-05,
+      "loss": 0.9793,
+      "step": 10013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0353701821466354e-05,
+      "loss": 0.896,
+      "step": 10014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035207253452389e-05,
+      "loss": 0.9441,
+      "step": 10015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0350443238223743e-05,
+      "loss": 0.9601,
+      "step": 10016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348813932609218e-05,
+      "loss": 0.8903,
+      "step": 10017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0347184617723613e-05,
+      "loss": 0.8954,
+      "step": 10018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.034555529361024e-05,
+      "loss": 0.8902,
+      "step": 10019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0343925960312402e-05,
+      "loss": 0.8615,
+      "step": 10020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0342296617873406e-05,
+      "loss": 0.9545,
+      "step": 10021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0340667266336553e-05,
+      "loss": 0.9998,
+      "step": 10022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339037905745155e-05,
+      "loss": 0.9691,
+      "step": 10023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0337408536142517e-05,
+      "loss": 0.8928,
+      "step": 10024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335779157571946e-05,
+      "loss": 0.4485,
+      "step": 10025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0334149770076747e-05,
+      "loss": 0.8964,
+      "step": 10026
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033252037370023e-05,
+      "loss": 1.005,
+      "step": 10027
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03308909684857e-05,
+      "loss": 0.9454,
+      "step": 10028
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0329261554476466e-05,
+      "loss": 0.8717,
+      "step": 10029
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0327632131715837e-05,
+      "loss": 0.9532,
+      "step": 10030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.032600270024712e-05,
+      "loss": 0.8767,
+      "step": 10031
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0324373260113625e-05,
+      "loss": 0.9916,
+      "step": 10032
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0322743811358655e-05,
+      "loss": 0.8811,
+      "step": 10033
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0321114354025525e-05,
+      "loss": 0.9667,
+      "step": 10034
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0319484888157545e-05,
+      "loss": 0.9556,
+      "step": 10035
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0317855413798018e-05,
+      "loss": 0.8705,
+      "step": 10036
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0316225930990258e-05,
+      "loss": 0.9257,
+      "step": 10037
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0314596439777575e-05,
+      "loss": 0.9118,
+      "step": 10038
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312966940203278e-05,
+      "loss": 0.9925,
+      "step": 10039
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0311337432310675e-05,
+      "loss": 0.9378,
+      "step": 10040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0309707916143078e-05,
+      "loss": 0.8893,
+      "step": 10041
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0308078391743804e-05,
+      "loss": 0.8665,
+      "step": 10042
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306448859156155e-05,
+      "loss": 0.88,
+      "step": 10043
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304819318423444e-05,
+      "loss": 0.8712,
+      "step": 10044
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0303189769588986e-05,
+      "loss": 0.9495,
+      "step": 10045
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0301560212696087e-05,
+      "loss": 0.8701,
+      "step": 10046
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0299930647788064e-05,
+      "loss": 0.9087,
+      "step": 10047
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0298301074908228e-05,
+      "loss": 0.935,
+      "step": 10048
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296671494099887e-05,
+      "loss": 0.9168,
+      "step": 10049
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.029504190540636e-05,
+      "loss": 1.0116,
+      "step": 10050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0293412308870954e-05,
+      "loss": 0.9497,
+      "step": 10051
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291782704536982e-05,
+      "loss": 0.9088,
+      "step": 10052
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0290153092447763e-05,
+      "loss": 0.8494,
+      "step": 10053
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0288523472646603e-05,
+      "loss": 0.9892,
+      "step": 10054
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0286893845176823e-05,
+      "loss": 1.0112,
+      "step": 10055
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028526421008173e-05,
+      "loss": 0.9936,
+      "step": 10056
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0283634567404634e-05,
+      "loss": 0.8364,
+      "step": 10057
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0282004917188863e-05,
+      "loss": 0.9351,
+      "step": 10058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0280375259477719e-05,
+      "loss": 0.9797,
+      "step": 10059
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278745594314525e-05,
+      "loss": 0.9787,
+      "step": 10060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0277115921742591e-05,
+      "loss": 0.8888,
+      "step": 10061
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027548624180523e-05,
+      "loss": 0.865,
+      "step": 10062
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0273856554545758e-05,
+      "loss": 0.4038,
+      "step": 10063
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0272226860007496e-05,
+      "loss": 0.9052,
+      "step": 10064
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027059715823375e-05,
+      "loss": 0.9258,
+      "step": 10065
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0268967449267846e-05,
+      "loss": 0.8886,
+      "step": 10066
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0267337733153088e-05,
+      "loss": 0.4204,
+      "step": 10067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265708009932804e-05,
+      "loss": 1.0128,
+      "step": 10068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0264078279650303e-05,
+      "loss": 0.9273,
+      "step": 10069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0262448542348899e-05,
+      "loss": 0.852,
+      "step": 10070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0260818798071918e-05,
+      "loss": 0.9126,
+      "step": 10071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259189046862667e-05,
+      "loss": 0.9015,
+      "step": 10072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0257559288764471e-05,
+      "loss": 0.9608,
+      "step": 10073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0255929523820645e-05,
+      "loss": 0.9511,
+      "step": 10074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0254299752074498e-05,
+      "loss": 0.9156,
+      "step": 10075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0252669973569356e-05,
+      "loss": 0.9204,
+      "step": 10076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251040188348538e-05,
+      "loss": 0.9051,
+      "step": 10077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0249410396455354e-05,
+      "loss": 0.9261,
+      "step": 10078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024778059793313e-05,
+      "loss": 0.8686,
+      "step": 10079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024615079282518e-05,
+      "loss": 0.8889,
+      "step": 10080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0244520981174822e-05,
+      "loss": 0.9215,
+      "step": 10081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0242891163025377e-05,
+      "loss": 0.4494,
+      "step": 10082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0241261338420159e-05,
+      "loss": 0.9212,
+      "step": 10083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0239631507402494e-05,
+      "loss": 0.965,
+      "step": 10084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0238001670015696e-05,
+      "loss": 0.9547,
+      "step": 10085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0236371826303086e-05,
+      "loss": 0.9431,
+      "step": 10086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.023474197630798e-05,
+      "loss": 0.9984,
+      "step": 10087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0233112120073703e-05,
+      "loss": 0.4486,
+      "step": 10088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0231482257643573e-05,
+      "loss": 0.9129,
+      "step": 10089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0229852389060905e-05,
+      "loss": 0.8677,
+      "step": 10090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0228222514369026e-05,
+      "loss": 0.9532,
+      "step": 10091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0226592633611254e-05,
+      "loss": 0.9761,
+      "step": 10092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224962746830905e-05,
+      "loss": 0.952,
+      "step": 10093
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0223332854071307e-05,
+      "loss": 0.8604,
+      "step": 10094
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0221702955375776e-05,
+      "loss": 0.9346,
+      "step": 10095
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0220073050787631e-05,
+      "loss": 0.9732,
+      "step": 10096
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0218443140350201e-05,
+      "loss": 0.9453,
+      "step": 10097
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0216813224106797e-05,
+      "loss": 0.8882,
+      "step": 10098
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215183302100743e-05,
+      "loss": 0.9172,
+      "step": 10099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0213553374375369e-05,
+      "loss": 1.0018,
+      "step": 10100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211923440973986e-05,
+      "loss": 0.9423,
+      "step": 10101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0210293501939922e-05,
+      "loss": 0.9546,
+      "step": 10102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0208663557316496e-05,
+      "loss": 0.9497,
+      "step": 10103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0207033607147033e-05,
+      "loss": 0.9666,
+      "step": 10104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020540365147485e-05,
+      "loss": 0.9479,
+      "step": 10105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0203773690343273e-05,
+      "loss": 0.9637,
+      "step": 10106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0202143723795626e-05,
+      "loss": 0.9189,
+      "step": 10107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0200513751875227e-05,
+      "loss": 0.9915,
+      "step": 10108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198883774625405e-05,
+      "loss": 0.8824,
+      "step": 10109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0197253792089477e-05,
+      "loss": 0.974,
+      "step": 10110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0195623804310766e-05,
+      "loss": 0.9537,
+      "step": 10111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0193993811332603e-05,
+      "loss": 0.9297,
+      "step": 10112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0192363813198306e-05,
+      "loss": 0.9321,
+      "step": 10113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190733809951194e-05,
+      "loss": 1.086,
+      "step": 10114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0189103801634599e-05,
+      "loss": 0.8735,
+      "step": 10115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0187473788291837e-05,
+      "loss": 0.8367,
+      "step": 10116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018584376996624e-05,
+      "loss": 0.965,
+      "step": 10117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0184213746701126e-05,
+      "loss": 0.9075,
+      "step": 10118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0182583718539821e-05,
+      "loss": 0.8916,
+      "step": 10119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018095368552565e-05,
+      "loss": 0.9445,
+      "step": 10120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0179323647701935e-05,
+      "loss": 0.8733,
+      "step": 10121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0177693605112003e-05,
+      "loss": 0.918,
+      "step": 10122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0176063557799178e-05,
+      "loss": 0.9,
+      "step": 10123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0174433505806784e-05,
+      "loss": 0.8965,
+      "step": 10124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0172803449178148e-05,
+      "loss": 0.8654,
+      "step": 10125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0171173387956594e-05,
+      "loss": 0.9124,
+      "step": 10126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0169543322185447e-05,
+      "loss": 0.9325,
+      "step": 10127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016791325190803e-05,
+      "loss": 0.9435,
+      "step": 10128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016628317716767e-05,
+      "loss": 0.9495,
+      "step": 10129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0164653098007698e-05,
+      "loss": 0.9838,
+      "step": 10130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163023014471431e-05,
+      "loss": 0.8183,
+      "step": 10131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0161392926602198e-05,
+      "loss": 0.9782,
+      "step": 10132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0159762834443327e-05,
+      "loss": 0.9367,
+      "step": 10133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158132738038143e-05,
+      "loss": 0.901,
+      "step": 10134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156502637429968e-05,
+      "loss": 0.9498,
+      "step": 10135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0154872532662134e-05,
+      "loss": 0.9111,
+      "step": 10136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0153242423777964e-05,
+      "loss": 0.9349,
+      "step": 10137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0151612310820789e-05,
+      "loss": 0.929,
+      "step": 10138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0149982193833927e-05,
+      "loss": 1.0121,
+      "step": 10139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014835207286071e-05,
+      "loss": 0.9,
+      "step": 10140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0146721947944466e-05,
+      "loss": 0.4424,
+      "step": 10141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014509181912852e-05,
+      "loss": 0.9206,
+      "step": 10142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0143461686456202e-05,
+      "loss": 0.909,
+      "step": 10143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0141831549970834e-05,
+      "loss": 0.9327,
+      "step": 10144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0140201409715745e-05,
+      "loss": 0.9343,
+      "step": 10145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013857126573426e-05,
+      "loss": 0.8907,
+      "step": 10146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0136941118069715e-05,
+      "loss": 0.9891,
+      "step": 10147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0135310966765425e-05,
+      "loss": 0.8994,
+      "step": 10148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133680811864728e-05,
+      "loss": 0.9346,
+      "step": 10149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0132050653410947e-05,
+      "loss": 0.9157,
+      "step": 10150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013042049144741e-05,
+      "loss": 0.942,
+      "step": 10151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0128790326017447e-05,
+      "loss": 0.9169,
+      "step": 10152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0127160157164379e-05,
+      "loss": 0.9236,
+      "step": 10153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0125529984931545e-05,
+      "loss": 0.8805,
+      "step": 10154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0123899809362265e-05,
+      "loss": 0.9628,
+      "step": 10155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122269630499872e-05,
+      "loss": 0.8571,
+      "step": 10156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0120639448387688e-05,
+      "loss": 0.9703,
+      "step": 10157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0119009263069047e-05,
+      "loss": 0.8855,
+      "step": 10158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0117379074587277e-05,
+      "loss": 0.9275,
+      "step": 10159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0115748882985703e-05,
+      "loss": 1.0537,
+      "step": 10160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0114118688307656e-05,
+      "loss": 0.9331,
+      "step": 10161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0112488490596468e-05,
+      "loss": 0.9196,
+      "step": 10162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110858289895463e-05,
+      "loss": 0.9303,
+      "step": 10163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0109228086247969e-05,
+      "loss": 0.9406,
+      "step": 10164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107597879697319e-05,
+      "loss": 0.8672,
+      "step": 10165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0105967670286839e-05,
+      "loss": 0.9398,
+      "step": 10166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010433745805986e-05,
+      "loss": 1.0004,
+      "step": 10167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010270724305971e-05,
+      "loss": 0.9539,
+      "step": 10168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010107702532972e-05,
+      "loss": 0.9551,
+      "step": 10169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0099446804913217e-05,
+      "loss": 0.9228,
+      "step": 10170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0097816581853532e-05,
+      "loss": 0.4131,
+      "step": 10171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0096186356193995e-05,
+      "loss": 0.9615,
+      "step": 10172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094556127977933e-05,
+      "loss": 0.9974,
+      "step": 10173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092925897248678e-05,
+      "loss": 0.9654,
+      "step": 10174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0091295664049557e-05,
+      "loss": 0.9166,
+      "step": 10175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0089665428423903e-05,
+      "loss": 0.9518,
+      "step": 10176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0088035190415044e-05,
+      "loss": 0.9386,
+      "step": 10177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.008640495006631e-05,
+      "loss": 0.9528,
+      "step": 10178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0084774707421032e-05,
+      "loss": 0.9568,
+      "step": 10179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083144462522541e-05,
+      "loss": 0.916,
+      "step": 10180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081514215414158e-05,
+      "loss": 0.95,
+      "step": 10181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0079883966139227e-05,
+      "loss": 0.8356,
+      "step": 10182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0078253714741071e-05,
+      "loss": 0.9101,
+      "step": 10183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0076623461263017e-05,
+      "loss": 0.9045,
+      "step": 10184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0074993205748403e-05,
+      "loss": 0.934,
+      "step": 10185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073362948240552e-05,
+      "loss": 0.851,
+      "step": 10186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0071732688782798e-05,
+      "loss": 0.9205,
+      "step": 10187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0070102427418474e-05,
+      "loss": 0.8767,
+      "step": 10188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0068472164190903e-05,
+      "loss": 0.9206,
+      "step": 10189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0066841899143424e-05,
+      "loss": 0.8714,
+      "step": 10190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006521163231936e-05,
+      "loss": 0.9211,
+      "step": 10191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006358136376205e-05,
+      "loss": 0.8482,
+      "step": 10192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0061951093514816e-05,
+      "loss": 0.9266,
+      "step": 10193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0060320821620992e-05,
+      "loss": 0.865,
+      "step": 10194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0058690548123916e-05,
+      "loss": 0.9655,
+      "step": 10195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0057060273066908e-05,
+      "loss": 0.9092,
+      "step": 10196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00554299964933e-05,
+      "loss": 0.8874,
+      "step": 10197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0053799718446431e-05,
+      "loss": 0.8722,
+      "step": 10198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0052169438969621e-05,
+      "loss": 0.8933,
+      "step": 10199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0050539158106213e-05,
+      "loss": 0.9001,
+      "step": 10200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004890887589953e-05,
+      "loss": 0.432,
+      "step": 10201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0047278592392904e-05,
+      "loss": 0.9404,
+      "step": 10202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004564830762967e-05,
+      "loss": 0.8641,
+      "step": 10203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0044018021653152e-05,
+      "loss": 0.9532,
+      "step": 10204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0042387734506687e-05,
+      "loss": 0.8702,
+      "step": 10205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0040757446233604e-05,
+      "loss": 0.9523,
+      "step": 10206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0039127156877235e-05,
+      "loss": 0.8357,
+      "step": 10207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0037496866480911e-05,
+      "loss": 0.8908,
+      "step": 10208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035866575087963e-05,
+      "loss": 0.9529,
+      "step": 10209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0034236282741724e-05,
+      "loss": 0.9017,
+      "step": 10210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003260598948552e-05,
+      "loss": 0.899,
+      "step": 10211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0030975695362688e-05,
+      "loss": 1.0218,
+      "step": 10212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0029345400416559e-05,
+      "loss": 0.938,
+      "step": 10213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002771510469046e-05,
+      "loss": 0.9821,
+      "step": 10214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0026084808227727e-05,
+      "loss": 0.982,
+      "step": 10215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002445451107169e-05,
+      "loss": 0.3736,
+      "step": 10216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022824213265677e-05,
+      "loss": 0.8588,
+      "step": 10217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0021193914853024e-05,
+      "loss": 0.9014,
+      "step": 10218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0019563615877063e-05,
+      "loss": 0.882,
+      "step": 10219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001793331638112e-05,
+      "loss": 0.8963,
+      "step": 10220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0016303016408535e-05,
+      "loss": 0.9327,
+      "step": 10221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0014672716002631e-05,
+      "loss": 1.0077,
+      "step": 10222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0013042415206742e-05,
+      "loss": 0.9126,
+      "step": 10223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011412114064203e-05,
+      "loss": 0.9536,
+      "step": 10224
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009781812618339e-05,
+      "loss": 0.8646,
+      "step": 10225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0008151510912492e-05,
+      "loss": 0.8057,
+      "step": 10226
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0006521208989981e-05,
+      "loss": 0.9695,
+      "step": 10227
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0004890906894149e-05,
+      "loss": 0.9225,
+      "step": 10228
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003260604668319e-05,
+      "loss": 0.9447,
+      "step": 10229
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0001630302355825e-05,
+      "loss": 0.8453,
+      "step": 10230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1e-05,
+      "loss": 0.9707,
+      "step": 10231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998369697644178e-06,
+      "loss": 0.9033,
+      "step": 10232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.996739395331685e-06,
+      "loss": 0.8865,
+      "step": 10233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.995109093105856e-06,
+      "loss": 0.8633,
+      "step": 10234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99347879101002e-06,
+      "loss": 0.8873,
+      "step": 10235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.991848489087513e-06,
+      "loss": 1.0023,
+      "step": 10236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990218187381661e-06,
+      "loss": 0.9496,
+      "step": 10237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988587885935797e-06,
+      "loss": 0.9317,
+      "step": 10238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.986957584793258e-06,
+      "loss": 0.9791,
+      "step": 10239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985327283997372e-06,
+      "loss": 0.8921,
+      "step": 10240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.983696983591468e-06,
+      "loss": 0.4249,
+      "step": 10241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982066683618881e-06,
+      "loss": 0.9237,
+      "step": 10242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.980436384122938e-06,
+      "loss": 0.8459,
+      "step": 10243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.978806085146979e-06,
+      "loss": 0.9512,
+      "step": 10244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977175786734327e-06,
+      "loss": 0.9967,
+      "step": 10245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975545488928315e-06,
+      "loss": 0.9281,
+      "step": 10246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.973915191772278e-06,
+      "loss": 0.92,
+      "step": 10247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972284895309543e-06,
+      "loss": 0.8667,
+      "step": 10248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970654599583442e-06,
+      "loss": 0.9292,
+      "step": 10249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969024304637313e-06,
+      "loss": 0.8995,
+      "step": 10250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96739401051448e-06,
+      "loss": 0.9417,
+      "step": 10251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96576371725828e-06,
+      "loss": 0.9747,
+      "step": 10252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96413342491204e-06,
+      "loss": 0.8603,
+      "step": 10253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96250313351909e-06,
+      "loss": 0.9598,
+      "step": 10254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.960872843122769e-06,
+      "loss": 0.9683,
+      "step": 10255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959242553766398e-06,
+      "loss": 0.9517,
+      "step": 10256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957612265493318e-06,
+      "loss": 1.035,
+      "step": 10257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.955981978346851e-06,
+      "loss": 0.8541,
+      "step": 10258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.954351692370337e-06,
+      "loss": 0.8784,
+      "step": 10259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9527214076071e-06,
+      "loss": 0.9706,
+      "step": 10260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.951091124100473e-06,
+      "loss": 0.9287,
+      "step": 10261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.949460841893787e-06,
+      "loss": 0.9219,
+      "step": 10262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.947830561030379e-06,
+      "loss": 0.9202,
+      "step": 10263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946200281553572e-06,
+      "loss": 0.9457,
+      "step": 10264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944570003506702e-06,
+      "loss": 0.9977,
+      "step": 10265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942939726933097e-06,
+      "loss": 0.8261,
+      "step": 10266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94130945187609e-06,
+      "loss": 0.8247,
+      "step": 10267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93967917837901e-06,
+      "loss": 0.9074,
+      "step": 10268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938048906485187e-06,
+      "loss": 0.9467,
+      "step": 10269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936418636237956e-06,
+      "loss": 0.9104,
+      "step": 10270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.934788367680642e-06,
+      "loss": 1.0539,
+      "step": 10271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93315810085658e-06,
+      "loss": 0.8951,
+      "step": 10272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.931527835809097e-06,
+      "loss": 1.0187,
+      "step": 10273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.929897572581531e-06,
+      "loss": 0.9039,
+      "step": 10274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.928267311217204e-06,
+      "loss": 0.9019,
+      "step": 10275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926637051759451e-06,
+      "loss": 0.9125,
+      "step": 10276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9250067942516e-06,
+      "loss": 0.9334,
+      "step": 10277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.923376538736985e-06,
+      "loss": 0.8662,
+      "step": 10278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921746285258932e-06,
+      "loss": 0.9315,
+      "step": 10279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920116033860776e-06,
+      "loss": 1.0064,
+      "step": 10280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918485784585844e-06,
+      "loss": 0.9919,
+      "step": 10281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916855537477464e-06,
+      "loss": 0.8735,
+      "step": 10282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.915225292578973e-06,
+      "loss": 0.8741,
+      "step": 10283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913595049933692e-06,
+      "loss": 0.965,
+      "step": 10284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.911964809584956e-06,
+      "loss": 0.9488,
+      "step": 10285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910334571576098e-06,
+      "loss": 0.9676,
+      "step": 10286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.908704335950444e-06,
+      "loss": 0.8876,
+      "step": 10287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907074102751325e-06,
+      "loss": 0.9216,
+      "step": 10288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.905443872022072e-06,
+      "loss": 0.8533,
+      "step": 10289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903813643806008e-06,
+      "loss": 1.0301,
+      "step": 10290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902183418146471e-06,
+      "loss": 0.9965,
+      "step": 10291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.900553195086786e-06,
+      "loss": 1.0479,
+      "step": 10292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.898922974670283e-06,
+      "loss": 0.9077,
+      "step": 10293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.897292756940296e-06,
+      "loss": 0.8642,
+      "step": 10294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.895662541940143e-06,
+      "loss": 0.9889,
+      "step": 10295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894032329713163e-06,
+      "loss": 0.8465,
+      "step": 10296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892402120302684e-06,
+      "loss": 0.4535,
+      "step": 10297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890771913752033e-06,
+      "loss": 0.9173,
+      "step": 10298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889141710104542e-06,
+      "loss": 1.0262,
+      "step": 10299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887511509403534e-06,
+      "loss": 0.8562,
+      "step": 10300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.885881311692345e-06,
+      "loss": 0.4306,
+      "step": 10301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8842511170143e-06,
+      "loss": 0.8883,
+      "step": 10302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882620925412727e-06,
+      "loss": 0.9001,
+      "step": 10303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880990736930956e-06,
+      "loss": 0.9176,
+      "step": 10304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879360551612315e-06,
+      "loss": 0.9017,
+      "step": 10305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877730369500133e-06,
+      "loss": 0.9571,
+      "step": 10306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876100190637738e-06,
+      "loss": 0.9678,
+      "step": 10307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.874470015068455e-06,
+      "loss": 0.9143,
+      "step": 10308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87283984283562e-06,
+      "loss": 0.9137,
+      "step": 10309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.871209673982557e-06,
+      "loss": 0.9467,
+      "step": 10310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869579508552592e-06,
+      "loss": 0.984,
+      "step": 10311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.867949346589056e-06,
+      "loss": 0.9904,
+      "step": 10312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866319188135273e-06,
+      "loss": 0.9367,
+      "step": 10313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.864689033234577e-06,
+      "loss": 0.9022,
+      "step": 10314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.86305888193029e-06,
+      "loss": 0.9644,
+      "step": 10315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.861428734265742e-06,
+      "loss": 0.9777,
+      "step": 10316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85979859028426e-06,
+      "loss": 0.933,
+      "step": 10317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.858168450029171e-06,
+      "loss": 0.9097,
+      "step": 10318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.856538313543804e-06,
+      "loss": 0.8821,
+      "step": 10319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85490818087148e-06,
+      "loss": 0.9463,
+      "step": 10320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.853278052055532e-06,
+      "loss": 0.8595,
+      "step": 10321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85164792713929e-06,
+      "loss": 0.9402,
+      "step": 10322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.850017806166076e-06,
+      "loss": 0.8769,
+      "step": 10323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848387689179215e-06,
+      "loss": 0.9389,
+      "step": 10324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.846757576222038e-06,
+      "loss": 0.915,
+      "step": 10325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845127467337868e-06,
+      "loss": 0.9257,
+      "step": 10326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.843497362570034e-06,
+      "loss": 0.9431,
+      "step": 10327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841867261961862e-06,
+      "loss": 0.8621,
+      "step": 10328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.840237165556678e-06,
+      "loss": 0.8889,
+      "step": 10329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838607073397807e-06,
+      "loss": 0.9502,
+      "step": 10330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.836976985528574e-06,
+      "loss": 0.922,
+      "step": 10331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.835346901992304e-06,
+      "loss": 0.9051,
+      "step": 10332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.833716822832329e-06,
+      "loss": 0.8567,
+      "step": 10333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.832086748091971e-06,
+      "loss": 0.8982,
+      "step": 10334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830456677814556e-06,
+      "loss": 0.9707,
+      "step": 10335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.828826612043408e-06,
+      "loss": 0.9514,
+      "step": 10336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.827196550821853e-06,
+      "loss": 0.8905,
+      "step": 10337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82556649419322e-06,
+      "loss": 0.8981,
+      "step": 10338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823936442200826e-06,
+      "loss": 0.9288,
+      "step": 10339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.822306394888002e-06,
+      "loss": 0.8764,
+      "step": 10340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820676352298069e-06,
+      "loss": 0.9302,
+      "step": 10341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.819046314474356e-06,
+      "loss": 0.9067,
+      "step": 10342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.817416281460186e-06,
+      "loss": 0.9492,
+      "step": 10343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.815786253298877e-06,
+      "loss": 0.9072,
+      "step": 10344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814156230033762e-06,
+      "loss": 0.8851,
+      "step": 10345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812526211708164e-06,
+      "loss": 0.8893,
+      "step": 10346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810896198365404e-06,
+      "loss": 0.9626,
+      "step": 10347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80926619004881e-06,
+      "loss": 0.919,
+      "step": 10348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807636186801698e-06,
+      "loss": 0.9192,
+      "step": 10349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.806006188667399e-06,
+      "loss": 0.9095,
+      "step": 10350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804376195689236e-06,
+      "loss": 0.9133,
+      "step": 10351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802746207910528e-06,
+      "loss": 0.8167,
+      "step": 10352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8011162253746e-06,
+      "loss": 0.896,
+      "step": 10353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.799486248124775e-06,
+      "loss": 0.9862,
+      "step": 10354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.797856276204376e-06,
+      "loss": 0.8505,
+      "step": 10355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.796226309656728e-06,
+      "loss": 0.9213,
+      "step": 10356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.794596348525152e-06,
+      "loss": 0.9048,
+      "step": 10357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79296639285297e-06,
+      "loss": 0.9265,
+      "step": 10358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791336442683507e-06,
+      "loss": 0.8722,
+      "step": 10359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78970649806008e-06,
+      "loss": 0.93,
+      "step": 10360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788076559026015e-06,
+      "loss": 0.9279,
+      "step": 10361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.786446625624634e-06,
+      "loss": 0.9293,
+      "step": 10362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.784816697899258e-06,
+      "loss": 0.8966,
+      "step": 10363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.783186775893206e-06,
+      "loss": 0.8926,
+      "step": 10364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.781556859649806e-06,
+      "loss": 1.0094,
+      "step": 10365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.779926949212372e-06,
+      "loss": 0.8817,
+      "step": 10366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.778297044624226e-06,
+      "loss": 0.9722,
+      "step": 10367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.776667145928695e-06,
+      "loss": 0.9557,
+      "step": 10368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775037253169095e-06,
+      "loss": 0.9242,
+      "step": 10369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.773407366388748e-06,
+      "loss": 0.8747,
+      "step": 10370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771777485630976e-06,
+      "loss": 0.9698,
+      "step": 10371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.770147610939098e-06,
+      "loss": 0.8756,
+      "step": 10372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768517742356432e-06,
+      "loss": 0.9397,
+      "step": 10373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.766887879926302e-06,
+      "loss": 0.895,
+      "step": 10374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.765258023692023e-06,
+      "loss": 0.891,
+      "step": 10375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.763628173696919e-06,
+      "loss": 0.8643,
+      "step": 10376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.761998329984307e-06,
+      "loss": 0.9732,
+      "step": 10377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76036849259751e-06,
+      "loss": 0.8465,
+      "step": 10378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.758738661579841e-06,
+      "loss": 0.9084,
+      "step": 10379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757108836974626e-06,
+      "loss": 0.989,
+      "step": 10380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75547901882518e-06,
+      "loss": 0.9013,
+      "step": 10381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753849207174823e-06,
+      "loss": 0.8825,
+      "step": 10382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.752219402066872e-06,
+      "loss": 0.9167,
+      "step": 10383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.750589603544647e-06,
+      "loss": 0.9138,
+      "step": 10384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748959811651465e-06,
+      "loss": 0.8953,
+      "step": 10385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.747330026430646e-06,
+      "loss": 0.925,
+      "step": 10386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.745700247925507e-06,
+      "loss": 0.8702,
+      "step": 10387
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.744070476179361e-06,
+      "loss": 0.93,
+      "step": 10388
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742440711235534e-06,
+      "loss": 0.8375,
+      "step": 10389
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740810953137335e-06,
+      "loss": 1.0118,
+      "step": 10390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.739181201928082e-06,
+      "loss": 0.8813,
+      "step": 10391
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737551457651099e-06,
+      "loss": 0.9607,
+      "step": 10392
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7359217203497e-06,
+      "loss": 0.8149,
+      "step": 10393
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734291990067198e-06,
+      "loss": 0.8691,
+      "step": 10394
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.732662266846914e-06,
+      "loss": 0.9295,
+      "step": 10395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.731032550732159e-06,
+      "loss": 0.8917,
+      "step": 10396
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729402841766253e-06,
+      "loss": 0.9394,
+      "step": 10397
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.727773139992508e-06,
+      "loss": 0.8617,
+      "step": 10398
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.726143445454245e-06,
+      "loss": 0.9215,
+      "step": 10399
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.724513758194775e-06,
+      "loss": 0.9579,
+      "step": 10400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.722884078257414e-06,
+      "loss": 0.9526,
+      "step": 10401
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72125440568548e-06,
+      "loss": 0.8435,
+      "step": 10402
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71962474052228e-06,
+      "loss": 0.8604,
+      "step": 10403
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.717995082811137e-06,
+      "loss": 0.9519,
+      "step": 10404
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716365432595366e-06,
+      "loss": 0.9496,
+      "step": 10405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.714735789918274e-06,
+      "loss": 0.9105,
+      "step": 10406
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71310615482318e-06,
+      "loss": 0.8932,
+      "step": 10407
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7114765273534e-06,
+      "loss": 0.9134,
+      "step": 10408
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70984690755224e-06,
+      "loss": 0.9348,
+      "step": 10409
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70821729546302e-06,
+      "loss": 1.0303,
+      "step": 10410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70658769112905e-06,
+      "loss": 1.004,
+      "step": 10411
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.704958094593646e-06,
+      "loss": 0.9996,
+      "step": 10412
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703328505900115e-06,
+      "loss": 0.8779,
+      "step": 10413
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.701698925091779e-06,
+      "loss": 0.8623,
+      "step": 10414
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700069352211936e-06,
+      "loss": 0.8918,
+      "step": 10415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.698439787303916e-06,
+      "loss": 0.9096,
+      "step": 10416
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.696810230411018e-06,
+      "loss": 1.0143,
+      "step": 10417
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.695180681576559e-06,
+      "loss": 0.9153,
+      "step": 10418
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693551140843848e-06,
+      "loss": 1.0275,
+      "step": 10419
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.6919216082562e-06,
+      "loss": 0.9209,
+      "step": 10420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690292083856923e-06,
+      "loss": 0.4207,
+      "step": 10421
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688662567689327e-06,
+      "loss": 0.9856,
+      "step": 10422
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.687033059796727e-06,
+      "loss": 0.9503,
+      "step": 10423
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.685403560222428e-06,
+      "loss": 0.8992,
+      "step": 10424
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.683774069009745e-06,
+      "loss": 0.9499,
+      "step": 10425
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.682144586201984e-06,
+      "loss": 0.8761,
+      "step": 10426
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.680515111842457e-06,
+      "loss": 0.9054,
+      "step": 10427
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.678885645974477e-06,
+      "loss": 1.0142,
+      "step": 10428
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.677256188641347e-06,
+      "loss": 0.8995,
+      "step": 10429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.675626739886378e-06,
+      "loss": 0.9413,
+      "step": 10430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673997299752882e-06,
+      "loss": 1.0349,
+      "step": 10431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.672367868284164e-06,
+      "loss": 0.9815,
+      "step": 10432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.670738445523536e-06,
+      "loss": 0.9082,
+      "step": 10433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669109031514303e-06,
+      "loss": 0.4688,
+      "step": 10434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667479626299774e-06,
+      "loss": 0.8644,
+      "step": 10435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665850229923258e-06,
+      "loss": 0.9355,
+      "step": 10436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664220842428059e-06,
+      "loss": 0.8482,
+      "step": 10437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662591463857483e-06,
+      "loss": 0.936,
+      "step": 10438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.660962094254847e-06,
+      "loss": 0.9291,
+      "step": 10439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.659332733663447e-06,
+      "loss": 0.9614,
+      "step": 10440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.657703382126597e-06,
+      "loss": 0.8983,
+      "step": 10441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656074039687601e-06,
+      "loss": 0.8889,
+      "step": 10442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.654444706389761e-06,
+      "loss": 0.904,
+      "step": 10443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652815382276389e-06,
+      "loss": 0.8642,
+      "step": 10444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.651186067390785e-06,
+      "loss": 0.8586,
+      "step": 10445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649556761776258e-06,
+      "loss": 0.8484,
+      "step": 10446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647927465476111e-06,
+      "loss": 0.448,
+      "step": 10447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.646298178533651e-06,
+      "loss": 0.8713,
+      "step": 10448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644668900992183e-06,
+      "loss": 0.9319,
+      "step": 10449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.643039632895004e-06,
+      "loss": 0.9139,
+      "step": 10450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641410374285428e-06,
+      "loss": 0.9112,
+      "step": 10451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.639781125206757e-06,
+      "loss": 0.854,
+      "step": 10452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.638151885702288e-06,
+      "loss": 0.8882,
+      "step": 10453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63652265581533e-06,
+      "loss": 0.9487,
+      "step": 10454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634893435589186e-06,
+      "loss": 0.9207,
+      "step": 10455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.633264225067157e-06,
+      "loss": 0.8705,
+      "step": 10456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631635024292548e-06,
+      "loss": 0.8679,
+      "step": 10457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.630005833308655e-06,
+      "loss": 0.9236,
+      "step": 10458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.628376652158788e-06,
+      "loss": 0.9895,
+      "step": 10459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626747480886243e-06,
+      "loss": 0.7981,
+      "step": 10460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.625118319534324e-06,
+      "loss": 0.8998,
+      "step": 10461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62348916814633e-06,
+      "loss": 0.9563,
+      "step": 10462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.621860026765568e-06,
+      "loss": 0.926,
+      "step": 10463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62023089543533e-06,
+      "loss": 0.9741,
+      "step": 10464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.618601774198926e-06,
+      "loss": 0.9275,
+      "step": 10465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.616972663099648e-06,
+      "loss": 1.0359,
+      "step": 10466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6153435621808e-06,
+      "loss": 0.8729,
+      "step": 10467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61371447148568e-06,
+      "loss": 0.8575,
+      "step": 10468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.612085391057591e-06,
+      "loss": 0.969,
+      "step": 10469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610456320939828e-06,
+      "loss": 0.9614,
+      "step": 10470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60882726117569e-06,
+      "loss": 0.9035,
+      "step": 10471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.607198211808477e-06,
+      "loss": 0.9414,
+      "step": 10472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.605569172881486e-06,
+      "loss": 0.9468,
+      "step": 10473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.603940144438014e-06,
+      "loss": 0.9821,
+      "step": 10474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.602311126521363e-06,
+      "loss": 0.9845,
+      "step": 10475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.600682119174829e-06,
+      "loss": 0.9028,
+      "step": 10476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59905312244171e-06,
+      "loss": 0.9431,
+      "step": 10477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5974241363653e-06,
+      "loss": 0.8966,
+      "step": 10478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.595795160988893e-06,
+      "loss": 1.0119,
+      "step": 10479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.594166196355794e-06,
+      "loss": 0.9485,
+      "step": 10480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59253724250929e-06,
+      "loss": 0.9369,
+      "step": 10481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.590908299492683e-06,
+      "loss": 0.9026,
+      "step": 10482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589279367349264e-06,
+      "loss": 0.9257,
+      "step": 10483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.587650446122332e-06,
+      "loss": 0.9116,
+      "step": 10484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586021535855177e-06,
+      "loss": 0.8857,
+      "step": 10485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5843926365911e-06,
+      "loss": 0.8736,
+      "step": 10486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.582763748373392e-06,
+      "loss": 0.9093,
+      "step": 10487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.581134871245349e-06,
+      "loss": 0.8318,
+      "step": 10488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57950600525026e-06,
+      "loss": 0.9331,
+      "step": 10489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.577877150431422e-06,
+      "loss": 0.8861,
+      "step": 10490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.576248306832128e-06,
+      "loss": 0.9228,
+      "step": 10491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.574619474495668e-06,
+      "loss": 0.9661,
+      "step": 10492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.572990653465337e-06,
+      "loss": 0.9312,
+      "step": 10493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.571361843784426e-06,
+      "loss": 0.9703,
+      "step": 10494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.569733045496231e-06,
+      "loss": 0.8803,
+      "step": 10495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.568104258644036e-06,
+      "loss": 0.8704,
+      "step": 10496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.566475483271135e-06,
+      "loss": 0.9111,
+      "step": 10497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.564846719420826e-06,
+      "loss": 0.8614,
+      "step": 10498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563217967136393e-06,
+      "loss": 0.91,
+      "step": 10499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.561589226461126e-06,
+      "loss": 0.96,
+      "step": 10500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55996049743832e-06,
+      "loss": 0.9606,
+      "step": 10501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558331780111257e-06,
+      "loss": 0.9349,
+      "step": 10502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556703074523234e-06,
+      "loss": 0.9486,
+      "step": 10503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.555074380717535e-06,
+      "loss": 0.9104,
+      "step": 10504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.553445698737452e-06,
+      "loss": 0.9279,
+      "step": 10505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.551817028626273e-06,
+      "loss": 0.9244,
+      "step": 10506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.550188370427284e-06,
+      "loss": 0.8846,
+      "step": 10507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.548559724183776e-06,
+      "loss": 0.9199,
+      "step": 10508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54693108993903e-06,
+      "loss": 0.9463,
+      "step": 10509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.545302467736342e-06,
+      "loss": 0.8255,
+      "step": 10510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543673857618997e-06,
+      "loss": 0.9145,
+      "step": 10511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.542045259630277e-06,
+      "loss": 0.4858,
+      "step": 10512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.540416673813471e-06,
+      "loss": 0.9027,
+      "step": 10513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.538788100211866e-06,
+      "loss": 0.9211,
+      "step": 10514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537159538868746e-06,
+      "loss": 0.9654,
+      "step": 10515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535530989827397e-06,
+      "loss": 0.9915,
+      "step": 10516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533902453131103e-06,
+      "loss": 0.9123,
+      "step": 10517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532273928823151e-06,
+      "loss": 0.8688,
+      "step": 10518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530645416946825e-06,
+      "loss": 0.9147,
+      "step": 10519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.529016917545405e-06,
+      "loss": 0.8746,
+      "step": 10520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.527388430662174e-06,
+      "loss": 0.9466,
+      "step": 10521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.525759956340425e-06,
+      "loss": 0.928,
+      "step": 10522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.524131494623432e-06,
+      "loss": 0.9694,
+      "step": 10523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.522503045554481e-06,
+      "loss": 0.9419,
+      "step": 10524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.520874609176852e-06,
+      "loss": 0.9066,
+      "step": 10525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519246185533832e-06,
+      "loss": 0.9959,
+      "step": 10526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5176177746687e-06,
+      "loss": 0.9039,
+      "step": 10527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.515989376624733e-06,
+      "loss": 0.949,
+      "step": 10528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514360991445218e-06,
+      "loss": 0.8954,
+      "step": 10529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.512732619173431e-06,
+      "loss": 0.9007,
+      "step": 10530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.511104259852656e-06,
+      "loss": 0.9383,
+      "step": 10531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509475913526171e-06,
+      "loss": 0.9064,
+      "step": 10532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.507847580237251e-06,
+      "loss": 0.9523,
+      "step": 10533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506219260029187e-06,
+      "loss": 1.0098,
+      "step": 10534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50459095294525e-06,
+      "loss": 0.9649,
+      "step": 10535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502962659028717e-06,
+      "loss": 0.9221,
+      "step": 10536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.501334378322873e-06,
+      "loss": 0.8722,
+      "step": 10537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.499706110870988e-06,
+      "loss": 0.8377,
+      "step": 10538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.498077856716347e-06,
+      "loss": 0.9382,
+      "step": 10539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496449615902222e-06,
+      "loss": 0.9454,
+      "step": 10540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.494821388471888e-06,
+      "loss": 0.8621,
+      "step": 10541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49319317446863e-06,
+      "loss": 0.9188,
+      "step": 10542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491564973935714e-06,
+      "loss": 0.9334,
+      "step": 10543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489936786916417e-06,
+      "loss": 0.9467,
+      "step": 10544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.488308613454024e-06,
+      "loss": 0.891,
+      "step": 10545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4866804535918e-06,
+      "loss": 0.9409,
+      "step": 10546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485052307373026e-06,
+      "loss": 0.9736,
+      "step": 10547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.483424174840974e-06,
+      "loss": 0.964,
+      "step": 10548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.481796056038916e-06,
+      "loss": 0.9129,
+      "step": 10549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480167951010127e-06,
+      "loss": 0.9993,
+      "step": 10550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47853985979788e-06,
+      "loss": 1.0013,
+      "step": 10551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476911782445449e-06,
+      "loss": 0.8654,
+      "step": 10552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.475283718996103e-06,
+      "loss": 0.9153,
+      "step": 10553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47365566949312e-06,
+      "loss": 0.9196,
+      "step": 10554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.472027633979767e-06,
+      "loss": 0.4601,
+      "step": 10555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470399612499313e-06,
+      "loss": 0.9206,
+      "step": 10556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468771605095032e-06,
+      "loss": 0.8878,
+      "step": 10557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.467143611810201e-06,
+      "loss": 0.9082,
+      "step": 10558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46551563268808e-06,
+      "loss": 0.902,
+      "step": 10559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463887667771946e-06,
+      "loss": 1.0315,
+      "step": 10560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.462259717105065e-06,
+      "loss": 0.8924,
+      "step": 10561
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.460631780730705e-06,
+      "loss": 0.9063,
+      "step": 10562
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.459003858692138e-06,
+      "loss": 0.8876,
+      "step": 10563
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457375951032627e-06,
+      "loss": 0.3723,
+      "step": 10564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455748057795448e-06,
+      "loss": 0.9464,
+      "step": 10565
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45412017902386e-06,
+      "loss": 0.9278,
+      "step": 10566
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452492314761137e-06,
+      "loss": 0.9909,
+      "step": 10567
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450864465050539e-06,
+      "loss": 0.9617,
+      "step": 10568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44923662993534e-06,
+      "loss": 0.8587,
+      "step": 10569
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4476088094588e-06,
+      "loss": 0.8776,
+      "step": 10570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44598100366419e-06,
+      "loss": 0.9341,
+      "step": 10571
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44435321259477e-06,
+      "loss": 0.9448,
+      "step": 10572
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442725436293808e-06,
+      "loss": 0.899,
+      "step": 10573
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441097674804566e-06,
+      "loss": 0.996,
+      "step": 10574
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439469928170313e-06,
+      "loss": 0.9763,
+      "step": 10575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437842196434308e-06,
+      "loss": 0.9536,
+      "step": 10576
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436214479639814e-06,
+      "loss": 0.9297,
+      "step": 10577
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434586777830096e-06,
+      "loss": 0.8923,
+      "step": 10578
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.432959091048414e-06,
+      "loss": 0.9515,
+      "step": 10579
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.431331419338028e-06,
+      "loss": 0.9701,
+      "step": 10580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42970376274221e-06,
+      "loss": 0.8859,
+      "step": 10581
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428076121304213e-06,
+      "loss": 0.9635,
+      "step": 10582
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4264484950673e-06,
+      "loss": 0.9022,
+      "step": 10583
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.424820884074731e-06,
+      "loss": 0.4199,
+      "step": 10584
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.423193288369764e-06,
+      "loss": 0.8315,
+      "step": 10585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421565707995664e-06,
+      "loss": 0.9575,
+      "step": 10586
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.419938142995684e-06,
+      "loss": 0.459,
+      "step": 10587
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418310593413087e-06,
+      "loss": 0.3698,
+      "step": 10588
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41668305929113e-06,
+      "loss": 0.8938,
+      "step": 10589
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415055540673071e-06,
+      "loss": 0.4618,
+      "step": 10590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413428037602166e-06,
+      "loss": 0.9194,
+      "step": 10591
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.411800550121672e-06,
+      "loss": 0.946,
+      "step": 10592
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41017307827485e-06,
+      "loss": 0.9993,
+      "step": 10593
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.408545622104957e-06,
+      "loss": 0.9142,
+      "step": 10594
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.406918181655242e-06,
+      "loss": 0.9278,
+      "step": 10595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.405290756968966e-06,
+      "loss": 0.9942,
+      "step": 10596
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.403663348089384e-06,
+      "loss": 0.8337,
+      "step": 10597
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402035955059746e-06,
+      "loss": 0.9359,
+      "step": 10598
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400408577923312e-06,
+      "loss": 0.9275,
+      "step": 10599
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39878121672333e-06,
+      "loss": 0.9227,
+      "step": 10600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39715387150306e-06,
+      "loss": 0.9708,
+      "step": 10601
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39552654230575e-06,
+      "loss": 0.8528,
+      "step": 10602
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.393899229174651e-06,
+      "loss": 0.9167,
+      "step": 10603
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.392271932153022e-06,
+      "loss": 0.9234,
+      "step": 10604
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.390644651284112e-06,
+      "loss": 0.9238,
+      "step": 10605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38901738661117e-06,
+      "loss": 0.9395,
+      "step": 10606
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.387390138177447e-06,
+      "loss": 0.868,
+      "step": 10607
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.385762906026195e-06,
+      "loss": 0.8512,
+      "step": 10608
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384135690200665e-06,
+      "loss": 0.5045,
+      "step": 10609
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.382508490744104e-06,
+      "loss": 0.9057,
+      "step": 10610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.380881307699761e-06,
+      "loss": 0.9035,
+      "step": 10611
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379254141110887e-06,
+      "loss": 1.0418,
+      "step": 10612
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.377626991020728e-06,
+      "loss": 0.8731,
+      "step": 10613
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.375999857472534e-06,
+      "loss": 0.8969,
+      "step": 10614
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374372740509548e-06,
+      "loss": 0.8861,
+      "step": 10615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37274564017502e-06,
+      "loss": 0.8554,
+      "step": 10616
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.3711185565122e-06,
+      "loss": 0.9469,
+      "step": 10617
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.369491489564331e-06,
+      "loss": 0.9757,
+      "step": 10618
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367864439374654e-06,
+      "loss": 0.8914,
+      "step": 10619
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.366237405986422e-06,
+      "loss": 0.9221,
+      "step": 10620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.364610389442875e-06,
+      "loss": 0.9802,
+      "step": 10621
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362983389787258e-06,
+      "loss": 0.9038,
+      "step": 10622
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.361356407062814e-06,
+      "loss": 0.8921,
+      "step": 10623
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.359729441312789e-06,
+      "loss": 0.9147,
+      "step": 10624
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.358102492580425e-06,
+      "loss": 0.9238,
+      "step": 10625
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.356475560908959e-06,
+      "loss": 0.9703,
+      "step": 10626
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354848646341638e-06,
+      "loss": 0.9503,
+      "step": 10627
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353221748921707e-06,
+      "loss": 0.8879,
+      "step": 10628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.3515948686924e-06,
+      "loss": 0.865,
+      "step": 10629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.349968005696965e-06,
+      "loss": 0.9537,
+      "step": 10630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348341159978634e-06,
+      "loss": 0.8867,
+      "step": 10631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346714331580652e-06,
+      "loss": 0.9484,
+      "step": 10632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34508752054626e-06,
+      "loss": 1.0091,
+      "step": 10633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34346072691869e-06,
+      "loss": 0.8671,
+      "step": 10634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.341833950741186e-06,
+      "loss": 0.9599,
+      "step": 10635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.340207192056982e-06,
+      "loss": 0.9016,
+      "step": 10636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338580450909319e-06,
+      "loss": 0.8858,
+      "step": 10637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.336953727341432e-06,
+      "loss": 0.8766,
+      "step": 10638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.335327021396554e-06,
+      "loss": 0.8953,
+      "step": 10639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.333700333117926e-06,
+      "loss": 0.9034,
+      "step": 10640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.332073662548785e-06,
+      "loss": 0.924,
+      "step": 10641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330447009732361e-06,
+      "loss": 1.0347,
+      "step": 10642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328820374711893e-06,
+      "loss": 0.7905,
+      "step": 10643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.327193757530611e-06,
+      "loss": 0.9288,
+      "step": 10644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325567158231752e-06,
+      "loss": 0.4086,
+      "step": 10645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32394057685855e-06,
+      "loss": 0.9416,
+      "step": 10646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32231401345423e-06,
+      "loss": 0.8844,
+      "step": 10647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320687468062033e-06,
+      "loss": 0.903,
+      "step": 10648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.319060940725183e-06,
+      "loss": 0.9045,
+      "step": 10649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317434431486921e-06,
+      "loss": 0.9289,
+      "step": 10650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.315807940390466e-06,
+      "loss": 0.8864,
+      "step": 10651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314181467479057e-06,
+      "loss": 0.9408,
+      "step": 10652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.312555012795925e-06,
+      "loss": 1.0361,
+      "step": 10653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.310928576384293e-06,
+      "loss": 0.9807,
+      "step": 10654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309302158287393e-06,
+      "loss": 0.9792,
+      "step": 10655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307675758548454e-06,
+      "loss": 0.9641,
+      "step": 10656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.306049377210698e-06,
+      "loss": 0.7808,
+      "step": 10657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.304423014317362e-06,
+      "loss": 0.8795,
+      "step": 10658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.302796669911667e-06,
+      "loss": 1.0277,
+      "step": 10659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.301170344036839e-06,
+      "loss": 0.8672,
+      "step": 10660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299544036736105e-06,
+      "loss": 0.8559,
+      "step": 10661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297917748052688e-06,
+      "loss": 0.9312,
+      "step": 10662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296291478029818e-06,
+      "loss": 0.9696,
+      "step": 10663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294665226710717e-06,
+      "loss": 0.9452,
+      "step": 10664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293038994138607e-06,
+      "loss": 0.926,
+      "step": 10665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291412780356716e-06,
+      "loss": 0.8128,
+      "step": 10666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.289786585408265e-06,
+      "loss": 1.0092,
+      "step": 10667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.288160409336472e-06,
+      "loss": 0.8536,
+      "step": 10668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.286534252184564e-06,
+      "loss": 0.9672,
+      "step": 10669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.284908113995759e-06,
+      "loss": 0.9313,
+      "step": 10670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283281994813282e-06,
+      "loss": 0.9324,
+      "step": 10671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28165589468035e-06,
+      "loss": 0.9417,
+      "step": 10672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280029813640183e-06,
+      "loss": 0.9191,
+      "step": 10673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.278403751735998e-06,
+      "loss": 0.8466,
+      "step": 10674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.276777709011025e-06,
+      "loss": 0.9376,
+      "step": 10675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.275151685508468e-06,
+      "loss": 0.9912,
+      "step": 10676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.273525681271556e-06,
+      "loss": 0.8423,
+      "step": 10677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.271899696343498e-06,
+      "loss": 0.9054,
+      "step": 10678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.270273730767517e-06,
+      "loss": 0.8877,
+      "step": 10679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268647784586825e-06,
+      "loss": 0.9393,
+      "step": 10680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267021857844642e-06,
+      "loss": 0.9083,
+      "step": 10681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.265395950584181e-06,
+      "loss": 0.9012,
+      "step": 10682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.263770062848654e-06,
+      "loss": 0.9137,
+      "step": 10683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26214419468128e-06,
+      "loss": 0.9863,
+      "step": 10684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.260518346125269e-06,
+      "loss": 0.8974,
+      "step": 10685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.258892517223832e-06,
+      "loss": 0.8682,
+      "step": 10686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257266708020191e-06,
+      "loss": 0.856,
+      "step": 10687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255640918557554e-06,
+      "loss": 0.9091,
+      "step": 10688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.254015148879126e-06,
+      "loss": 1.0092,
+      "step": 10689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.252389399028127e-06,
+      "loss": 0.9897,
+      "step": 10690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.250763669047762e-06,
+      "loss": 1.0184,
+      "step": 10691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249137958981245e-06,
+      "loss": 0.9279,
+      "step": 10692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247512268871782e-06,
+      "loss": 0.964,
+      "step": 10693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.245886598762585e-06,
+      "loss": 0.9254,
+      "step": 10694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244260948696862e-06,
+      "loss": 0.8538,
+      "step": 10695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.242635318717815e-06,
+      "loss": 0.9257,
+      "step": 10696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24100970886866e-06,
+      "loss": 0.8999,
+      "step": 10697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.239384119192595e-06,
+      "loss": 0.9833,
+      "step": 10698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.237758549732836e-06,
+      "loss": 0.9061,
+      "step": 10699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236133000532584e-06,
+      "loss": 0.8536,
+      "step": 10700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234507471635043e-06,
+      "loss": 0.9035,
+      "step": 10701
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.232881963083421e-06,
+      "loss": 0.9387,
+      "step": 10702
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.231256474920922e-06,
+      "loss": 0.9205,
+      "step": 10703
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229631007190744e-06,
+      "loss": 0.8424,
+      "step": 10704
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.228005559936096e-06,
+      "loss": 0.9932,
+      "step": 10705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.226380133200177e-06,
+      "loss": 0.8883,
+      "step": 10706
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.224754727026191e-06,
+      "loss": 0.819,
+      "step": 10707
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22312934145734e-06,
+      "loss": 0.9432,
+      "step": 10708
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221503976536824e-06,
+      "loss": 0.9732,
+      "step": 10709
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219878632307838e-06,
+      "loss": 0.9403,
+      "step": 10710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218253308813592e-06,
+      "loss": 0.9416,
+      "step": 10711
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21662800609728e-06,
+      "loss": 0.8431,
+      "step": 10712
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2150027242021e-06,
+      "loss": 0.8982,
+      "step": 10713
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21337746317125e-06,
+      "loss": 0.9817,
+      "step": 10714
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211752223047932e-06,
+      "loss": 0.9508,
+      "step": 10715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.210127003875339e-06,
+      "loss": 0.8557,
+      "step": 10716
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.208501805696666e-06,
+      "loss": 0.9184,
+      "step": 10717
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.206876628555113e-06,
+      "loss": 0.9747,
+      "step": 10718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.205251472493871e-06,
+      "loss": 0.9191,
+      "step": 10719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.203626337556139e-06,
+      "loss": 0.9566,
+      "step": 10720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.202001223785107e-06,
+      "loss": 0.4347,
+      "step": 10721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20037613122397e-06,
+      "loss": 0.886,
+      "step": 10722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.198751059915926e-06,
+      "loss": 0.9286,
+      "step": 10723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197126009904163e-06,
+      "loss": 0.909,
+      "step": 10724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195500981231872e-06,
+      "loss": 0.9665,
+      "step": 10725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19387597394225e-06,
+      "loss": 0.9281,
+      "step": 10726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19225098807848e-06,
+      "loss": 0.9127,
+      "step": 10727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190626023683759e-06,
+      "loss": 0.8411,
+      "step": 10728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18900108080127e-06,
+      "loss": 0.9068,
+      "step": 10729
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18737615947421e-06,
+      "loss": 0.8734,
+      "step": 10730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.185751259745764e-06,
+      "loss": 0.9113,
+      "step": 10731
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.184126381659116e-06,
+      "loss": 0.8691,
+      "step": 10732
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182501525257457e-06,
+      "loss": 0.9312,
+      "step": 10733
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180876690583975e-06,
+      "loss": 0.8495,
+      "step": 10734
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.179251877681855e-06,
+      "loss": 0.9323,
+      "step": 10735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.177627086594284e-06,
+      "loss": 0.9551,
+      "step": 10736
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176002317364447e-06,
+      "loss": 0.8818,
+      "step": 10737
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.174377570035523e-06,
+      "loss": 0.8998,
+      "step": 10738
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.172752844650704e-06,
+      "loss": 0.9081,
+      "step": 10739
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171128141253167e-06,
+      "loss": 0.9347,
+      "step": 10740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1695034598861e-06,
+      "loss": 0.9861,
+      "step": 10741
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16787880059268e-06,
+      "loss": 0.8913,
+      "step": 10742
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166254163416093e-06,
+      "loss": 1.0109,
+      "step": 10743
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.164629548399518e-06,
+      "loss": 0.8929,
+      "step": 10744
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.163004955586131e-06,
+      "loss": 0.9063,
+      "step": 10745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16138038501912e-06,
+      "loss": 0.8668,
+      "step": 10746
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.159755836741661e-06,
+      "loss": 0.9534,
+      "step": 10747
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15813131079693e-06,
+      "loss": 0.9185,
+      "step": 10748
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.156506807228112e-06,
+      "loss": 0.9151,
+      "step": 10749
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.154882326078374e-06,
+      "loss": 1.0166,
+      "step": 10750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153257867390903e-06,
+      "loss": 0.9777,
+      "step": 10751
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15163343120887e-06,
+      "loss": 0.8031,
+      "step": 10752
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15000901757545e-06,
+      "loss": 0.9646,
+      "step": 10753
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148384626533822e-06,
+      "loss": 0.8689,
+      "step": 10754
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.146760258127154e-06,
+      "loss": 0.8277,
+      "step": 10755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145135912398628e-06,
+      "loss": 0.8805,
+      "step": 10756
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.143511589391407e-06,
+      "loss": 0.5055,
+      "step": 10757
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141887289148675e-06,
+      "loss": 0.9207,
+      "step": 10758
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.140263011713597e-06,
+      "loss": 0.8563,
+      "step": 10759
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138638757129348e-06,
+      "loss": 0.4258,
+      "step": 10760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137014525439093e-06,
+      "loss": 0.912,
+      "step": 10761
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.135390316686012e-06,
+      "loss": 0.9017,
+      "step": 10762
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.133766130913264e-06,
+      "loss": 0.8651,
+      "step": 10763
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132141968164026e-06,
+      "loss": 0.841,
+      "step": 10764
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130517828481464e-06,
+      "loss": 0.9293,
+      "step": 10765
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.128893711908743e-06,
+      "loss": 0.9052,
+      "step": 10766
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127269618489033e-06,
+      "loss": 0.9168,
+      "step": 10767
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125645548265498e-06,
+      "loss": 0.8866,
+      "step": 10768
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.124021501281303e-06,
+      "loss": 0.8773,
+      "step": 10769
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.122397477579623e-06,
+      "loss": 0.907,
+      "step": 10770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.120773477203611e-06,
+      "loss": 1.0046,
+      "step": 10771
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11914950019644e-06,
+      "loss": 0.9785,
+      "step": 10772
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117525546601268e-06,
+      "loss": 0.9151,
+      "step": 10773
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.115901616461256e-06,
+      "loss": 0.8905,
+      "step": 10774
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.114277709819572e-06,
+      "loss": 0.9794,
+      "step": 10775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112653826719372e-06,
+      "loss": 0.8317,
+      "step": 10776
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.111029967203825e-06,
+      "loss": 0.8391,
+      "step": 10777
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.109406131316081e-06,
+      "loss": 0.8773,
+      "step": 10778
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10778231909931e-06,
+      "loss": 0.8447,
+      "step": 10779
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.106158530596664e-06,
+      "loss": 0.9044,
+      "step": 10780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104534765851299e-06,
+      "loss": 0.9823,
+      "step": 10781
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10291102490638e-06,
+      "loss": 0.9221,
+      "step": 10782
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.101287307805063e-06,
+      "loss": 0.9431,
+      "step": 10783
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0996636145905e-06,
+      "loss": 0.8784,
+      "step": 10784
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098039945305854e-06,
+      "loss": 1.0038,
+      "step": 10785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.096416299994274e-06,
+      "loss": 0.9402,
+      "step": 10786
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.094792678698916e-06,
+      "loss": 1.0413,
+      "step": 10787
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093169081462936e-06,
+      "loss": 0.9992,
+      "step": 10788
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091545508329485e-06,
+      "loss": 0.8603,
+      "step": 10789
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.089921959341718e-06,
+      "loss": 0.8906,
+      "step": 10790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.088298434542785e-06,
+      "loss": 0.9574,
+      "step": 10791
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086674933975834e-06,
+      "loss": 0.4445,
+      "step": 10792
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.085051457684026e-06,
+      "loss": 0.8831,
+      "step": 10793
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.083428005710506e-06,
+      "loss": 0.9975,
+      "step": 10794
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08180457809842e-06,
+      "loss": 0.9613,
+      "step": 10795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08018117489092e-06,
+      "loss": 0.8742,
+      "step": 10796
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078557796131155e-06,
+      "loss": 0.9243,
+      "step": 10797
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076934441862271e-06,
+      "loss": 0.9041,
+      "step": 10798
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.075311112127414e-06,
+      "loss": 0.8446,
+      "step": 10799
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.073687806969734e-06,
+      "loss": 0.8826,
+      "step": 10800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072064526432373e-06,
+      "loss": 0.9286,
+      "step": 10801
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070441270558477e-06,
+      "loss": 0.8801,
+      "step": 10802
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06881803939119e-06,
+      "loss": 0.9386,
+      "step": 10803
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067194832973654e-06,
+      "loss": 0.9325,
+      "step": 10804
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065571651349015e-06,
+      "loss": 0.8867,
+      "step": 10805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063948494560416e-06,
+      "loss": 0.8825,
+      "step": 10806
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.062325362650998e-06,
+      "loss": 0.4576,
+      "step": 10807
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060702255663898e-06,
+      "loss": 1.0008,
+      "step": 10808
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.059079173642261e-06,
+      "loss": 0.8997,
+      "step": 10809
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.057456116629224e-06,
+      "loss": 0.9542,
+      "step": 10810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.055833084667929e-06,
+      "loss": 0.8177,
+      "step": 10811
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054210077801509e-06,
+      "loss": 0.9414,
+      "step": 10812
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.052587096073108e-06,
+      "loss": 1.0192,
+      "step": 10813
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050964139525859e-06,
+      "loss": 0.9316,
+      "step": 10814
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.049341208202898e-06,
+      "loss": 0.8784,
+      "step": 10815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04771830214736e-06,
+      "loss": 0.9375,
+      "step": 10816
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.046095421402385e-06,
+      "loss": 0.907,
+      "step": 10817
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044472566011103e-06,
+      "loss": 0.9056,
+      "step": 10818
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042849736016652e-06,
+      "loss": 0.9775,
+      "step": 10819
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.041226931462158e-06,
+      "loss": 0.9806,
+      "step": 10820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.03960415239076e-06,
+      "loss": 0.9296,
+      "step": 10821
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037981398845586e-06,
+      "loss": 0.8254,
+      "step": 10822
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036358670869766e-06,
+      "loss": 0.9215,
+      "step": 10823
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.034735968506434e-06,
+      "loss": 0.8323,
+      "step": 10824
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033113291798716e-06,
+      "loss": 0.8955,
+      "step": 10825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031490640789744e-06,
+      "loss": 0.902,
+      "step": 10826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.029868015522643e-06,
+      "loss": 1.0002,
+      "step": 10827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028245416040541e-06,
+      "loss": 0.9615,
+      "step": 10828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.026622842386568e-06,
+      "loss": 0.9078,
+      "step": 10829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.025000294603849e-06,
+      "loss": 0.8388,
+      "step": 10830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.023377772735506e-06,
+      "loss": 0.9505,
+      "step": 10831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02175527682467e-06,
+      "loss": 0.8933,
+      "step": 10832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020132806914456e-06,
+      "loss": 0.89,
+      "step": 10833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.018510363047999e-06,
+      "loss": 0.9585,
+      "step": 10834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.016887945268413e-06,
+      "loss": 0.9074,
+      "step": 10835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.015265553618821e-06,
+      "loss": 0.9465,
+      "step": 10836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.013643188142349e-06,
+      "loss": 0.7917,
+      "step": 10837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.012020848882112e-06,
+      "loss": 0.9856,
+      "step": 10838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010398535881236e-06,
+      "loss": 0.9709,
+      "step": 10839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00877624918283e-06,
+      "loss": 0.9386,
+      "step": 10840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.007153988830022e-06,
+      "loss": 0.8652,
+      "step": 10841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005531754865929e-06,
+      "loss": 0.9094,
+      "step": 10842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003909547333666e-06,
+      "loss": 0.9198,
+      "step": 10843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00228736627635e-06,
+      "loss": 0.834,
+      "step": 10844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000665211737096e-06,
+      "loss": 0.8725,
+      "step": 10845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.999043083759016e-06,
+      "loss": 0.8966,
+      "step": 10846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.997420982385234e-06,
+      "loss": 0.949,
+      "step": 10847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.995798907658852e-06,
+      "loss": 0.8765,
+      "step": 10848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994176859622991e-06,
+      "loss": 0.9056,
+      "step": 10849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99255483832076e-06,
+      "loss": 0.956,
+      "step": 10850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.990932843795265e-06,
+      "loss": 0.9426,
+      "step": 10851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989310876089628e-06,
+      "loss": 0.9562,
+      "step": 10852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.987688935246954e-06,
+      "loss": 0.8958,
+      "step": 10853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98606702131035e-06,
+      "loss": 0.8809,
+      "step": 10854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.984445134322928e-06,
+      "loss": 0.8909,
+      "step": 10855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.982823274327794e-06,
+      "loss": 0.9221,
+      "step": 10856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.981201441368055e-06,
+      "loss": 0.9475,
+      "step": 10857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97957963548682e-06,
+      "loss": 0.9612,
+      "step": 10858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97795785672719e-06,
+      "loss": 0.9455,
+      "step": 10859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.976336105132277e-06,
+      "loss": 0.9589,
+      "step": 10860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974714380745176e-06,
+      "loss": 0.9201,
+      "step": 10861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.973092683609001e-06,
+      "loss": 0.9018,
+      "step": 10862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.971471013766843e-06,
+      "loss": 0.9728,
+      "step": 10863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.969849371261816e-06,
+      "loss": 0.8513,
+      "step": 10864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968227756137014e-06,
+      "loss": 0.9566,
+      "step": 10865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.96660616843554e-06,
+      "loss": 0.8854,
+      "step": 10866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964984608200494e-06,
+      "loss": 0.9835,
+      "step": 10867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.963363075474978e-06,
+      "loss": 0.8954,
+      "step": 10868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.961741570302084e-06,
+      "loss": 0.8854,
+      "step": 10869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.960120092724916e-06,
+      "loss": 0.8314,
+      "step": 10870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.958498642786566e-06,
+      "loss": 0.8937,
+      "step": 10871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.956877220530133e-06,
+      "loss": 0.9077,
+      "step": 10872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.955255825998713e-06,
+      "loss": 0.8648,
+      "step": 10873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.953634459235399e-06,
+      "loss": 0.9315,
+      "step": 10874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.952013120283284e-06,
+      "loss": 0.8416,
+      "step": 10875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.950391809185464e-06,
+      "loss": 0.9797,
+      "step": 10876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.948770525985034e-06,
+      "loss": 0.9155,
+      "step": 10877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.947149270725081e-06,
+      "loss": 1.0042,
+      "step": 10878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945528043448698e-06,
+      "loss": 0.9068,
+      "step": 10879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943906844198976e-06,
+      "loss": 0.9175,
+      "step": 10880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.942285673019005e-06,
+      "loss": 0.9576,
+      "step": 10881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94066452995187e-06,
+      "loss": 0.9414,
+      "step": 10882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.939043415040663e-06,
+      "loss": 0.9967,
+      "step": 10883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937422328328472e-06,
+      "loss": 0.8287,
+      "step": 10884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.93580126985838e-06,
+      "loss": 0.9015,
+      "step": 10885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.934180239673475e-06,
+      "loss": 0.9354,
+      "step": 10886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.932559237816839e-06,
+      "loss": 0.9294,
+      "step": 10887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.930938264331561e-06,
+      "loss": 0.9048,
+      "step": 10888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929317319260727e-06,
+      "loss": 0.9688,
+      "step": 10889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927696402647411e-06,
+      "loss": 0.9044,
+      "step": 10890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.926075514534703e-06,
+      "loss": 0.9972,
+      "step": 10891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924454654965682e-06,
+      "loss": 0.9784,
+      "step": 10892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922833823983426e-06,
+      "loss": 0.9014,
+      "step": 10893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.921213021631018e-06,
+      "loss": 0.9184,
+      "step": 10894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919592247951534e-06,
+      "loss": 0.9225,
+      "step": 10895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917971502988055e-06,
+      "loss": 0.9126,
+      "step": 10896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.916350786783658e-06,
+      "loss": 0.997,
+      "step": 10897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914730099381418e-06,
+      "loss": 0.8935,
+      "step": 10898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.913109440824412e-06,
+      "loss": 0.8672,
+      "step": 10899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.911488811155718e-06,
+      "loss": 0.9261,
+      "step": 10900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.909868210418407e-06,
+      "loss": 0.8621,
+      "step": 10901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.908247638655553e-06,
+      "loss": 0.9335,
+      "step": 10902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906627095910231e-06,
+      "loss": 0.9207,
+      "step": 10903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.905006582225513e-06,
+      "loss": 0.9408,
+      "step": 10904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.903386097644471e-06,
+      "loss": 0.4198,
+      "step": 10905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90176564221017e-06,
+      "loss": 0.9504,
+      "step": 10906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.900145215965688e-06,
+      "loss": 0.904,
+      "step": 10907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898524818954085e-06,
+      "loss": 0.9613,
+      "step": 10908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89690445121844e-06,
+      "loss": 0.9515,
+      "step": 10909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.895284112801807e-06,
+      "loss": 0.9451,
+      "step": 10910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.893663803747264e-06,
+      "loss": 0.8702,
+      "step": 10911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.892043524097877e-06,
+      "loss": 0.4693,
+      "step": 10912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890423273896706e-06,
+      "loss": 0.9527,
+      "step": 10913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888803053186817e-06,
+      "loss": 0.8741,
+      "step": 10914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887182862011272e-06,
+      "loss": 0.9292,
+      "step": 10915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.885562700413138e-06,
+      "loss": 0.9586,
+      "step": 10916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.883942568435474e-06,
+      "loss": 0.8391,
+      "step": 10917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88232246612134e-06,
+      "loss": 0.8498,
+      "step": 10918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880702393513801e-06,
+      "loss": 0.8697,
+      "step": 10919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879082350655916e-06,
+      "loss": 0.9206,
+      "step": 10920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877462337590736e-06,
+      "loss": 0.9569,
+      "step": 10921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.875842354361325e-06,
+      "loss": 0.9002,
+      "step": 10922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.874222401010745e-06,
+      "loss": 0.9378,
+      "step": 10923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872602477582044e-06,
+      "loss": 1.0253,
+      "step": 10924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.870982584118285e-06,
+      "loss": 0.8978,
+      "step": 10925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869362720662519e-06,
+      "loss": 0.9556,
+      "step": 10926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867742887257798e-06,
+      "loss": 0.966,
+      "step": 10927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.866123083947182e-06,
+      "loss": 0.4372,
+      "step": 10928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.864503310773715e-06,
+      "loss": 0.8635,
+      "step": 10929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.862883567780457e-06,
+      "loss": 0.9532,
+      "step": 10930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86126385501045e-06,
+      "loss": 0.8799,
+      "step": 10931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.859644172506753e-06,
+      "loss": 0.8931,
+      "step": 10932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858024520312413e-06,
+      "loss": 0.8898,
+      "step": 10933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.856404898470471e-06,
+      "loss": 0.8754,
+      "step": 10934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.854785307023983e-06,
+      "loss": 0.9315,
+      "step": 10935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853165746015997e-06,
+      "loss": 0.8626,
+      "step": 10936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851546215489553e-06,
+      "loss": 0.7899,
+      "step": 10937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8499267154877e-06,
+      "loss": 0.9221,
+      "step": 10938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84830724605348e-06,
+      "loss": 0.8845,
+      "step": 10939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84668780722994e-06,
+      "loss": 0.9257,
+      "step": 10940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84506839906012e-06,
+      "loss": 0.8511,
+      "step": 10941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.843449021587063e-06,
+      "loss": 0.9617,
+      "step": 10942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84182967485381e-06,
+      "loss": 0.884,
+      "step": 10943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8402103589034e-06,
+      "loss": 0.9559,
+      "step": 10944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.838591073778878e-06,
+      "loss": 0.933,
+      "step": 10945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.836971819523273e-06,
+      "loss": 0.8883,
+      "step": 10946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835352596179629e-06,
+      "loss": 0.9384,
+      "step": 10947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.833733403790987e-06,
+      "loss": 0.9148,
+      "step": 10948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83211424240038e-06,
+      "loss": 0.8142,
+      "step": 10949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830495112050839e-06,
+      "loss": 0.8966,
+      "step": 10950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.828876012785404e-06,
+      "loss": 0.8949,
+      "step": 10951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827256944647107e-06,
+      "loss": 0.8242,
+      "step": 10952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825637907678982e-06,
+      "loss": 0.9978,
+      "step": 10953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82401890192406e-06,
+      "loss": 0.9705,
+      "step": 10954
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.822399927425369e-06,
+      "loss": 0.4331,
+      "step": 10955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.820780984225948e-06,
+      "loss": 0.8275,
+      "step": 10956
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819162072368816e-06,
+      "loss": 0.9692,
+      "step": 10957
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.817543191897007e-06,
+      "loss": 0.9167,
+      "step": 10958
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.815924342853555e-06,
+      "loss": 0.9784,
+      "step": 10959
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814305525281478e-06,
+      "loss": 0.9482,
+      "step": 10960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81268673922381e-06,
+      "loss": 0.3911,
+      "step": 10961
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81106798472357e-06,
+      "loss": 0.8506,
+      "step": 10962
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.809449261823784e-06,
+      "loss": 0.9608,
+      "step": 10963
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80783057056748e-06,
+      "loss": 0.8774,
+      "step": 10964
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806211910997674e-06,
+      "loss": 0.8671,
+      "step": 10965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804593283157395e-06,
+      "loss": 0.9208,
+      "step": 10966
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.802974687089658e-06,
+      "loss": 0.8253,
+      "step": 10967
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80135612283749e-06,
+      "loss": 0.9297,
+      "step": 10968
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.799737590443904e-06,
+      "loss": 0.959,
+      "step": 10969
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.798119089951924e-06,
+      "loss": 0.9196,
+      "step": 10970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.796500621404564e-06,
+      "loss": 0.4163,
+      "step": 10971
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.794882184844846e-06,
+      "loss": 0.8692,
+      "step": 10972
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793263780315782e-06,
+      "loss": 0.9722,
+      "step": 10973
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791645407860388e-06,
+      "loss": 0.9205,
+      "step": 10974
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790027067521681e-06,
+      "loss": 0.9243,
+      "step": 10975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.788408759342671e-06,
+      "loss": 0.8968,
+      "step": 10976
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.786790483366373e-06,
+      "loss": 0.8565,
+      "step": 10977
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785172239635796e-06,
+      "loss": 0.9062,
+      "step": 10978
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.783554028193957e-06,
+      "loss": 0.8299,
+      "step": 10979
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78193584908386e-06,
+      "loss": 0.8549,
+      "step": 10980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780317702348515e-06,
+      "loss": 0.9016,
+      "step": 10981
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.778699588030936e-06,
+      "loss": 0.9299,
+      "step": 10982
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.777081506174127e-06,
+      "loss": 0.8459,
+      "step": 10983
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775463456821092e-06,
+      "loss": 0.9654,
+      "step": 10984
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.773845440014845e-06,
+      "loss": 0.9055,
+      "step": 10985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.772227455798381e-06,
+      "loss": 0.9675,
+      "step": 10986
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.770609504214712e-06,
+      "loss": 0.9218,
+      "step": 10987
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768991585306835e-06,
+      "loss": 0.9916,
+      "step": 10988
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76737369911776e-06,
+      "loss": 0.9718,
+      "step": 10989
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.765755845690482e-06,
+      "loss": 0.8385,
+      "step": 10990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764138025068003e-06,
+      "loss": 0.9154,
+      "step": 10991
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.762520237293327e-06,
+      "loss": 0.9415,
+      "step": 10992
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.760902482409443e-06,
+      "loss": 0.885,
+      "step": 10993
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75928476045936e-06,
+      "loss": 0.9554,
+      "step": 10994
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757667071486073e-06,
+      "loss": 0.8843,
+      "step": 10995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756049415532573e-06,
+      "loss": 0.9507,
+      "step": 10996
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.754431792641862e-06,
+      "loss": 0.845,
+      "step": 10997
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75281420285693e-06,
+      "loss": 0.9108,
+      "step": 10998
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.751196646220772e-06,
+      "loss": 0.8801,
+      "step": 10999
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749579122776381e-06,
+      "loss": 0.8879,
+      "step": 11000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.747961632566748e-06,
+      "loss": 0.9676,
+      "step": 11001
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746344175634867e-06,
+      "loss": 0.4214,
+      "step": 11002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.744726752023725e-06,
+      "loss": 0.9477,
+      "step": 11003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.743109361776312e-06,
+      "loss": 0.8397,
+      "step": 11004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.741492004935613e-06,
+      "loss": 0.8696,
+      "step": 11005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.739874681544623e-06,
+      "loss": 0.8768,
+      "step": 11006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.738257391646323e-06,
+      "loss": 0.9787,
+      "step": 11007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.736640135283704e-06,
+      "loss": 0.9006,
+      "step": 11008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735022912499745e-06,
+      "loss": 0.4203,
+      "step": 11009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.733405723337433e-06,
+      "loss": 0.9672,
+      "step": 11010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73178856783975e-06,
+      "loss": 0.8683,
+      "step": 11011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.730171446049677e-06,
+      "loss": 0.9511,
+      "step": 11012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.728554358010199e-06,
+      "loss": 0.8962,
+      "step": 11013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726937303764291e-06,
+      "loss": 0.7869,
+      "step": 11014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725320283354939e-06,
+      "loss": 0.9005,
+      "step": 11015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.723703296825116e-06,
+      "loss": 0.9461,
+      "step": 11016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.722086344217799e-06,
+      "loss": 0.9325,
+      "step": 11017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72046942557597e-06,
+      "loss": 0.9446,
+      "step": 11018
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.718852540942605e-06,
+      "loss": 0.9577,
+      "step": 11019
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717235690360672e-06,
+      "loss": 0.8796,
+      "step": 11020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.715618873873152e-06,
+      "loss": 0.3968,
+      "step": 11021
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714002091523015e-06,
+      "loss": 0.9157,
+      "step": 11022
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.712385343353235e-06,
+      "loss": 0.9109,
+      "step": 11023
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71076862940678e-06,
+      "loss": 0.9051,
+      "step": 11024
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709151949726623e-06,
+      "loss": 0.9383,
+      "step": 11025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.707535304355735e-06,
+      "loss": 0.9711,
+      "step": 11026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705918693337077e-06,
+      "loss": 0.9,
+      "step": 11027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.704302116713628e-06,
+      "loss": 1.0039,
+      "step": 11028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702685574528343e-06,
+      "loss": 0.9703,
+      "step": 11029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.701069066824195e-06,
+      "loss": 0.9219,
+      "step": 11030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699452593644151e-06,
+      "loss": 0.8467,
+      "step": 11031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.697836155031172e-06,
+      "loss": 0.878,
+      "step": 11032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696219751028217e-06,
+      "loss": 0.9256,
+      "step": 11033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.694603381678257e-06,
+      "loss": 0.9083,
+      "step": 11034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692987047024243e-06,
+      "loss": 0.8395,
+      "step": 11035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.691370747109144e-06,
+      "loss": 0.9474,
+      "step": 11036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.689754481975915e-06,
+      "loss": 0.932,
+      "step": 11037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688138251667515e-06,
+      "loss": 0.8908,
+      "step": 11038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.686522056226903e-06,
+      "loss": 0.8488,
+      "step": 11039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68490589569703e-06,
+      "loss": 0.9148,
+      "step": 11040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68328977012086e-06,
+      "loss": 0.8398,
+      "step": 11041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.681673679541347e-06,
+      "loss": 0.9038,
+      "step": 11042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.680057624001436e-06,
+      "loss": 0.9248,
+      "step": 11043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.67844160354409e-06,
+      "loss": 0.9249,
+      "step": 11044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676825618212256e-06,
+      "loss": 0.9094,
+      "step": 11045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675209668048886e-06,
+      "loss": 0.8318,
+      "step": 11046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.673593753096931e-06,
+      "loss": 0.8863,
+      "step": 11047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671977873399336e-06,
+      "loss": 0.9573,
+      "step": 11048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670362028999056e-06,
+      "loss": 0.9381,
+      "step": 11049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668746219939031e-06,
+      "loss": 0.9336,
+      "step": 11050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.667130446262214e-06,
+      "loss": 0.9542,
+      "step": 11051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.665514708011544e-06,
+      "loss": 0.8615,
+      "step": 11052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663899005229972e-06,
+      "loss": 0.9035,
+      "step": 11053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.662283337960437e-06,
+      "loss": 0.9314,
+      "step": 11054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660667706245885e-06,
+      "loss": 0.9285,
+      "step": 11055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.659052110129254e-06,
+      "loss": 0.9321,
+      "step": 11056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.657436549653486e-06,
+      "loss": 0.9381,
+      "step": 11057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655821024861522e-06,
+      "loss": 0.8869,
+      "step": 11058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654205535796301e-06,
+      "loss": 0.8456,
+      "step": 11059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65259008250076e-06,
+      "loss": 0.9621,
+      "step": 11060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650974665017833e-06,
+      "loss": 0.9475,
+      "step": 11061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649359283390461e-06,
+      "loss": 0.9886,
+      "step": 11062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.647743937661576e-06,
+      "loss": 0.9161,
+      "step": 11063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.646128627874109e-06,
+      "loss": 0.8438,
+      "step": 11064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.644513354071e-06,
+      "loss": 0.9047,
+      "step": 11065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642898116295179e-06,
+      "loss": 0.9211,
+      "step": 11066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.641282914589577e-06,
+      "loss": 0.9396,
+      "step": 11067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.639667748997121e-06,
+      "loss": 0.8515,
+      "step": 11068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638052619560743e-06,
+      "loss": 0.8908,
+      "step": 11069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.636437526323372e-06,
+      "loss": 0.972,
+      "step": 11070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634822469327933e-06,
+      "loss": 0.8014,
+      "step": 11071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.633207448617356e-06,
+      "loss": 0.9937,
+      "step": 11072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.631592464234562e-06,
+      "loss": 0.9389,
+      "step": 11073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.629977516222477e-06,
+      "loss": 0.8835,
+      "step": 11074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.628362604624025e-06,
+      "loss": 0.9249,
+      "step": 11075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.626747729482127e-06,
+      "loss": 0.9971,
+      "step": 11076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.625132890839706e-06,
+      "loss": 0.9126,
+      "step": 11077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.623518088739685e-06,
+      "loss": 0.9209,
+      "step": 11078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621903323224979e-06,
+      "loss": 0.976,
+      "step": 11079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.620288594338511e-06,
+      "loss": 0.9394,
+      "step": 11080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.618673902123196e-06,
+      "loss": 0.8755,
+      "step": 11081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61705924662195e-06,
+      "loss": 0.9329,
+      "step": 11082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.615444627877691e-06,
+      "loss": 0.9187,
+      "step": 11083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613830045933332e-06,
+      "loss": 0.9217,
+      "step": 11084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.612215500831787e-06,
+      "loss": 0.9637,
+      "step": 11085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.610600992615968e-06,
+      "loss": 0.9626,
+      "step": 11086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.608986521328791e-06,
+      "loss": 0.8888,
+      "step": 11087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.607372087013157e-06,
+      "loss": 0.8853,
+      "step": 11088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.605757689711987e-06,
+      "loss": 0.9791,
+      "step": 11089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604143329468185e-06,
+      "loss": 0.9362,
+      "step": 11090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60252900632466e-06,
+      "loss": 0.9837,
+      "step": 11091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600914720324315e-06,
+      "loss": 0.9039,
+      "step": 11092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.599300471510063e-06,
+      "loss": 0.7818,
+      "step": 11093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597686259924802e-06,
+      "loss": 0.8579,
+      "step": 11094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.596072085611438e-06,
+      "loss": 0.9728,
+      "step": 11095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.594457948612878e-06,
+      "loss": 0.9702,
+      "step": 11096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592843848972017e-06,
+      "loss": 0.9357,
+      "step": 11097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591229786731761e-06,
+      "loss": 0.8498,
+      "step": 11098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589615761935006e-06,
+      "loss": 0.9638,
+      "step": 11099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.588001774624652e-06,
+      "loss": 0.8832,
+      "step": 11100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.586387824843602e-06,
+      "loss": 0.967,
+      "step": 11101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58477391263475e-06,
+      "loss": 0.9684,
+      "step": 11102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.583160038040988e-06,
+      "loss": 0.919,
+      "step": 11103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581546201105217e-06,
+      "loss": 0.9343,
+      "step": 11104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.579932401870326e-06,
+      "loss": 0.9601,
+      "step": 11105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57831864037921e-06,
+      "loss": 0.8743,
+      "step": 11106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.576704916674761e-06,
+      "loss": 0.8939,
+      "step": 11107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57509123079987e-06,
+      "loss": 0.9101,
+      "step": 11108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573477582797427e-06,
+      "loss": 0.9048,
+      "step": 11109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57186397271032e-06,
+      "loss": 0.9112,
+      "step": 11110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570250400581433e-06,
+      "loss": 0.9192,
+      "step": 11111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.568636866453664e-06,
+      "loss": 0.9984,
+      "step": 11112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567023370369888e-06,
+      "loss": 0.8359,
+      "step": 11113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.565409912372998e-06,
+      "loss": 0.8899,
+      "step": 11114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56379649250587e-06,
+      "loss": 0.867,
+      "step": 11115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.562183110811396e-06,
+      "loss": 0.9227,
+      "step": 11116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560569767332452e-06,
+      "loss": 0.8918,
+      "step": 11117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558956462111916e-06,
+      "loss": 0.9583,
+      "step": 11118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557343195192674e-06,
+      "loss": 0.9324,
+      "step": 11119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5557299666176e-06,
+      "loss": 0.801,
+      "step": 11120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554116776429577e-06,
+      "loss": 0.9102,
+      "step": 11121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552503624671479e-06,
+      "loss": 0.9025,
+      "step": 11122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.550890511386176e-06,
+      "loss": 0.9267,
+      "step": 11123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549277436616551e-06,
+      "loss": 0.396,
+      "step": 11124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547664400405477e-06,
+      "loss": 0.9499,
+      "step": 11125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546051402795822e-06,
+      "loss": 0.8416,
+      "step": 11126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544438443830461e-06,
+      "loss": 0.9019,
+      "step": 11127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.542825523552263e-06,
+      "loss": 0.8958,
+      "step": 11128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.541212642004103e-06,
+      "loss": 0.8959,
+      "step": 11129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53959979922884e-06,
+      "loss": 0.852,
+      "step": 11130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537986995269348e-06,
+      "loss": 0.8533,
+      "step": 11131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.536374230168492e-06,
+      "loss": 0.9549,
+      "step": 11132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534761503969135e-06,
+      "loss": 0.9827,
+      "step": 11133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.533148816714148e-06,
+      "loss": 0.9831,
+      "step": 11134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531536168446384e-06,
+      "loss": 0.8595,
+      "step": 11135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529923559208713e-06,
+      "loss": 0.9135,
+      "step": 11136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.528310989044e-06,
+      "loss": 0.9116,
+      "step": 11137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.526698457995098e-06,
+      "loss": 0.9467,
+      "step": 11138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.525085966104866e-06,
+      "loss": 0.865,
+      "step": 11139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523473513416167e-06,
+      "loss": 0.8889,
+      "step": 11140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521861099971854e-06,
+      "loss": 0.8912,
+      "step": 11141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.520248725814785e-06,
+      "loss": 0.9086,
+      "step": 11142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518636390987815e-06,
+      "loss": 0.8778,
+      "step": 11143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517024095533799e-06,
+      "loss": 0.9115,
+      "step": 11144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.515411839495589e-06,
+      "loss": 0.9981,
+      "step": 11145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.513799622916034e-06,
+      "loss": 0.9504,
+      "step": 11146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.512187445837988e-06,
+      "loss": 0.9963,
+      "step": 11147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510575308304302e-06,
+      "loss": 0.9931,
+      "step": 11148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508963210357822e-06,
+      "loss": 0.8342,
+      "step": 11149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5073511520414e-06,
+      "loss": 0.9894,
+      "step": 11150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505739133397876e-06,
+      "loss": 0.9256,
+      "step": 11151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.504127154470101e-06,
+      "loss": 0.789,
+      "step": 11152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.502515215300918e-06,
+      "loss": 0.9382,
+      "step": 11153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.500903315933169e-06,
+      "loss": 0.8822,
+      "step": 11154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.4992914564097e-06,
+      "loss": 1.0021,
+      "step": 11155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.497679636773346e-06,
+      "loss": 0.836,
+      "step": 11156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.496067857066957e-06,
+      "loss": 0.918,
+      "step": 11157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49445611733336e-06,
+      "loss": 0.8776,
+      "step": 11158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492844417615405e-06,
+      "loss": 0.9183,
+      "step": 11159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49123275795592e-06,
+      "loss": 0.8781,
+      "step": 11160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48962113839775e-06,
+      "loss": 0.9091,
+      "step": 11161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488009558983722e-06,
+      "loss": 0.8984,
+      "step": 11162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486398019756676e-06,
+      "loss": 0.791,
+      "step": 11163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.484786520759438e-06,
+      "loss": 0.9873,
+      "step": 11164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.483175062034847e-06,
+      "loss": 0.4139,
+      "step": 11165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48156364362573e-06,
+      "loss": 0.879,
+      "step": 11166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479952265574916e-06,
+      "loss": 0.9125,
+      "step": 11167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478340927925235e-06,
+      "loss": 0.8228,
+      "step": 11168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476729630719514e-06,
+      "loss": 0.8826,
+      "step": 11169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.475118374000576e-06,
+      "loss": 0.9458,
+      "step": 11170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473507157811254e-06,
+      "loss": 0.9081,
+      "step": 11171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47189598219437e-06,
+      "loss": 0.8808,
+      "step": 11172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.470284847192744e-06,
+      "loss": 0.9187,
+      "step": 11173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.468673752849201e-06,
+      "loss": 0.8491,
+      "step": 11174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.467062699206557e-06,
+      "loss": 0.9492,
+      "step": 11175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.465451686307639e-06,
+      "loss": 0.9762,
+      "step": 11176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46384071419526e-06,
+      "loss": 0.8489,
+      "step": 11177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.462229782912244e-06,
+      "loss": 0.8674,
+      "step": 11178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460618892501402e-06,
+      "loss": 0.8916,
+      "step": 11179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45900804300555e-06,
+      "loss": 0.8339,
+      "step": 11180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.457397234467507e-06,
+      "loss": 0.9507,
+      "step": 11181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45578646693008e-06,
+      "loss": 0.9196,
+      "step": 11182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.454175740436086e-06,
+      "loss": 0.8959,
+      "step": 11183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.452565055028338e-06,
+      "loss": 0.97,
+      "step": 11184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45095441074964e-06,
+      "loss": 0.8642,
+      "step": 11185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.44934380764281e-06,
+      "loss": 0.5,
+      "step": 11186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447733245750647e-06,
+      "loss": 0.892,
+      "step": 11187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446122725115962e-06,
+      "loss": 0.9328,
+      "step": 11188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444512245781561e-06,
+      "loss": 0.9351,
+      "step": 11189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.442901807790246e-06,
+      "loss": 0.9271,
+      "step": 11190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.441291411184825e-06,
+      "loss": 0.9246,
+      "step": 11191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439681056008095e-06,
+      "loss": 0.9464,
+      "step": 11192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.438070742302864e-06,
+      "loss": 0.912,
+      "step": 11193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.436460470111924e-06,
+      "loss": 0.4072,
+      "step": 11194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.434850239478082e-06,
+      "loss": 0.9402,
+      "step": 11195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.433240050444134e-06,
+      "loss": 0.8541,
+      "step": 11196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431629903052876e-06,
+      "loss": 0.9194,
+      "step": 11197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430019797347102e-06,
+      "loss": 0.8841,
+      "step": 11198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42840973336961e-06,
+      "loss": 0.9011,
+      "step": 11199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.426799711163193e-06,
+      "loss": 0.9298,
+      "step": 11200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.425189730770642e-06,
+      "loss": 0.9111,
+      "step": 11201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42357979223475e-06,
+      "loss": 0.9013,
+      "step": 11202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.421969895598305e-06,
+      "loss": 0.9177,
+      "step": 11203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420360040904099e-06,
+      "loss": 0.8841,
+      "step": 11204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41875022819492e-06,
+      "loss": 0.9867,
+      "step": 11205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41714045751355e-06,
+      "loss": 0.888,
+      "step": 11206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.415530728902781e-06,
+      "loss": 0.8706,
+      "step": 11207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413921042405397e-06,
+      "loss": 0.9298,
+      "step": 11208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41231139806418e-06,
+      "loss": 0.9283,
+      "step": 11209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410701795921912e-06,
+      "loss": 0.9007,
+      "step": 11210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409092236021374e-06,
+      "loss": 0.9141,
+      "step": 11211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40748271840535e-06,
+      "loss": 0.925,
+      "step": 11212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405873243116616e-06,
+      "loss": 0.9316,
+      "step": 11213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40426381019795e-06,
+      "loss": 0.905,
+      "step": 11214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402654419692131e-06,
+      "loss": 0.9239,
+      "step": 11215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40104507164193e-06,
+      "loss": 0.9588,
+      "step": 11216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.399435766090123e-06,
+      "loss": 0.8642,
+      "step": 11217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.397826503079489e-06,
+      "loss": 0.927,
+      "step": 11218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396217282652795e-06,
+      "loss": 0.8868,
+      "step": 11219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.394608104852817e-06,
+      "loss": 0.4399,
+      "step": 11220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.39299896972232e-06,
+      "loss": 0.9342,
+      "step": 11221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.391389877304076e-06,
+      "loss": 0.8768,
+      "step": 11222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.389780827640852e-06,
+      "loss": 0.9158,
+      "step": 11223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.388171820775413e-06,
+      "loss": 0.9211,
+      "step": 11224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.386562856750527e-06,
+      "loss": 0.8892,
+      "step": 11225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384953935608958e-06,
+      "loss": 0.9041,
+      "step": 11226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383345057393468e-06,
+      "loss": 0.8342,
+      "step": 11227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381736222146821e-06,
+      "loss": 0.8801,
+      "step": 11228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.380127429911772e-06,
+      "loss": 0.9731,
+      "step": 11229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37851868073109e-06,
+      "loss": 0.9305,
+      "step": 11230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37690997464753e-06,
+      "loss": 0.9714,
+      "step": 11231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.375301311703849e-06,
+      "loss": 1.0278,
+      "step": 11232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.373692691942803e-06,
+      "loss": 0.9584,
+      "step": 11233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.372084115407147e-06,
+      "loss": 0.436,
+      "step": 11234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37047558213964e-06,
+      "loss": 0.9037,
+      "step": 11235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.368867092183028e-06,
+      "loss": 0.8589,
+      "step": 11236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.367258645580064e-06,
+      "loss": 0.9279,
+      "step": 11237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.365650242373502e-06,
+      "loss": 1.0483,
+      "step": 11238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36404188260609e-06,
+      "loss": 0.9817,
+      "step": 11239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.362433566320577e-06,
+      "loss": 0.9101,
+      "step": 11240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360825293559706e-06,
+      "loss": 0.8826,
+      "step": 11241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359217064366231e-06,
+      "loss": 0.3924,
+      "step": 11242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.357608878782891e-06,
+      "loss": 0.9621,
+      "step": 11243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35600073685243e-06,
+      "loss": 0.9608,
+      "step": 11244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.354392638617593e-06,
+      "loss": 0.9265,
+      "step": 11245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35278458412112e-06,
+      "loss": 0.9175,
+      "step": 11246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.351176573405752e-06,
+      "loss": 0.8715,
+      "step": 11247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349568606514228e-06,
+      "loss": 0.9301,
+      "step": 11248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347960683489285e-06,
+      "loss": 1.0036,
+      "step": 11249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34635280437366e-06,
+      "loss": 0.8642,
+      "step": 11250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34474496921009e-06,
+      "loss": 0.9924,
+      "step": 11251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.343137178041305e-06,
+      "loss": 0.9172,
+      "step": 11252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34152943091004e-06,
+      "loss": 0.912,
+      "step": 11253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.339921727859032e-06,
+      "loss": 0.8746,
+      "step": 11254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338314068931009e-06,
+      "loss": 0.8532,
+      "step": 11255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.336706454168701e-06,
+      "loss": 0.9228,
+      "step": 11256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.335098883614835e-06,
+      "loss": 0.9456,
+      "step": 11257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.333491357312137e-06,
+      "loss": 0.9828,
+      "step": 11258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.331883875303338e-06,
+      "loss": 0.9385,
+      "step": 11259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330276437631158e-06,
+      "loss": 0.909,
+      "step": 11260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.328669044338326e-06,
+      "loss": 0.9054,
+      "step": 11261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32706169546756e-06,
+      "loss": 0.9386,
+      "step": 11262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.325454391061586e-06,
+      "loss": 0.4148,
+      "step": 11263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.323847131163121e-06,
+      "loss": 0.8642,
+      "step": 11264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32223991581488e-06,
+      "loss": 0.9439,
+      "step": 11265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32063274505959e-06,
+      "loss": 0.9537,
+      "step": 11266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319025618939962e-06,
+      "loss": 0.8618,
+      "step": 11267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317418537498714e-06,
+      "loss": 0.8599,
+      "step": 11268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.315811500778561e-06,
+      "loss": 0.9835,
+      "step": 11269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.314204508822216e-06,
+      "loss": 0.8874,
+      "step": 11270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312597561672386e-06,
+      "loss": 0.9041,
+      "step": 11271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310990659371788e-06,
+      "loss": 0.4869,
+      "step": 11272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.309383801963128e-06,
+      "loss": 0.9395,
+      "step": 11273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.307776989489116e-06,
+      "loss": 0.9223,
+      "step": 11274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30617022199246e-06,
+      "loss": 0.898,
+      "step": 11275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30456349951586e-06,
+      "loss": 0.9192,
+      "step": 11276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.302956822102031e-06,
+      "loss": 0.9809,
+      "step": 11277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301350189793671e-06,
+      "loss": 0.8865,
+      "step": 11278
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.299743602633481e-06,
+      "loss": 0.786,
+      "step": 11279
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.298137060664167e-06,
+      "loss": 0.9145,
+      "step": 11280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.296530563928424e-06,
+      "loss": 0.9324,
+      "step": 11281
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.294924112468953e-06,
+      "loss": 0.8372,
+      "step": 11282
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.293317706328453e-06,
+      "loss": 0.924,
+      "step": 11283
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.291711345549619e-06,
+      "loss": 0.912,
+      "step": 11284
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290105030175146e-06,
+      "loss": 0.9243,
+      "step": 11285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.288498760247727e-06,
+      "loss": 0.8093,
+      "step": 11286
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.286892535810059e-06,
+      "loss": 0.8793,
+      "step": 11287
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.285286356904827e-06,
+      "loss": 0.9127,
+      "step": 11288
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.283680223574726e-06,
+      "loss": 0.9102,
+      "step": 11289
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.282074135862447e-06,
+      "loss": 0.9924,
+      "step": 11290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280468093810677e-06,
+      "loss": 0.9033,
+      "step": 11291
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.278862097462095e-06,
+      "loss": 0.837,
+      "step": 11292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.277256146859398e-06,
+      "loss": 0.9875,
+      "step": 11293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.275650242045263e-06,
+      "loss": 0.8804,
+      "step": 11294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274044383062376e-06,
+      "loss": 0.8916,
+      "step": 11295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.272438569953416e-06,
+      "loss": 0.9393,
+      "step": 11296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.270832802761069e-06,
+      "loss": 0.9234,
+      "step": 11297
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269227081528011e-06,
+      "loss": 0.9304,
+      "step": 11298
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.267621406296918e-06,
+      "loss": 0.9082,
+      "step": 11299
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266015777110468e-06,
+      "loss": 0.9019,
+      "step": 11300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26441019401134e-06,
+      "loss": 0.9544,
+      "step": 11301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.262804657042208e-06,
+      "loss": 0.9425,
+      "step": 11302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261199166245744e-06,
+      "loss": 0.8303,
+      "step": 11303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25959372166462e-06,
+      "loss": 0.8523,
+      "step": 11304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.257988323341509e-06,
+      "loss": 0.9084,
+      "step": 11305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.256382971319076e-06,
+      "loss": 0.4178,
+      "step": 11306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254777665639993e-06,
+      "loss": 0.9121,
+      "step": 11307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253172406346927e-06,
+      "loss": 0.9125,
+      "step": 11308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.251567193482543e-06,
+      "loss": 0.9407,
+      "step": 11309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.249962027089506e-06,
+      "loss": 0.9222,
+      "step": 11310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24835690721048e-06,
+      "loss": 0.8993,
+      "step": 11311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246751833888122e-06,
+      "loss": 0.9331,
+      "step": 11312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.245146807165103e-06,
+      "loss": 0.9084,
+      "step": 11313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.243541827084078e-06,
+      "loss": 0.8842,
+      "step": 11314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.241936893687703e-06,
+      "loss": 0.975,
+      "step": 11315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24033200701864e-06,
+      "loss": 0.9773,
+      "step": 11316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23872716711954e-06,
+      "loss": 0.8406,
+      "step": 11317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237122374033063e-06,
+      "loss": 0.8892,
+      "step": 11318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23551762780186e-06,
+      "loss": 0.8523,
+      "step": 11319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233912928468579e-06,
+      "loss": 0.8803,
+      "step": 11320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.232308276075879e-06,
+      "loss": 0.9833,
+      "step": 11321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.230703670666404e-06,
+      "loss": 0.8867,
+      "step": 11322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.229099112282807e-06,
+      "loss": 0.9028,
+      "step": 11323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.227494600967729e-06,
+      "loss": 0.9297,
+      "step": 11324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225890136763821e-06,
+      "loss": 0.9147,
+      "step": 11325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22428571971373e-06,
+      "loss": 0.8734,
+      "step": 11326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.222681349860096e-06,
+      "loss": 0.8537,
+      "step": 11327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22107702724556e-06,
+      "loss": 0.8823,
+      "step": 11328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.219472751912767e-06,
+      "loss": 0.9424,
+      "step": 11329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217868523904352e-06,
+      "loss": 1.0288,
+      "step": 11330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21626434326296e-06,
+      "loss": 0.8759,
+      "step": 11331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214660210031221e-06,
+      "loss": 0.9943,
+      "step": 11332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.213056124251777e-06,
+      "loss": 0.9446,
+      "step": 11333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.211452085967259e-06,
+      "loss": 0.9132,
+      "step": 11334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.209848095220303e-06,
+      "loss": 0.98,
+      "step": 11335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208244152053535e-06,
+      "loss": 0.888,
+      "step": 11336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206640256509598e-06,
+      "loss": 0.9435,
+      "step": 11337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20503640863111e-06,
+      "loss": 0.7895,
+      "step": 11338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203432608460708e-06,
+      "loss": 0.9211,
+      "step": 11339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.201828856041014e-06,
+      "loss": 0.908,
+      "step": 11340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200225151414654e-06,
+      "loss": 0.8451,
+      "step": 11341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.198621494624256e-06,
+      "loss": 0.9362,
+      "step": 11342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.197017885712439e-06,
+      "loss": 0.9839,
+      "step": 11343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19541432472183e-06,
+      "loss": 0.9004,
+      "step": 11344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193810811695044e-06,
+      "loss": 0.9271,
+      "step": 11345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.192207346674707e-06,
+      "loss": 0.8588,
+      "step": 11346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19060392970343e-06,
+      "loss": 0.8664,
+      "step": 11347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.189000560823837e-06,
+      "loss": 0.8692,
+      "step": 11348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18739724007854e-06,
+      "loss": 0.8213,
+      "step": 11349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185793967510155e-06,
+      "loss": 0.869,
+      "step": 11350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184190743161293e-06,
+      "loss": 0.961,
+      "step": 11351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18258756707457e-06,
+      "loss": 0.9405,
+      "step": 11352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180984439292589e-06,
+      "loss": 0.9601,
+      "step": 11353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179381359857968e-06,
+      "loss": 0.9913,
+      "step": 11354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.177778328813311e-06,
+      "loss": 0.9343,
+      "step": 11355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.176175346201224e-06,
+      "loss": 0.9277,
+      "step": 11356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174572412064315e-06,
+      "loss": 0.7956,
+      "step": 11357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.172969526445183e-06,
+      "loss": 0.9136,
+      "step": 11358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171366689386433e-06,
+      "loss": 0.9407,
+      "step": 11359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16976390093067e-06,
+      "loss": 0.9791,
+      "step": 11360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.168161161120492e-06,
+      "loss": 0.9838,
+      "step": 11361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166558469998502e-06,
+      "loss": 0.8956,
+      "step": 11362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.164955827607291e-06,
+      "loss": 0.9187,
+      "step": 11363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163353233989457e-06,
+      "loss": 0.8643,
+      "step": 11364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1617506891876e-06,
+      "loss": 0.8788,
+      "step": 11365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.160148193244305e-06,
+      "loss": 0.9101,
+      "step": 11366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.158545746202175e-06,
+      "loss": 0.9307,
+      "step": 11367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156943348103794e-06,
+      "loss": 0.8565,
+      "step": 11368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.155340998991753e-06,
+      "loss": 0.9902,
+      "step": 11369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153738698908644e-06,
+      "loss": 0.834,
+      "step": 11370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.152136447897046e-06,
+      "loss": 0.8984,
+      "step": 11371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.150534245999553e-06,
+      "loss": 0.8607,
+      "step": 11372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14893209325875e-06,
+      "loss": 0.9417,
+      "step": 11373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.147329989717214e-06,
+      "loss": 0.9896,
+      "step": 11374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145727935417536e-06,
+      "loss": 0.8715,
+      "step": 11375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14412593040229e-06,
+      "loss": 0.9891,
+      "step": 11376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142523974714055e-06,
+      "loss": 0.9297,
+      "step": 11377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.140922068395414e-06,
+      "loss": 0.8943,
+      "step": 11378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13932021148894e-06,
+      "loss": 0.8969,
+      "step": 11379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.137718404037209e-06,
+      "loss": 0.9212,
+      "step": 11380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.136116646082795e-06,
+      "loss": 0.9348,
+      "step": 11381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.134514937668276e-06,
+      "loss": 0.8877,
+      "step": 11382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.132913278836213e-06,
+      "loss": 0.9845,
+      "step": 11383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131311669629187e-06,
+      "loss": 0.8736,
+      "step": 11384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12971011008976e-06,
+      "loss": 0.8603,
+      "step": 11385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.128108600260506e-06,
+      "loss": 0.9548,
+      "step": 11386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126507140183983e-06,
+      "loss": 0.9344,
+      "step": 11387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124905729902766e-06,
+      "loss": 0.9156,
+      "step": 11388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12330436945941e-06,
+      "loss": 0.9129,
+      "step": 11389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12170305889648e-06,
+      "loss": 0.9318,
+      "step": 11390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12010179825654e-06,
+      "loss": 0.9175,
+      "step": 11391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.118500587582143e-06,
+      "loss": 0.941,
+      "step": 11392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.116899426915855e-06,
+      "loss": 0.9992,
+      "step": 11393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.115298316300227e-06,
+      "loss": 0.9164,
+      "step": 11394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113697255777814e-06,
+      "loss": 0.9518,
+      "step": 11395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.112096245391181e-06,
+      "loss": 0.9589,
+      "step": 11396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11049528518287e-06,
+      "loss": 0.9619,
+      "step": 11397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.108894375195438e-06,
+      "loss": 1.0271,
+      "step": 11398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.107293515471433e-06,
+      "loss": 0.8921,
+      "step": 11399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105692706053403e-06,
+      "loss": 0.8853,
+      "step": 11400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1040919469839e-06,
+      "loss": 1.0173,
+      "step": 11401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.102491238305465e-06,
+      "loss": 0.8887,
+      "step": 11402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10089058006065e-06,
+      "loss": 0.8995,
+      "step": 11403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09928997229199e-06,
+      "loss": 0.9737,
+      "step": 11404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097689415042033e-06,
+      "loss": 0.8827,
+      "step": 11405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.096088908353316e-06,
+      "loss": 0.8749,
+      "step": 11406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.094488452268386e-06,
+      "loss": 1.0678,
+      "step": 11407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.092888046829772e-06,
+      "loss": 0.8471,
+      "step": 11408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09128769208002e-06,
+      "loss": 0.9149,
+      "step": 11409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.089687388061659e-06,
+      "loss": 0.8842,
+      "step": 11410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088087134817227e-06,
+      "loss": 0.9843,
+      "step": 11411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.086486932389257e-06,
+      "loss": 0.8947,
+      "step": 11412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084886780820273e-06,
+      "loss": 0.9473,
+      "step": 11413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083286680152816e-06,
+      "loss": 0.9584,
+      "step": 11414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.081686630429407e-06,
+      "loss": 0.9286,
+      "step": 11415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.08008663169258e-06,
+      "loss": 0.902,
+      "step": 11416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.078486683984857e-06,
+      "loss": 0.4316,
+      "step": 11417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076886787348758e-06,
+      "loss": 0.9378,
+      "step": 11418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.075286941826814e-06,
+      "loss": 0.9431,
+      "step": 11419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.073687147461548e-06,
+      "loss": 0.895,
+      "step": 11420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072087404295475e-06,
+      "loss": 0.8909,
+      "step": 11421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.070487712371119e-06,
+      "loss": 0.9532,
+      "step": 11422
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.068888071730994e-06,
+      "loss": 0.8657,
+      "step": 11423
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067288482417619e-06,
+      "loss": 0.9054,
+      "step": 11424
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.06568894447351e-06,
+      "loss": 0.923,
+      "step": 11425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064089457941178e-06,
+      "loss": 0.8928,
+      "step": 11426
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.06249002286314e-06,
+      "loss": 0.9156,
+      "step": 11427
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.0608906392819e-06,
+      "loss": 0.8228,
+      "step": 11428
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059291307239976e-06,
+      "loss": 0.9128,
+      "step": 11429
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057692026779867e-06,
+      "loss": 0.9068,
+      "step": 11430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.056092797944088e-06,
+      "loss": 0.8996,
+      "step": 11431
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054493620775146e-06,
+      "loss": 0.9589,
+      "step": 11432
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052894495315539e-06,
+      "loss": 0.4448,
+      "step": 11433
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051295421607773e-06,
+      "loss": 0.4435,
+      "step": 11434
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04969639969435e-06,
+      "loss": 0.8856,
+      "step": 11435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.048097429617767e-06,
+      "loss": 0.9091,
+      "step": 11436
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046498511420527e-06,
+      "loss": 0.9553,
+      "step": 11437
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044899645145125e-06,
+      "loss": 0.9605,
+      "step": 11438
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.043300830834058e-06,
+      "loss": 1.0318,
+      "step": 11439
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041702068529819e-06,
+      "loss": 0.9737,
+      "step": 11440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.040103358274902e-06,
+      "loss": 0.8257,
+      "step": 11441
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.038504700111796e-06,
+      "loss": 0.948,
+      "step": 11442
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036906094083001e-06,
+      "loss": 0.8515,
+      "step": 11443
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035307540230997e-06,
+      "loss": 0.957,
+      "step": 11444
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033709038598277e-06,
+      "loss": 0.9456,
+      "step": 11445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032110589227324e-06,
+      "loss": 0.928,
+      "step": 11446
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.030512192160622e-06,
+      "loss": 1.0575,
+      "step": 11447
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02891384744066e-06,
+      "loss": 0.8999,
+      "step": 11448
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.027315555109914e-06,
+      "loss": 0.9152,
+      "step": 11449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.025717315210868e-06,
+      "loss": 1.0089,
+      "step": 11450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024119127786001e-06,
+      "loss": 0.8708,
+      "step": 11451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.022520992877792e-06,
+      "loss": 0.8897,
+      "step": 11452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020922910528717e-06,
+      "loss": 0.9066,
+      "step": 11453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.019324880781246e-06,
+      "loss": 0.8352,
+      "step": 11454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.017726903677861e-06,
+      "loss": 0.8784,
+      "step": 11455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01612897926103e-06,
+      "loss": 0.9481,
+      "step": 11456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.014531107573226e-06,
+      "loss": 1.0298,
+      "step": 11457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012933288656919e-06,
+      "loss": 0.9866,
+      "step": 11458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.011335522554573e-06,
+      "loss": 0.9251,
+      "step": 11459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.009737809308658e-06,
+      "loss": 0.8609,
+      "step": 11460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.008140148961642e-06,
+      "loss": 1.0006,
+      "step": 11461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006542541555982e-06,
+      "loss": 0.8734,
+      "step": 11462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.004944987134147e-06,
+      "loss": 0.8835,
+      "step": 11463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003347485738594e-06,
+      "loss": 0.8998,
+      "step": 11464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.001750037411782e-06,
+      "loss": 0.8461,
+      "step": 11465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.000152642196175e-06,
+      "loss": 0.8658,
+      "step": 11466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.998555300134228e-06,
+      "loss": 0.9776,
+      "step": 11467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.996958011268393e-06,
+      "loss": 0.9446,
+      "step": 11468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.995360775641128e-06,
+      "loss": 0.8708,
+      "step": 11469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.993763593294882e-06,
+      "loss": 0.8784,
+      "step": 11470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99216646427211e-06,
+      "loss": 0.9155,
+      "step": 11471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99056938861526e-06,
+      "loss": 0.914,
+      "step": 11472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988972366366782e-06,
+      "loss": 0.8846,
+      "step": 11473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.987375397569121e-06,
+      "loss": 0.8971,
+      "step": 11474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985778482264721e-06,
+      "loss": 0.8993,
+      "step": 11475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98418162049603e-06,
+      "loss": 0.8925,
+      "step": 11476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.982584812305488e-06,
+      "loss": 0.8848,
+      "step": 11477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980988057735537e-06,
+      "loss": 0.8944,
+      "step": 11478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97939135682862e-06,
+      "loss": 0.8655,
+      "step": 11479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.977794709627171e-06,
+      "loss": 0.9529,
+      "step": 11480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.976198116173633e-06,
+      "loss": 0.8598,
+      "step": 11481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.974601576510435e-06,
+      "loss": 0.9037,
+      "step": 11482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.973005090680013e-06,
+      "loss": 1.0387,
+      "step": 11483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.971408658724802e-06,
+      "loss": 0.9284,
+      "step": 11484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969812280687231e-06,
+      "loss": 0.9126,
+      "step": 11485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.968215956609733e-06,
+      "loss": 0.9225,
+      "step": 11486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.966619686534733e-06,
+      "loss": 0.884,
+      "step": 11487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965023470504658e-06,
+      "loss": 0.8182,
+      "step": 11488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963427308561934e-06,
+      "loss": 0.9423,
+      "step": 11489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.961831200748989e-06,
+      "loss": 0.8645,
+      "step": 11490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96023514710824e-06,
+      "loss": 0.8651,
+      "step": 11491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.958639147682114e-06,
+      "loss": 0.8182,
+      "step": 11492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957043202513026e-06,
+      "loss": 0.8069,
+      "step": 11493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9554473116434e-06,
+      "loss": 0.8781,
+      "step": 11494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.953851475115647e-06,
+      "loss": 0.903,
+      "step": 11495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952255692972184e-06,
+      "loss": 0.4276,
+      "step": 11496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.950659965255429e-06,
+      "loss": 0.8786,
+      "step": 11497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.949064292007788e-06,
+      "loss": 0.8785,
+      "step": 11498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.947468673271677e-06,
+      "loss": 0.9362,
+      "step": 11499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.945873109089503e-06,
+      "loss": 1.0168,
+      "step": 11500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944277599503675e-06,
+      "loss": 0.9019,
+      "step": 11501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942682144556605e-06,
+      "loss": 0.8848,
+      "step": 11502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.941086744290691e-06,
+      "loss": 0.9209,
+      "step": 11503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93949139874834e-06,
+      "loss": 0.8301,
+      "step": 11504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937896107971957e-06,
+      "loss": 0.9145,
+      "step": 11505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.936300872003937e-06,
+      "loss": 0.9168,
+      "step": 11506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.934705690886683e-06,
+      "loss": 0.959,
+      "step": 11507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933110564662594e-06,
+      "loss": 0.8906,
+      "step": 11508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931515493374067e-06,
+      "loss": 0.9694,
+      "step": 11509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929920477063496e-06,
+      "loss": 0.9177,
+      "step": 11510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.928325515773271e-06,
+      "loss": 0.8751,
+      "step": 11511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92673060954579e-06,
+      "loss": 0.9283,
+      "step": 11512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925135758423439e-06,
+      "loss": 0.9773,
+      "step": 11513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.923540962448609e-06,
+      "loss": 0.8905,
+      "step": 11514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.921946221663693e-06,
+      "loss": 0.9581,
+      "step": 11515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.920351536111073e-06,
+      "loss": 0.7877,
+      "step": 11516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.918756905833132e-06,
+      "loss": 0.8253,
+      "step": 11517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.917162330872257e-06,
+      "loss": 0.8804,
+      "step": 11518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.915567811270826e-06,
+      "loss": 0.9935,
+      "step": 11519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.913973347071224e-06,
+      "loss": 0.9153,
+      "step": 11520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.912378938315825e-06,
+      "loss": 0.9845,
+      "step": 11521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910784585047014e-06,
+      "loss": 0.8816,
+      "step": 11522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.909190287307162e-06,
+      "loss": 0.8664,
+      "step": 11523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90759604513864e-06,
+      "loss": 0.911,
+      "step": 11524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.906001858583825e-06,
+      "loss": 0.8709,
+      "step": 11525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.904407727685093e-06,
+      "loss": 0.9735,
+      "step": 11526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90281365248481e-06,
+      "loss": 0.9021,
+      "step": 11527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.901219633025343e-06,
+      "loss": 0.8828,
+      "step": 11528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899625669349062e-06,
+      "loss": 0.9083,
+      "step": 11529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.898031761498333e-06,
+      "loss": 0.8596,
+      "step": 11530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89643790951552e-06,
+      "loss": 0.9933,
+      "step": 11531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.894844113442982e-06,
+      "loss": 0.9132,
+      "step": 11532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893250373323084e-06,
+      "loss": 0.8396,
+      "step": 11533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891656689198184e-06,
+      "loss": 0.8213,
+      "step": 11534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890063061110645e-06,
+      "loss": 0.9043,
+      "step": 11535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.888469489102813e-06,
+      "loss": 0.9103,
+      "step": 11536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.886875973217057e-06,
+      "loss": 0.8847,
+      "step": 11537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88528251349572e-06,
+      "loss": 0.964,
+      "step": 11538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.883689109981162e-06,
+      "loss": 0.9943,
+      "step": 11539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882095762715727e-06,
+      "loss": 0.9713,
+      "step": 11540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880502471741772e-06,
+      "loss": 0.919,
+      "step": 11541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878909237101637e-06,
+      "loss": 0.8576,
+      "step": 11542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.877316058837674e-06,
+      "loss": 0.892,
+      "step": 11543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.875722936992225e-06,
+      "loss": 0.9025,
+      "step": 11544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.874129871607634e-06,
+      "loss": 0.9368,
+      "step": 11545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872536862726245e-06,
+      "loss": 0.9515,
+      "step": 11546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870943910390392e-06,
+      "loss": 0.9553,
+      "step": 11547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.869351014642419e-06,
+      "loss": 0.8584,
+      "step": 11548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.867758175524666e-06,
+      "loss": 0.896,
+      "step": 11549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866165393079463e-06,
+      "loss": 1.0095,
+      "step": 11550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.864572667349149e-06,
+      "loss": 0.8876,
+      "step": 11551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862979998376054e-06,
+      "loss": 0.8811,
+      "step": 11552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861387386202507e-06,
+      "loss": 0.8845,
+      "step": 11553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.859794830870845e-06,
+      "loss": 0.9294,
+      "step": 11554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858202332423389e-06,
+      "loss": 0.9343,
+      "step": 11555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85660989090247e-06,
+      "loss": 0.9164,
+      "step": 11556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.855017506350412e-06,
+      "loss": 0.9852,
+      "step": 11557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853425178809541e-06,
+      "loss": 0.9569,
+      "step": 11558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.851832908322175e-06,
+      "loss": 0.8801,
+      "step": 11559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850240694930634e-06,
+      "loss": 0.9281,
+      "step": 11560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.848648538677241e-06,
+      "loss": 0.9663,
+      "step": 11561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.847056439604317e-06,
+      "loss": 0.96,
+      "step": 11562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84546439775417e-06,
+      "loss": 0.9257,
+      "step": 11563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84387241316912e-06,
+      "loss": 0.8182,
+      "step": 11564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.842280485891482e-06,
+      "loss": 0.8931,
+      "step": 11565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84068861596356e-06,
+      "loss": 0.9366,
+      "step": 11566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83909680342767e-06,
+      "loss": 0.9321,
+      "step": 11567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.837505048326119e-06,
+      "loss": 0.9375,
+      "step": 11568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.835913350701215e-06,
+      "loss": 1.0154,
+      "step": 11569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83432171059526e-06,
+      "loss": 0.937,
+      "step": 11570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832730128050562e-06,
+      "loss": 0.8479,
+      "step": 11571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.831138603109418e-06,
+      "loss": 0.9967,
+      "step": 11572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829547135814139e-06,
+      "loss": 0.9731,
+      "step": 11573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.827955726207013e-06,
+      "loss": 0.9423,
+      "step": 11574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826364374330348e-06,
+      "loss": 0.8052,
+      "step": 11575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824773080226432e-06,
+      "loss": 0.9173,
+      "step": 11576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.823181843937567e-06,
+      "loss": 0.9727,
+      "step": 11577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821590665506039e-06,
+      "loss": 0.9161,
+      "step": 11578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.819999544974144e-06,
+      "loss": 0.8926,
+      "step": 11579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.818408482384172e-06,
+      "loss": 0.8676,
+      "step": 11580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81681747777841e-06,
+      "loss": 0.8029,
+      "step": 11581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.815226531199146e-06,
+      "loss": 0.9425,
+      "step": 11582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.813635642688663e-06,
+      "loss": 0.8772,
+      "step": 11583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81204481228925e-06,
+      "loss": 0.9507,
+      "step": 11584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810454040043188e-06,
+      "loss": 0.922,
+      "step": 11585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808863325992757e-06,
+      "loss": 0.9199,
+      "step": 11586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807272670180234e-06,
+      "loss": 0.8554,
+      "step": 11587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.805682072647903e-06,
+      "loss": 0.4402,
+      "step": 11588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804091533438032e-06,
+      "loss": 0.8476,
+      "step": 11589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802501052592904e-06,
+      "loss": 0.8946,
+      "step": 11590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800910630154785e-06,
+      "loss": 0.8638,
+      "step": 11591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799320266165951e-06,
+      "loss": 0.9751,
+      "step": 11592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.797729960668673e-06,
+      "loss": 0.8517,
+      "step": 11593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.796139713705214e-06,
+      "loss": 0.9627,
+      "step": 11594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.794549525317842e-06,
+      "loss": 0.9923,
+      "step": 11595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79295939554883e-06,
+      "loss": 0.8432,
+      "step": 11596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791369324440434e-06,
+      "loss": 0.8546,
+      "step": 11597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78977931203492e-06,
+      "loss": 0.9185,
+      "step": 11598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.788189358374547e-06,
+      "loss": 0.8032,
+      "step": 11599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.786599463501577e-06,
+      "loss": 0.9107,
+      "step": 11600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785009627458264e-06,
+      "loss": 0.9034,
+      "step": 11601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.783419850286864e-06,
+      "loss": 1.0364,
+      "step": 11602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.781830132029635e-06,
+      "loss": 0.9041,
+      "step": 11603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780240472728827e-06,
+      "loss": 0.8232,
+      "step": 11604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.778650872426692e-06,
+      "loss": 0.9206,
+      "step": 11605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77706133116548e-06,
+      "loss": 0.8113,
+      "step": 11606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.775471848987434e-06,
+      "loss": 0.8422,
+      "step": 11607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77388242593481e-06,
+      "loss": 0.9729,
+      "step": 11608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77229306204985e-06,
+      "loss": 0.9218,
+      "step": 11609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770703757374793e-06,
+      "loss": 1.0211,
+      "step": 11610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.769114511951887e-06,
+      "loss": 0.9229,
+      "step": 11611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.767525325823366e-06,
+      "loss": 0.9526,
+      "step": 11612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.765936199031475e-06,
+      "loss": 0.8276,
+      "step": 11613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.764347131618449e-06,
+      "loss": 0.87,
+      "step": 11614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76275812362652e-06,
+      "loss": 0.898,
+      "step": 11615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.761169175097927e-06,
+      "loss": 0.9271,
+      "step": 11616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.7595802860749e-06,
+      "loss": 0.9316,
+      "step": 11617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75799145659967e-06,
+      "loss": 0.8721,
+      "step": 11618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756402686714462e-06,
+      "loss": 0.8007,
+      "step": 11619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.754813976461512e-06,
+      "loss": 0.8057,
+      "step": 11620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.753225325883044e-06,
+      "loss": 0.8535,
+      "step": 11621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751636735021279e-06,
+      "loss": 0.8756,
+      "step": 11622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750048203918442e-06,
+      "loss": 0.861,
+      "step": 11623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.748459732616755e-06,
+      "loss": 0.9005,
+      "step": 11624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746871321158435e-06,
+      "loss": 0.8196,
+      "step": 11625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745282969585704e-06,
+      "loss": 0.8496,
+      "step": 11626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.743694677940773e-06,
+      "loss": 0.914,
+      "step": 11627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742106446265864e-06,
+      "loss": 0.8478,
+      "step": 11628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.740518274603188e-06,
+      "loss": 0.8927,
+      "step": 11629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738930162994953e-06,
+      "loss": 0.8857,
+      "step": 11630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.737342111483367e-06,
+      "loss": 0.9504,
+      "step": 11631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.735754120110653e-06,
+      "loss": 0.8814,
+      "step": 11632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.734166188919003e-06,
+      "loss": 0.9197,
+      "step": 11633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73257831795063e-06,
+      "loss": 0.9356,
+      "step": 11634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.730990507247736e-06,
+      "loss": 0.9645,
+      "step": 11635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72940275685252e-06,
+      "loss": 0.8386,
+      "step": 11636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.727815066807188e-06,
+      "loss": 0.9025,
+      "step": 11637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726227437153934e-06,
+      "loss": 0.8713,
+      "step": 11638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724639867934961e-06,
+      "loss": 0.923,
+      "step": 11639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.723052359192459e-06,
+      "loss": 0.8997,
+      "step": 11640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721464910968628e-06,
+      "loss": 0.8891,
+      "step": 11641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.719877523305656e-06,
+      "loss": 0.8786,
+      "step": 11642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718290196245731e-06,
+      "loss": 0.981,
+      "step": 11643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716702929831049e-06,
+      "loss": 0.8864,
+      "step": 11644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715115724103797e-06,
+      "loss": 0.9528,
+      "step": 11645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713528579106158e-06,
+      "loss": 0.8783,
+      "step": 11646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71194149488032e-06,
+      "loss": 0.8739,
+      "step": 11647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.710354471468461e-06,
+      "loss": 0.8692,
+      "step": 11648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.708767508912768e-06,
+      "loss": 0.8716,
+      "step": 11649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707180607255418e-06,
+      "loss": 0.923,
+      "step": 11650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705593766538587e-06,
+      "loss": 0.9585,
+      "step": 11651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.704006986804453e-06,
+      "loss": 0.9427,
+      "step": 11652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702420268095192e-06,
+      "loss": 0.9011,
+      "step": 11653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.700833610452972e-06,
+      "loss": 0.4509,
+      "step": 11654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.699247013919974e-06,
+      "loss": 1.0252,
+      "step": 11655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697660478538363e-06,
+      "loss": 0.851,
+      "step": 11656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.696074004350303e-06,
+      "loss": 1.0102,
+      "step": 11657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694487591397968e-06,
+      "loss": 0.9299,
+      "step": 11658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.692901239723517e-06,
+      "loss": 0.9146,
+      "step": 11659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.691314949369119e-06,
+      "loss": 0.9495,
+      "step": 11660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68972872037693e-06,
+      "loss": 0.873,
+      "step": 11661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688142552789116e-06,
+      "loss": 0.8938,
+      "step": 11662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686556446647832e-06,
+      "loss": 0.8455,
+      "step": 11663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.684970401995233e-06,
+      "loss": 0.4643,
+      "step": 11664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683384418873479e-06,
+      "loss": 0.9876,
+      "step": 11665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.681798497324717e-06,
+      "loss": 0.8995,
+      "step": 11666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.680212637391106e-06,
+      "loss": 0.8973,
+      "step": 11667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.678626839114796e-06,
+      "loss": 0.9556,
+      "step": 11668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67704110253793e-06,
+      "loss": 0.8394,
+      "step": 11669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.675455427702661e-06,
+      "loss": 0.9171,
+      "step": 11670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673869814651133e-06,
+      "loss": 0.8904,
+      "step": 11671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.672284263425485e-06,
+      "loss": 0.8847,
+      "step": 11672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.670698774067866e-06,
+      "loss": 0.9476,
+      "step": 11673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66911334662041e-06,
+      "loss": 0.8853,
+      "step": 11674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667527981125262e-06,
+      "loss": 0.8854,
+      "step": 11675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.665942677624554e-06,
+      "loss": 0.427,
+      "step": 11676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664357436160427e-06,
+      "loss": 0.9285,
+      "step": 11677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662772256775006e-06,
+      "loss": 1.0215,
+      "step": 11678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.661187139510434e-06,
+      "loss": 0.8887,
+      "step": 11679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659602084408835e-06,
+      "loss": 0.9148,
+      "step": 11680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65801709151234e-06,
+      "loss": 0.9378,
+      "step": 11681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656432160863077e-06,
+      "loss": 0.9289,
+      "step": 11682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.654847292503169e-06,
+      "loss": 1.003,
+      "step": 11683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653262486474745e-06,
+      "loss": 0.8976,
+      "step": 11684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65167774281992e-06,
+      "loss": 1.0356,
+      "step": 11685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.650093061580819e-06,
+      "loss": 0.8541,
+      "step": 11686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64850844279956e-06,
+      "loss": 0.8245,
+      "step": 11687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.646923886518263e-06,
+      "loss": 0.8564,
+      "step": 11688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645339392779042e-06,
+      "loss": 0.879,
+      "step": 11689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643754961624005e-06,
+      "loss": 0.9878,
+      "step": 11690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.642170593095276e-06,
+      "loss": 0.8612,
+      "step": 11691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.640586287234959e-06,
+      "loss": 1.0164,
+      "step": 11692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639002044085161e-06,
+      "loss": 0.9142,
+      "step": 11693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.637417863687994e-06,
+      "loss": 0.967,
+      "step": 11694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635833746085562e-06,
+      "loss": 0.9113,
+      "step": 11695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63424969131997e-06,
+      "loss": 0.8836,
+      "step": 11696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.632665699433317e-06,
+      "loss": 0.8903,
+      "step": 11697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.631081770467708e-06,
+      "loss": 0.8745,
+      "step": 11698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.62949790446524e-06,
+      "loss": 0.9418,
+      "step": 11699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.627914101468008e-06,
+      "loss": 0.9633,
+      "step": 11700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.626330361518113e-06,
+      "loss": 0.9291,
+      "step": 11701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.624746684657641e-06,
+      "loss": 0.9841,
+      "step": 11702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.623163070928692e-06,
+      "loss": 1.0063,
+      "step": 11703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.621579520373355e-06,
+      "loss": 0.894,
+      "step": 11704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.619996033033719e-06,
+      "loss": 0.9497,
+      "step": 11705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6184126089518674e-06,
+      "loss": 0.8975,
+      "step": 11706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.616829248169891e-06,
+      "loss": 0.9406,
+      "step": 11707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615245950729869e-06,
+      "loss": 0.8894,
+      "step": 11708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.613662716673889e-06,
+      "loss": 0.9406,
+      "step": 11709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612079546044026e-06,
+      "loss": 0.8423,
+      "step": 11710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610496438882362e-06,
+      "loss": 0.9295,
+      "step": 11711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608913395230975e-06,
+      "loss": 0.9549,
+      "step": 11712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.607330415131935e-06,
+      "loss": 0.8521,
+      "step": 11713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605747498627322e-06,
+      "loss": 0.934,
+      "step": 11714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.604164645759208e-06,
+      "loss": 0.8181,
+      "step": 11715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602581856569659e-06,
+      "loss": 0.84,
+      "step": 11716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600999131100748e-06,
+      "loss": 0.9224,
+      "step": 11717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599416469394539e-06,
+      "loss": 0.4172,
+      "step": 11718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5978338714931e-06,
+      "loss": 0.9035,
+      "step": 11719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5962513374384924e-06,
+      "loss": 0.4531,
+      "step": 11720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.594668867272779e-06,
+      "loss": 0.9341,
+      "step": 11721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593086461038019e-06,
+      "loss": 0.9321,
+      "step": 11722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591504118776273e-06,
+      "loss": 0.9913,
+      "step": 11723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589921840529597e-06,
+      "loss": 0.8807,
+      "step": 11724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5883396263400424e-06,
+      "loss": 0.9142,
+      "step": 11725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586757476249667e-06,
+      "loss": 0.9173,
+      "step": 11726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.585175390300525e-06,
+      "loss": 0.8887,
+      "step": 11727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.583593368534663e-06,
+      "loss": 0.8788,
+      "step": 11728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582011410994127e-06,
+      "loss": 0.9866,
+      "step": 11729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.580429517720969e-06,
+      "loss": 0.8447,
+      "step": 11730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5788476887572275e-06,
+      "loss": 0.9031,
+      "step": 11731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577265924144952e-06,
+      "loss": 0.8861,
+      "step": 11732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.575684223926182e-06,
+      "loss": 1.0034,
+      "step": 11733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5741025881429534e-06,
+      "loss": 0.9414,
+      "step": 11734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.57252101683731e-06,
+      "loss": 0.9458,
+      "step": 11735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.570939510051283e-06,
+      "loss": 0.9818,
+      "step": 11736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.569358067826908e-06,
+      "loss": 0.9624,
+      "step": 11737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5677766902062234e-06,
+      "loss": 0.8483,
+      "step": 11738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.566195377231254e-06,
+      "loss": 0.9417,
+      "step": 11739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.564614128944034e-06,
+      "loss": 0.8885,
+      "step": 11740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563032945386589e-06,
+      "loss": 0.8875,
+      "step": 11741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561451826600943e-06,
+      "loss": 0.8861,
+      "step": 11742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559870772629126e-06,
+      "loss": 0.9053,
+      "step": 11743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.558289783513153e-06,
+      "loss": 0.9796,
+      "step": 11744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.556708859295051e-06,
+      "loss": 0.4372,
+      "step": 11745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555128000016836e-06,
+      "loss": 0.8686,
+      "step": 11746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5535472057205275e-06,
+      "loss": 0.9356,
+      "step": 11747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.55196647644814e-06,
+      "loss": 0.8948,
+      "step": 11748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.550385812241684e-06,
+      "loss": 0.8732,
+      "step": 11749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.548805213143178e-06,
+      "loss": 0.9108,
+      "step": 11750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547224679194631e-06,
+      "loss": 1.0079,
+      "step": 11751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.545644210438049e-06,
+      "loss": 0.8682,
+      "step": 11752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544063806915442e-06,
+      "loss": 0.9507,
+      "step": 11753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542483468668816e-06,
+      "loss": 0.9689,
+      "step": 11754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.54090319574017e-06,
+      "loss": 0.9377,
+      "step": 11755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5393229881715105e-06,
+      "loss": 1.001,
+      "step": 11756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537742846004832e-06,
+      "loss": 0.8686,
+      "step": 11757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.536162769282141e-06,
+      "loss": 0.9424,
+      "step": 11758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.534582758045426e-06,
+      "loss": 0.8863,
+      "step": 11759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.533002812336687e-06,
+      "loss": 0.8564,
+      "step": 11760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.531422932197913e-06,
+      "loss": 0.9367,
+      "step": 11761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.529843117671101e-06,
+      "loss": 0.9182,
+      "step": 11762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.528263368798237e-06,
+      "loss": 0.8857,
+      "step": 11763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.52668368562131e-06,
+      "loss": 0.9427,
+      "step": 11764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.525104068182305e-06,
+      "loss": 0.8773,
+      "step": 11765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.523524516523209e-06,
+      "loss": 0.8887,
+      "step": 11766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521945030686002e-06,
+      "loss": 0.8687,
+      "step": 11767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.520365610712667e-06,
+      "loss": 0.9543,
+      "step": 11768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.518786256645181e-06,
+      "loss": 0.9099,
+      "step": 11769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.517206968525521e-06,
+      "loss": 0.9134,
+      "step": 11770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515627746395667e-06,
+      "loss": 0.8512,
+      "step": 11771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514048590297585e-06,
+      "loss": 0.8491,
+      "step": 11772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.512469500273256e-06,
+      "loss": 0.9136,
+      "step": 11773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.510890476364647e-06,
+      "loss": 0.9543,
+      "step": 11774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509311518613724e-06,
+      "loss": 0.9162,
+      "step": 11775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.507732627062461e-06,
+      "loss": 0.9167,
+      "step": 11776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.506153801752816e-06,
+      "loss": 0.9568,
+      "step": 11777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5045750427267526e-06,
+      "loss": 0.8893,
+      "step": 11778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.502996350026237e-06,
+      "loss": 0.9251,
+      "step": 11779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.501417723693225e-06,
+      "loss": 0.9083,
+      "step": 11780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499839163769677e-06,
+      "loss": 0.958,
+      "step": 11781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4982606702975505e-06,
+      "loss": 0.8043,
+      "step": 11782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496682243318795e-06,
+      "loss": 1.0011,
+      "step": 11783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.495103882875363e-06,
+      "loss": 0.4064,
+      "step": 11784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.493525589009215e-06,
+      "loss": 0.9269,
+      "step": 11785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491947361762292e-06,
+      "loss": 1.0081,
+      "step": 11786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490369201176545e-06,
+      "loss": 0.8663,
+      "step": 11787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488791107293916e-06,
+      "loss": 1.0011,
+      "step": 11788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.487213080156355e-06,
+      "loss": 0.8387,
+      "step": 11789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4856351198057985e-06,
+      "loss": 0.8542,
+      "step": 11790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.484057226284187e-06,
+      "loss": 0.9085,
+      "step": 11791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.482479399633464e-06,
+      "loss": 0.9261,
+      "step": 11792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4809016398955615e-06,
+      "loss": 0.8742,
+      "step": 11793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479323947112418e-06,
+      "loss": 0.8655,
+      "step": 11794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.477746321325963e-06,
+      "loss": 0.8851,
+      "step": 11795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476168762578127e-06,
+      "loss": 0.8232,
+      "step": 11796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.474591270910847e-06,
+      "loss": 0.8254,
+      "step": 11797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4730138463660486e-06,
+      "loss": 0.9214,
+      "step": 11798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471436488985653e-06,
+      "loss": 0.9435,
+      "step": 11799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.469859198811589e-06,
+      "loss": 0.9733,
+      "step": 11800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.468281975885776e-06,
+      "loss": 0.8442,
+      "step": 11801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466704820250139e-06,
+      "loss": 0.8824,
+      "step": 11802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.465127731946594e-06,
+      "loss": 0.9535,
+      "step": 11803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463550711017058e-06,
+      "loss": 0.4323,
+      "step": 11804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461973757503448e-06,
+      "loss": 0.9492,
+      "step": 11805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.460396871447675e-06,
+      "loss": 0.9852,
+      "step": 11806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4588200528916554e-06,
+      "loss": 0.9324,
+      "step": 11807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457243301877291e-06,
+      "loss": 0.9238,
+      "step": 11808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4556666184464985e-06,
+      "loss": 1.0563,
+      "step": 11809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.454090002641183e-06,
+      "loss": 0.9269,
+      "step": 11810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452513454503249e-06,
+      "loss": 0.8936,
+      "step": 11811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450936974074594e-06,
+      "loss": 0.939,
+      "step": 11812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.449360561397125e-06,
+      "loss": 0.4435,
+      "step": 11813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447784216512739e-06,
+      "loss": 0.8866,
+      "step": 11814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.446207939463335e-06,
+      "loss": 0.9112,
+      "step": 11815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.444631730290805e-06,
+      "loss": 0.8918,
+      "step": 11816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443055589037046e-06,
+      "loss": 0.965,
+      "step": 11817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.44147951574395e-06,
+      "loss": 0.9037,
+      "step": 11818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439903510453405e-06,
+      "loss": 0.9596,
+      "step": 11819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.438327573207298e-06,
+      "loss": 0.9472,
+      "step": 11820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.436751704047523e-06,
+      "loss": 0.9118,
+      "step": 11821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435175903015956e-06,
+      "loss": 0.9223,
+      "step": 11822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.433600170154489e-06,
+      "loss": 0.8955,
+      "step": 11823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4320245055049955e-06,
+      "loss": 0.9827,
+      "step": 11824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.430448909109359e-06,
+      "loss": 0.8977,
+      "step": 11825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.428873381009456e-06,
+      "loss": 0.9283,
+      "step": 11826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42729792124716e-06,
+      "loss": 0.8889,
+      "step": 11827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4257225298643495e-06,
+      "loss": 0.8927,
+      "step": 11828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4241472069028915e-06,
+      "loss": 0.8969,
+      "step": 11829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.422571952404662e-06,
+      "loss": 0.9822,
+      "step": 11830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420996766411521e-06,
+      "loss": 0.9689,
+      "step": 11831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.419421648965345e-06,
+      "loss": 0.8714,
+      "step": 11832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.417846600107992e-06,
+      "loss": 0.8321,
+      "step": 11833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41627161988133e-06,
+      "loss": 1.021,
+      "step": 11834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414696708327214e-06,
+      "loss": 0.8619,
+      "step": 11835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.413121865487509e-06,
+      "loss": 0.8879,
+      "step": 11836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41154709140407e-06,
+      "loss": 0.8669,
+      "step": 11837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.409972386118753e-06,
+      "loss": 1.011,
+      "step": 11838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4083977496734135e-06,
+      "loss": 0.8558,
+      "step": 11839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.406823182109899e-06,
+      "loss": 0.8339,
+      "step": 11840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4052486834700655e-06,
+      "loss": 0.872,
+      "step": 11841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4036742537957566e-06,
+      "loss": 0.9013,
+      "step": 11842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.402099893128818e-06,
+      "loss": 0.9222,
+      "step": 11843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4005256015111e-06,
+      "loss": 0.8267,
+      "step": 11844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398951378984442e-06,
+      "loss": 0.9202,
+      "step": 11845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.397377225590687e-06,
+      "loss": 0.8078,
+      "step": 11846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3958031413716745e-06,
+      "loss": 0.4528,
+      "step": 11847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.394229126369236e-06,
+      "loss": 0.9378,
+      "step": 11848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.392655180625215e-06,
+      "loss": 0.9498,
+      "step": 11849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.39108130418144e-06,
+      "loss": 0.4704,
+      "step": 11850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3895074970797445e-06,
+      "loss": 0.9394,
+      "step": 11851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3879337593619586e-06,
+      "loss": 0.9437,
+      "step": 11852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3863600910699085e-06,
+      "loss": 0.9147,
+      "step": 11853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384786492245424e-06,
+      "loss": 1.0302,
+      "step": 11854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383212962930324e-06,
+      "loss": 0.919,
+      "step": 11855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381639503166436e-06,
+      "loss": 0.8299,
+      "step": 11856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.380066112995581e-06,
+      "loss": 0.4451,
+      "step": 11857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.378492792459576e-06,
+      "loss": 0.9317,
+      "step": 11858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.376919541600241e-06,
+      "loss": 0.9175,
+      "step": 11859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.375346360459387e-06,
+      "loss": 0.8286,
+      "step": 11860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.373773249078827e-06,
+      "loss": 0.4665,
+      "step": 11861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.372200207500377e-06,
+      "loss": 0.8587,
+      "step": 11862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.370627235765844e-06,
+      "loss": 0.9363,
+      "step": 11863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3690543339170365e-06,
+      "loss": 0.8841,
+      "step": 11864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.367481501995758e-06,
+      "loss": 0.991,
+      "step": 11865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365908740043817e-06,
+      "loss": 0.9247,
+      "step": 11866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.364336048103009e-06,
+      "loss": 0.8779,
+      "step": 11867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.362763426215142e-06,
+      "loss": 0.9546,
+      "step": 11868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.361190874422009e-06,
+      "loss": 0.9716,
+      "step": 11869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.359618392765413e-06,
+      "loss": 0.87,
+      "step": 11870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.358045981287141e-06,
+      "loss": 0.9463,
+      "step": 11871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3564736400289904e-06,
+      "loss": 0.8879,
+      "step": 11872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3549013690327495e-06,
+      "loss": 0.9158,
+      "step": 11873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.353329168340213e-06,
+      "loss": 0.8985,
+      "step": 11874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3517570379931635e-06,
+      "loss": 0.9379,
+      "step": 11875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.350184978033386e-06,
+      "loss": 0.8937,
+      "step": 11876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.348612988502666e-06,
+      "loss": 0.9485,
+      "step": 11877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347041069442783e-06,
+      "loss": 0.9431,
+      "step": 11878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345469220895516e-06,
+      "loss": 0.8739,
+      "step": 11879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.34389744290265e-06,
+      "loss": 0.8706,
+      "step": 11880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.342325735505957e-06,
+      "loss": 0.917,
+      "step": 11881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.340754098747207e-06,
+      "loss": 0.9019,
+      "step": 11882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33918253266818e-06,
+      "loss": 0.8438,
+      "step": 11883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.337611037310639e-06,
+      "loss": 1.0248,
+      "step": 11884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.336039612716358e-06,
+      "loss": 0.9515,
+      "step": 11885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3344682589271e-06,
+      "loss": 0.938,
+      "step": 11886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332896975984634e-06,
+      "loss": 0.9162,
+      "step": 11887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.331325763930719e-06,
+      "loss": 0.8745,
+      "step": 11888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3297546228071144e-06,
+      "loss": 0.8724,
+      "step": 11889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3281835526555815e-06,
+      "loss": 0.8066,
+      "step": 11890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326612553517882e-06,
+      "loss": 0.9441,
+      "step": 11891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.325041625435766e-06,
+      "loss": 0.9742,
+      "step": 11892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.32347076845099e-06,
+      "loss": 0.8689,
+      "step": 11893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.321899982605303e-06,
+      "loss": 0.9917,
+      "step": 11894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.320329267940456e-06,
+      "loss": 0.9288,
+      "step": 11895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318758624498198e-06,
+      "loss": 0.9477,
+      "step": 11896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317188052320271e-06,
+      "loss": 0.9713,
+      "step": 11897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.315617551448424e-06,
+      "loss": 0.916,
+      "step": 11898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.314047121924393e-06,
+      "loss": 0.8593,
+      "step": 11899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.312476763789926e-06,
+      "loss": 0.8633,
+      "step": 11900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.310906477086757e-06,
+      "loss": 0.9226,
+      "step": 11901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.309336261856617e-06,
+      "loss": 0.952,
+      "step": 11902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3077661181412485e-06,
+      "loss": 0.873,
+      "step": 11903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.306196045982386e-06,
+      "loss": 0.978,
+      "step": 11904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.304626045421751e-06,
+      "loss": 0.8605,
+      "step": 11905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.30305611650108e-06,
+      "loss": 0.9019,
+      "step": 11906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.301486259262096e-06,
+      "loss": 0.9116,
+      "step": 11907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.299916473746527e-06,
+      "loss": 0.9665,
+      "step": 11908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.298346759996095e-06,
+      "loss": 0.934,
+      "step": 11909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.296777118052517e-06,
+      "loss": 0.9451,
+      "step": 11910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.295207547957519e-06,
+      "loss": 0.9091,
+      "step": 11911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293638049752813e-06,
+      "loss": 0.9146,
+      "step": 11912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.292068623480119e-06,
+      "loss": 0.9195,
+      "step": 11913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.290499269181144e-06,
+      "loss": 0.8768,
+      "step": 11914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.288929986897604e-06,
+      "loss": 0.9225,
+      "step": 11915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2873607766712125e-06,
+      "loss": 0.93,
+      "step": 11916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285791638543673e-06,
+      "loss": 0.8719,
+      "step": 11917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284222572556691e-06,
+      "loss": 0.873,
+      "step": 11918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.282653578751973e-06,
+      "loss": 0.9621,
+      "step": 11919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281084657171216e-06,
+      "loss": 0.9271,
+      "step": 11920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.279515807856126e-06,
+      "loss": 0.9896,
+      "step": 11921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.277947030848397e-06,
+      "loss": 0.7951,
+      "step": 11922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.276378326189729e-06,
+      "loss": 0.9148,
+      "step": 11923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.274809693921814e-06,
+      "loss": 0.9638,
+      "step": 11924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2732411340863444e-06,
+      "loss": 0.9287,
+      "step": 11925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.271672646725008e-06,
+      "loss": 0.864,
+      "step": 11926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2701042318795e-06,
+      "loss": 0.7913,
+      "step": 11927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.268535889591502e-06,
+      "loss": 0.9158,
+      "step": 11928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266967619902703e-06,
+      "loss": 0.9293,
+      "step": 11929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.265399422854783e-06,
+      "loss": 0.8654,
+      "step": 11930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.263831298489421e-06,
+      "loss": 0.9583,
+      "step": 11931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2622632468483e-06,
+      "loss": 0.9225,
+      "step": 11932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260695267973093e-06,
+      "loss": 0.969,
+      "step": 11933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.25912736190548e-06,
+      "loss": 0.9555,
+      "step": 11934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2575595286871296e-06,
+      "loss": 0.9145,
+      "step": 11935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.255991768359716e-06,
+      "loss": 0.8905,
+      "step": 11936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2544240809649065e-06,
+      "loss": 0.917,
+      "step": 11937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.252856466544367e-06,
+      "loss": 0.9638,
+      "step": 11938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251288925139765e-06,
+      "loss": 0.903,
+      "step": 11939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249721456792766e-06,
+      "loss": 1.0046,
+      "step": 11940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.24815406154503e-06,
+      "loss": 0.9433,
+      "step": 11941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.246586739438217e-06,
+      "loss": 0.827,
+      "step": 11942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.245019490513981e-06,
+      "loss": 0.9719,
+      "step": 11943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243452314813983e-06,
+      "loss": 0.8453,
+      "step": 11944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.241885212379874e-06,
+      "loss": 0.9815,
+      "step": 11945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.240318183253304e-06,
+      "loss": 0.9296,
+      "step": 11946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.238751227475928e-06,
+      "loss": 0.4173,
+      "step": 11947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.237184345089388e-06,
+      "loss": 0.8552,
+      "step": 11948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.235617536135333e-06,
+      "loss": 0.8396,
+      "step": 11949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234050800655405e-06,
+      "loss": 0.9385,
+      "step": 11950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2324841386912505e-06,
+      "loss": 0.8515,
+      "step": 11951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230917550284504e-06,
+      "loss": 0.8779,
+      "step": 11952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.22935103547681e-06,
+      "loss": 0.9027,
+      "step": 11953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227784594309799e-06,
+      "loss": 0.964,
+      "step": 11954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226218226825108e-06,
+      "loss": 0.9097,
+      "step": 11955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224651933064368e-06,
+      "loss": 0.937,
+      "step": 11956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.223085713069211e-06,
+      "loss": 0.9786,
+      "step": 11957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.221519566881265e-06,
+      "loss": 0.8781,
+      "step": 11958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.219953494542153e-06,
+      "loss": 0.9314,
+      "step": 11959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218387496093504e-06,
+      "loss": 0.8959,
+      "step": 11960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.216821571576935e-06,
+      "loss": 0.8947,
+      "step": 11961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215255721034071e-06,
+      "loss": 0.8654,
+      "step": 11962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2136899445065325e-06,
+      "loss": 0.7843,
+      "step": 11963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.212124242035929e-06,
+      "loss": 0.9164,
+      "step": 11964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.210558613663883e-06,
+      "loss": 0.8771,
+      "step": 11965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.208993059432003e-06,
+      "loss": 0.9183,
+      "step": 11966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.207427579381898e-06,
+      "loss": 0.9472,
+      "step": 11967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.205862173555181e-06,
+      "loss": 0.9522,
+      "step": 11968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.204296841993452e-06,
+      "loss": 0.8505,
+      "step": 11969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.202731584738323e-06,
+      "loss": 0.8884,
+      "step": 11970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.201166401831394e-06,
+      "loss": 0.9377,
+      "step": 11971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1996012933142645e-06,
+      "loss": 0.9078,
+      "step": 11972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.19803625922853e-06,
+      "loss": 0.8182,
+      "step": 11973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.196471299615797e-06,
+      "loss": 0.8505,
+      "step": 11974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1949064145176524e-06,
+      "loss": 0.9294,
+      "step": 11975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.193341603975694e-06,
+      "loss": 0.8784,
+      "step": 11976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191776868031508e-06,
+      "loss": 0.8816,
+      "step": 11977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.190212206726688e-06,
+      "loss": 0.9198,
+      "step": 11978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1886476201028176e-06,
+      "loss": 0.8014,
+      "step": 11979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1870831082014815e-06,
+      "loss": 0.85,
+      "step": 11980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185518671064266e-06,
+      "loss": 0.9026,
+      "step": 11981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183954308732745e-06,
+      "loss": 0.9394,
+      "step": 11982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.182390021248508e-06,
+      "loss": 0.8149,
+      "step": 11983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180825808653122e-06,
+      "loss": 0.9044,
+      "step": 11984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179261670988165e-06,
+      "loss": 0.8748,
+      "step": 11985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177697608295215e-06,
+      "loss": 0.951,
+      "step": 11986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.17613362061584e-06,
+      "loss": 0.8815,
+      "step": 11987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.174569707991605e-06,
+      "loss": 0.95,
+      "step": 11988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173005870464083e-06,
+      "loss": 0.9556,
+      "step": 11989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.171442108074833e-06,
+      "loss": 0.8799,
+      "step": 11990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1698784208654235e-06,
+      "loss": 0.8813,
+      "step": 11991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168314808877413e-06,
+      "loss": 0.8335,
+      "step": 11992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.16675127215236e-06,
+      "loss": 0.8322,
+      "step": 11993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.165187810731824e-06,
+      "loss": 0.9107,
+      "step": 11994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163624424657356e-06,
+      "loss": 0.778,
+      "step": 11995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1620611139705145e-06,
+      "loss": 0.8393,
+      "step": 11996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160497878712842e-06,
+      "loss": 0.8785,
+      "step": 11997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158934718925895e-06,
+      "loss": 0.8139,
+      "step": 11998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.15737163465122e-06,
+      "loss": 0.8219,
+      "step": 11999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.155808625930362e-06,
+      "loss": 0.8927,
+      "step": 12000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.15424569280486e-06,
+      "loss": 0.9568,
+      "step": 12001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.152682835316258e-06,
+      "loss": 0.8326,
+      "step": 12002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1511200535060935e-06,
+      "loss": 0.7943,
+      "step": 12003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1495573474159055e-06,
+      "loss": 0.8565,
+      "step": 12004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.147994717087225e-06,
+      "loss": 0.904,
+      "step": 12005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.146432162561591e-06,
+      "loss": 0.9092,
+      "step": 12006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144869683880531e-06,
+      "loss": 0.8625,
+      "step": 12007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.143307281085571e-06,
+      "loss": 0.8974,
+      "step": 12008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14174495421824e-06,
+      "loss": 0.8634,
+      "step": 12009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.140182703320067e-06,
+      "loss": 0.8254,
+      "step": 12010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.138620528432568e-06,
+      "loss": 0.9494,
+      "step": 12011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137058429597271e-06,
+      "loss": 0.9479,
+      "step": 12012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.135496406855688e-06,
+      "loss": 0.8782,
+      "step": 12013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133934460249343e-06,
+      "loss": 0.855,
+      "step": 12014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132372589819745e-06,
+      "loss": 0.9336,
+      "step": 12015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1308107956084075e-06,
+      "loss": 0.9356,
+      "step": 12016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129249077656844e-06,
+      "loss": 0.8654,
+      "step": 12017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.127687436006559e-06,
+      "loss": 0.862,
+      "step": 12018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.126125870699063e-06,
+      "loss": 0.8824,
+      "step": 12019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124564381775856e-06,
+      "loss": 0.8723,
+      "step": 12020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123002969278446e-06,
+      "loss": 1.0205,
+      "step": 12021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.121441633248331e-06,
+      "loss": 0.9519,
+      "step": 12022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.119880373727011e-06,
+      "loss": 0.9165,
+      "step": 12023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11831919075598e-06,
+      "loss": 0.9259,
+      "step": 12024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116758084376733e-06,
+      "loss": 0.8572,
+      "step": 12025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.115197054630765e-06,
+      "loss": 0.8282,
+      "step": 12026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.113636101559565e-06,
+      "loss": 0.9024,
+      "step": 12027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11207522520462e-06,
+      "loss": 0.9415,
+      "step": 12028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.110514425607416e-06,
+      "loss": 1.042,
+      "step": 12029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.10895370280944e-06,
+      "loss": 0.899,
+      "step": 12030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107393056852171e-06,
+      "loss": 0.9639,
+      "step": 12031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.105832487777089e-06,
+      "loss": 0.8274,
+      "step": 12032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.104271995625678e-06,
+      "loss": 0.925,
+      "step": 12033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1027115804394075e-06,
+      "loss": 0.4323,
+      "step": 12034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101151242259756e-06,
+      "loss": 0.9008,
+      "step": 12035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.099590981128195e-06,
+      "loss": 0.7868,
+      "step": 12036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.098030797086191e-06,
+      "loss": 0.834,
+      "step": 12037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096470690175215e-06,
+      "loss": 0.9291,
+      "step": 12038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.094910660436732e-06,
+      "loss": 0.9533,
+      "step": 12039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.093350707912205e-06,
+      "loss": 0.902,
+      "step": 12040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091790832643095e-06,
+      "loss": 0.9339,
+      "step": 12041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.090231034670867e-06,
+      "loss": 0.8659,
+      "step": 12042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0886713140369725e-06,
+      "loss": 0.9618,
+      "step": 12043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.087111670782867e-06,
+      "loss": 0.8859,
+      "step": 12044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085552104950008e-06,
+      "loss": 0.8671,
+      "step": 12045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083992616579846e-06,
+      "loss": 0.8951,
+      "step": 12046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0824332057138305e-06,
+      "loss": 0.8978,
+      "step": 12047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.080873872393409e-06,
+      "loss": 0.9078,
+      "step": 12048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0793146166600265e-06,
+      "loss": 0.9053,
+      "step": 12049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.077755438555123e-06,
+      "loss": 0.877,
+      "step": 12050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.076196338120145e-06,
+      "loss": 1.018,
+      "step": 12051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0746373153965266e-06,
+      "loss": 0.941,
+      "step": 12052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.073078370425709e-06,
+      "loss": 0.9485,
+      "step": 12053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0715195032491235e-06,
+      "loss": 0.916,
+      "step": 12054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.069960713908208e-06,
+      "loss": 0.9604,
+      "step": 12055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.068402002444386e-06,
+      "loss": 0.9141,
+      "step": 12056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066843368899092e-06,
+      "loss": 0.9198,
+      "step": 12057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.065284813313753e-06,
+      "loss": 0.9575,
+      "step": 12058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.063726335729793e-06,
+      "loss": 0.9339,
+      "step": 12059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.06216793618863e-06,
+      "loss": 0.9705,
+      "step": 12060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.06060961473169e-06,
+      "loss": 0.4475,
+      "step": 12061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.059051371400388e-06,
+      "loss": 0.8563,
+      "step": 12062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.057493206236143e-06,
+      "loss": 1.0131,
+      "step": 12063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.055935119280369e-06,
+      "loss": 0.9208,
+      "step": 12064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0543771105744755e-06,
+      "loss": 0.8427,
+      "step": 12065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052819180159876e-06,
+      "loss": 0.9393,
+      "step": 12066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.051261328077973e-06,
+      "loss": 0.8159,
+      "step": 12067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.049703554370175e-06,
+      "loss": 0.8677,
+      "step": 12068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0481458590778905e-06,
+      "loss": 0.4619,
+      "step": 12069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.046588242242519e-06,
+      "loss": 0.9229,
+      "step": 12070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045030703905457e-06,
+      "loss": 0.8274,
+      "step": 12071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0434732441081055e-06,
+      "loss": 0.8988,
+      "step": 12072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.041915862891855e-06,
+      "loss": 0.9152,
+      "step": 12073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0403585602981065e-06,
+      "loss": 0.8514,
+      "step": 12074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.038801336368244e-06,
+      "loss": 0.9498,
+      "step": 12075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.037244191143662e-06,
+      "loss": 0.8534,
+      "step": 12076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0356871246657464e-06,
+      "loss": 0.9247,
+      "step": 12077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.034130136975879e-06,
+      "loss": 0.9934,
+      "step": 12078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.032573228115442e-06,
+      "loss": 0.9037,
+      "step": 12079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.031016398125825e-06,
+      "loss": 0.8498,
+      "step": 12080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0294596470484e-06,
+      "loss": 0.8909,
+      "step": 12081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.027902974924545e-06,
+      "loss": 0.9811,
+      "step": 12082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0263463817956345e-06,
+      "loss": 0.9184,
+      "step": 12083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.024789867703041e-06,
+      "loss": 0.8734,
+      "step": 12084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023233432688136e-06,
+      "loss": 0.9754,
+      "step": 12085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.021677076792284e-06,
+      "loss": 0.8183,
+      "step": 12086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020120800056855e-06,
+      "loss": 0.879,
+      "step": 12087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0185646025232116e-06,
+      "loss": 0.8501,
+      "step": 12088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.017008484232717e-06,
+      "loss": 0.9464,
+      "step": 12089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015452445226727e-06,
+      "loss": 0.9101,
+      "step": 12090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.013896485546602e-06,
+      "loss": 0.895,
+      "step": 12091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012340605233701e-06,
+      "loss": 0.9589,
+      "step": 12092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.010784804329375e-06,
+      "loss": 0.9151,
+      "step": 12093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.009229082874972e-06,
+      "loss": 0.8793,
+      "step": 12094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.007673440911846e-06,
+      "loss": 0.9772,
+      "step": 12095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006117878481342e-06,
+      "loss": 0.8453,
+      "step": 12096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004562395624808e-06,
+      "loss": 0.9108,
+      "step": 12097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.003006992383582e-06,
+      "loss": 0.856,
+      "step": 12098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.001451668799006e-06,
+      "loss": 0.9064,
+      "step": 12099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9998964249124225e-06,
+      "loss": 0.8595,
+      "step": 12100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9983412607651635e-06,
+      "loss": 0.8645,
+      "step": 12101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.996786176398567e-06,
+      "loss": 0.8448,
+      "step": 12102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99523117185396e-06,
+      "loss": 0.8788,
+      "step": 12103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.993676247172679e-06,
+      "loss": 0.8579,
+      "step": 12104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.992121402396053e-06,
+      "loss": 0.938,
+      "step": 12105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990566637565403e-06,
+      "loss": 0.9116,
+      "step": 12106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.989011952722054e-06,
+      "loss": 0.9633,
+      "step": 12107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.987457347907329e-06,
+      "loss": 0.8733,
+      "step": 12108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985902823162545e-06,
+      "loss": 0.9382,
+      "step": 12109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984348378529025e-06,
+      "loss": 0.8548,
+      "step": 12110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9827940140480776e-06,
+      "loss": 0.9359,
+      "step": 12111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9812397297610224e-06,
+      "loss": 0.9306,
+      "step": 12112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979685525709167e-06,
+      "loss": 0.8324,
+      "step": 12113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.978131401933818e-06,
+      "loss": 0.9528,
+      "step": 12114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.976577358476284e-06,
+      "loss": 1.0019,
+      "step": 12115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9750233953778745e-06,
+      "loss": 0.8844,
+      "step": 12116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973469512679887e-06,
+      "loss": 1.0176,
+      "step": 12117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.971915710423625e-06,
+      "loss": 0.8965,
+      "step": 12118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970361988650382e-06,
+      "loss": 0.8884,
+      "step": 12119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.968808347401459e-06,
+      "loss": 0.9269,
+      "step": 12120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9672547867181496e-06,
+      "loss": 0.9184,
+      "step": 12121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965701306641741e-06,
+      "loss": 0.8441,
+      "step": 12122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964147907213528e-06,
+      "loss": 0.9029,
+      "step": 12123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962594588474796e-06,
+      "loss": 0.941,
+      "step": 12124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9610413504668306e-06,
+      "loss": 0.9089,
+      "step": 12125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.959488193230916e-06,
+      "loss": 0.9195,
+      "step": 12126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95793511680833e-06,
+      "loss": 0.8431,
+      "step": 12127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956382121240355e-06,
+      "loss": 0.9266,
+      "step": 12128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9548292065682695e-06,
+      "loss": 0.8962,
+      "step": 12129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.953276372833344e-06,
+      "loss": 0.9269,
+      "step": 12130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951723620076854e-06,
+      "loss": 0.9214,
+      "step": 12131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.950170948340067e-06,
+      "loss": 0.9143,
+      "step": 12132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948618357664256e-06,
+      "loss": 0.8622,
+      "step": 12133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947065848090684e-06,
+      "loss": 0.9075,
+      "step": 12134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9455134196606115e-06,
+      "loss": 0.9464,
+      "step": 12135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943961072415306e-06,
+      "loss": 0.9043,
+      "step": 12136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.942408806396024e-06,
+      "loss": 0.8742,
+      "step": 12137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.940856621644021e-06,
+      "loss": 0.9256,
+      "step": 12138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939304518200558e-06,
+      "loss": 0.8731,
+      "step": 12139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.937752496106886e-06,
+      "loss": 0.9538,
+      "step": 12140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.936200555404255e-06,
+      "loss": 0.9254,
+      "step": 12141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.934648696133914e-06,
+      "loss": 0.8479,
+      "step": 12142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93309691833711e-06,
+      "loss": 0.9448,
+      "step": 12143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931545222055087e-06,
+      "loss": 0.4128,
+      "step": 12144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929993607329087e-06,
+      "loss": 0.9142,
+      "step": 12145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92844207420035e-06,
+      "loss": 0.9789,
+      "step": 12146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9268906227101175e-06,
+      "loss": 0.8676,
+      "step": 12147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9253392528996185e-06,
+      "loss": 0.8895,
+      "step": 12148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9237879648100935e-06,
+      "loss": 0.9726,
+      "step": 12149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.922236758482767e-06,
+      "loss": 0.9535,
+      "step": 12150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920685633958875e-06,
+      "loss": 0.9846,
+      "step": 12151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919134591279641e-06,
+      "loss": 0.8745,
+      "step": 12152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917583630486292e-06,
+      "loss": 0.913,
+      "step": 12153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916032751620051e-06,
+      "loss": 0.9582,
+      "step": 12154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.914481954722136e-06,
+      "loss": 0.9078,
+      "step": 12155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9129312398337666e-06,
+      "loss": 0.8084,
+      "step": 12156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.91138060699616e-06,
+      "loss": 0.9171,
+      "step": 12157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909830056250527e-06,
+      "loss": 0.8997,
+      "step": 12158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9082795876380845e-06,
+      "loss": 0.9643,
+      "step": 12159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.906729201200038e-06,
+      "loss": 0.9045,
+      "step": 12160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9051788969775955e-06,
+      "loss": 0.9414,
+      "step": 12161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903628675011962e-06,
+      "loss": 0.8794,
+      "step": 12162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.902078535344345e-06,
+      "loss": 0.9035,
+      "step": 12163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.900528478015941e-06,
+      "loss": 0.8706,
+      "step": 12164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.898978503067952e-06,
+      "loss": 0.9414,
+      "step": 12165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89742861054157e-06,
+      "loss": 0.9188,
+      "step": 12166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.895878800477996e-06,
+      "loss": 0.8556,
+      "step": 12167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.894329072918416e-06,
+      "loss": 0.8818,
+      "step": 12168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.892779427904021e-06,
+      "loss": 0.9472,
+      "step": 12169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.891229865476002e-06,
+      "loss": 0.8981,
+      "step": 12170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88968038567554e-06,
+      "loss": 0.9991,
+      "step": 12171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.888130988543824e-06,
+      "loss": 0.8496,
+      "step": 12172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88658167412203e-06,
+      "loss": 0.9701,
+      "step": 12173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.885032442451336e-06,
+      "loss": 0.9133,
+      "step": 12174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883483293572928e-06,
+      "loss": 0.9531,
+      "step": 12175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.881934227527973e-06,
+      "loss": 0.8406,
+      "step": 12176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.880385244357645e-06,
+      "loss": 0.9588,
+      "step": 12177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.878836344103115e-06,
+      "loss": 1.0145,
+      "step": 12178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.877287526805547e-06,
+      "loss": 0.8992,
+      "step": 12179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875738792506115e-06,
+      "loss": 0.8864,
+      "step": 12180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.874190141245973e-06,
+      "loss": 0.851,
+      "step": 12181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.872641573066291e-06,
+      "loss": 0.8213,
+      "step": 12182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.871093088008223e-06,
+      "loss": 0.9565,
+      "step": 12183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.869544686112925e-06,
+      "loss": 0.8322,
+      "step": 12184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.867996367421557e-06,
+      "loss": 0.897,
+      "step": 12185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.866448131975264e-06,
+      "loss": 0.9491,
+      "step": 12186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8648999798152024e-06,
+      "loss": 0.9483,
+      "step": 12187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863351910982521e-06,
+      "loss": 0.9479,
+      "step": 12188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.861803925518362e-06,
+      "loss": 0.8843,
+      "step": 12189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.860256023463871e-06,
+      "loss": 0.8618,
+      "step": 12190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8587082048601905e-06,
+      "loss": 0.9549,
+      "step": 12191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857160469748456e-06,
+      "loss": 0.7915,
+      "step": 12192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.855612818169808e-06,
+      "loss": 0.9035,
+      "step": 12193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.854065250165378e-06,
+      "loss": 0.9091,
+      "step": 12194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.852517765776303e-06,
+      "loss": 0.8724,
+      "step": 12195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850970365043712e-06,
+      "loss": 0.8997,
+      "step": 12196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849423048008727e-06,
+      "loss": 0.8538,
+      "step": 12197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.847875814712482e-06,
+      "loss": 0.8872,
+      "step": 12198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.846328665196098e-06,
+      "loss": 0.8666,
+      "step": 12199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.844781599500695e-06,
+      "loss": 0.8877,
+      "step": 12200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8432346176673955e-06,
+      "loss": 0.8506,
+      "step": 12201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8416877197373135e-06,
+      "loss": 0.8736,
+      "step": 12202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.840140905751565e-06,
+      "loss": 0.886,
+      "step": 12203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.838594175751264e-06,
+      "loss": 0.824,
+      "step": 12204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.837047529777516e-06,
+      "loss": 0.8354,
+      "step": 12205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.835500967871436e-06,
+      "loss": 0.8715,
+      "step": 12206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833954490074121e-06,
+      "loss": 0.8274,
+      "step": 12207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.832408096426685e-06,
+      "loss": 0.9075,
+      "step": 12208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.830861786970218e-06,
+      "loss": 0.9589,
+      "step": 12209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829315561745829e-06,
+      "loss": 0.9204,
+      "step": 12210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.82776942079461e-06,
+      "loss": 0.8646,
+      "step": 12211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8262233641576595e-06,
+      "loss": 0.8249,
+      "step": 12212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824677391876065e-06,
+      "loss": 0.9353,
+      "step": 12213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.823131503990919e-06,
+      "loss": 0.8771,
+      "step": 12214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821585700543308e-06,
+      "loss": 0.9057,
+      "step": 12215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820039981574322e-06,
+      "loss": 0.8692,
+      "step": 12216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.818494347125041e-06,
+      "loss": 0.8536,
+      "step": 12217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816948797236545e-06,
+      "loss": 0.8923,
+      "step": 12218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815403331949915e-06,
+      "loss": 0.9764,
+      "step": 12219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.813857951306227e-06,
+      "loss": 0.9568,
+      "step": 12220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812312655346552e-06,
+      "loss": 0.8914,
+      "step": 12221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.810767444111972e-06,
+      "loss": 0.8277,
+      "step": 12222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.809222317643547e-06,
+      "loss": 0.9247,
+      "step": 12223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8076772759823514e-06,
+      "loss": 0.9426,
+      "step": 12224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.806132319169448e-06,
+      "loss": 0.8357,
+      "step": 12225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804587447245899e-06,
+      "loss": 0.438,
+      "step": 12226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803042660252766e-06,
+      "loss": 0.961,
+      "step": 12227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.801497958231107e-06,
+      "loss": 0.9181,
+      "step": 12228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.799953341221981e-06,
+      "loss": 0.8381,
+      "step": 12229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798408809266438e-06,
+      "loss": 0.8891,
+      "step": 12230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796864362405534e-06,
+      "loss": 0.9291,
+      "step": 12231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.795320000680318e-06,
+      "loss": 0.8888,
+      "step": 12232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.793775724131831e-06,
+      "loss": 0.9191,
+      "step": 12233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792231532801127e-06,
+      "loss": 0.9121,
+      "step": 12234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790687426729245e-06,
+      "loss": 0.8771,
+      "step": 12235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.789143405957226e-06,
+      "loss": 0.9471,
+      "step": 12236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.787599470526109e-06,
+      "loss": 0.913,
+      "step": 12237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7860556204769275e-06,
+      "loss": 0.9932,
+      "step": 12238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78451185585072e-06,
+      "loss": 0.8242,
+      "step": 12239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782968176688514e-06,
+      "loss": 0.8953,
+      "step": 12240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.781424583031338e-06,
+      "loss": 0.9231,
+      "step": 12241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.779881074920221e-06,
+      "loss": 0.986,
+      "step": 12242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.778337652396187e-06,
+      "loss": 0.865,
+      "step": 12243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.776794315500259e-06,
+      "loss": 0.8706,
+      "step": 12244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.775251064273454e-06,
+      "loss": 0.8522,
+      "step": 12245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.773707898756797e-06,
+      "loss": 0.9646,
+      "step": 12246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.772164818991297e-06,
+      "loss": 0.8561,
+      "step": 12247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.770621825017971e-06,
+      "loss": 0.9159,
+      "step": 12248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769078916877827e-06,
+      "loss": 0.9343,
+      "step": 12249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.767536094611877e-06,
+      "loss": 0.9432,
+      "step": 12250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.765993358261125e-06,
+      "loss": 0.8799,
+      "step": 12251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.764450707866577e-06,
+      "loss": 0.9599,
+      "step": 12252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.762908143469235e-06,
+      "loss": 0.9487,
+      "step": 12253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7613656651100936e-06,
+      "loss": 0.8818,
+      "step": 12254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.759823272830157e-06,
+      "loss": 0.8887,
+      "step": 12255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.758280966670414e-06,
+      "loss": 0.9397,
+      "step": 12256
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75673874667186e-06,
+      "loss": 0.9183,
+      "step": 12257
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75519661287549e-06,
+      "loss": 1.0108,
+      "step": 12258
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7536545653222855e-06,
+      "loss": 0.862,
+      "step": 12259
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.752112604053236e-06,
+      "loss": 0.4021,
+      "step": 12260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.750570729109326e-06,
+      "loss": 0.8688,
+      "step": 12261
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7490289405315314e-06,
+      "loss": 0.8542,
+      "step": 12262
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.747487238360837e-06,
+      "loss": 0.9245,
+      "step": 12263
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7459456226382145e-06,
+      "loss": 0.9037,
+      "step": 12264
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744404093404644e-06,
+      "loss": 0.9214,
+      "step": 12265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742862650701094e-06,
+      "loss": 0.9527,
+      "step": 12266
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7413212945685324e-06,
+      "loss": 0.8982,
+      "step": 12267
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.739780025047929e-06,
+      "loss": 0.8961,
+      "step": 12268
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738238842180249e-06,
+      "loss": 0.8823,
+      "step": 12269
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.736697746006456e-06,
+      "loss": 0.8863,
+      "step": 12270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.735156736567512e-06,
+      "loss": 0.871,
+      "step": 12271
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.73361581390437e-06,
+      "loss": 0.8613,
+      "step": 12272
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732074978057992e-06,
+      "loss": 0.9176,
+      "step": 12273
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7305342290693285e-06,
+      "loss": 0.9202,
+      "step": 12274
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728993566979329e-06,
+      "loss": 0.9332,
+      "step": 12275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.727452991828947e-06,
+      "loss": 0.8888,
+      "step": 12276
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.725912503659124e-06,
+      "loss": 0.9781,
+      "step": 12277
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72437210251081e-06,
+      "loss": 0.8327,
+      "step": 12278
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.722831788424943e-06,
+      "loss": 0.949,
+      "step": 12279
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72129156144246e-06,
+      "loss": 0.8971,
+      "step": 12280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.719751421604309e-06,
+      "loss": 0.4008,
+      "step": 12281
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.718211368951417e-06,
+      "loss": 0.9132,
+      "step": 12282
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.716671403524718e-06,
+      "loss": 0.8409,
+      "step": 12283
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.715131525365145e-06,
+      "loss": 0.8851,
+      "step": 12284
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.713591734513622e-06,
+      "loss": 0.9271,
+      "step": 12285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.712052031011079e-06,
+      "loss": 0.8544,
+      "step": 12286
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7105124148984376e-06,
+      "loss": 0.8185,
+      "step": 12287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.708972886216618e-06,
+      "loss": 0.9268,
+      "step": 12288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.707433445006544e-06,
+      "loss": 0.9245,
+      "step": 12289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.705894091309124e-06,
+      "loss": 0.9062,
+      "step": 12290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70435482516528e-06,
+      "loss": 0.9432,
+      "step": 12291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.702815646615916e-06,
+      "loss": 0.9152,
+      "step": 12292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7012765557019475e-06,
+      "loss": 1.0274,
+      "step": 12293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699737552464285e-06,
+      "loss": 0.7838,
+      "step": 12294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.698198636943827e-06,
+      "loss": 0.464,
+      "step": 12295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6966598091814785e-06,
+      "loss": 0.8979,
+      "step": 12296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.69512106921814e-06,
+      "loss": 0.9421,
+      "step": 12297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693582417094706e-06,
+      "loss": 0.9494,
+      "step": 12298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6920438528520794e-06,
+      "loss": 0.9189,
+      "step": 12299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6905053765311445e-06,
+      "loss": 0.8661,
+      "step": 12300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.688966988172801e-06,
+      "loss": 0.8947,
+      "step": 12301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.687428687817933e-06,
+      "loss": 0.8873,
+      "step": 12302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.685890475507424e-06,
+      "loss": 0.9161,
+      "step": 12303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68435235128216e-06,
+      "loss": 0.9067,
+      "step": 12304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.682814315183028e-06,
+      "loss": 0.8705,
+      "step": 12305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6812763672509e-06,
+      "loss": 0.875,
+      "step": 12306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679738507526659e-06,
+      "loss": 0.7936,
+      "step": 12307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6782007360511735e-06,
+      "loss": 0.8869,
+      "step": 12308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6766630528653206e-06,
+      "loss": 0.9851,
+      "step": 12309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6751254580099675e-06,
+      "loss": 0.8619,
+      "step": 12310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673587951525982e-06,
+      "loss": 0.4416,
+      "step": 12311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.672050533454231e-06,
+      "loss": 0.8844,
+      "step": 12312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.670513203835574e-06,
+      "loss": 0.9324,
+      "step": 12313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668975962710876e-06,
+      "loss": 0.8778,
+      "step": 12314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66743881012099e-06,
+      "loss": 0.8775,
+      "step": 12315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665901746106772e-06,
+      "loss": 0.7718,
+      "step": 12316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66436477070908e-06,
+      "loss": 0.8046,
+      "step": 12317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662827883968765e-06,
+      "loss": 0.8358,
+      "step": 12318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66129108592667e-06,
+      "loss": 0.8535,
+      "step": 12319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.659754376623647e-06,
+      "loss": 0.9429,
+      "step": 12320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6582177561005365e-06,
+      "loss": 0.917,
+      "step": 12321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656681224398182e-06,
+      "loss": 0.9325,
+      "step": 12322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.655144781557425e-06,
+      "loss": 0.9693,
+      "step": 12323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653608427619096e-06,
+      "loss": 0.9079,
+      "step": 12324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652072162624034e-06,
+      "loss": 0.8636,
+      "step": 12325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.650535986613069e-06,
+      "loss": 0.9782,
+      "step": 12326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648999899627031e-06,
+      "loss": 0.9289,
+      "step": 12327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.647463901706751e-06,
+      "loss": 0.9203,
+      "step": 12328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64592799289305e-06,
+      "loss": 0.8231,
+      "step": 12329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644392173226755e-06,
+      "loss": 0.9339,
+      "step": 12330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.642856442748682e-06,
+      "loss": 0.9175,
+      "step": 12331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6413208014996484e-06,
+      "loss": 0.8801,
+      "step": 12332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.639785249520476e-06,
+      "loss": 0.913,
+      "step": 12333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638249786851971e-06,
+      "loss": 0.4079,
+      "step": 12334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636714413534948e-06,
+      "loss": 0.9452,
+      "step": 12335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6351791296102165e-06,
+      "loss": 0.9493,
+      "step": 12336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.633643935118578e-06,
+      "loss": 0.9016,
+      "step": 12337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632108830100841e-06,
+      "loss": 0.9,
+      "step": 12338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.630573814597801e-06,
+      "loss": 0.9072,
+      "step": 12339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629038888650261e-06,
+      "loss": 0.8716,
+      "step": 12340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.627504052299021e-06,
+      "loss": 0.8352,
+      "step": 12341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.625969305584869e-06,
+      "loss": 0.9086,
+      "step": 12342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624434648548602e-06,
+      "loss": 0.8985,
+      "step": 12343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.622900081231007e-06,
+      "loss": 0.9496,
+      "step": 12344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621365603672868e-06,
+      "loss": 0.9232,
+      "step": 12345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619831215914974e-06,
+      "loss": 0.8914,
+      "step": 12346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.618296917998105e-06,
+      "loss": 0.9109,
+      "step": 12347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616762709963043e-06,
+      "loss": 0.9178,
+      "step": 12348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.615228591850562e-06,
+      "loss": 0.9312,
+      "step": 12349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613694563701443e-06,
+      "loss": 0.9883,
+      "step": 12350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.612160625556449e-06,
+      "loss": 0.7652,
+      "step": 12351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.61062677745636e-06,
+      "loss": 0.864,
+      "step": 12352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.609093019441939e-06,
+      "loss": 0.8301,
+      "step": 12353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.607559351553955e-06,
+      "loss": 0.924,
+      "step": 12354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.606025773833167e-06,
+      "loss": 0.9607,
+      "step": 12355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.604492286320339e-06,
+      "loss": 0.9087,
+      "step": 12356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.602958889056226e-06,
+      "loss": 0.8483,
+      "step": 12357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.601425582081589e-06,
+      "loss": 0.9068,
+      "step": 12358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599892365437177e-06,
+      "loss": 0.8603,
+      "step": 12359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.598359239163741e-06,
+      "loss": 0.9012,
+      "step": 12360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596826203302034e-06,
+      "loss": 0.9505,
+      "step": 12361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.595293257892798e-06,
+      "loss": 1.012,
+      "step": 12362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.593760402976775e-06,
+      "loss": 0.9189,
+      "step": 12363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592227638594715e-06,
+      "loss": 0.9955,
+      "step": 12364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.590694964787353e-06,
+      "loss": 0.8855,
+      "step": 12365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.589162381595422e-06,
+      "loss": 0.9381,
+      "step": 12366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587629889059664e-06,
+      "loss": 0.9455,
+      "step": 12367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.586097487220801e-06,
+      "loss": 0.8883,
+      "step": 12368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584565176119571e-06,
+      "loss": 0.8773,
+      "step": 12369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.583032955796695e-06,
+      "loss": 0.9234,
+      "step": 12370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581500826292904e-06,
+      "loss": 0.9415,
+      "step": 12371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579968787648916e-06,
+      "loss": 0.8484,
+      "step": 12372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5784368399054485e-06,
+      "loss": 0.8167,
+      "step": 12373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576904983103226e-06,
+      "loss": 0.8873,
+      "step": 12374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575373217282954e-06,
+      "loss": 0.9104,
+      "step": 12375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.573841542485353e-06,
+      "loss": 0.8594,
+      "step": 12376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572309958751133e-06,
+      "loss": 0.9163,
+      "step": 12377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.570778466120997e-06,
+      "loss": 0.88,
+      "step": 12378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.569247064635655e-06,
+      "loss": 0.8551,
+      "step": 12379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.56771575433581e-06,
+      "loss": 0.8963,
+      "step": 12380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.566184535262155e-06,
+      "loss": 0.9276,
+      "step": 12381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.564653407455397e-06,
+      "loss": 0.9227,
+      "step": 12382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.563122370956225e-06,
+      "loss": 0.9859,
+      "step": 12383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.561591425805338e-06,
+      "loss": 0.8653,
+      "step": 12384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.560060572043423e-06,
+      "loss": 0.9504,
+      "step": 12385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.558529809711165e-06,
+      "loss": 0.9016,
+      "step": 12386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.556999138849256e-06,
+      "loss": 0.8859,
+      "step": 12387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.555468559498382e-06,
+      "loss": 0.9531,
+      "step": 12388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.553938071699215e-06,
+      "loss": 0.4433,
+      "step": 12389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5524076754924405e-06,
+      "loss": 0.8473,
+      "step": 12390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550877370918731e-06,
+      "loss": 0.9352,
+      "step": 12391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549347158018764e-06,
+      "loss": 0.8838,
+      "step": 12392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.547817036833208e-06,
+      "loss": 0.8771,
+      "step": 12393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5462870074027314e-06,
+      "loss": 0.9414,
+      "step": 12394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544757069768003e-06,
+      "loss": 0.9215,
+      "step": 12395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.543227223969684e-06,
+      "loss": 0.8928,
+      "step": 12396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.541697470048438e-06,
+      "loss": 0.9635,
+      "step": 12397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.54016780804492e-06,
+      "loss": 0.8889,
+      "step": 12398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538638237999793e-06,
+      "loss": 0.8402,
+      "step": 12399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53710875995371e-06,
+      "loss": 0.8974,
+      "step": 12400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53557937394732e-06,
+      "loss": 0.8741,
+      "step": 12401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534050080021272e-06,
+      "loss": 0.8819,
+      "step": 12402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.532520878216217e-06,
+      "loss": 0.8581,
+      "step": 12403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530991768572794e-06,
+      "loss": 0.9693,
+      "step": 12404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529462751131651e-06,
+      "loss": 0.9067,
+      "step": 12405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.527933825933421e-06,
+      "loss": 0.7949,
+      "step": 12406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.526404993018749e-06,
+      "loss": 0.9078,
+      "step": 12407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524876252428263e-06,
+      "loss": 0.8549,
+      "step": 12408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.523347604202594e-06,
+      "loss": 0.8779,
+      "step": 12409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.521819048382375e-06,
+      "loss": 0.947,
+      "step": 12410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5202905850082375e-06,
+      "loss": 0.9671,
+      "step": 12411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5187622141207984e-06,
+      "loss": 0.9366,
+      "step": 12412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.517233935760687e-06,
+      "loss": 0.4692,
+      "step": 12413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.515705749968519e-06,
+      "loss": 0.9122,
+      "step": 12414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514177656784911e-06,
+      "loss": 0.8365,
+      "step": 12415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512649656250483e-06,
+      "loss": 0.8556,
+      "step": 12416
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.51112174840584e-06,
+      "loss": 0.4311,
+      "step": 12417
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.509593933291599e-06,
+      "loss": 0.8482,
+      "step": 12418
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.508066210948363e-06,
+      "loss": 0.925,
+      "step": 12419
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5065385814167414e-06,
+      "loss": 0.9266,
+      "step": 12420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505011044737334e-06,
+      "loss": 0.8394,
+      "step": 12421
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.503483600950737e-06,
+      "loss": 0.849,
+      "step": 12422
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501956250097555e-06,
+      "loss": 0.9277,
+      "step": 12423
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.500428992218383e-06,
+      "loss": 0.4249,
+      "step": 12424
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.498901827353809e-06,
+      "loss": 0.9292,
+      "step": 12425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.497374755544428e-06,
+      "loss": 0.8991,
+      "step": 12426
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4958477768308235e-06,
+      "loss": 0.902,
+      "step": 12427
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.494320891253585e-06,
+      "loss": 0.8761,
+      "step": 12428
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.492794098853292e-06,
+      "loss": 0.8773,
+      "step": 12429
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4912673996705265e-06,
+      "loss": 0.898,
+      "step": 12430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.489740793745867e-06,
+      "loss": 0.9287,
+      "step": 12431
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.488214281119887e-06,
+      "loss": 0.9355,
+      "step": 12432
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486687861833161e-06,
+      "loss": 0.9004,
+      "step": 12433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.485161535926257e-06,
+      "loss": 0.8349,
+      "step": 12434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483635303439747e-06,
+      "loss": 0.8229,
+      "step": 12435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.482109164414194e-06,
+      "loss": 0.9452,
+      "step": 12436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.480583118890162e-06,
+      "loss": 0.8765,
+      "step": 12437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479057166908212e-06,
+      "loss": 0.8887,
+      "step": 12438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4775313085089e-06,
+      "loss": 0.4264,
+      "step": 12439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476005543732783e-06,
+      "loss": 0.433,
+      "step": 12440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.474479872620415e-06,
+      "loss": 0.8711,
+      "step": 12441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.472954295212346e-06,
+      "loss": 0.8595,
+      "step": 12442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4714288115491204e-06,
+      "loss": 0.9511,
+      "step": 12443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46990342167129e-06,
+      "loss": 0.8369,
+      "step": 12444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46837812561939e-06,
+      "loss": 0.9099,
+      "step": 12445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466852923433968e-06,
+      "loss": 0.9196,
+      "step": 12446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465327815155563e-06,
+      "loss": 0.8562,
+      "step": 12447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.463802800824706e-06,
+      "loss": 0.9063,
+      "step": 12448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.462277880481933e-06,
+      "loss": 0.8921,
+      "step": 12449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460753054167773e-06,
+      "loss": 0.9355,
+      "step": 12450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.459228321922752e-06,
+      "loss": 0.9188,
+      "step": 12451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457703683787403e-06,
+      "loss": 0.9161,
+      "step": 12452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456179139802241e-06,
+      "loss": 0.9029,
+      "step": 12453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.454654690007791e-06,
+      "loss": 0.857,
+      "step": 12454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45313033444457e-06,
+      "loss": 0.8932,
+      "step": 12455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451606073153094e-06,
+      "loss": 0.9144,
+      "step": 12456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.450081906173873e-06,
+      "loss": 0.8763,
+      "step": 12457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448557833547424e-06,
+      "loss": 0.9941,
+      "step": 12458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.447033855314251e-06,
+      "loss": 0.8494,
+      "step": 12459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445509971514861e-06,
+      "loss": 0.9618,
+      "step": 12460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4439861821897546e-06,
+      "loss": 0.4351,
+      "step": 12461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.442462487379435e-06,
+      "loss": 0.9418,
+      "step": 12462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.440938887124401e-06,
+      "loss": 0.9461,
+      "step": 12463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.439415381465144e-06,
+      "loss": 0.872,
+      "step": 12464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.437891970442161e-06,
+      "loss": 0.3936,
+      "step": 12465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436368654095939e-06,
+      "loss": 0.9498,
+      "step": 12466
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434845432466972e-06,
+      "loss": 0.9161,
+      "step": 12467
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4333223055957376e-06,
+      "loss": 0.7389,
+      "step": 12468
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.431799273522722e-06,
+      "loss": 0.9728,
+      "step": 12469
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43027633628841e-06,
+      "loss": 0.9219,
+      "step": 12470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.428753493933275e-06,
+      "loss": 0.9175,
+      "step": 12471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.427230746497793e-06,
+      "loss": 0.873,
+      "step": 12472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425708094022438e-06,
+      "loss": 0.9024,
+      "step": 12473
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.424185536547679e-06,
+      "loss": 0.8766,
+      "step": 12474
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.422663074113986e-06,
+      "loss": 0.8761,
+      "step": 12475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.42114070676182e-06,
+      "loss": 0.9577,
+      "step": 12476
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41961843453165e-06,
+      "loss": 0.8835,
+      "step": 12477
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418096257463931e-06,
+      "loss": 0.8418,
+      "step": 12478
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416574175599121e-06,
+      "loss": 0.879,
+      "step": 12479
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41505218897768e-06,
+      "loss": 0.9287,
+      "step": 12480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.413530297640052e-06,
+      "loss": 0.8615,
+      "step": 12481
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412008501626695e-06,
+      "loss": 0.8669,
+      "step": 12482
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410486800978057e-06,
+      "loss": 0.9423,
+      "step": 12483
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40896519573458e-06,
+      "loss": 1.0232,
+      "step": 12484
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.407443685936703e-06,
+      "loss": 0.7367,
+      "step": 12485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.405922271624874e-06,
+      "loss": 1.0011,
+      "step": 12486
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.404400952839522e-06,
+      "loss": 0.9163,
+      "step": 12487
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40287972962109e-06,
+      "loss": 0.9058,
+      "step": 12488
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.401358602010004e-06,
+      "loss": 0.9253,
+      "step": 12489
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.399837570046698e-06,
+      "loss": 0.9531,
+      "step": 12490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398316633771598e-06,
+      "loss": 0.7743,
+      "step": 12491
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.396795793225126e-06,
+      "loss": 0.9291,
+      "step": 12492
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.395275048447704e-06,
+      "loss": 0.9425,
+      "step": 12493
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.393754399479759e-06,
+      "loss": 0.9015,
+      "step": 12494
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392233846361701e-06,
+      "loss": 0.9173,
+      "step": 12495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.390713389133949e-06,
+      "loss": 0.8971,
+      "step": 12496
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389193027836911e-06,
+      "loss": 0.854,
+      "step": 12497
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3876727625110005e-06,
+      "loss": 0.8552,
+      "step": 12498
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.386152593196622e-06,
+      "loss": 0.9753,
+      "step": 12499
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384632519934179e-06,
+      "loss": 0.9175,
+      "step": 12500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.383112542764075e-06,
+      "loss": 0.8701,
+      "step": 12501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.381592661726708e-06,
+      "loss": 0.9332,
+      "step": 12502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3800728768624765e-06,
+      "loss": 0.4261,
+      "step": 12503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.37855318821177e-06,
+      "loss": 0.9035,
+      "step": 12504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3770335958149856e-06,
+      "loss": 0.899,
+      "step": 12505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.375514099712512e-06,
+      "loss": 0.808,
+      "step": 12506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3739946999447355e-06,
+      "loss": 0.886,
+      "step": 12507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372475396552036e-06,
+      "loss": 0.9062,
+      "step": 12508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.370956189574798e-06,
+      "loss": 0.8949,
+      "step": 12509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3694370790534e-06,
+      "loss": 0.9453,
+      "step": 12510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.367918065028218e-06,
+      "loss": 0.9465,
+      "step": 12511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.366399147539626e-06,
+      "loss": 0.8952,
+      "step": 12512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3648803266279934e-06,
+      "loss": 0.8837,
+      "step": 12513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.363361602333692e-06,
+      "loss": 0.8956,
+      "step": 12514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.361842974697084e-06,
+      "loss": 0.7883,
+      "step": 12515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360324443758531e-06,
+      "loss": 0.8747,
+      "step": 12516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.358806009558402e-06,
+      "loss": 0.9341,
+      "step": 12517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357287672137049e-06,
+      "loss": 1.0375,
+      "step": 12518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.355769431534832e-06,
+      "loss": 0.9326,
+      "step": 12519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.354251287792101e-06,
+      "loss": 0.8941,
+      "step": 12520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.352733240949205e-06,
+      "loss": 0.9533,
+      "step": 12521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351215291046495e-06,
+      "loss": 0.9314,
+      "step": 12522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349697438124314e-06,
+      "loss": 0.8825,
+      "step": 12523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.348179682223008e-06,
+      "loss": 0.9839,
+      "step": 12524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.346662023382912e-06,
+      "loss": 0.8711,
+      "step": 12525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.34514446164437e-06,
+      "loss": 0.939,
+      "step": 12526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3436269970477135e-06,
+      "loss": 0.9646,
+      "step": 12527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.342109629633273e-06,
+      "loss": 0.9411,
+      "step": 12528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.340592359441381e-06,
+      "loss": 0.9624,
+      "step": 12529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339075186512366e-06,
+      "loss": 0.9018,
+      "step": 12530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337558110886552e-06,
+      "loss": 0.9715,
+      "step": 12531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336041132604261e-06,
+      "loss": 0.8783,
+      "step": 12532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.334524251705812e-06,
+      "loss": 0.9505,
+      "step": 12533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333007468231521e-06,
+      "loss": 0.8496,
+      "step": 12534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331490782221705e-06,
+      "loss": 0.8789,
+      "step": 12535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3299741937166714e-06,
+      "loss": 0.8794,
+      "step": 12536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.328457702756734e-06,
+      "loss": 0.901,
+      "step": 12537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3269413093821975e-06,
+      "loss": 0.8577,
+      "step": 12538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325425013633366e-06,
+      "loss": 0.8924,
+      "step": 12539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323908815550537e-06,
+      "loss": 0.9319,
+      "step": 12540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3223927151740184e-06,
+      "loss": 0.8941,
+      "step": 12541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3208767125440985e-06,
+      "loss": 0.9138,
+      "step": 12542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.319360807701076e-06,
+      "loss": 0.8718,
+      "step": 12543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317845000685237e-06,
+      "loss": 0.9418,
+      "step": 12544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.316329291536874e-06,
+      "loss": 0.8359,
+      "step": 12545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.31481368029627e-06,
+      "loss": 0.8481,
+      "step": 12546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.313298167003712e-06,
+      "loss": 0.9276,
+      "step": 12547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3117827516994765e-06,
+      "loss": 0.914,
+      "step": 12548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.310267434423842e-06,
+      "loss": 0.8455,
+      "step": 12549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.308752215217087e-06,
+      "loss": 0.8891,
+      "step": 12550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30723709411948e-06,
+      "loss": 0.8775,
+      "step": 12551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305722071171292e-06,
+      "loss": 0.9441,
+      "step": 12552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.304207146412797e-06,
+      "loss": 0.9084,
+      "step": 12553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.302692319884251e-06,
+      "loss": 0.8629,
+      "step": 12554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301177591625924e-06,
+      "loss": 0.97,
+      "step": 12555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2996629616780725e-06,
+      "loss": 0.8404,
+      "step": 12556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2981484300809505e-06,
+      "loss": 0.9345,
+      "step": 12557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2966339968748165e-06,
+      "loss": 0.4203,
+      "step": 12558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29511966209992e-06,
+      "loss": 0.9706,
+      "step": 12559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293605425796514e-06,
+      "loss": 0.9571,
+      "step": 12560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.292091288004842e-06,
+      "loss": 0.8745,
+      "step": 12561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290577248765146e-06,
+      "loss": 0.8066,
+      "step": 12562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.289063308117673e-06,
+      "loss": 0.8576,
+      "step": 12563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2875494661026545e-06,
+      "loss": 0.8149,
+      "step": 12564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.286035722760334e-06,
+      "loss": 0.988,
+      "step": 12565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.284522078130943e-06,
+      "loss": 0.9431,
+      "step": 12566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283008532254711e-06,
+      "loss": 0.951,
+      "step": 12567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2814950851718695e-06,
+      "loss": 0.9308,
+      "step": 12568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27998173692264e-06,
+      "loss": 0.8489,
+      "step": 12569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278468487547247e-06,
+      "loss": 0.8756,
+      "step": 12570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2769553370859115e-06,
+      "loss": 0.9471,
+      "step": 12571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27544228557885e-06,
+      "loss": 0.8964,
+      "step": 12572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27392933306628e-06,
+      "loss": 0.9365,
+      "step": 12573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.272416479588411e-06,
+      "loss": 0.9328,
+      "step": 12574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.270903725185453e-06,
+      "loss": 0.867,
+      "step": 12575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.269391069897619e-06,
+      "loss": 0.8326,
+      "step": 12576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267878513765109e-06,
+      "loss": 0.9047,
+      "step": 12577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.266366056828123e-06,
+      "loss": 0.8787,
+      "step": 12578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2648536991268675e-06,
+      "loss": 0.8417,
+      "step": 12579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2633414407015315e-06,
+      "loss": 0.917,
+      "step": 12580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261829281592313e-06,
+      "loss": 0.9386,
+      "step": 12581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260317221839405e-06,
+      "loss": 0.3751,
+      "step": 12582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.25880526148299e-06,
+      "loss": 0.889,
+      "step": 12583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2572934005632625e-06,
+      "loss": 0.894,
+      "step": 12584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2557816391203976e-06,
+      "loss": 0.9127,
+      "step": 12585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254269977194584e-06,
+      "loss": 0.8502,
+      "step": 12586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.252758414825992e-06,
+      "loss": 0.8279,
+      "step": 12587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251246952054803e-06,
+      "loss": 0.8948,
+      "step": 12588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.249735588921191e-06,
+      "loss": 0.7289,
+      "step": 12589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248224325465323e-06,
+      "loss": 0.9207,
+      "step": 12590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.246713161727367e-06,
+      "loss": 0.9551,
+      "step": 12591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.245202097747489e-06,
+      "loss": 0.8891,
+      "step": 12592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24369113356585e-06,
+      "loss": 0.8562,
+      "step": 12593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.242180269222612e-06,
+      "loss": 0.829,
+      "step": 12594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240669504757928e-06,
+      "loss": 0.8851,
+      "step": 12595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.239158840211958e-06,
+      "loss": 0.9839,
+      "step": 12596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23764827562485e-06,
+      "loss": 0.8611,
+      "step": 12597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236137811036753e-06,
+      "loss": 0.9713,
+      "step": 12598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.234627446487811e-06,
+      "loss": 0.8406,
+      "step": 12599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.233117182018176e-06,
+      "loss": 0.9384,
+      "step": 12600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231607017667982e-06,
+      "loss": 0.9516,
+      "step": 12601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.230096953477372e-06,
+      "loss": 0.9103,
+      "step": 12602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228586989486476e-06,
+      "loss": 0.9126,
+      "step": 12603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.227077125735432e-06,
+      "loss": 0.9796,
+      "step": 12604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.225567362264371e-06,
+      "loss": 0.9414,
+      "step": 12605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2240576991134146e-06,
+      "loss": 0.8811,
+      "step": 12606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222548136322694e-06,
+      "loss": 0.929,
+      "step": 12607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.221038673932329e-06,
+      "loss": 0.821,
+      "step": 12608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219529311982442e-06,
+      "loss": 0.855,
+      "step": 12609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.218020050513148e-06,
+      "loss": 0.8819,
+      "step": 12610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.216510889564557e-06,
+      "loss": 0.9472,
+      "step": 12611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.215001829176789e-06,
+      "loss": 0.8809,
+      "step": 12612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2134928693899496e-06,
+      "loss": 0.8895,
+      "step": 12613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.211984010244146e-06,
+      "loss": 0.9122,
+      "step": 12614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.21047525177948e-06,
+      "loss": 0.3814,
+      "step": 12615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.208966594036052e-06,
+      "loss": 0.8742,
+      "step": 12616
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.207458037053966e-06,
+      "loss": 0.8728,
+      "step": 12617
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.205949580873311e-06,
+      "loss": 0.8493,
+      "step": 12618
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.204441225534183e-06,
+      "loss": 0.9847,
+      "step": 12619
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202932971076675e-06,
+      "loss": 0.9112,
+      "step": 12620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201424817540869e-06,
+      "loss": 0.465,
+      "step": 12621
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199916764966856e-06,
+      "loss": 0.9008,
+      "step": 12622
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.198408813394709e-06,
+      "loss": 0.8544,
+      "step": 12623
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196900962864517e-06,
+      "loss": 0.7963,
+      "step": 12624
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.195393213416356e-06,
+      "loss": 0.9121,
+      "step": 12625
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.193885565090298e-06,
+      "loss": 0.9193,
+      "step": 12626
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.192378017926412e-06,
+      "loss": 0.9134,
+      "step": 12627
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1908705719647735e-06,
+      "loss": 0.8208,
+      "step": 12628
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1893632272454415e-06,
+      "loss": 0.9163,
+      "step": 12629
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.187855983808485e-06,
+      "loss": 0.8412,
+      "step": 12630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.186348841693962e-06,
+      "loss": 0.8539,
+      "step": 12631
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.18484180094193e-06,
+      "loss": 0.88,
+      "step": 12632
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183334861592447e-06,
+      "loss": 0.9022,
+      "step": 12633
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.181828023685561e-06,
+      "loss": 0.7977,
+      "step": 12634
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180321287261326e-06,
+      "loss": 0.9074,
+      "step": 12635
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.178814652359793e-06,
+      "loss": 0.8959,
+      "step": 12636
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177308119020997e-06,
+      "loss": 0.883,
+      "step": 12637
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17580168728499e-06,
+      "loss": 0.8755,
+      "step": 12638
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174295357191807e-06,
+      "loss": 0.9066,
+      "step": 12639
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17278912878148e-06,
+      "loss": 0.9383,
+      "step": 12640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.171283002094048e-06,
+      "loss": 0.9305,
+      "step": 12641
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169776977169542e-06,
+      "loss": 0.8831,
+      "step": 12642
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16827105404799e-06,
+      "loss": 0.9289,
+      "step": 12643
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1667652327694135e-06,
+      "loss": 0.9232,
+      "step": 12644
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.165259513373845e-06,
+      "loss": 0.898,
+      "step": 12645
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1637538959012925e-06,
+      "loss": 0.9404,
+      "step": 12646
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.162248380391783e-06,
+      "loss": 0.9462,
+      "step": 12647
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160742966885328e-06,
+      "loss": 0.904,
+      "step": 12648
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159237655421941e-06,
+      "loss": 1.0058,
+      "step": 12649
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.15773244604163e-06,
+      "loss": 0.8435,
+      "step": 12650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.156227338784404e-06,
+      "loss": 0.8169,
+      "step": 12651
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154722333690264e-06,
+      "loss": 0.9225,
+      "step": 12652
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.153217430799212e-06,
+      "loss": 0.9303,
+      "step": 12653
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.151712630151249e-06,
+      "loss": 0.425,
+      "step": 12654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150207931786367e-06,
+      "loss": 0.8899,
+      "step": 12655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.148703335744563e-06,
+      "loss": 0.9068,
+      "step": 12656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147198842065824e-06,
+      "loss": 0.8888,
+      "step": 12657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.145694450790137e-06,
+      "loss": 0.9608,
+      "step": 12658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144190161957494e-06,
+      "loss": 1.0061,
+      "step": 12659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1426859756078725e-06,
+      "loss": 0.8648,
+      "step": 12660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.14118189178125e-06,
+      "loss": 0.9255,
+      "step": 12661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139677910517609e-06,
+      "loss": 0.8904,
+      "step": 12662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138174031856916e-06,
+      "loss": 1.0072,
+      "step": 12663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.13667025583915e-06,
+      "loss": 0.8827,
+      "step": 12664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135166582504275e-06,
+      "loss": 0.7844,
+      "step": 12665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1336630118922594e-06,
+      "loss": 0.8693,
+      "step": 12666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132159544043063e-06,
+      "loss": 0.8434,
+      "step": 12667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.13065617899665e-06,
+      "loss": 0.9277,
+      "step": 12668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129152916792977e-06,
+      "loss": 0.8988,
+      "step": 12669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.127649757471994e-06,
+      "loss": 0.8816,
+      "step": 12670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126146701073661e-06,
+      "loss": 0.9573,
+      "step": 12671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124643747637926e-06,
+      "loss": 0.8802,
+      "step": 12672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1231408972047316e-06,
+      "loss": 0.9324,
+      "step": 12673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.121638149814027e-06,
+      "loss": 0.9455,
+      "step": 12674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1201355055057486e-06,
+      "loss": 0.9221,
+      "step": 12675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118632964319837e-06,
+      "loss": 0.8768,
+      "step": 12676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117130526296231e-06,
+      "loss": 0.9161,
+      "step": 12677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.115628191474858e-06,
+      "loss": 0.8622,
+      "step": 12678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114125959895654e-06,
+      "loss": 0.8489,
+      "step": 12679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.112623831598543e-06,
+      "loss": 0.9119,
+      "step": 12680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1111218066234505e-06,
+      "loss": 0.9438,
+      "step": 12681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109619885010294e-06,
+      "loss": 0.9649,
+      "step": 12682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.108118066799004e-06,
+      "loss": 0.9691,
+      "step": 12683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106616352029489e-06,
+      "loss": 0.9245,
+      "step": 12684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1051147407416645e-06,
+      "loss": 0.8707,
+      "step": 12685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.103613232975441e-06,
+      "loss": 0.9539,
+      "step": 12686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.102111828770729e-06,
+      "loss": 0.9519,
+      "step": 12687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1006105281674344e-06,
+      "loss": 0.871,
+      "step": 12688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.099109331205455e-06,
+      "loss": 0.8386,
+      "step": 12689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.097608237924697e-06,
+      "loss": 0.926,
+      "step": 12690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0961072483650526e-06,
+      "loss": 0.8673,
+      "step": 12691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.094606362566422e-06,
+      "loss": 0.9278,
+      "step": 12692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09310558056869e-06,
+      "loss": 0.9804,
+      "step": 12693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09160490241175e-06,
+      "loss": 0.8619,
+      "step": 12694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090104328135491e-06,
+      "loss": 0.8796,
+      "step": 12695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0886038577797924e-06,
+      "loss": 0.9255,
+      "step": 12696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.087103491384534e-06,
+      "loss": 0.8847,
+      "step": 12697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0856032289896e-06,
+      "loss": 0.9023,
+      "step": 12698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.084103070634857e-06,
+      "loss": 0.8971,
+      "step": 12699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082603016360185e-06,
+      "loss": 0.9176,
+      "step": 12700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081103066205451e-06,
+      "loss": 0.8713,
+      "step": 12701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0796032202105195e-06,
+      "loss": 0.939,
+      "step": 12702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078103478415258e-06,
+      "loss": 0.888,
+      "step": 12703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076603840859527e-06,
+      "loss": 0.979,
+      "step": 12704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.075104307583181e-06,
+      "loss": 0.9429,
+      "step": 12705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073604878626085e-06,
+      "loss": 0.7844,
+      "step": 12706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.072105554028083e-06,
+      "loss": 0.7924,
+      "step": 12707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.070606333829033e-06,
+      "loss": 0.834,
+      "step": 12708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069107218068778e-06,
+      "loss": 0.9216,
+      "step": 12709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0676082067871634e-06,
+      "loss": 0.8558,
+      "step": 12710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.066109300024031e-06,
+      "loss": 0.9463,
+      "step": 12711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0646104978192196e-06,
+      "loss": 0.9755,
+      "step": 12712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.063111800212569e-06,
+      "loss": 0.8952,
+      "step": 12713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.061613207243908e-06,
+      "loss": 0.9136,
+      "step": 12714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060114718953072e-06,
+      "loss": 0.9252,
+      "step": 12715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058616335379886e-06,
+      "loss": 0.95,
+      "step": 12716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.057118056564173e-06,
+      "loss": 0.9123,
+      "step": 12717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.055619882545762e-06,
+      "loss": 0.8873,
+      "step": 12718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054121813364471e-06,
+      "loss": 0.8845,
+      "step": 12719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.052623849060113e-06,
+      "loss": 0.8258,
+      "step": 12720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051125989672506e-06,
+      "loss": 0.8443,
+      "step": 12721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049628235241459e-06,
+      "loss": 0.9002,
+      "step": 12722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.048130585806784e-06,
+      "loss": 0.8521,
+      "step": 12723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.046633041408285e-06,
+      "loss": 0.9541,
+      "step": 12724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.045135602085761e-06,
+      "loss": 0.864,
+      "step": 12725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0436382678790195e-06,
+      "loss": 0.9031,
+      "step": 12726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.042141038827851e-06,
+      "loss": 0.8573,
+      "step": 12727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.040643914972057e-06,
+      "loss": 0.8918,
+      "step": 12728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.03914689635142e-06,
+      "loss": 0.9169,
+      "step": 12729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.03764998300574e-06,
+      "loss": 0.9652,
+      "step": 12730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.036153174974795e-06,
+      "loss": 0.9135,
+      "step": 12731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034656472298374e-06,
+      "loss": 0.8158,
+      "step": 12732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033159875016252e-06,
+      "loss": 0.9594,
+      "step": 12733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031663383168214e-06,
+      "loss": 0.7955,
+      "step": 12734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.030166996794027e-06,
+      "loss": 0.394,
+      "step": 12735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.028670715933471e-06,
+      "loss": 0.8914,
+      "step": 12736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027174540626311e-06,
+      "loss": 0.9186,
+      "step": 12737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.025678470912312e-06,
+      "loss": 0.9384,
+      "step": 12738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.024182506831242e-06,
+      "loss": 0.3885,
+      "step": 12739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.022686648422858e-06,
+      "loss": 0.9327,
+      "step": 12740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021190895726919e-06,
+      "loss": 0.8325,
+      "step": 12741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.019695248783186e-06,
+      "loss": 0.8681,
+      "step": 12742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018199707631405e-06,
+      "loss": 0.901,
+      "step": 12743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.01670427231133e-06,
+      "loss": 0.9222,
+      "step": 12744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015208942862706e-06,
+      "loss": 0.8506,
+      "step": 12745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013713719325276e-06,
+      "loss": 0.8821,
+      "step": 12746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.012218601738784e-06,
+      "loss": 0.844,
+      "step": 12747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010723590142964e-06,
+      "loss": 0.8407,
+      "step": 12748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009228684577558e-06,
+      "loss": 0.8768,
+      "step": 12749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007733885082295e-06,
+      "loss": 0.9369,
+      "step": 12750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.006239191696905e-06,
+      "loss": 0.9311,
+      "step": 12751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004744604461113e-06,
+      "loss": 0.9927,
+      "step": 12752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.00325012341465e-06,
+      "loss": 0.9071,
+      "step": 12753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0017557485972335e-06,
+      "loss": 0.928,
+      "step": 12754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.000261480048583e-06,
+      "loss": 0.8467,
+      "step": 12755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.998767317808414e-06,
+      "loss": 0.9094,
+      "step": 12756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997273261916441e-06,
+      "loss": 0.8258,
+      "step": 12757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995779312412372e-06,
+      "loss": 0.9325,
+      "step": 12758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.994285469335914e-06,
+      "loss": 0.9189,
+      "step": 12759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.992791732726777e-06,
+      "loss": 0.8802,
+      "step": 12760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991298102624656e-06,
+      "loss": 0.879,
+      "step": 12761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989804579069255e-06,
+      "loss": 0.8595,
+      "step": 12762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988311162100266e-06,
+      "loss": 0.8945,
+      "step": 12763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.986817851757383e-06,
+      "loss": 0.8363,
+      "step": 12764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985324648080302e-06,
+      "loss": 0.8594,
+      "step": 12765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983831551108706e-06,
+      "loss": 0.3559,
+      "step": 12766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.982338560882279e-06,
+      "loss": 0.4443,
+      "step": 12767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.980845677440705e-06,
+      "loss": 0.8406,
+      "step": 12768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.979352900823663e-06,
+      "loss": 0.9067,
+      "step": 12769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.97786023107083e-06,
+      "loss": 0.958,
+      "step": 12770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.976367668221876e-06,
+      "loss": 0.8787,
+      "step": 12771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.974875212316476e-06,
+      "loss": 0.8581,
+      "step": 12772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973382863394296e-06,
+      "loss": 0.9216,
+      "step": 12773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.971890621494998e-06,
+      "loss": 0.8391,
+      "step": 12774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.970398486658249e-06,
+      "loss": 0.8867,
+      "step": 12775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.968906458923702e-06,
+      "loss": 0.9228,
+      "step": 12776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96741453833102e-06,
+      "loss": 0.9215,
+      "step": 12777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9659227249198555e-06,
+      "loss": 0.8223,
+      "step": 12778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9644310187298575e-06,
+      "loss": 0.9048,
+      "step": 12779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.962939419800673e-06,
+      "loss": 0.9562,
+      "step": 12780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961447928171948e-06,
+      "loss": 0.8505,
+      "step": 12781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959956543883324e-06,
+      "loss": 0.9808,
+      "step": 12782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.958465266974443e-06,
+      "loss": 0.8941,
+      "step": 12783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.956974097484935e-06,
+      "loss": 0.8114,
+      "step": 12784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.955483035454444e-06,
+      "loss": 0.9377,
+      "step": 12785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.953992080922592e-06,
+      "loss": 0.897,
+      "step": 12786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.952501233929007e-06,
+      "loss": 0.4065,
+      "step": 12787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.951010494513316e-06,
+      "loss": 0.9689,
+      "step": 12788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9495198627151455e-06,
+      "loss": 0.8271,
+      "step": 12789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.948029338574109e-06,
+      "loss": 0.9104,
+      "step": 12790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.946538922129826e-06,
+      "loss": 0.9104,
+      "step": 12791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.945048613421906e-06,
+      "loss": 0.4973,
+      "step": 12792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.943558412489967e-06,
+      "loss": 0.9078,
+      "step": 12793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.94206831937361e-06,
+      "loss": 0.8877,
+      "step": 12794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9405783341124435e-06,
+      "loss": 0.8714,
+      "step": 12795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.939088456746069e-06,
+      "loss": 0.9086,
+      "step": 12796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937598687314084e-06,
+      "loss": 0.8994,
+      "step": 12797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9361090258560875e-06,
+      "loss": 0.9029,
+      "step": 12798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934619472411671e-06,
+      "loss": 0.9324,
+      "step": 12799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.933130027020422e-06,
+      "loss": 0.9252,
+      "step": 12800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9316406897219344e-06,
+      "loss": 0.8658,
+      "step": 12801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9301514605557935e-06,
+      "loss": 0.8897,
+      "step": 12802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9286623395615764e-06,
+      "loss": 0.9372,
+      "step": 12803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.927173326778865e-06,
+      "loss": 0.8832,
+      "step": 12804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925684422247236e-06,
+      "loss": 0.8614,
+      "step": 12805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.92419562600626e-06,
+      "loss": 0.905,
+      "step": 12806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.922706938095512e-06,
+      "loss": 0.8963,
+      "step": 12807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.921218358554553e-06,
+      "loss": 0.9461,
+      "step": 12808
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919729887422955e-06,
+      "loss": 0.8659,
+      "step": 12809
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9182415247402736e-06,
+      "loss": 0.9296,
+      "step": 12810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.916753270546069e-06,
+      "loss": 0.8208,
+      "step": 12811
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.915265124879903e-06,
+      "loss": 0.8376,
+      "step": 12812
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913777087781323e-06,
+      "loss": 0.8868,
+      "step": 12813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.912289159289884e-06,
+      "loss": 0.8966,
+      "step": 12814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910801339445128e-06,
+      "loss": 0.9471,
+      "step": 12815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9093136282866014e-06,
+      "loss": 0.8504,
+      "step": 12816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.907826025853849e-06,
+      "loss": 0.8638,
+      "step": 12817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906338532186404e-06,
+      "loss": 0.7937,
+      "step": 12818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904851147323809e-06,
+      "loss": 0.3988,
+      "step": 12819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.90336387130559e-06,
+      "loss": 0.9089,
+      "step": 12820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901876704171283e-06,
+      "loss": 0.896,
+      "step": 12821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900389645960414e-06,
+      "loss": 0.9261,
+      "step": 12822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8989026967124996e-06,
+      "loss": 0.8508,
+      "step": 12823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8974158564670705e-06,
+      "loss": 0.9064,
+      "step": 12824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.895929125263645e-06,
+      "loss": 0.9271,
+      "step": 12825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.894442503141733e-06,
+      "loss": 0.9298,
+      "step": 12826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892955990140854e-06,
+      "loss": 0.9476,
+      "step": 12827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8914695863005136e-06,
+      "loss": 0.8907,
+      "step": 12828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889983291660215e-06,
+      "loss": 0.926,
+      "step": 12829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.888497106259471e-06,
+      "loss": 0.8724,
+      "step": 12830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887011030137775e-06,
+      "loss": 0.8451,
+      "step": 12831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88552506333463e-06,
+      "loss": 0.4041,
+      "step": 12832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884039205889529e-06,
+      "loss": 0.9148,
+      "step": 12833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.882553457841966e-06,
+      "loss": 0.8814,
+      "step": 12834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.881067819231425e-06,
+      "loss": 0.9365,
+      "step": 12835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879582290097402e-06,
+      "loss": 0.9117,
+      "step": 12836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878096870479373e-06,
+      "loss": 0.8412,
+      "step": 12837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.876611560416825e-06,
+      "loss": 0.8708,
+      "step": 12838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8751263599492295e-06,
+      "loss": 0.8632,
+      "step": 12839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.873641269116066e-06,
+      "loss": 0.848,
+      "step": 12840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.872156287956803e-06,
+      "loss": 0.8111,
+      "step": 12841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870671416510913e-06,
+      "loss": 0.8919,
+      "step": 12842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.869186654817861e-06,
+      "loss": 0.8679,
+      "step": 12843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.867702002917108e-06,
+      "loss": 0.9058,
+      "step": 12844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866217460848118e-06,
+      "loss": 0.9443,
+      "step": 12845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.864733028650344e-06,
+      "loss": 0.8709,
+      "step": 12846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.863248706363243e-06,
+      "loss": 0.9576,
+      "step": 12847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8617644940262696e-06,
+      "loss": 0.964,
+      "step": 12848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.86028039167887e-06,
+      "loss": 0.8483,
+      "step": 12849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.858796399360487e-06,
+      "loss": 0.9017,
+      "step": 12850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857312517110569e-06,
+      "loss": 0.921,
+      "step": 12851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85582874496855e-06,
+      "loss": 0.8248,
+      "step": 12852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854345082973871e-06,
+      "loss": 0.8638,
+      "step": 12853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852861531165966e-06,
+      "loss": 0.9091,
+      "step": 12854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.851378089584264e-06,
+      "loss": 0.8087,
+      "step": 12855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8498947582681955e-06,
+      "loss": 0.8898,
+      "step": 12856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.848411537257182e-06,
+      "loss": 0.9111,
+      "step": 12857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.84692842659065e-06,
+      "loss": 0.9274,
+      "step": 12858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8454454263080144e-06,
+      "loss": 0.9089,
+      "step": 12859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.843962536448695e-06,
+      "loss": 0.9218,
+      "step": 12860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8424797570521075e-06,
+      "loss": 0.9267,
+      "step": 12861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840997088157658e-06,
+      "loss": 0.9524,
+      "step": 12862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.839514529804757e-06,
+      "loss": 0.8871,
+      "step": 12863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838032082032808e-06,
+      "loss": 0.8432,
+      "step": 12864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836549744881212e-06,
+      "loss": 0.8614,
+      "step": 12865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.835067518389368e-06,
+      "loss": 0.8756,
+      "step": 12866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833585402596677e-06,
+      "loss": 0.948,
+      "step": 12867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.832103397542522e-06,
+      "loss": 0.8922,
+      "step": 12868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.830621503266301e-06,
+      "loss": 0.8611,
+      "step": 12869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8291397198073965e-06,
+      "loss": 0.8134,
+      "step": 12870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.827658047205195e-06,
+      "loss": 0.9617,
+      "step": 12871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826176485499077e-06,
+      "loss": 0.8508,
+      "step": 12872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.824695034728426e-06,
+      "loss": 0.9022,
+      "step": 12873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823213694932608e-06,
+      "loss": 0.9252,
+      "step": 12874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.821732466151e-06,
+      "loss": 0.9281,
+      "step": 12875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82025134842297e-06,
+      "loss": 0.9132,
+      "step": 12876
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8187703417878905e-06,
+      "loss": 0.8446,
+      "step": 12877
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.817289446285115e-06,
+      "loss": 0.9363,
+      "step": 12878
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81580866195401e-06,
+      "loss": 0.9411,
+      "step": 12879
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814327988833934e-06,
+      "loss": 0.8803,
+      "step": 12880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812847426964239e-06,
+      "loss": 0.9512,
+      "step": 12881
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81136697638427e-06,
+      "loss": 0.3445,
+      "step": 12882
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.809886637133393e-06,
+      "loss": 1.0033,
+      "step": 12883
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.808406409250938e-06,
+      "loss": 0.8916,
+      "step": 12884
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8069262927762535e-06,
+      "loss": 0.8973,
+      "step": 12885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805446287748684e-06,
+      "loss": 0.8939,
+      "step": 12886
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.803966394207557e-06,
+      "loss": 0.8975,
+      "step": 12887
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8024866121922105e-06,
+      "loss": 0.8849,
+      "step": 12888
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801006941741977e-06,
+      "loss": 0.875,
+      "step": 12889
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799527382896186e-06,
+      "loss": 0.8366,
+      "step": 12890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798047935694158e-06,
+      "loss": 0.9295,
+      "step": 12891
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.796568600175214e-06,
+      "loss": 0.8774,
+      "step": 12892
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7950893763786815e-06,
+      "loss": 0.9144,
+      "step": 12893
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.793610264343863e-06,
+      "loss": 0.9298,
+      "step": 12894
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.792131264110083e-06,
+      "loss": 0.8862,
+      "step": 12895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.790652375716653e-06,
+      "loss": 0.8858,
+      "step": 12896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789173599202869e-06,
+      "loss": 0.8794,
+      "step": 12897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787694934608044e-06,
+      "loss": 0.7915,
+      "step": 12898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786216381971479e-06,
+      "loss": 0.8722,
+      "step": 12899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7847379413324665e-06,
+      "loss": 0.9574,
+      "step": 12900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.783259612730306e-06,
+      "loss": 0.8572,
+      "step": 12901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7817813962042875e-06,
+      "loss": 0.9241,
+      "step": 12902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780303291793708e-06,
+      "loss": 0.8419,
+      "step": 12903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.778825299537842e-06,
+      "loss": 0.9148,
+      "step": 12904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777347419475978e-06,
+      "loss": 0.8551,
+      "step": 12905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7758696516473985e-06,
+      "loss": 0.905,
+      "step": 12906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.774391996091378e-06,
+      "loss": 0.9913,
+      "step": 12907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.772914452847194e-06,
+      "loss": 0.9764,
+      "step": 12908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771437021954118e-06,
+      "loss": 0.4585,
+      "step": 12909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.769959703451413e-06,
+      "loss": 0.8832,
+      "step": 12910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.768482497378348e-06,
+      "loss": 0.8778,
+      "step": 12911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767005403774189e-06,
+      "loss": 0.8684,
+      "step": 12912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765528422678188e-06,
+      "loss": 0.8619,
+      "step": 12913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.764051554129606e-06,
+      "loss": 0.8525,
+      "step": 12914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.762574798167695e-06,
+      "loss": 0.8648,
+      "step": 12915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7610981548317115e-06,
+      "loss": 0.9058,
+      "step": 12916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.759621624160893e-06,
+      "loss": 0.8751,
+      "step": 12917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7581452061944845e-06,
+      "loss": 0.8638,
+      "step": 12918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7566689009717405e-06,
+      "loss": 0.8979,
+      "step": 12919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755192708531885e-06,
+      "loss": 0.8993,
+      "step": 12920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753716628914161e-06,
+      "loss": 0.8698,
+      "step": 12921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.752240662157802e-06,
+      "loss": 0.8984,
+      "step": 12922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750764808302032e-06,
+      "loss": 0.782,
+      "step": 12923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.749289067386079e-06,
+      "loss": 0.904,
+      "step": 12924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.747813439449169e-06,
+      "loss": 0.8568,
+      "step": 12925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746337924530525e-06,
+      "loss": 0.8939,
+      "step": 12926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744862522669357e-06,
+      "loss": 0.4014,
+      "step": 12927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7433872339048824e-06,
+      "loss": 0.9252,
+      "step": 12928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.741912058276319e-06,
+      "loss": 0.8714,
+      "step": 12929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740436995822861e-06,
+      "loss": 0.9286,
+      "step": 12930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.738962046583727e-06,
+      "loss": 0.9996,
+      "step": 12931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7374872105981184e-06,
+      "loss": 0.9044,
+      "step": 12932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736012487905229e-06,
+      "loss": 0.9115,
+      "step": 12933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734537878544257e-06,
+      "loss": 0.9264,
+      "step": 12934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.733063382554401e-06,
+      "loss": 0.8655,
+      "step": 12935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.731588999974842e-06,
+      "loss": 0.841,
+      "step": 12936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7301147308447725e-06,
+      "loss": 0.9013,
+      "step": 12937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.728640575203377e-06,
+      "loss": 0.8709,
+      "step": 12938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.727166533089841e-06,
+      "loss": 0.8382,
+      "step": 12939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725692604543335e-06,
+      "loss": 0.8803,
+      "step": 12940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.724218789603036e-06,
+      "loss": 0.8483,
+      "step": 12941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.72274508830812e-06,
+      "loss": 0.958,
+      "step": 12942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.721271500697754e-06,
+      "loss": 0.8384,
+      "step": 12943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7197980268111045e-06,
+      "loss": 0.8482,
+      "step": 12944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.718324666687338e-06,
+      "loss": 0.9537,
+      "step": 12945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71685142036561e-06,
+      "loss": 0.9017,
+      "step": 12946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.715378287885078e-06,
+      "loss": 0.9172,
+      "step": 12947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713905269284902e-06,
+      "loss": 0.8876,
+      "step": 12948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712432364604223e-06,
+      "loss": 0.8831,
+      "step": 12949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.710959573882197e-06,
+      "loss": 0.9695,
+      "step": 12950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709486897157968e-06,
+      "loss": 0.9064,
+      "step": 12951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.708014334470678e-06,
+      "loss": 0.891,
+      "step": 12952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.706541885859459e-06,
+      "loss": 0.8728,
+      "step": 12953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.705069551363456e-06,
+      "loss": 0.877,
+      "step": 12954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.703597331021804e-06,
+      "loss": 0.9093,
+      "step": 12955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.702125224873624e-06,
+      "loss": 0.8801,
+      "step": 12956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.700653232958047e-06,
+      "loss": 0.9164,
+      "step": 12957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.699181355314199e-06,
+      "loss": 0.9013,
+      "step": 12958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697709591981196e-06,
+      "loss": 0.9044,
+      "step": 12959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696237942998158e-06,
+      "loss": 0.8854,
+      "step": 12960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.694766408404204e-06,
+      "loss": 0.4247,
+      "step": 12961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.693294988238438e-06,
+      "loss": 0.9265,
+      "step": 12962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691823682539973e-06,
+      "loss": 0.833,
+      "step": 12963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.690352491347913e-06,
+      "loss": 0.8588,
+      "step": 12964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.688881414701364e-06,
+      "loss": 0.8569,
+      "step": 12965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687410452639421e-06,
+      "loss": 0.8303,
+      "step": 12966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685939605201183e-06,
+      "loss": 0.8556,
+      "step": 12967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.684468872425748e-06,
+      "loss": 0.3883,
+      "step": 12968
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682998254352197e-06,
+      "loss": 0.8378,
+      "step": 12969
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681527751019622e-06,
+      "loss": 0.9184,
+      "step": 12970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680057362467111e-06,
+      "loss": 0.9552,
+      "step": 12971
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.678587088733737e-06,
+      "loss": 0.98,
+      "step": 12972
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.677116929858584e-06,
+      "loss": 0.8875,
+      "step": 12973
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675646885880725e-06,
+      "loss": 0.8815,
+      "step": 12974
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6741769568392365e-06,
+      "loss": 0.8731,
+      "step": 12975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.672707142773179e-06,
+      "loss": 0.9286,
+      "step": 12976
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671237443721627e-06,
+      "loss": 0.8705,
+      "step": 12977
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.669767859723636e-06,
+      "loss": 1.0003,
+      "step": 12978
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668298390818272e-06,
+      "loss": 0.8327,
+      "step": 12979
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6668290370445885e-06,
+      "loss": 0.9159,
+      "step": 12980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.665359798441646e-06,
+      "loss": 0.9331,
+      "step": 12981
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.663890675048483e-06,
+      "loss": 0.965,
+      "step": 12982
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.662421666904154e-06,
+      "loss": 0.8672,
+      "step": 12983
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660952774047708e-06,
+      "loss": 0.8055,
+      "step": 12984
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659483996518177e-06,
+      "loss": 0.9723,
+      "step": 12985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.658015334354604e-06,
+      "loss": 0.9913,
+      "step": 12986
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6565467875960244e-06,
+      "loss": 0.9291,
+      "step": 12987
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.655078356281475e-06,
+      "loss": 0.9205,
+      "step": 12988
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.653610040449973e-06,
+      "loss": 0.7939,
+      "step": 12989
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652141840140559e-06,
+      "loss": 0.8818,
+      "step": 12990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6506737553922454e-06,
+      "loss": 0.4283,
+      "step": 12991
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.649205786244056e-06,
+      "loss": 0.9863,
+      "step": 12992
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647737932735007e-06,
+      "loss": 0.9225,
+      "step": 12993
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646270194904118e-06,
+      "loss": 0.8014,
+      "step": 12994
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644802572790391e-06,
+      "loss": 0.8994,
+      "step": 12995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.643335066432837e-06,
+      "loss": 0.9035,
+      "step": 12996
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.641867675870465e-06,
+      "loss": 0.9844,
+      "step": 12997
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640400401142269e-06,
+      "loss": 1.0503,
+      "step": 12998
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638933242287251e-06,
+      "loss": 0.8921,
+      "step": 12999
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.637466199344407e-06,
+      "loss": 0.9257,
+      "step": 13000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.635999272352728e-06,
+      "loss": 0.9638,
+      "step": 13001
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.634532461351205e-06,
+      "loss": 0.8834,
+      "step": 13002
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.633065766378823e-06,
+      "loss": 0.917,
+      "step": 13003
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.63159918747457e-06,
+      "loss": 0.9525,
+      "step": 13004
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.630132724677417e-06,
+      "loss": 1.0225,
+      "step": 13005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628666378026345e-06,
+      "loss": 0.9279,
+      "step": 13006
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6272001475603325e-06,
+      "loss": 0.9594,
+      "step": 13007
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6257340333183415e-06,
+      "loss": 0.9667,
+      "step": 13008
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.624268035339344e-06,
+      "loss": 0.9085,
+      "step": 13009
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.622802153662309e-06,
+      "loss": 0.8914,
+      "step": 13010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.621336388326189e-06,
+      "loss": 0.9559,
+      "step": 13011
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.619870739369943e-06,
+      "loss": 0.8786,
+      "step": 13012
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618405206832539e-06,
+      "loss": 0.8852,
+      "step": 13013
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616939790752916e-06,
+      "loss": 0.8265,
+      "step": 13014
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615474491170027e-06,
+      "loss": 0.9171,
+      "step": 13015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614009308122818e-06,
+      "loss": 0.9222,
+      "step": 13016
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6125442416502375e-06,
+      "loss": 0.9999,
+      "step": 13017
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611079291791215e-06,
+      "loss": 0.8839,
+      "step": 13018
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6096144585846935e-06,
+      "loss": 0.823,
+      "step": 13019
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608149742069608e-06,
+      "loss": 0.9574,
+      "step": 13020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606685142284883e-06,
+      "loss": 1.0276,
+      "step": 13021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.60522065926945e-06,
+      "loss": 0.8911,
+      "step": 13022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603756293062232e-06,
+      "loss": 0.9389,
+      "step": 13023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.60229204370215e-06,
+      "loss": 0.9416,
+      "step": 13024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.600827911228124e-06,
+      "loss": 0.822,
+      "step": 13025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.599363895679071e-06,
+      "loss": 0.3959,
+      "step": 13026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.597899997093896e-06,
+      "loss": 0.94,
+      "step": 13027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.596436215511511e-06,
+      "loss": 0.882,
+      "step": 13028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594972550970823e-06,
+      "loss": 0.9306,
+      "step": 13029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.593509003510735e-06,
+      "loss": 0.8789,
+      "step": 13030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5920455731701435e-06,
+      "loss": 0.9096,
+      "step": 13031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.590582259987947e-06,
+      "loss": 0.7415,
+      "step": 13032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589119064003039e-06,
+      "loss": 0.8786,
+      "step": 13033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.587655985254308e-06,
+      "loss": 0.8723,
+      "step": 13034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.58619302378064e-06,
+      "loss": 0.8628,
+      "step": 13035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.584730179620919e-06,
+      "loss": 0.7952,
+      "step": 13036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.583267452814029e-06,
+      "loss": 0.9098,
+      "step": 13037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.581804843398844e-06,
+      "loss": 0.8956,
+      "step": 13038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5803423514142454e-06,
+      "loss": 0.8663,
+      "step": 13039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.578879976899095e-06,
+      "loss": 0.9021,
+      "step": 13040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.577417719892266e-06,
+      "loss": 0.7852,
+      "step": 13041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.575955580432622e-06,
+      "loss": 0.956,
+      "step": 13042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.574493558559031e-06,
+      "loss": 0.8884,
+      "step": 13043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5730316543103415e-06,
+      "loss": 0.862,
+      "step": 13044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5715698677254145e-06,
+      "loss": 0.8854,
+      "step": 13045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570108198843108e-06,
+      "loss": 0.8422,
+      "step": 13046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568646647702263e-06,
+      "loss": 0.9354,
+      "step": 13047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5671852143417225e-06,
+      "loss": 0.8535,
+      "step": 13048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.565723898800346e-06,
+      "loss": 1.0644,
+      "step": 13049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564262701116958e-06,
+      "loss": 0.8813,
+      "step": 13050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562801621330402e-06,
+      "loss": 0.4634,
+      "step": 13051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.561340659479513e-06,
+      "loss": 0.9424,
+      "step": 13052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559879815603122e-06,
+      "loss": 0.8593,
+      "step": 13053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55841908974005e-06,
+      "loss": 0.8491,
+      "step": 13054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.556958481929128e-06,
+      "loss": 0.9402,
+      "step": 13055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.555497992209178e-06,
+      "loss": 0.9404,
+      "step": 13056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55403762061901e-06,
+      "loss": 0.7602,
+      "step": 13057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.552577367197446e-06,
+      "loss": 0.8151,
+      "step": 13058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.551117231983295e-06,
+      "loss": 0.9177,
+      "step": 13059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549657215015367e-06,
+      "loss": 0.8838,
+      "step": 13060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5481973163324686e-06,
+      "loss": 0.9658,
+      "step": 13061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5467375359734036e-06,
+      "loss": 0.8327,
+      "step": 13062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.545277873976967e-06,
+      "loss": 0.889,
+      "step": 13063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.543818330381954e-06,
+      "loss": 0.9513,
+      "step": 13064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.542358905227161e-06,
+      "loss": 0.7816,
+      "step": 13065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.540899598551381e-06,
+      "loss": 0.8711,
+      "step": 13066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.539440410393395e-06,
+      "loss": 0.9042,
+      "step": 13067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537981340791987e-06,
+      "loss": 0.9602,
+      "step": 13068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5365223897859425e-06,
+      "loss": 0.8813,
+      "step": 13069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.535063557414031e-06,
+      "loss": 0.9043,
+      "step": 13070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533604843715028e-06,
+      "loss": 0.8803,
+      "step": 13071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.532146248727715e-06,
+      "loss": 0.9337,
+      "step": 13072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530687772490846e-06,
+      "loss": 0.9628,
+      "step": 13073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.529229415043194e-06,
+      "loss": 0.88,
+      "step": 13074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.52777117642352e-06,
+      "loss": 0.9429,
+      "step": 13075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.526313056670577e-06,
+      "loss": 0.9511,
+      "step": 13076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.524855055823124e-06,
+      "loss": 0.9098,
+      "step": 13077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523397173919911e-06,
+      "loss": 1.0096,
+      "step": 13078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.521939410999695e-06,
+      "loss": 0.9535,
+      "step": 13079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520481767101209e-06,
+      "loss": 0.9182,
+      "step": 13080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5190242422632024e-06,
+      "loss": 0.4252,
+      "step": 13081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5175668365244175e-06,
+      "loss": 0.9046,
+      "step": 13082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516109549923579e-06,
+      "loss": 0.8964,
+      "step": 13083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.51465238249943e-06,
+      "loss": 0.8744,
+      "step": 13084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.513195334290703e-06,
+      "loss": 0.8999,
+      "step": 13085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511738405336116e-06,
+      "loss": 0.8263,
+      "step": 13086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.510281595674396e-06,
+      "loss": 0.8743,
+      "step": 13087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.508824905344267e-06,
+      "loss": 0.9193,
+      "step": 13088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507368334384438e-06,
+      "loss": 0.8361,
+      "step": 13089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.505911882833629e-06,
+      "loss": 0.9084,
+      "step": 13090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5044555507305484e-06,
+      "loss": 0.8999,
+      "step": 13091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.50299933811391e-06,
+      "loss": 0.8923,
+      "step": 13092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.501543245022407e-06,
+      "loss": 0.9525,
+      "step": 13093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5000872714947496e-06,
+      "loss": 0.9107,
+      "step": 13094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498631417569631e-06,
+      "loss": 0.918,
+      "step": 13095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49717568328575e-06,
+      "loss": 0.8955,
+      "step": 13096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.495720068681795e-06,
+      "loss": 0.8426,
+      "step": 13097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49426457379646e-06,
+      "loss": 0.9599,
+      "step": 13098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492809198668424e-06,
+      "loss": 1.0315,
+      "step": 13099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49135394333637e-06,
+      "loss": 0.8552,
+      "step": 13100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.48989880783898e-06,
+      "loss": 0.9133,
+      "step": 13101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.488443792214933e-06,
+      "loss": 0.8967,
+      "step": 13102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.486988896502894e-06,
+      "loss": 0.8818,
+      "step": 13103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.485534120741534e-06,
+      "loss": 0.9702,
+      "step": 13104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.484079464969526e-06,
+      "loss": 0.9196,
+      "step": 13105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482624929225524e-06,
+      "loss": 0.9044,
+      "step": 13106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481170513548189e-06,
+      "loss": 0.9057,
+      "step": 13107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.479716217976188e-06,
+      "loss": 0.8904,
+      "step": 13108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4782620425481615e-06,
+      "loss": 0.8669,
+      "step": 13109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476807987302768e-06,
+      "loss": 0.8387,
+      "step": 13110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.475354052278654e-06,
+      "loss": 0.9629,
+      "step": 13111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.47390023751446e-06,
+      "loss": 0.9096,
+      "step": 13112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4724465430488275e-06,
+      "loss": 0.9498,
+      "step": 13113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470992968920394e-06,
+      "loss": 0.9465,
+      "step": 13114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4695395151678e-06,
+      "loss": 0.9421,
+      "step": 13115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.468086181829666e-06,
+      "loss": 0.8978,
+      "step": 13116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466632968944628e-06,
+      "loss": 0.8914,
+      "step": 13117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.465179876551305e-06,
+      "loss": 0.8573,
+      "step": 13118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463726904688324e-06,
+      "loss": 0.9466,
+      "step": 13119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4622740533943e-06,
+      "loss": 0.9653,
+      "step": 13120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.460821322707852e-06,
+      "loss": 0.924,
+      "step": 13121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.459368712667586e-06,
+      "loss": 0.9266,
+      "step": 13122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.457916223312114e-06,
+      "loss": 0.8824,
+      "step": 13123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456463854680045e-06,
+      "loss": 0.9541,
+      "step": 13124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.455011606809975e-06,
+      "loss": 0.8564,
+      "step": 13125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.453559479740502e-06,
+      "loss": 0.9456,
+      "step": 13126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452107473510228e-06,
+      "loss": 0.9236,
+      "step": 13127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.450655588157748e-06,
+      "loss": 0.8433,
+      "step": 13128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44920382372164e-06,
+      "loss": 0.8877,
+      "step": 13129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4477521802405e-06,
+      "loss": 0.8914,
+      "step": 13130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4463006577529056e-06,
+      "loss": 0.9584,
+      "step": 13131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44484925629744e-06,
+      "loss": 0.9291,
+      "step": 13132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4433979759126765e-06,
+      "loss": 0.8623,
+      "step": 13133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4419468166371966e-06,
+      "loss": 0.915,
+      "step": 13134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44049577850956e-06,
+      "loss": 0.9287,
+      "step": 13135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.439044861568339e-06,
+      "loss": 0.896,
+      "step": 13136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4375940658521e-06,
+      "loss": 0.8899,
+      "step": 13137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.436143391399395e-06,
+      "loss": 0.871,
+      "step": 13138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4346928382487874e-06,
+      "loss": 0.8423,
+      "step": 13139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43324240643883e-06,
+      "loss": 0.9017,
+      "step": 13140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.431792096008077e-06,
+      "loss": 0.8308,
+      "step": 13141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430341906995064e-06,
+      "loss": 0.8046,
+      "step": 13142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428891839438351e-06,
+      "loss": 0.8132,
+      "step": 13143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.427441893376474e-06,
+      "loss": 0.8244,
+      "step": 13144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425992068847965e-06,
+      "loss": 0.8614,
+      "step": 13145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.424542365891363e-06,
+      "loss": 0.8842,
+      "step": 13146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4230927845452045e-06,
+      "loss": 0.9454,
+      "step": 13147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.421643324848007e-06,
+      "loss": 0.9329,
+      "step": 13148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.420193986838303e-06,
+      "loss": 0.889,
+      "step": 13149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.418744770554609e-06,
+      "loss": 0.8149,
+      "step": 13150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417295676035452e-06,
+      "loss": 0.8914,
+      "step": 13151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415846703319339e-06,
+      "loss": 0.8864,
+      "step": 13152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.414397852444783e-06,
+      "loss": 0.387,
+      "step": 13153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.412949123450297e-06,
+      "loss": 0.9441,
+      "step": 13154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411500516374384e-06,
+      "loss": 0.8945,
+      "step": 13155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410052031255546e-06,
+      "loss": 0.3875,
+      "step": 13156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4086036681322874e-06,
+      "loss": 0.9479,
+      "step": 13157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407155427043095e-06,
+      "loss": 0.9002,
+      "step": 13158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405707308026465e-06,
+      "loss": 0.8413,
+      "step": 13159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.404259311120893e-06,
+      "loss": 0.8444,
+      "step": 13160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402811436364856e-06,
+      "loss": 0.8946,
+      "step": 13161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401363683796839e-06,
+      "loss": 0.9271,
+      "step": 13162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399916053455323e-06,
+      "loss": 0.8786,
+      "step": 13163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3984685453787895e-06,
+      "loss": 0.8905,
+      "step": 13164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397021159605702e-06,
+      "loss": 0.9026,
+      "step": 13165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.395573896174536e-06,
+      "loss": 0.8514,
+      "step": 13166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.394126755123756e-06,
+      "loss": 0.4376,
+      "step": 13167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.392679736491828e-06,
+      "loss": 0.8215,
+      "step": 13168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391232840317209e-06,
+      "loss": 0.867,
+      "step": 13169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.389786066638362e-06,
+      "loss": 0.9882,
+      "step": 13170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3883394154937305e-06,
+      "loss": 0.8761,
+      "step": 13171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.386892886921773e-06,
+      "loss": 0.8433,
+      "step": 13172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.385446480960936e-06,
+      "loss": 0.9403,
+      "step": 13173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384000197649658e-06,
+      "loss": 0.8698,
+      "step": 13174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382554037026382e-06,
+      "loss": 0.968,
+      "step": 13175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.381107999129547e-06,
+      "loss": 0.8632,
+      "step": 13176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3796620839975856e-06,
+      "loss": 0.971,
+      "step": 13177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.378216291668929e-06,
+      "loss": 0.9077,
+      "step": 13178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376770622182008e-06,
+      "loss": 0.9338,
+      "step": 13179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.375325075575242e-06,
+      "loss": 0.8313,
+      "step": 13180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.373879651887051e-06,
+      "loss": 0.4012,
+      "step": 13181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.372434351155858e-06,
+      "loss": 0.9726,
+      "step": 13182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.370989173420076e-06,
+      "loss": 0.8447,
+      "step": 13183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369544118718113e-06,
+      "loss": 0.9561,
+      "step": 13184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.368099187088378e-06,
+      "loss": 0.4431,
+      "step": 13185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366654378569281e-06,
+      "loss": 0.9693,
+      "step": 13186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365209693199214e-06,
+      "loss": 0.8968,
+      "step": 13187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.36376513101658e-06,
+      "loss": 0.9674,
+      "step": 13188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.362320692059774e-06,
+      "loss": 0.9153,
+      "step": 13189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.360876376367187e-06,
+      "loss": 0.4386,
+      "step": 13190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.359432183977207e-06,
+      "loss": 0.9001,
+      "step": 13191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357988114928221e-06,
+      "loss": 0.8252,
+      "step": 13192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.356544169258613e-06,
+      "loss": 0.9503,
+      "step": 13193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355100347006752e-06,
+      "loss": 0.8863,
+      "step": 13194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.353656648211021e-06,
+      "loss": 0.9397,
+      "step": 13195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.352213072909791e-06,
+      "loss": 0.9189,
+      "step": 13196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.350769621141427e-06,
+      "loss": 0.9142,
+      "step": 13197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.349326292944297e-06,
+      "loss": 0.8619,
+      "step": 13198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.347883088356762e-06,
+      "loss": 0.9057,
+      "step": 13199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346440007417185e-06,
+      "loss": 0.8449,
+      "step": 13200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.344997050163909e-06,
+      "loss": 0.9707,
+      "step": 13201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343554216635305e-06,
+      "loss": 0.9212,
+      "step": 13202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.342111506869707e-06,
+      "loss": 0.84,
+      "step": 13203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.340668920905464e-06,
+      "loss": 0.8337,
+      "step": 13204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.339226458780921e-06,
+      "loss": 0.8849,
+      "step": 13205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.337784120534419e-06,
+      "loss": 0.8929,
+      "step": 13206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.336341906204287e-06,
+      "loss": 0.9301,
+      "step": 13207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.33489981582886e-06,
+      "loss": 0.8966,
+      "step": 13208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.333457849446472e-06,
+      "loss": 0.8582,
+      "step": 13209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3320160070954395e-06,
+      "loss": 0.8607,
+      "step": 13210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.330574288814092e-06,
+      "loss": 0.8835,
+      "step": 13211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329132694640746e-06,
+      "loss": 0.86,
+      "step": 13212
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.327691224613719e-06,
+      "loss": 0.8952,
+      "step": 13213
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326249878771322e-06,
+      "loss": 0.883,
+      "step": 13214
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.324808657151869e-06,
+      "loss": 0.8815,
+      "step": 13215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.323367559793661e-06,
+      "loss": 0.9126,
+      "step": 13216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3219265867350005e-06,
+      "loss": 0.4399,
+      "step": 13217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.320485738014187e-06,
+      "loss": 0.8651,
+      "step": 13218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3190450136695236e-06,
+      "loss": 0.9004,
+      "step": 13219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.317604413739292e-06,
+      "loss": 0.885,
+      "step": 13220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.316163938261789e-06,
+      "loss": 0.9124,
+      "step": 13221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314723587275304e-06,
+      "loss": 0.8584,
+      "step": 13222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.313283360818109e-06,
+      "loss": 0.888,
+      "step": 13223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311843258928489e-06,
+      "loss": 0.8863,
+      "step": 13224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310403281644723e-06,
+      "loss": 0.9129,
+      "step": 13225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.308963429005081e-06,
+      "loss": 0.8955,
+      "step": 13226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.307523701047832e-06,
+      "loss": 0.8885,
+      "step": 13227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3060840978112485e-06,
+      "loss": 0.9561,
+      "step": 13228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3046446193335855e-06,
+      "loss": 0.9033,
+      "step": 13229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.303205265653104e-06,
+      "loss": 0.8998,
+      "step": 13230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.301766036808065e-06,
+      "loss": 0.909,
+      "step": 13231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.300326932836721e-06,
+      "loss": 0.9441,
+      "step": 13232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298887953777315e-06,
+      "loss": 0.9998,
+      "step": 13233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2974490996680996e-06,
+      "loss": 0.8765,
+      "step": 13234
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2960103705473196e-06,
+      "loss": 0.8912,
+      "step": 13235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.294571766453207e-06,
+      "loss": 0.9546,
+      "step": 13236
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.293133287424e-06,
+      "loss": 0.8917,
+      "step": 13237
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.291694933497942e-06,
+      "loss": 0.9441,
+      "step": 13238
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2902567047132505e-06,
+      "loss": 0.9403,
+      "step": 13239
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288818601108157e-06,
+      "loss": 0.9045,
+      "step": 13240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.287380622720886e-06,
+      "loss": 0.8746,
+      "step": 13241
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.285942769589658e-06,
+      "loss": 0.8758,
+      "step": 13242
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.284505041752684e-06,
+      "loss": 0.8677,
+      "step": 13243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.283067439248179e-06,
+      "loss": 0.8986,
+      "step": 13244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281629962114358e-06,
+      "loss": 0.9233,
+      "step": 13245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.280192610389422e-06,
+      "loss": 0.8693,
+      "step": 13246
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.278755384111575e-06,
+      "loss": 0.9533,
+      "step": 13247
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277318283319017e-06,
+      "loss": 0.9335,
+      "step": 13248
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275881308049946e-06,
+      "loss": 0.8576,
+      "step": 13249
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.274444458342554e-06,
+      "loss": 0.9324,
+      "step": 13250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273007734235036e-06,
+      "loss": 0.799,
+      "step": 13251
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.271571135765568e-06,
+      "loss": 0.9715,
+      "step": 13252
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270134662972341e-06,
+      "loss": 0.8731,
+      "step": 13253
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.26869831589353e-06,
+      "loss": 0.8776,
+      "step": 13254
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.26726209456732e-06,
+      "loss": 0.9564,
+      "step": 13255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2658259990318734e-06,
+      "loss": 0.8826,
+      "step": 13256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.264390029325366e-06,
+      "loss": 0.8833,
+      "step": 13257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262954185485965e-06,
+      "loss": 0.9149,
+      "step": 13258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.26151846755183e-06,
+      "loss": 0.9335,
+      "step": 13259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.260082875561117e-06,
+      "loss": 0.914,
+      "step": 13260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2586474095519956e-06,
+      "loss": 0.9056,
+      "step": 13261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.257212069562608e-06,
+      "loss": 0.8782,
+      "step": 13262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255776855631105e-06,
+      "loss": 0.8579,
+      "step": 13263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.254341767795639e-06,
+      "loss": 0.8482,
+      "step": 13264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.252906806094345e-06,
+      "loss": 0.8811,
+      "step": 13265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251471970565367e-06,
+      "loss": 0.9577,
+      "step": 13266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.250037261246841e-06,
+      "loss": 0.8183,
+      "step": 13267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.248602678176902e-06,
+      "loss": 0.874,
+      "step": 13268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247168221393675e-06,
+      "loss": 0.8732,
+      "step": 13269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.245733890935287e-06,
+      "loss": 0.9579,
+      "step": 13270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244299686839867e-06,
+      "loss": 0.8384,
+      "step": 13271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242865609145522e-06,
+      "loss": 0.8693,
+      "step": 13272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241431657890379e-06,
+      "loss": 0.896,
+      "step": 13273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.239997833112552e-06,
+      "loss": 0.9034,
+      "step": 13274
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.238564134850142e-06,
+      "loss": 0.9554,
+      "step": 13275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237130563141261e-06,
+      "loss": 0.8885,
+      "step": 13276
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235697118024012e-06,
+      "loss": 0.8358,
+      "step": 13277
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.234263799536488e-06,
+      "loss": 0.9041,
+      "step": 13278
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.232830607716791e-06,
+      "loss": 0.8523,
+      "step": 13279
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231397542603011e-06,
+      "loss": 0.9058,
+      "step": 13280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229964604233242e-06,
+      "loss": 0.9324,
+      "step": 13281
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228531792645561e-06,
+      "loss": 0.8455,
+      "step": 13282
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.227099107878058e-06,
+      "loss": 0.9251,
+      "step": 13283
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.225666549968806e-06,
+      "loss": 0.9375,
+      "step": 13284
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.224234118955886e-06,
+      "loss": 0.8827,
+      "step": 13285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.22280181487737e-06,
+      "loss": 0.9495,
+      "step": 13286
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221369637771327e-06,
+      "loss": 0.9238,
+      "step": 13287
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.21993758767582e-06,
+      "loss": 0.8398,
+      "step": 13288
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218505664628911e-06,
+      "loss": 0.9247,
+      "step": 13289
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.217073868668661e-06,
+      "loss": 0.8189,
+      "step": 13290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.215642199833129e-06,
+      "loss": 0.7885,
+      "step": 13291
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214210658160359e-06,
+      "loss": 0.8085,
+      "step": 13292
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.212779243688404e-06,
+      "loss": 0.9222,
+      "step": 13293
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.211347956455313e-06,
+      "loss": 0.904,
+      "step": 13294
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2099167964991195e-06,
+      "loss": 0.4499,
+      "step": 13295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208485763857863e-06,
+      "loss": 0.8534,
+      "step": 13296
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.20705485856959e-06,
+      "loss": 0.8685,
+      "step": 13297
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.205624080672322e-06,
+      "loss": 0.8906,
+      "step": 13298
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204193430204089e-06,
+      "loss": 0.9205,
+      "step": 13299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202762907202923e-06,
+      "loss": 0.909,
+      "step": 13300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201332511706833e-06,
+      "loss": 0.9526,
+      "step": 13301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.199902243753847e-06,
+      "loss": 0.8452,
+      "step": 13302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1984721033819775e-06,
+      "loss": 0.9635,
+      "step": 13303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.197042090629238e-06,
+      "loss": 0.9275,
+      "step": 13304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.195612205533631e-06,
+      "loss": 0.91,
+      "step": 13305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.194182448133163e-06,
+      "loss": 0.9135,
+      "step": 13306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.192752818465838e-06,
+      "loss": 0.8965,
+      "step": 13307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.191323316569653e-06,
+      "loss": 0.9786,
+      "step": 13308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1898939424826024e-06,
+      "loss": 0.8088,
+      "step": 13309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1884646962426785e-06,
+      "loss": 0.4004,
+      "step": 13310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.187035577887867e-06,
+      "loss": 0.8443,
+      "step": 13311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.185606587456151e-06,
+      "loss": 0.895,
+      "step": 13312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.184177724985517e-06,
+      "loss": 0.9072,
+      "step": 13313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182748990513937e-06,
+      "loss": 0.8934,
+      "step": 13314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181320384079386e-06,
+      "loss": 0.9159,
+      "step": 13315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.179891905719836e-06,
+      "loss": 0.7835,
+      "step": 13316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1784635554732565e-06,
+      "loss": 0.9087,
+      "step": 13317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.177035333377607e-06,
+      "loss": 0.8948,
+      "step": 13318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.175607239470844e-06,
+      "loss": 0.8851,
+      "step": 13319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.17417927379094e-06,
+      "loss": 0.9302,
+      "step": 13320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.172751436375835e-06,
+      "loss": 0.8136,
+      "step": 13321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.171323727263484e-06,
+      "loss": 0.8232,
+      "step": 13322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.169896146491836e-06,
+      "loss": 0.8707,
+      "step": 13323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.168468694098829e-06,
+      "loss": 0.8965,
+      "step": 13324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.167041370122405e-06,
+      "loss": 0.8791,
+      "step": 13325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1656141746005075e-06,
+      "loss": 0.8994,
+      "step": 13326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.164187107571059e-06,
+      "loss": 0.838,
+      "step": 13327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.162760169071994e-06,
+      "loss": 0.8652,
+      "step": 13328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161333359141238e-06,
+      "loss": 0.893,
+      "step": 13329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159906677816721e-06,
+      "loss": 0.8847,
+      "step": 13330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158480125136347e-06,
+      "loss": 0.9258,
+      "step": 13331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.157053701138046e-06,
+      "loss": 0.8471,
+      "step": 13332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.155627405859731e-06,
+      "loss": 0.9198,
+      "step": 13333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.154201239339302e-06,
+      "loss": 0.882,
+      "step": 13334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.15277520161467e-06,
+      "loss": 0.9695,
+      "step": 13335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.151349292723742e-06,
+      "loss": 0.8882,
+      "step": 13336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149923512704406e-06,
+      "loss": 0.9098,
+      "step": 13337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.148497861594567e-06,
+      "loss": 0.8982,
+      "step": 13338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.147072339432111e-06,
+      "loss": 0.8447,
+      "step": 13339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.145646946254935e-06,
+      "loss": 0.8924,
+      "step": 13340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.144221682100914e-06,
+      "loss": 0.9388,
+      "step": 13341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.142796547007935e-06,
+      "loss": 0.8069,
+      "step": 13342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.141371541013877e-06,
+      "loss": 0.8782,
+      "step": 13343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139946664156614e-06,
+      "loss": 0.9524,
+      "step": 13344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1385219164740194e-06,
+      "loss": 0.9276,
+      "step": 13345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137097298003962e-06,
+      "loss": 0.4252,
+      "step": 13346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135672808784301e-06,
+      "loss": 0.8567,
+      "step": 13347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134248448852901e-06,
+      "loss": 0.9189,
+      "step": 13348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132824218247625e-06,
+      "loss": 0.8813,
+      "step": 13349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.131400117006318e-06,
+      "loss": 0.7813,
+      "step": 13350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.129976145166835e-06,
+      "loss": 0.9452,
+      "step": 13351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1285523027670225e-06,
+      "loss": 0.8718,
+      "step": 13352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127128589844732e-06,
+      "loss": 0.9406,
+      "step": 13353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.125705006437794e-06,
+      "loss": 0.8554,
+      "step": 13354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12428155258405e-06,
+      "loss": 0.8484,
+      "step": 13355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.122858228321333e-06,
+      "loss": 0.8939,
+      "step": 13356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121435033687474e-06,
+      "loss": 0.8932,
+      "step": 13357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.120011968720302e-06,
+      "loss": 0.931,
+      "step": 13358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.118589033457641e-06,
+      "loss": 0.9422,
+      "step": 13359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117166227937303e-06,
+      "loss": 0.8953,
+      "step": 13360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.115743552197112e-06,
+      "loss": 0.9269,
+      "step": 13361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114321006274881e-06,
+      "loss": 0.844,
+      "step": 13362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.112898590208415e-06,
+      "loss": 0.8558,
+      "step": 13363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111476304035523e-06,
+      "loss": 0.9929,
+      "step": 13364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.110054147794008e-06,
+      "loss": 0.8776,
+      "step": 13365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108632121521669e-06,
+      "loss": 0.8121,
+      "step": 13366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1072102252563004e-06,
+      "loss": 0.87,
+      "step": 13367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1057884590356966e-06,
+      "loss": 0.8423,
+      "step": 13368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.10436682289765e-06,
+      "loss": 0.8476,
+      "step": 13369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.102945316879937e-06,
+      "loss": 0.8205,
+      "step": 13370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1015239410203454e-06,
+      "loss": 0.9587,
+      "step": 13371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.100102695356657e-06,
+      "loss": 0.8802,
+      "step": 13372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.098681579926637e-06,
+      "loss": 0.9247,
+      "step": 13373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.097260594768065e-06,
+      "loss": 0.8472,
+      "step": 13374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0958397399187095e-06,
+      "loss": 0.885,
+      "step": 13375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.09441901541633e-06,
+      "loss": 0.9064,
+      "step": 13376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0929984212986895e-06,
+      "loss": 0.9224,
+      "step": 13377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.091577957603547e-06,
+      "loss": 0.9476,
+      "step": 13378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0901576243686554e-06,
+      "loss": 0.8134,
+      "step": 13379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.088737421631767e-06,
+      "loss": 0.8395,
+      "step": 13380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0873173494306296e-06,
+      "loss": 0.9506,
+      "step": 13381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085897407802988e-06,
+      "loss": 0.8424,
+      "step": 13382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.084477596786579e-06,
+      "loss": 0.8945,
+      "step": 13383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0830579164191414e-06,
+      "loss": 0.8803,
+      "step": 13384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081638366738413e-06,
+      "loss": 0.8935,
+      "step": 13385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080218947782114e-06,
+      "loss": 0.8337,
+      "step": 13386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.078799659587977e-06,
+      "loss": 0.8718,
+      "step": 13387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077380502193725e-06,
+      "loss": 0.8417,
+      "step": 13388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075961475637081e-06,
+      "loss": 0.9282,
+      "step": 13389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074542579955748e-06,
+      "loss": 0.8408,
+      "step": 13390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073123815187457e-06,
+      "loss": 0.9289,
+      "step": 13391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071705181369904e-06,
+      "loss": 0.968,
+      "step": 13392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.070286678540799e-06,
+      "loss": 0.9428,
+      "step": 13393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.068868306737842e-06,
+      "loss": 0.9213,
+      "step": 13394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.067450065998739e-06,
+      "loss": 0.9053,
+      "step": 13395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.066031956361175e-06,
+      "loss": 0.8528,
+      "step": 13396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.064613977862847e-06,
+      "loss": 0.8514,
+      "step": 13397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0631961305414466e-06,
+      "loss": 0.9374,
+      "step": 13398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.06177841443465e-06,
+      "loss": 0.93,
+      "step": 13399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060360829580145e-06,
+      "loss": 0.8623,
+      "step": 13400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.058943376015606e-06,
+      "loss": 0.8422,
+      "step": 13401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057526053778709e-06,
+      "loss": 0.8802,
+      "step": 13402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056108862907126e-06,
+      "loss": 0.9312,
+      "step": 13403
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.054691803438525e-06,
+      "loss": 0.9229,
+      "step": 13404
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0532748754105645e-06,
+      "loss": 0.9251,
+      "step": 13405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051858078860907e-06,
+      "loss": 0.9361,
+      "step": 13406
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.050441413827212e-06,
+      "loss": 0.7664,
+      "step": 13407
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049024880347135e-06,
+      "loss": 0.9268,
+      "step": 13408
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.047608478458319e-06,
+      "loss": 0.8796,
+      "step": 13409
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.046192208198411e-06,
+      "loss": 0.9072,
+      "step": 13410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044776069605062e-06,
+      "loss": 0.994,
+      "step": 13411
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.043360062715902e-06,
+      "loss": 0.9048,
+      "step": 13412
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.04194418756857e-06,
+      "loss": 0.8578,
+      "step": 13413
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.040528444200698e-06,
+      "loss": 0.8731,
+      "step": 13414
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.039112832649917e-06,
+      "loss": 0.8893,
+      "step": 13415
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.037697352953851e-06,
+      "loss": 0.4217,
+      "step": 13416
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0362820051501216e-06,
+      "loss": 0.929,
+      "step": 13417
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.034866789276352e-06,
+      "loss": 1.0034,
+      "step": 13418
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.033451705370147e-06,
+      "loss": 0.7899,
+      "step": 13419
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.032036753469124e-06,
+      "loss": 0.4089,
+      "step": 13420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030621933610893e-06,
+      "loss": 0.8506,
+      "step": 13421
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0292072458330535e-06,
+      "loss": 0.8187,
+      "step": 13422
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.027792690173208e-06,
+      "loss": 0.819,
+      "step": 13423
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0263782666689565e-06,
+      "loss": 0.8828,
+      "step": 13424
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.024963975357883e-06,
+      "loss": 0.9551,
+      "step": 13425
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0235498162775885e-06,
+      "loss": 0.9216,
+      "step": 13426
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0221357894656605e-06,
+      "loss": 0.9039,
+      "step": 13427
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.020721894959675e-06,
+      "loss": 0.3994,
+      "step": 13428
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019308132797214e-06,
+      "loss": 0.9257,
+      "step": 13429
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.017894503015855e-06,
+      "loss": 0.9746,
+      "step": 13430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016481005653171e-06,
+      "loss": 0.86,
+      "step": 13431
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.015067640746729e-06,
+      "loss": 0.4016,
+      "step": 13432
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.013654408334094e-06,
+      "loss": 0.908,
+      "step": 13433
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.012241308452837e-06,
+      "loss": 0.8972,
+      "step": 13434
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010828341140502e-06,
+      "loss": 0.8669,
+      "step": 13435
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.009415506434654e-06,
+      "loss": 0.8969,
+      "step": 13436
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00800280437284e-06,
+      "loss": 0.9237,
+      "step": 13437
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006590234992612e-06,
+      "loss": 0.9209,
+      "step": 13438
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005177798331511e-06,
+      "loss": 0.8583,
+      "step": 13439
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.003765494427084e-06,
+      "loss": 0.8837,
+      "step": 13440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00235332331686e-06,
+      "loss": 0.8746,
+      "step": 13441
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000941285038376e-06,
+      "loss": 0.9276,
+      "step": 13442
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.999529379629164e-06,
+      "loss": 0.9636,
+      "step": 13443
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.998117607126755e-06,
+      "loss": 0.8997,
+      "step": 13444
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996705967568662e-06,
+      "loss": 0.898,
+      "step": 13445
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99529446099241e-06,
+      "loss": 0.8856,
+      "step": 13446
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99388308743552e-06,
+      "loss": 0.8881,
+      "step": 13447
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.992471846935495e-06,
+      "loss": 0.8933,
+      "step": 13448
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991060739529846e-06,
+      "loss": 0.8851,
+      "step": 13449
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989649765256089e-06,
+      "loss": 0.8051,
+      "step": 13450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9882389241517156e-06,
+      "loss": 0.8172,
+      "step": 13451
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986828216254228e-06,
+      "loss": 0.8369,
+      "step": 13452
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9854176416011246e-06,
+      "loss": 0.8741,
+      "step": 13453
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984007200229889e-06,
+      "loss": 0.9167,
+      "step": 13454
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.982596892178012e-06,
+      "loss": 0.8821,
+      "step": 13455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9811867174829806e-06,
+      "loss": 0.9356,
+      "step": 13456
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979776676182277e-06,
+      "loss": 0.912,
+      "step": 13457
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.978366768313373e-06,
+      "loss": 0.85,
+      "step": 13458
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976956993913745e-06,
+      "loss": 0.9084,
+      "step": 13459
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975547353020867e-06,
+      "loss": 0.9501,
+      "step": 13460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974137845672193e-06,
+      "loss": 0.8872,
+      "step": 13461
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.972728471905201e-06,
+      "loss": 0.8416,
+      "step": 13462
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971319231757346e-06,
+      "loss": 0.8403,
+      "step": 13463
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.96991012526608e-06,
+      "loss": 0.9607,
+      "step": 13464
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968501152468857e-06,
+      "loss": 0.8713,
+      "step": 13465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967092313403128e-06,
+      "loss": 0.8455,
+      "step": 13466
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965683608106341e-06,
+      "loss": 0.8917,
+      "step": 13467
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964275036615929e-06,
+      "loss": 0.4418,
+      "step": 13468
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.962866598969336e-06,
+      "loss": 0.8804,
+      "step": 13469
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961458295203999e-06,
+      "loss": 0.8743,
+      "step": 13470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.960050125357343e-06,
+      "loss": 0.9231,
+      "step": 13471
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.958642089466799e-06,
+      "loss": 0.9372,
+      "step": 13472
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957234187569788e-06,
+      "loss": 0.8855,
+      "step": 13473
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.955826419703736e-06,
+      "loss": 0.9006,
+      "step": 13474
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9544187859060544e-06,
+      "loss": 0.9186,
+      "step": 13475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.953011286214163e-06,
+      "loss": 0.9316,
+      "step": 13476
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.951603920665465e-06,
+      "loss": 0.9731,
+      "step": 13477
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.950196689297368e-06,
+      "loss": 0.8788,
+      "step": 13478
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948789592147275e-06,
+      "loss": 0.8775,
+      "step": 13479
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.947382629252591e-06,
+      "loss": 0.8959,
+      "step": 13480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945975800650701e-06,
+      "loss": 1.0007,
+      "step": 13481
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.944569106379003e-06,
+      "loss": 0.9129,
+      "step": 13482
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.943162546474886e-06,
+      "loss": 0.9071,
+      "step": 13483
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.941756120975727e-06,
+      "loss": 0.9052,
+      "step": 13484
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9403498299189166e-06,
+      "loss": 0.9239,
+      "step": 13485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938943673341834e-06,
+      "loss": 0.8349,
+      "step": 13486
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937537651281844e-06,
+      "loss": 0.8924,
+      "step": 13487
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.93613176377632e-06,
+      "loss": 0.8446,
+      "step": 13488
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934726010862636e-06,
+      "loss": 0.8913,
+      "step": 13489
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.933320392578145e-06,
+      "loss": 0.9674,
+      "step": 13490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9319149089602106e-06,
+      "loss": 0.8669,
+      "step": 13491
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930509560046189e-06,
+      "loss": 0.9312,
+      "step": 13492
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.929104345873439e-06,
+      "loss": 0.9958,
+      "step": 13493
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927699266479299e-06,
+      "loss": 0.8784,
+      "step": 13494
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.926294321901119e-06,
+      "loss": 0.9068,
+      "step": 13495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.924889512176243e-06,
+      "loss": 0.9308,
+      "step": 13496
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.923484837342006e-06,
+      "loss": 0.8877,
+      "step": 13497
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.922080297435745e-06,
+      "loss": 0.9871,
+      "step": 13498
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920675892494793e-06,
+      "loss": 0.907,
+      "step": 13499
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.919271622556472e-06,
+      "loss": 0.8964,
+      "step": 13500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.917867487658107e-06,
+      "loss": 0.9035,
+      "step": 13501
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916463487837024e-06,
+      "loss": 0.9519,
+      "step": 13502
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.915059623130532e-06,
+      "loss": 0.8904,
+      "step": 13503
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9136558935759486e-06,
+      "loss": 0.956,
+      "step": 13504
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912252299210581e-06,
+      "loss": 0.8738,
+      "step": 13505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910848840071741e-06,
+      "loss": 0.9066,
+      "step": 13506
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.909445516196723e-06,
+      "loss": 1.0313,
+      "step": 13507
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908042327622825e-06,
+      "loss": 0.9175,
+      "step": 13508
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906639274387354e-06,
+      "loss": 0.835,
+      "step": 13509
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90523635652759e-06,
+      "loss": 0.9278,
+      "step": 13510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903833574080825e-06,
+      "loss": 0.9028,
+      "step": 13511
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9024309270843475e-06,
+      "loss": 0.9129,
+      "step": 13512
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901028415575428e-06,
+      "loss": 0.8747,
+      "step": 13513
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899626039591352e-06,
+      "loss": 0.8327,
+      "step": 13514
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89822379916939e-06,
+      "loss": 0.8825,
+      "step": 13515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.896821694346816e-06,
+      "loss": 0.3799,
+      "step": 13516
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8954197251608906e-06,
+      "loss": 0.9431,
+      "step": 13517
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.894017891648877e-06,
+      "loss": 0.9166,
+      "step": 13518
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892616193848041e-06,
+      "loss": 0.9027,
+      "step": 13519
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.891214631795626e-06,
+      "loss": 0.7991,
+      "step": 13520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.889813205528895e-06,
+      "loss": 0.7988,
+      "step": 13521
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888411915085096e-06,
+      "loss": 0.8869,
+      "step": 13522
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.887010760501466e-06,
+      "loss": 0.9731,
+      "step": 13523
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885609741815251e-06,
+      "loss": 0.88,
+      "step": 13524
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8842088590636925e-06,
+      "loss": 0.9249,
+      "step": 13525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.882808112284015e-06,
+      "loss": 0.9203,
+      "step": 13526
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881407501513453e-06,
+      "loss": 0.7736,
+      "step": 13527
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.880007026789235e-06,
+      "loss": 0.8524,
+      "step": 13528
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.878606688148586e-06,
+      "loss": 0.9074,
+      "step": 13529
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877206485628718e-06,
+      "loss": 0.9178,
+      "step": 13530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.875806419266851e-06,
+      "loss": 0.8237,
+      "step": 13531
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874406489100196e-06,
+      "loss": 0.935,
+      "step": 13532
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873006695165963e-06,
+      "loss": 0.4149,
+      "step": 13533
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.871607037501357e-06,
+      "loss": 0.8975,
+      "step": 13534
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.870207516143582e-06,
+      "loss": 0.9006,
+      "step": 13535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86880813112983e-06,
+      "loss": 0.8614,
+      "step": 13536
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8674088824972964e-06,
+      "loss": 0.9169,
+      "step": 13537
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.866009770283177e-06,
+      "loss": 0.8912,
+      "step": 13538
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864610794524649e-06,
+      "loss": 0.9644,
+      "step": 13539
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.863211955258902e-06,
+      "loss": 0.9291,
+      "step": 13540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861813252523114e-06,
+      "loss": 0.8169,
+      "step": 13541
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8604146863544666e-06,
+      "loss": 0.919,
+      "step": 13542
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.859016256790123e-06,
+      "loss": 0.8913,
+      "step": 13543
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.857617963867254e-06,
+      "loss": 0.9989,
+      "step": 13544
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856219807623027e-06,
+      "loss": 0.8902,
+      "step": 13545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854821788094602e-06,
+      "loss": 0.9331,
+      "step": 13546
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.853423905319139e-06,
+      "loss": 0.8652,
+      "step": 13547
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8520261593337945e-06,
+      "loss": 0.8793,
+      "step": 13548
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8506285501757096e-06,
+      "loss": 0.9487,
+      "step": 13549
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849231077882038e-06,
+      "loss": 0.845,
+      "step": 13550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.847833742489926e-06,
+      "loss": 0.8236,
+      "step": 13551
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.846436544036505e-06,
+      "loss": 0.9058,
+      "step": 13552
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845039482558913e-06,
+      "loss": 0.4763,
+      "step": 13553
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.843642558094285e-06,
+      "loss": 0.9114,
+      "step": 13554
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842245770679749e-06,
+      "loss": 0.922,
+      "step": 13555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840849120352429e-06,
+      "loss": 0.9056,
+      "step": 13556
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.839452607149448e-06,
+      "loss": 0.8693,
+      "step": 13557
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838056231107926e-06,
+      "loss": 0.8748,
+      "step": 13558
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.836659992264969e-06,
+      "loss": 0.8973,
+      "step": 13559
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.835263890657694e-06,
+      "loss": 0.9445,
+      "step": 13560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8338679263232095e-06,
+      "loss": 0.9085,
+      "step": 13561
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.832472099298611e-06,
+      "loss": 1.0035,
+      "step": 13562
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831076409621003e-06,
+      "loss": 0.9876,
+      "step": 13563
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.829680857327479e-06,
+      "loss": 0.8598,
+      "step": 13564
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.828285442455138e-06,
+      "loss": 0.9607,
+      "step": 13565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826890165041059e-06,
+      "loss": 0.8869,
+      "step": 13566
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.825495025122331e-06,
+      "loss": 0.8969,
+      "step": 13567
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.824100022736036e-06,
+      "loss": 0.9513,
+      "step": 13568
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82270515791925e-06,
+      "loss": 0.9409,
+      "step": 13569
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8213104307090486e-06,
+      "loss": 0.9691,
+      "step": 13570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819915841142505e-06,
+      "loss": 0.9419,
+      "step": 13571
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818521389256678e-06,
+      "loss": 0.9258,
+      "step": 13572
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8171270750886344e-06,
+      "loss": 0.875,
+      "step": 13573
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8157328986754385e-06,
+      "loss": 0.8803,
+      "step": 13574
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.814338860054138e-06,
+      "loss": 0.8567,
+      "step": 13575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812944959261786e-06,
+      "loss": 0.9753,
+      "step": 13576
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.811551196335434e-06,
+      "loss": 0.9518,
+      "step": 13577
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810157571312129e-06,
+      "loss": 0.8671,
+      "step": 13578
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8087640842289e-06,
+      "loss": 0.7692,
+      "step": 13579
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8073707351228005e-06,
+      "loss": 0.9122,
+      "step": 13580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.805977524030854e-06,
+      "loss": 0.8859,
+      "step": 13581
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804584450990091e-06,
+      "loss": 0.9077,
+      "step": 13582
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.803191516037541e-06,
+      "loss": 0.9063,
+      "step": 13583
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801798719210228e-06,
+      "loss": 0.924,
+      "step": 13584
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800406060545164e-06,
+      "loss": 0.8801,
+      "step": 13585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799013540079369e-06,
+      "loss": 0.8284,
+      "step": 13586
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797621157849858e-06,
+      "loss": 0.7973,
+      "step": 13587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.796228913893632e-06,
+      "loss": 0.8713,
+      "step": 13588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.794836808247696e-06,
+      "loss": 0.9109,
+      "step": 13589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.793444840949054e-06,
+      "loss": 0.8911,
+      "step": 13590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.792053012034702e-06,
+      "loss": 0.8681,
+      "step": 13591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.790661321541632e-06,
+      "loss": 0.8929,
+      "step": 13592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.789269769506839e-06,
+      "loss": 0.85,
+      "step": 13593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7878783559673005e-06,
+      "loss": 0.9534,
+      "step": 13594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.786487080960001e-06,
+      "loss": 0.8269,
+      "step": 13595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.785095944521923e-06,
+      "loss": 0.8844,
+      "step": 13596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783704946690041e-06,
+      "loss": 0.983,
+      "step": 13597
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.782314087501321e-06,
+      "loss": 0.8072,
+      "step": 13598
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7809233669927325e-06,
+      "loss": 0.4274,
+      "step": 13599
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.779532785201244e-06,
+      "loss": 0.8851,
+      "step": 13600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778142342163808e-06,
+      "loss": 0.9446,
+      "step": 13601
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776752037917385e-06,
+      "loss": 0.9209,
+      "step": 13602
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775361872498927e-06,
+      "loss": 0.8458,
+      "step": 13603
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.773971845945383e-06,
+      "loss": 0.9814,
+      "step": 13604
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772581958293699e-06,
+      "loss": 0.8834,
+      "step": 13605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7711922095808174e-06,
+      "loss": 0.8718,
+      "step": 13606
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769802599843678e-06,
+      "loss": 0.8175,
+      "step": 13607
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.768413129119207e-06,
+      "loss": 0.8647,
+      "step": 13608
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.767023797444341e-06,
+      "loss": 0.9957,
+      "step": 13609
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76563460485601e-06,
+      "loss": 0.8976,
+      "step": 13610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.764245551391129e-06,
+      "loss": 0.8884,
+      "step": 13611
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762856637086622e-06,
+      "loss": 0.8808,
+      "step": 13612
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.761467861979403e-06,
+      "loss": 0.8486,
+      "step": 13613
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760079226106386e-06,
+      "loss": 0.8297,
+      "step": 13614
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758690729504478e-06,
+      "loss": 0.841,
+      "step": 13615
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.757302372210588e-06,
+      "loss": 0.8774,
+      "step": 13616
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7559141542616095e-06,
+      "loss": 0.948,
+      "step": 13617
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.754526075694445e-06,
+      "loss": 0.9478,
+      "step": 13618
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753138136545984e-06,
+      "loss": 0.8686,
+      "step": 13619
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751750336853123e-06,
+      "loss": 0.9338,
+      "step": 13620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.750362676652742e-06,
+      "loss": 0.8891,
+      "step": 13621
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.748975155981724e-06,
+      "loss": 0.9022,
+      "step": 13622
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7475877748769535e-06,
+      "loss": 0.8177,
+      "step": 13623
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.746200533375298e-06,
+      "loss": 0.8053,
+      "step": 13624
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.744813431513631e-06,
+      "loss": 0.9065,
+      "step": 13625
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.743426469328821e-06,
+      "loss": 0.9175,
+      "step": 13626
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7420396468577315e-06,
+      "loss": 0.9398,
+      "step": 13627
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740652964137224e-06,
+      "loss": 0.9529,
+      "step": 13628
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7392664212041575e-06,
+      "loss": 0.8458,
+      "step": 13629
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737880018095377e-06,
+      "loss": 0.9619,
+      "step": 13630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7364937548477354e-06,
+      "loss": 0.9135,
+      "step": 13631
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735107631498078e-06,
+      "loss": 0.9076,
+      "step": 13632
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.733721648083251e-06,
+      "loss": 0.9005,
+      "step": 13633
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7323358046400844e-06,
+      "loss": 0.7908,
+      "step": 13634
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730950101205416e-06,
+      "loss": 0.8982,
+      "step": 13635
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729564537816081e-06,
+      "loss": 0.8842,
+      "step": 13636
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.728179114508895e-06,
+      "loss": 0.9533,
+      "step": 13637
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726793831320684e-06,
+      "loss": 0.9213,
+      "step": 13638
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.725408688288279e-06,
+      "loss": 0.8185,
+      "step": 13639
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.724023685448482e-06,
+      "loss": 0.9414,
+      "step": 13640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72263882283811e-06,
+      "loss": 0.886,
+      "step": 13641
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721254100493975e-06,
+      "loss": 0.8829,
+      "step": 13642
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.719869518452873e-06,
+      "loss": 0.9008,
+      "step": 13643
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.718485076751609e-06,
+      "loss": 0.8661,
+      "step": 13644
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717100775426979e-06,
+      "loss": 0.8538,
+      "step": 13645
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.71571661451578e-06,
+      "loss": 0.8605,
+      "step": 13646
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7143325940547964e-06,
+      "loss": 0.9324,
+      "step": 13647
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.712948714080814e-06,
+      "loss": 0.9246,
+      "step": 13648
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.711564974630621e-06,
+      "loss": 0.8284,
+      "step": 13649
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710181375740983e-06,
+      "loss": 0.827,
+      "step": 13650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7087979174486866e-06,
+      "loss": 0.8938,
+      "step": 13651
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.707414599790502e-06,
+      "loss": 0.8702,
+      "step": 13652
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706031422803189e-06,
+      "loss": 0.9178,
+      "step": 13653
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704648386523514e-06,
+      "loss": 0.9586,
+      "step": 13654
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.703265490988237e-06,
+      "loss": 0.9045,
+      "step": 13655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701882736234118e-06,
+      "loss": 0.8517,
+      "step": 13656
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.700500122297901e-06,
+      "loss": 0.867,
+      "step": 13657
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6991176492163386e-06,
+      "loss": 0.9311,
+      "step": 13658
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.697735317026178e-06,
+      "loss": 0.8821,
+      "step": 13659
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.696353125764152e-06,
+      "loss": 0.967,
+      "step": 13660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.694971075467005e-06,
+      "loss": 0.9321,
+      "step": 13661
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693589166171466e-06,
+      "loss": 0.8903,
+      "step": 13662
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692207397914268e-06,
+      "loss": 0.9168,
+      "step": 13663
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690825770732135e-06,
+      "loss": 0.9352,
+      "step": 13664
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6894442846617916e-06,
+      "loss": 0.8641,
+      "step": 13665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6880629397399514e-06,
+      "loss": 0.8813,
+      "step": 13666
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6866817360033316e-06,
+      "loss": 0.94,
+      "step": 13667
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6853006734886415e-06,
+      "loss": 0.8898,
+      "step": 13668
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683919752232595e-06,
+      "loss": 0.8096,
+      "step": 13669
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682538972271885e-06,
+      "loss": 0.8949,
+      "step": 13670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.681158333643216e-06,
+      "loss": 0.9307,
+      "step": 13671
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679777836383289e-06,
+      "loss": 0.9174,
+      "step": 13672
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6783974805287825e-06,
+      "loss": 0.8238,
+      "step": 13673
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.677017266116396e-06,
+      "loss": 0.9804,
+      "step": 13674
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675637193182816e-06,
+      "loss": 0.854,
+      "step": 13675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674257261764714e-06,
+      "loss": 0.9197,
+      "step": 13676
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.672877471898771e-06,
+      "loss": 0.9118,
+      "step": 13677
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.671497823621666e-06,
+      "loss": 0.8878,
+      "step": 13678
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670118316970058e-06,
+      "loss": 0.8915,
+      "step": 13679
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668738951980618e-06,
+      "loss": 0.9131,
+      "step": 13680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6673597286900075e-06,
+      "loss": 0.8881,
+      "step": 13681
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.66598064713489e-06,
+      "loss": 0.8917,
+      "step": 13682
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.66460170735191e-06,
+      "loss": 0.9697,
+      "step": 13683
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663222909377724e-06,
+      "loss": 0.419,
+      "step": 13684
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.661844253248976e-06,
+      "loss": 0.8725,
+      "step": 13685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.660465739002312e-06,
+      "loss": 0.355,
+      "step": 13686
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6590873666743705e-06,
+      "loss": 0.4079,
+      "step": 13687
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6577091363017915e-06,
+      "loss": 1.054,
+      "step": 13688
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.656331047921198e-06,
+      "loss": 0.9151,
+      "step": 13689
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.654953101569221e-06,
+      "loss": 0.9127,
+      "step": 13690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65357529728249e-06,
+      "loss": 0.9377,
+      "step": 13691
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.652197635097618e-06,
+      "loss": 0.8382,
+      "step": 13692
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.650820115051226e-06,
+      "loss": 0.9097,
+      "step": 13693
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6494427371799244e-06,
+      "loss": 0.8831,
+      "step": 13694
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6480655015203295e-06,
+      "loss": 0.4188,
+      "step": 13695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646688408109035e-06,
+      "loss": 0.9659,
+      "step": 13696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645311456982647e-06,
+      "loss": 0.8783,
+      "step": 13697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.643934648177771e-06,
+      "loss": 0.9294,
+      "step": 13698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.642557981730992e-06,
+      "loss": 0.8506,
+      "step": 13699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641181457678902e-06,
+      "loss": 0.938,
+      "step": 13700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.639805076058092e-06,
+      "loss": 0.9108,
+      "step": 13701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638428836905139e-06,
+      "loss": 0.9192,
+      "step": 13702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.637052740256623e-06,
+      "loss": 0.9183,
+      "step": 13703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63567678614912e-06,
+      "loss": 0.7873,
+      "step": 13704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634300974619206e-06,
+      "loss": 0.8175,
+      "step": 13705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63292530570344e-06,
+      "loss": 0.858,
+      "step": 13706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631549779438389e-06,
+      "loss": 0.8652,
+      "step": 13707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.630174395860617e-06,
+      "loss": 0.8668,
+      "step": 13708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628799155006669e-06,
+      "loss": 0.8467,
+      "step": 13709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62742405691311e-06,
+      "loss": 0.9063,
+      "step": 13710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626049101616486e-06,
+      "loss": 0.8711,
+      "step": 13711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.624674289153336e-06,
+      "loss": 0.9455,
+      "step": 13712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6232996195602046e-06,
+      "loss": 0.8745,
+      "step": 13713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621925092873631e-06,
+      "loss": 0.9474,
+      "step": 13714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620550709130144e-06,
+      "loss": 0.8702,
+      "step": 13715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.619176468366274e-06,
+      "loss": 0.9608,
+      "step": 13716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.617802370618548e-06,
+      "loss": 0.8901,
+      "step": 13717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.616428415923493e-06,
+      "loss": 0.8647,
+      "step": 13718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615054604317617e-06,
+      "loss": 0.8879,
+      "step": 13719
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61368093583744e-06,
+      "loss": 0.866,
+      "step": 13720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.612307410519472e-06,
+      "loss": 0.8818,
+      "step": 13721
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.610934028400219e-06,
+      "loss": 0.904,
+      "step": 13722
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609560789516185e-06,
+      "loss": 0.8803,
+      "step": 13723
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.608187693903873e-06,
+      "loss": 0.9143,
+      "step": 13724
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60681474159977e-06,
+      "loss": 0.9021,
+      "step": 13725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605441932640371e-06,
+      "loss": 0.8721,
+      "step": 13726
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.604069267062169e-06,
+      "loss": 0.8093,
+      "step": 13727
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602696744901638e-06,
+      "loss": 0.9975,
+      "step": 13728
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.601324366195265e-06,
+      "loss": 0.9291,
+      "step": 13729
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.599952130979522e-06,
+      "loss": 0.9046,
+      "step": 13730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598580039290889e-06,
+      "loss": 1.0146,
+      "step": 13731
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.597208091165823e-06,
+      "loss": 0.7697,
+      "step": 13732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595836286640799e-06,
+      "loss": 0.8603,
+      "step": 13733
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.594464625752278e-06,
+      "loss": 0.8917,
+      "step": 13734
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593093108536708e-06,
+      "loss": 0.4329,
+      "step": 13735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.59172173503055e-06,
+      "loss": 0.9149,
+      "step": 13736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.590350505270253e-06,
+      "loss": 0.8643,
+      "step": 13737
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.58897941929226e-06,
+      "loss": 0.8488,
+      "step": 13738
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587608477133012e-06,
+      "loss": 0.9402,
+      "step": 13739
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.586237678828955e-06,
+      "loss": 0.8753,
+      "step": 13740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584867024416511e-06,
+      "loss": 0.7964,
+      "step": 13741
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583496513932119e-06,
+      "loss": 0.7967,
+      "step": 13742
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5821261474122025e-06,
+      "loss": 0.9349,
+      "step": 13743
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580755924893184e-06,
+      "loss": 0.9392,
+      "step": 13744
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.579385846411486e-06,
+      "loss": 0.9173,
+      "step": 13745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.57801591200352e-06,
+      "loss": 0.9584,
+      "step": 13746
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576646121705702e-06,
+      "loss": 0.8591,
+      "step": 13747
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575276475554433e-06,
+      "loss": 0.8701,
+      "step": 13748
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.57390697358612e-06,
+      "loss": 0.8884,
+      "step": 13749
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572537615837165e-06,
+      "loss": 0.767,
+      "step": 13750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571168402343959e-06,
+      "loss": 0.9089,
+      "step": 13751
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.569799333142894e-06,
+      "loss": 0.9096,
+      "step": 13752
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.568430408270363e-06,
+      "loss": 0.7845,
+      "step": 13753
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567061627762752e-06,
+      "loss": 0.8994,
+      "step": 13754
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565692991656434e-06,
+      "loss": 0.935,
+      "step": 13755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56432449998779e-06,
+      "loss": 0.8972,
+      "step": 13756
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562956152793192e-06,
+      "loss": 0.8721,
+      "step": 13757
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56158795010901e-06,
+      "loss": 0.9083,
+      "step": 13758
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560219891971609e-06,
+      "loss": 0.8919,
+      "step": 13759
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.558851978417355e-06,
+      "loss": 1.0283,
+      "step": 13760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.557484209482595e-06,
+      "loss": 0.8487,
+      "step": 13761
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556116585203692e-06,
+      "loss": 1.0127,
+      "step": 13762
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5547491056169935e-06,
+      "loss": 0.8942,
+      "step": 13763
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5533817707588425e-06,
+      "loss": 0.8847,
+      "step": 13764
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.552014580665582e-06,
+      "loss": 0.8751,
+      "step": 13765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550647535373554e-06,
+      "loss": 0.8266,
+      "step": 13766
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549280634919093e-06,
+      "loss": 0.8813,
+      "step": 13767
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.54791387933852e-06,
+      "loss": 0.876,
+      "step": 13768
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546547268668176e-06,
+      "loss": 0.4633,
+      "step": 13769
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545180802944375e-06,
+      "loss": 0.9734,
+      "step": 13770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.543814482203438e-06,
+      "loss": 0.8725,
+      "step": 13771
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542448306481679e-06,
+      "loss": 0.8348,
+      "step": 13772
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.541082275815416e-06,
+      "loss": 0.9048,
+      "step": 13773
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.539716390240948e-06,
+      "loss": 0.8746,
+      "step": 13774
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538350649794582e-06,
+      "loss": 0.8615,
+      "step": 13775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.53698505451262e-06,
+      "loss": 0.871,
+      "step": 13776
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535619604431354e-06,
+      "loss": 0.88,
+      "step": 13777
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.534254299587077e-06,
+      "loss": 0.833,
+      "step": 13778
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532889140016079e-06,
+      "loss": 0.9528,
+      "step": 13779
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.531524125754643e-06,
+      "loss": 0.8691,
+      "step": 13780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.53015925683905e-06,
+      "loss": 0.8863,
+      "step": 13781
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5287945333055774e-06,
+      "loss": 0.4502,
+      "step": 13782
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527429955190501e-06,
+      "loss": 0.855,
+      "step": 13783
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.526065522530082e-06,
+      "loss": 0.929,
+      "step": 13784
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.52470123536059e-06,
+      "loss": 0.8856,
+      "step": 13785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.523337093718288e-06,
+      "loss": 0.8391,
+      "step": 13786
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.52197309763943e-06,
+      "loss": 0.8724,
+      "step": 13787
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520609247160269e-06,
+      "loss": 0.8744,
+      "step": 13788
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51924554231706e-06,
+      "loss": 0.9485,
+      "step": 13789
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517881983146041e-06,
+      "loss": 0.8809,
+      "step": 13790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516518569683455e-06,
+      "loss": 0.9961,
+      "step": 13791
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515155301965549e-06,
+      "loss": 0.9123,
+      "step": 13792
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.513792180028547e-06,
+      "loss": 0.9064,
+      "step": 13793
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512429203908684e-06,
+      "loss": 0.9003,
+      "step": 13794
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511066373642185e-06,
+      "loss": 0.9821,
+      "step": 13795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.509703689265276e-06,
+      "loss": 0.8064,
+      "step": 13796
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.508341150814168e-06,
+      "loss": 0.9655,
+      "step": 13797
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5069787583250815e-06,
+      "loss": 0.8271,
+      "step": 13798
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50561651183423e-06,
+      "loss": 0.9302,
+      "step": 13799
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5042544113778116e-06,
+      "loss": 0.9368,
+      "step": 13800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502892456992035e-06,
+      "loss": 0.9355,
+      "step": 13801
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.501530648713099e-06,
+      "loss": 0.8998,
+      "step": 13802
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.500168986577197e-06,
+      "loss": 0.8384,
+      "step": 13803
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498807470620522e-06,
+      "loss": 0.9511,
+      "step": 13804
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.497446100879266e-06,
+      "loss": 0.8901,
+      "step": 13805
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.496084877389605e-06,
+      "loss": 0.9671,
+      "step": 13806
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494723800187721e-06,
+      "loss": 0.906,
+      "step": 13807
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493362869309791e-06,
+      "loss": 0.8286,
+      "step": 13808
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.49200208479199e-06,
+      "loss": 0.892,
+      "step": 13809
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4906414466704796e-06,
+      "loss": 0.8762,
+      "step": 13810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.489280954981428e-06,
+      "loss": 0.9199,
+      "step": 13811
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487920609760996e-06,
+      "loss": 0.9735,
+      "step": 13812
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4865604110453375e-06,
+      "loss": 0.8271,
+      "step": 13813
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485200358870605e-06,
+      "loss": 0.9272,
+      "step": 13814
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48384045327295e-06,
+      "loss": 1.0079,
+      "step": 13815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482480694288515e-06,
+      "loss": 0.9234,
+      "step": 13816
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4811210819534405e-06,
+      "loss": 0.9066,
+      "step": 13817
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.479761616303869e-06,
+      "loss": 0.9349,
+      "step": 13818
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4784022973759254e-06,
+      "loss": 0.9293,
+      "step": 13819
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.477043125205742e-06,
+      "loss": 0.8948,
+      "step": 13820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4756840998294445e-06,
+      "loss": 0.914,
+      "step": 13821
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.474325221283159e-06,
+      "loss": 0.9287,
+      "step": 13822
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4729664896029935e-06,
+      "loss": 0.9198,
+      "step": 13823
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.471607904825067e-06,
+      "loss": 0.8961,
+      "step": 13824
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.470249466985491e-06,
+      "loss": 0.8473,
+      "step": 13825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468891176120366e-06,
+      "loss": 0.827,
+      "step": 13826
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.467533032265792e-06,
+      "loss": 0.9304,
+      "step": 13827
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46617503545788e-06,
+      "loss": 0.8977,
+      "step": 13828
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46481718573271e-06,
+      "loss": 0.9271,
+      "step": 13829
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4634594831263775e-06,
+      "loss": 0.9297,
+      "step": 13830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.462101927674969e-06,
+      "loss": 0.892,
+      "step": 13831
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460744519414571e-06,
+      "loss": 0.8957,
+      "step": 13832
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.459387258381251e-06,
+      "loss": 0.9605,
+      "step": 13833
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4580301446110915e-06,
+      "loss": 0.8883,
+      "step": 13834
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.456673178140165e-06,
+      "loss": 0.965,
+      "step": 13835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.45531635900453e-06,
+      "loss": 0.846,
+      "step": 13836
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453959687240255e-06,
+      "loss": 0.4238,
+      "step": 13837
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4526031628834e-06,
+      "loss": 1.0161,
+      "step": 13838
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451246785970009e-06,
+      "loss": 0.9162,
+      "step": 13839
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449890556536146e-06,
+      "loss": 0.9398,
+      "step": 13840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.448534474617856e-06,
+      "loss": 0.8456,
+      "step": 13841
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.447178540251177e-06,
+      "loss": 0.8566,
+      "step": 13842
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4458227534721486e-06,
+      "loss": 0.8122,
+      "step": 13843
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444467114316809e-06,
+      "loss": 0.8959,
+      "step": 13844
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443111622821192e-06,
+      "loss": 0.8712,
+      "step": 13845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.441756279021319e-06,
+      "loss": 0.8648,
+      "step": 13846
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440401082953214e-06,
+      "loss": 0.8344,
+      "step": 13847
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.439046034652903e-06,
+      "loss": 0.8836,
+      "step": 13848
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437691134156393e-06,
+      "loss": 0.8774,
+      "step": 13849
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.436336381499701e-06,
+      "loss": 0.8512,
+      "step": 13850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434981776718833e-06,
+      "loss": 0.8918,
+      "step": 13851
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433627319849793e-06,
+      "loss": 0.9808,
+      "step": 13852
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.432273010928583e-06,
+      "loss": 0.8744,
+      "step": 13853
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4309188499912e-06,
+      "loss": 0.9481,
+      "step": 13854
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42956483707363e-06,
+      "loss": 0.822,
+      "step": 13855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.428210972211865e-06,
+      "loss": 0.9052,
+      "step": 13856
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.426857255441887e-06,
+      "loss": 0.9733,
+      "step": 13857
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.425503686799684e-06,
+      "loss": 0.8886,
+      "step": 13858
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424150266321222e-06,
+      "loss": 0.8985,
+      "step": 13859
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422796994042476e-06,
+      "loss": 0.7733,
+      "step": 13860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.421443869999422e-06,
+      "loss": 0.901,
+      "step": 13861
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420090894228011e-06,
+      "loss": 0.968,
+      "step": 13862
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.418738066764213e-06,
+      "loss": 0.3633,
+      "step": 13863
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.417385387643989e-06,
+      "loss": 0.8808,
+      "step": 13864
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41603285690328e-06,
+      "loss": 0.7714,
+      "step": 13865
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41468047457804e-06,
+      "loss": 0.9066,
+      "step": 13866
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.413328240704219e-06,
+      "loss": 0.8584,
+      "step": 13867
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411976155317748e-06,
+      "loss": 0.9065,
+      "step": 13868
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.410624218454569e-06,
+      "loss": 0.8125,
+      "step": 13869
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.409272430150614e-06,
+      "loss": 0.8961,
+      "step": 13870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4079207904418175e-06,
+      "loss": 0.8597,
+      "step": 13871
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.406569299364095e-06,
+      "loss": 0.8867,
+      "step": 13872
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405217956953373e-06,
+      "loss": 0.9381,
+      "step": 13873
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.403866763245567e-06,
+      "loss": 0.9024,
+      "step": 13874
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402515718276591e-06,
+      "loss": 0.8353,
+      "step": 13875
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.401164822082355e-06,
+      "loss": 0.9187,
+      "step": 13876
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.399814074698767e-06,
+      "loss": 0.912,
+      "step": 13877
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.398463476161721e-06,
+      "loss": 0.8984,
+      "step": 13878
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.397113026507118e-06,
+      "loss": 0.8642,
+      "step": 13879
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.395762725770852e-06,
+      "loss": 0.8508,
+      "step": 13880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.394412573988816e-06,
+      "loss": 0.9236,
+      "step": 13881
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.393062571196888e-06,
+      "loss": 0.8689,
+      "step": 13882
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.391712717430953e-06,
+      "loss": 1.0223,
+      "step": 13883
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.390363012726891e-06,
+      "loss": 0.9271,
+      "step": 13884
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.389013457120571e-06,
+      "loss": 0.8883,
+      "step": 13885
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.387664050647862e-06,
+      "loss": 0.8992,
+      "step": 13886
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.386314793344638e-06,
+      "loss": 0.9298,
+      "step": 13887
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.384965685246753e-06,
+      "loss": 0.8551,
+      "step": 13888
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.383616726390066e-06,
+      "loss": 0.9291,
+      "step": 13889
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.382267916810437e-06,
+      "loss": 0.9402,
+      "step": 13890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3809192565437065e-06,
+      "loss": 0.947,
+      "step": 13891
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379570745625724e-06,
+      "loss": 0.8491,
+      "step": 13892
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378222384092333e-06,
+      "loss": 0.9127,
+      "step": 13893
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3768741719793725e-06,
+      "loss": 0.887,
+      "step": 13894
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.375526109322672e-06,
+      "loss": 0.851,
+      "step": 13895
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374178196158063e-06,
+      "loss": 0.874,
+      "step": 13896
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.372830432521377e-06,
+      "loss": 0.8828,
+      "step": 13897
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.371482818448422e-06,
+      "loss": 0.8677,
+      "step": 13898
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.370135353975031e-06,
+      "loss": 1.0093,
+      "step": 13899
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.368788039137015e-06,
+      "loss": 0.8314,
+      "step": 13900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367440873970178e-06,
+      "loss": 0.8897,
+      "step": 13901
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36609385851033e-06,
+      "loss": 0.9604,
+      "step": 13902
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.364746992793275e-06,
+      "loss": 0.4036,
+      "step": 13903
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.363400276854806e-06,
+      "loss": 0.8875,
+      "step": 13904
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36205371073072e-06,
+      "loss": 0.8729,
+      "step": 13905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.360707294456807e-06,
+      "loss": 0.8672,
+      "step": 13906
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3593610280688585e-06,
+      "loss": 0.8619,
+      "step": 13907
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.358014911602646e-06,
+      "loss": 0.8479,
+      "step": 13908
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.356668945093955e-06,
+      "loss": 0.8617,
+      "step": 13909
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.355323128578557e-06,
+      "loss": 0.8967,
+      "step": 13910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.353977462092224e-06,
+      "loss": 0.9452,
+      "step": 13911
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.352631945670721e-06,
+      "loss": 0.9104,
+      "step": 13912
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.351286579349815e-06,
+      "loss": 0.8605,
+      "step": 13913
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.349941363165257e-06,
+      "loss": 0.8724,
+      "step": 13914
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.348596297152804e-06,
+      "loss": 0.8851,
+      "step": 13915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347251381348211e-06,
+      "loss": 0.8311,
+      "step": 13916
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3459066157872165e-06,
+      "loss": 0.8604,
+      "step": 13917
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.344562000505567e-06,
+      "loss": 0.9469,
+      "step": 13918
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.343217535538999e-06,
+      "loss": 0.9092,
+      "step": 13919
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.341873220923253e-06,
+      "loss": 0.82,
+      "step": 13920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3405290566940475e-06,
+      "loss": 0.9091,
+      "step": 13921
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.339185042887118e-06,
+      "loss": 0.9259,
+      "step": 13922
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.33784117953819e-06,
+      "loss": 0.8806,
+      "step": 13923
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.336497466682973e-06,
+      "loss": 0.9152,
+      "step": 13924
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3351539043571846e-06,
+      "loss": 0.9015,
+      "step": 13925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333810492596539e-06,
+      "loss": 0.9061,
+      "step": 13926
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.332467231436735e-06,
+      "loss": 0.8394,
+      "step": 13927
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.33112412091348e-06,
+      "loss": 0.9326,
+      "step": 13928
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.329781161062471e-06,
+      "loss": 0.383,
+      "step": 13929
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3284383519194075e-06,
+      "loss": 0.9389,
+      "step": 13930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.327095693519971e-06,
+      "loss": 0.9114,
+      "step": 13931
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.325753185899851e-06,
+      "loss": 0.8126,
+      "step": 13932
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.32441082909473e-06,
+      "loss": 0.9215,
+      "step": 13933
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3230686231402885e-06,
+      "loss": 0.9481,
+      "step": 13934
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321726568072199e-06,
+      "loss": 0.8975,
+      "step": 13935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320384663926135e-06,
+      "loss": 0.8959,
+      "step": 13936
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.319042910737756e-06,
+      "loss": 0.8272,
+      "step": 13937
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.31770130854273e-06,
+      "loss": 0.92,
+      "step": 13938
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3163598573767165e-06,
+      "loss": 0.839,
+      "step": 13939
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3150185572753624e-06,
+      "loss": 0.9431,
+      "step": 13940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.313677408274322e-06,
+      "loss": 0.9182,
+      "step": 13941
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312336410409241e-06,
+      "loss": 0.9837,
+      "step": 13942
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.310995563715767e-06,
+      "loss": 0.8001,
+      "step": 13943
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3096548682295304e-06,
+      "loss": 0.8579,
+      "step": 13944
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.308314323986168e-06,
+      "loss": 0.3806,
+      "step": 13945
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.30697393102131e-06,
+      "loss": 0.8521,
+      "step": 13946
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.305633689370583e-06,
+      "loss": 0.8201,
+      "step": 13947
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.304293599069609e-06,
+      "loss": 0.9573,
+      "step": 13948
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3029536601540086e-06,
+      "loss": 0.8532,
+      "step": 13949
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3016138726593905e-06,
+      "loss": 0.4087,
+      "step": 13950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3002742366213676e-06,
+      "loss": 0.9845,
+      "step": 13951
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.29893475207555e-06,
+      "loss": 0.8257,
+      "step": 13952
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297595419057531e-06,
+      "loss": 0.8974,
+      "step": 13953
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.296256237602912e-06,
+      "loss": 0.9613,
+      "step": 13954
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.29491720774729e-06,
+      "loss": 0.9347,
+      "step": 13955
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293578329526256e-06,
+      "loss": 0.8651,
+      "step": 13956
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.292239602975384e-06,
+      "loss": 0.943,
+      "step": 13957
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2909010281302734e-06,
+      "loss": 0.8824,
+      "step": 13958
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.289562605026489e-06,
+      "loss": 0.9349,
+      "step": 13959
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288224333699608e-06,
+      "loss": 0.8528,
+      "step": 13960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.286886214185201e-06,
+      "loss": 0.3824,
+      "step": 13961
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.285548246518837e-06,
+      "loss": 0.8191,
+      "step": 13962
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.28421043073607e-06,
+      "loss": 0.8897,
+      "step": 13963
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.282872766872462e-06,
+      "loss": 0.8906,
+      "step": 13964
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.28153525496357e-06,
+      "loss": 0.9413,
+      "step": 13965
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.280197895044937e-06,
+      "loss": 0.822,
+      "step": 13966
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.27886068715211e-06,
+      "loss": 0.9276,
+      "step": 13967
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2775236313206324e-06,
+      "loss": 0.4629,
+      "step": 13968
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2761867275860406e-06,
+      "loss": 0.9146,
+      "step": 13969
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274849975983869e-06,
+      "loss": 1.0189,
+      "step": 13970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2735133765496465e-06,
+      "loss": 0.8658,
+      "step": 13971
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272176929318901e-06,
+      "loss": 1.0017,
+      "step": 13972
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.270840634327147e-06,
+      "loss": 0.8498,
+      "step": 13973
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.269504491609906e-06,
+      "loss": 0.9205,
+      "step": 13974
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.268168501202695e-06,
+      "loss": 0.9209,
+      "step": 13975
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.266832663141015e-06,
+      "loss": 0.8056,
+      "step": 13976
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.265496977460375e-06,
+      "loss": 0.9342,
+      "step": 13977
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264161444196279e-06,
+      "loss": 0.8493,
+      "step": 13978
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.262826063384217e-06,
+      "loss": 0.8918,
+      "step": 13979
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2614908350596825e-06,
+      "loss": 0.4788,
+      "step": 13980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260155759258175e-06,
+      "loss": 0.9344,
+      "step": 13981
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.258820836015168e-06,
+      "loss": 0.9094,
+      "step": 13982
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.257486065366146e-06,
+      "loss": 0.8785,
+      "step": 13983
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.256151447346587e-06,
+      "loss": 0.915,
+      "step": 13984
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2548169819919645e-06,
+      "loss": 0.7921,
+      "step": 13985
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253482669337743e-06,
+      "loss": 0.8935,
+      "step": 13986
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.252148509419388e-06,
+      "loss": 0.3973,
+      "step": 13987
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.250814502272365e-06,
+      "loss": 0.9603,
+      "step": 13988
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249480647932121e-06,
+      "loss": 0.829,
+      "step": 13989
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.248146946434115e-06,
+      "loss": 0.8537,
+      "step": 13990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2468133978137945e-06,
+      "loss": 0.9179,
+      "step": 13991
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.245480002106602e-06,
+      "loss": 0.8326,
+      "step": 13992
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.24414675934798e-06,
+      "loss": 0.9289,
+      "step": 13993
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.242813669573366e-06,
+      "loss": 0.8932,
+      "step": 13994
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241480732818186e-06,
+      "loss": 0.8683,
+      "step": 13995
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240147949117871e-06,
+      "loss": 0.9141,
+      "step": 13996
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238815318507845e-06,
+      "loss": 0.859,
+      "step": 13997
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.237482841023533e-06,
+      "loss": 0.8801,
+      "step": 13998
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23615051670034e-06,
+      "loss": 0.9494,
+      "step": 13999
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.234818345573685e-06,
+      "loss": 0.4853,
+      "step": 14000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.233486327678976e-06,
+      "loss": 0.388,
+      "step": 14001
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.232154463051612e-06,
+      "loss": 0.9193,
+      "step": 14002
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.230822751726994e-06,
+      "loss": 0.9572,
+      "step": 14003
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229491193740519e-06,
+      "loss": 0.9169,
+      "step": 14004
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.228159789127577e-06,
+      "loss": 0.9072,
+      "step": 14005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.226828537923556e-06,
+      "loss": 0.8859,
+      "step": 14006
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.225497440163843e-06,
+      "loss": 0.9099,
+      "step": 14007
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.224166495883808e-06,
+      "loss": 0.8646,
+      "step": 14008
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222835705118832e-06,
+      "loss": 0.8873,
+      "step": 14009
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.221505067904285e-06,
+      "loss": 0.9239,
+      "step": 14010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2201745842755356e-06,
+      "loss": 0.8873,
+      "step": 14011
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.218844254267942e-06,
+      "loss": 0.9481,
+      "step": 14012
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.217514077916866e-06,
+      "loss": 0.9492,
+      "step": 14013
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216184055257665e-06,
+      "loss": 0.8252,
+      "step": 14014
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.214854186325682e-06,
+      "loss": 0.8976,
+      "step": 14015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.213524471156263e-06,
+      "loss": 0.8381,
+      "step": 14016
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2121949097847635e-06,
+      "loss": 0.9063,
+      "step": 14017
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210865502246508e-06,
+      "loss": 1.0016,
+      "step": 14018
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2095362485768365e-06,
+      "loss": 0.7972,
+      "step": 14019
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.208207148811079e-06,
+      "loss": 0.8628,
+      "step": 14020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.206878202984562e-06,
+      "loss": 0.823,
+      "step": 14021
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.205549411132604e-06,
+      "loss": 0.7858,
+      "step": 14022
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.204220773290525e-06,
+      "loss": 0.894,
+      "step": 14023
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.202892289493641e-06,
+      "loss": 0.869,
+      "step": 14024
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.201563959777256e-06,
+      "loss": 0.9307,
+      "step": 14025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.200235784176679e-06,
+      "loss": 0.9564,
+      "step": 14026
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198907762727213e-06,
+      "loss": 0.9064,
+      "step": 14027
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.197579895464148e-06,
+      "loss": 0.874,
+      "step": 14028
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.196252182422784e-06,
+      "loss": 0.8912,
+      "step": 14029
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.194924623638413e-06,
+      "loss": 0.9818,
+      "step": 14030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1935972191463115e-06,
+      "loss": 0.8807,
+      "step": 14031
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1922699689817645e-06,
+      "loss": 0.9258,
+      "step": 14032
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1909428731800495e-06,
+      "loss": 0.9148,
+      "step": 14033
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.18961593177644e-06,
+      "loss": 0.8714,
+      "step": 14034
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1882891448062015e-06,
+      "loss": 0.8509,
+      "step": 14035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.186962512304598e-06,
+      "loss": 0.8705,
+      "step": 14036
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185636034306898e-06,
+      "loss": 0.8377,
+      "step": 14037
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1843097108483465e-06,
+      "loss": 0.8417,
+      "step": 14038
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.182983541964196e-06,
+      "loss": 0.7771,
+      "step": 14039
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181657527689708e-06,
+      "loss": 0.8462,
+      "step": 14040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180331668060114e-06,
+      "loss": 0.9112,
+      "step": 14041
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.179005963110657e-06,
+      "loss": 0.9375,
+      "step": 14042
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177680412876577e-06,
+      "loss": 0.4333,
+      "step": 14043
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.176355017393099e-06,
+      "loss": 0.8452,
+      "step": 14044
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1750297766954524e-06,
+      "loss": 0.8705,
+      "step": 14045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173704690818862e-06,
+      "loss": 0.9031,
+      "step": 14046
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1723797597985505e-06,
+      "loss": 0.9255,
+      "step": 14047
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171054983669725e-06,
+      "loss": 0.8744,
+      "step": 14048
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169730362467601e-06,
+      "loss": 0.9083,
+      "step": 14049
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1684058962273885e-06,
+      "loss": 0.9215,
+      "step": 14050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.167081584984281e-06,
+      "loss": 0.8146,
+      "step": 14051
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.165757428773484e-06,
+      "loss": 0.8604,
+      "step": 14052
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164433427630198e-06,
+      "loss": 0.4298,
+      "step": 14053
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.163109581589602e-06,
+      "loss": 0.8746,
+      "step": 14054
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.161785890686886e-06,
+      "loss": 1.0143,
+      "step": 14055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160462354957238e-06,
+      "loss": 0.845,
+      "step": 14056
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.159138974435827e-06,
+      "loss": 0.8923,
+      "step": 14057
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.157815749157831e-06,
+      "loss": 0.9266,
+      "step": 14058
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.156492679158421e-06,
+      "loss": 0.9165,
+      "step": 14059
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155169764472763e-06,
+      "loss": 0.9056,
+      "step": 14060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.153847005136016e-06,
+      "loss": 0.8721,
+      "step": 14061
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.152524401183336e-06,
+      "loss": 0.879,
+      "step": 14062
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.151201952649881e-06,
+      "loss": 0.8762,
+      "step": 14063
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.149879659570798e-06,
+      "loss": 0.7996,
+      "step": 14064
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1485575219812325e-06,
+      "loss": 0.9366,
+      "step": 14065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.147235539916328e-06,
+      "loss": 0.8381,
+      "step": 14066
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145913713411215e-06,
+      "loss": 0.8878,
+      "step": 14067
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.144592042501029e-06,
+      "loss": 0.9133,
+      "step": 14068
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1432705272208995e-06,
+      "loss": 0.9637,
+      "step": 14069
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.141949167605954e-06,
+      "loss": 0.9025,
+      "step": 14070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140627963691305e-06,
+      "loss": 0.8542,
+      "step": 14071
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139306915512074e-06,
+      "loss": 0.8526,
+      "step": 14072
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.137986023103373e-06,
+      "loss": 0.8554,
+      "step": 14073
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.136665286500306e-06,
+      "loss": 0.7409,
+      "step": 14074
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135344705737976e-06,
+      "loss": 0.845,
+      "step": 14075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134024280851493e-06,
+      "loss": 0.9299,
+      "step": 14076
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.13270401187594e-06,
+      "loss": 0.8513,
+      "step": 14077
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.131383898846415e-06,
+      "loss": 0.8382,
+      "step": 14078
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.130063941798006e-06,
+      "loss": 0.8939,
+      "step": 14079
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.128744140765789e-06,
+      "loss": 0.8569,
+      "step": 14080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.127424495784849e-06,
+      "loss": 0.928,
+      "step": 14081
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.126105006890258e-06,
+      "loss": 0.9137,
+      "step": 14082
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12478567411709e-06,
+      "loss": 0.8689,
+      "step": 14083
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.123466497500405e-06,
+      "loss": 0.9468,
+      "step": 14084
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12214747707527e-06,
+      "loss": 0.8581,
+      "step": 14085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.120828612876745e-06,
+      "loss": 0.9089,
+      "step": 14086
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119509904939873e-06,
+      "loss": 0.9407,
+      "step": 14087
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.118191353299717e-06,
+      "loss": 0.8922,
+      "step": 14088
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1168729579913195e-06,
+      "loss": 0.8649,
+      "step": 14089
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115554719049715e-06,
+      "loss": 0.7859,
+      "step": 14090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.114236636509946e-06,
+      "loss": 0.9064,
+      "step": 14091
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.11291871040705e-06,
+      "loss": 0.8796,
+      "step": 14092
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.111600940776046e-06,
+      "loss": 0.8787,
+      "step": 14093
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.110283327651965e-06,
+      "loss": 0.8512,
+      "step": 14094
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1089658710698255e-06,
+      "loss": 0.8886,
+      "step": 14095
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1076485710646485e-06,
+      "loss": 0.8844,
+      "step": 14096
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.106331427671439e-06,
+      "loss": 0.9112,
+      "step": 14097
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.105014440925206e-06,
+      "loss": 0.8888,
+      "step": 14098
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.103697610860964e-06,
+      "loss": 0.8977,
+      "step": 14099
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.102380937513701e-06,
+      "loss": 0.9711,
+      "step": 14100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101064420918417e-06,
+      "loss": 0.8335,
+      "step": 14101
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099748061110106e-06,
+      "loss": 0.9518,
+      "step": 14102
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.09843185812375e-06,
+      "loss": 0.947,
+      "step": 14103
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097115811994335e-06,
+      "loss": 0.8895,
+      "step": 14104
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0957999227568444e-06,
+      "loss": 0.903,
+      "step": 14105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.094484190446245e-06,
+      "loss": 0.9238,
+      "step": 14106
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093168615097511e-06,
+      "loss": 0.9527,
+      "step": 14107
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.091853196745607e-06,
+      "loss": 0.8989,
+      "step": 14108
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.090537935425504e-06,
+      "loss": 0.9417,
+      "step": 14109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0892228311721434e-06,
+      "loss": 0.891,
+      "step": 14110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0879078840204945e-06,
+      "loss": 0.9061,
+      "step": 14111
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0865930940055046e-06,
+      "loss": 0.8817,
+      "step": 14112
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.085278461162113e-06,
+      "loss": 0.9334,
+      "step": 14113
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.083963985525266e-06,
+      "loss": 0.8284,
+      "step": 14114
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.082649667129901e-06,
+      "loss": 0.3503,
+      "step": 14115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.081335506010947e-06,
+      "loss": 0.9146,
+      "step": 14116
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0800215022033355e-06,
+      "loss": 0.9177,
+      "step": 14117
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0787076557419915e-06,
+      "loss": 0.8534,
+      "step": 14118
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.077393966661838e-06,
+      "loss": 0.864,
+      "step": 14119
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076080434997786e-06,
+      "loss": 0.7439,
+      "step": 14120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.074767060784749e-06,
+      "loss": 0.8216,
+      "step": 14121
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073453844057638e-06,
+      "loss": 0.9565,
+      "step": 14122
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.072140784851353e-06,
+      "loss": 0.8399,
+      "step": 14123
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.070827883200797e-06,
+      "loss": 0.9821,
+      "step": 14124
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.069515139140867e-06,
+      "loss": 0.8645,
+      "step": 14125
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0682025527064486e-06,
+      "loss": 0.8406,
+      "step": 14126
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.066890123932431e-06,
+      "loss": 0.805,
+      "step": 14127
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.065577852853702e-06,
+      "loss": 0.8579,
+      "step": 14128
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.064265739505131e-06,
+      "loss": 0.9491,
+      "step": 14129
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.062953783921599e-06,
+      "loss": 0.9103,
+      "step": 14130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.061641986137973e-06,
+      "loss": 0.9591,
+      "step": 14131
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.060330346189125e-06,
+      "loss": 0.8684,
+      "step": 14132
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0590188641099096e-06,
+      "loss": 0.8848,
+      "step": 14133
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0577075399351875e-06,
+      "loss": 0.435,
+      "step": 14134
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.056396373699811e-06,
+      "loss": 0.4009,
+      "step": 14135
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0550853654386315e-06,
+      "loss": 0.8888,
+      "step": 14136
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.053774515186494e-06,
+      "loss": 0.8741,
+      "step": 14137
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052463822978241e-06,
+      "loss": 0.8871,
+      "step": 14138
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0511532888487035e-06,
+      "loss": 0.9655,
+      "step": 14139
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049842912832718e-06,
+      "loss": 0.9264,
+      "step": 14140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.048532694965115e-06,
+      "loss": 0.8633,
+      "step": 14141
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.047222635280712e-06,
+      "loss": 0.9507,
+      "step": 14142
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045912733814333e-06,
+      "loss": 0.9881,
+      "step": 14143
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.044602990600795e-06,
+      "loss": 0.8401,
+      "step": 14144
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.043293405674909e-06,
+      "loss": 0.8739,
+      "step": 14145
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041983979071474e-06,
+      "loss": 0.8822,
+      "step": 14146
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.040674710825305e-06,
+      "loss": 0.8802,
+      "step": 14147
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039365600971199e-06,
+      "loss": 0.9101,
+      "step": 14148
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038056649543944e-06,
+      "loss": 0.9426,
+      "step": 14149
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0367478565783335e-06,
+      "loss": 0.83,
+      "step": 14150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0354392221091585e-06,
+      "loss": 0.7939,
+      "step": 14151
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.034130746171194e-06,
+      "loss": 0.8718,
+      "step": 14152
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.032822428799219e-06,
+      "loss": 0.8442,
+      "step": 14153
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031514270028013e-06,
+      "loss": 0.883,
+      "step": 14154
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.030206269892338e-06,
+      "loss": 0.9028,
+      "step": 14155
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.028898428426962e-06,
+      "loss": 0.9124,
+      "step": 14156
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027590745666645e-06,
+      "loss": 0.9108,
+      "step": 14157
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0262832216461454e-06,
+      "loss": 0.9101,
+      "step": 14158
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024975856400216e-06,
+      "loss": 0.8963,
+      "step": 14159
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0236686499636035e-06,
+      "loss": 0.8744,
+      "step": 14160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.022361602371056e-06,
+      "loss": 0.8679,
+      "step": 14161
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021054713657308e-06,
+      "loss": 0.8512,
+      "step": 14162
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.019747983857096e-06,
+      "loss": 0.8342,
+      "step": 14163
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018441413005156e-06,
+      "loss": 0.893,
+      "step": 14164
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0171350011362066e-06,
+      "loss": 0.8178,
+      "step": 14165
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.015828748284978e-06,
+      "loss": 0.8798,
+      "step": 14166
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014522654486185e-06,
+      "loss": 0.9322,
+      "step": 14167
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.013216719774548e-06,
+      "loss": 0.9125,
+      "step": 14168
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.011910944184766e-06,
+      "loss": 0.8472,
+      "step": 14169
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.01060532775156e-06,
+      "loss": 0.8294,
+      "step": 14170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009299870509619e-06,
+      "loss": 0.8653,
+      "step": 14171
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007994572493646e-06,
+      "loss": 1.0449,
+      "step": 14172
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.006689433738333e-06,
+      "loss": 0.8403,
+      "step": 14173
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.005384454278375e-06,
+      "loss": 0.9257,
+      "step": 14174
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004079634148447e-06,
+      "loss": 0.9163,
+      "step": 14175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.002774973383235e-06,
+      "loss": 0.8802,
+      "step": 14176
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.001470472017417e-06,
+      "loss": 0.4126,
+      "step": 14177
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000166130085661e-06,
+      "loss": 0.8384,
+      "step": 14178
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998861947622635e-06,
+      "loss": 0.95,
+      "step": 14179
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997557924663006e-06,
+      "loss": 0.7946,
+      "step": 14180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.996254061241431e-06,
+      "loss": 0.9275,
+      "step": 14181
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994950357392566e-06,
+      "loss": 0.8794,
+      "step": 14182
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.993646813151066e-06,
+      "loss": 0.4245,
+      "step": 14183
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.992343428551572e-06,
+      "loss": 0.8698,
+      "step": 14184
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991040203628725e-06,
+      "loss": 0.819,
+      "step": 14185
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989737138417169e-06,
+      "loss": 0.9206,
+      "step": 14186
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.988434232951538e-06,
+      "loss": 0.8672,
+      "step": 14187
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987131487266456e-06,
+      "loss": 0.3729,
+      "step": 14188
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.985828901396551e-06,
+      "loss": 0.8574,
+      "step": 14189
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984526475376449e-06,
+      "loss": 0.9288,
+      "step": 14190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.983224209240759e-06,
+      "loss": 0.8224,
+      "step": 14191
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.981922103024098e-06,
+      "loss": 0.9004,
+      "step": 14192
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980620156761075e-06,
+      "loss": 0.8672,
+      "step": 14193
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.979318370486292e-06,
+      "loss": 0.877,
+      "step": 14194
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.978016744234352e-06,
+      "loss": 0.8678,
+      "step": 14195
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.976715278039849e-06,
+      "loss": 0.8893,
+      "step": 14196
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9754139719373775e-06,
+      "loss": 0.9052,
+      "step": 14197
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974112825961518e-06,
+      "loss": 0.7713,
+      "step": 14198
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972811840146858e-06,
+      "loss": 0.9171,
+      "step": 14199
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.971511014527979e-06,
+      "loss": 0.8978,
+      "step": 14200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970210349139449e-06,
+      "loss": 0.9372,
+      "step": 14201
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.96890984401584e-06,
+      "loss": 0.9045,
+      "step": 14202
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.967609499191724e-06,
+      "loss": 0.9419,
+      "step": 14203
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.966309314701653e-06,
+      "loss": 0.868,
+      "step": 14204
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965009290580186e-06,
+      "loss": 0.9998,
+      "step": 14205
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963709426861887e-06,
+      "loss": 0.9304,
+      "step": 14206
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.962409723581294e-06,
+      "loss": 0.9263,
+      "step": 14207
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.961110180772955e-06,
+      "loss": 0.9143,
+      "step": 14208
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.95981079847141e-06,
+      "loss": 0.9286,
+      "step": 14209
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.958511576711199e-06,
+      "loss": 0.8784,
+      "step": 14210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957212515526846e-06,
+      "loss": 0.8559,
+      "step": 14211
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955913614952884e-06,
+      "loss": 0.8429,
+      "step": 14212
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.954614875023838e-06,
+      "loss": 0.9284,
+      "step": 14213
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953316295774221e-06,
+      "loss": 0.8507,
+      "step": 14214
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.952017877238551e-06,
+      "loss": 0.9271,
+      "step": 14215
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9507196194513375e-06,
+      "loss": 0.9003,
+      "step": 14216
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9494215224470875e-06,
+      "loss": 0.9766,
+      "step": 14217
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.948123586260304e-06,
+      "loss": 0.9473,
+      "step": 14218
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946825810925486e-06,
+      "loss": 0.9511,
+      "step": 14219
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9455281964771235e-06,
+      "loss": 0.9813,
+      "step": 14220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.944230742949704e-06,
+      "loss": 0.8852,
+      "step": 14221
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9429334503777165e-06,
+      "loss": 0.8714,
+      "step": 14222
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.941636318795643e-06,
+      "loss": 0.848,
+      "step": 14223
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.940339348237954e-06,
+      "loss": 0.9287,
+      "step": 14224
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.939042538739123e-06,
+      "loss": 0.8741,
+      "step": 14225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937745890333623e-06,
+      "loss": 0.8672,
+      "step": 14226
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.936449403055911e-06,
+      "loss": 1.003,
+      "step": 14227
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935153076940443e-06,
+      "loss": 0.857,
+      "step": 14228
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.933856912021688e-06,
+      "loss": 0.9166,
+      "step": 14229
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.932560908334085e-06,
+      "loss": 0.9225,
+      "step": 14230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.931265065912083e-06,
+      "loss": 0.9241,
+      "step": 14231
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929969384790128e-06,
+      "loss": 0.7583,
+      "step": 14232
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92867386500265e-06,
+      "loss": 0.8564,
+      "step": 14233
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9273785065840864e-06,
+      "loss": 0.8273,
+      "step": 14234
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.926083309568868e-06,
+      "loss": 0.9306,
+      "step": 14235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92478827399142e-06,
+      "loss": 0.9631,
+      "step": 14236
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923493399886158e-06,
+      "loss": 0.863,
+      "step": 14237
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922198687287502e-06,
+      "loss": 0.8814,
+      "step": 14238
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920904136229866e-06,
+      "loss": 0.9227,
+      "step": 14239
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919609746747649e-06,
+      "loss": 0.9633,
+      "step": 14240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9183155188752646e-06,
+      "loss": 0.9308,
+      "step": 14241
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91702145264711e-06,
+      "loss": 0.9719,
+      "step": 14242
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.915727548097574e-06,
+      "loss": 0.9032,
+      "step": 14243
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914433805261051e-06,
+      "loss": 0.8939,
+      "step": 14244
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9131402241719274e-06,
+      "loss": 0.9449,
+      "step": 14245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.911846804864589e-06,
+      "loss": 0.905,
+      "step": 14246
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.910553547373405e-06,
+      "loss": 0.9262,
+      "step": 14247
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909260451732753e-06,
+      "loss": 0.8416,
+      "step": 14248
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9079675179770046e-06,
+      "loss": 0.9112,
+      "step": 14249
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906674746140518e-06,
+      "loss": 0.9735,
+      "step": 14250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.905382136257659e-06,
+      "loss": 0.8767,
+      "step": 14251
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904089688362782e-06,
+      "loss": 0.9038,
+      "step": 14252
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902797402490238e-06,
+      "loss": 0.8792,
+      "step": 14253
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.901505278674376e-06,
+      "loss": 0.8352,
+      "step": 14254
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.900213316949542e-06,
+      "loss": 0.8779,
+      "step": 14255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898921517350067e-06,
+      "loss": 0.4401,
+      "step": 14256
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8976298799102905e-06,
+      "loss": 0.932,
+      "step": 14257
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.896338404664543e-06,
+      "loss": 0.8788,
+      "step": 14258
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8950470916471525e-06,
+      "loss": 0.8181,
+      "step": 14259
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8937559408924345e-06,
+      "loss": 0.8568,
+      "step": 14260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8924649524347095e-06,
+      "loss": 0.918,
+      "step": 14261
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891174126308293e-06,
+      "loss": 0.9511,
+      "step": 14262
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.88988346254749e-06,
+      "loss": 0.892,
+      "step": 14263
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888592961186601e-06,
+      "loss": 0.8924,
+      "step": 14264
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.887302622259939e-06,
+      "loss": 0.8507,
+      "step": 14265
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886012445801788e-06,
+      "loss": 0.8392,
+      "step": 14266
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.884722431846443e-06,
+      "loss": 0.8659,
+      "step": 14267
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8834325804281955e-06,
+      "loss": 1.0105,
+      "step": 14268
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882142891581321e-06,
+      "loss": 0.8643,
+      "step": 14269
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8808533653401e-06,
+      "loss": 0.865,
+      "step": 14270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8795640017388095e-06,
+      "loss": 0.8895,
+      "step": 14271
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.87827480081172e-06,
+      "loss": 0.8763,
+      "step": 14272
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.876985762593091e-06,
+      "loss": 0.8647,
+      "step": 14273
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875696887117187e-06,
+      "loss": 0.8917,
+      "step": 14274
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.874408174418271e-06,
+      "loss": 0.8627,
+      "step": 14275
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.87311962453058e-06,
+      "loss": 0.9525,
+      "step": 14276
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.871831237488376e-06,
+      "loss": 1.0004,
+      "step": 14277
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.870543013325902e-06,
+      "loss": 0.9055,
+      "step": 14278
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869254952077391e-06,
+      "loss": 0.8277,
+      "step": 14279
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867967053777082e-06,
+      "loss": 0.9041,
+      "step": 14280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.866679318459207e-06,
+      "loss": 0.8186,
+      "step": 14281
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.865391746157989e-06,
+      "loss": 0.4541,
+      "step": 14282
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864104336907651e-06,
+      "loss": 0.9331,
+      "step": 14283
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.862817090742413e-06,
+      "loss": 0.8151,
+      "step": 14284
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.86153000769649e-06,
+      "loss": 0.902,
+      "step": 14285
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.860243087804085e-06,
+      "loss": 0.9376,
+      "step": 14286
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8589563310994035e-06,
+      "loss": 0.8198,
+      "step": 14287
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857669737616656e-06,
+      "loss": 0.8096,
+      "step": 14288
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.856383307390028e-06,
+      "loss": 0.9222,
+      "step": 14289
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855097040453715e-06,
+      "loss": 0.8313,
+      "step": 14290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853810936841908e-06,
+      "loss": 0.9045,
+      "step": 14291
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.852524996588783e-06,
+      "loss": 0.9725,
+      "step": 14292
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851239219728522e-06,
+      "loss": 0.9021,
+      "step": 14293
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.849953606295301e-06,
+      "loss": 0.933,
+      "step": 14294
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848668156323292e-06,
+      "loss": 0.828,
+      "step": 14295
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.847382869846655e-06,
+      "loss": 0.859,
+      "step": 14296
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8460977468995545e-06,
+      "loss": 0.8668,
+      "step": 14297
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8448127875161505e-06,
+      "loss": 0.848,
+      "step": 14298
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.843527991730585e-06,
+      "loss": 0.8923,
+      "step": 14299
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842243359577018e-06,
+      "loss": 0.9104,
+      "step": 14300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8409588910895934e-06,
+      "loss": 0.7646,
+      "step": 14301
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.839674586302444e-06,
+      "loss": 0.9045,
+      "step": 14302
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.838390445249708e-06,
+      "loss": 0.9474,
+      "step": 14303
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83710646796552e-06,
+      "loss": 0.8565,
+      "step": 14304
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8358226544840005e-06,
+      "loss": 0.9049,
+      "step": 14305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.834539004839274e-06,
+      "loss": 0.8298,
+      "step": 14306
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83325551906546e-06,
+      "loss": 0.9499,
+      "step": 14307
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.831972197196674e-06,
+      "loss": 0.9047,
+      "step": 14308
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83068903926702e-06,
+      "loss": 0.8687,
+      "step": 14309
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.829406045310603e-06,
+      "loss": 0.9757,
+      "step": 14310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828123215361528e-06,
+      "loss": 0.9275,
+      "step": 14311
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.826840549453888e-06,
+      "loss": 0.8285,
+      "step": 14312
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.825558047621776e-06,
+      "loss": 0.8673,
+      "step": 14313
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824275709899282e-06,
+      "loss": 0.9061,
+      "step": 14314
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.822993536320484e-06,
+      "loss": 0.9046,
+      "step": 14315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.821711526919463e-06,
+      "loss": 0.8742,
+      "step": 14316
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.820429681730295e-06,
+      "loss": 0.9115,
+      "step": 14317
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.819148000787046e-06,
+      "loss": 0.7665,
+      "step": 14318
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817866484123785e-06,
+      "loss": 0.8901,
+      "step": 14319
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8165851317745705e-06,
+      "loss": 0.8508,
+      "step": 14320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.815303943773467e-06,
+      "loss": 0.9173,
+      "step": 14321
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8140229201545144e-06,
+      "loss": 0.8162,
+      "step": 14322
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8127420609517695e-06,
+      "loss": 0.8977,
+      "step": 14323
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8114613661992726e-06,
+      "loss": 0.8776,
+      "step": 14324
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8101808359310667e-06,
+      "loss": 0.9045,
+      "step": 14325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8089004701811828e-06,
+      "loss": 0.8862,
+      "step": 14326
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8076202689836585e-06,
+      "loss": 0.8234,
+      "step": 14327
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.806340232372511e-06,
+      "loss": 0.9618,
+      "step": 14328
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8050603603817658e-06,
+      "loss": 0.9416,
+      "step": 14329
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8037806530454445e-06,
+      "loss": 0.8458,
+      "step": 14330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.802501110397553e-06,
+      "loss": 0.9019,
+      "step": 14331
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8012217324721046e-06,
+      "loss": 0.9652,
+      "step": 14332
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7999425193031027e-06,
+      "loss": 0.8933,
+      "step": 14333
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798663470924551e-06,
+      "loss": 0.8395,
+      "step": 14334
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.797384587370434e-06,
+      "loss": 0.4206,
+      "step": 14335
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.796105868674754e-06,
+      "loss": 0.8805,
+      "step": 14336
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7948273148714985e-06,
+      "loss": 0.8903,
+      "step": 14337
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7935489259946434e-06,
+      "loss": 0.4029,
+      "step": 14338
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.792270702078168e-06,
+      "loss": 0.9455,
+      "step": 14339
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7909926431560517e-06,
+      "loss": 0.4273,
+      "step": 14340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7897147492622553e-06,
+      "loss": 0.9419,
+      "step": 14341
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7884370204307485e-06,
+      "loss": 0.9021,
+      "step": 14342
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.787159456695495e-06,
+      "loss": 0.8537,
+      "step": 14343
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7858820580904444e-06,
+      "loss": 0.9145,
+      "step": 14344
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7846048246495505e-06,
+      "loss": 0.8896,
+      "step": 14345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7833277564067627e-06,
+      "loss": 0.7791,
+      "step": 14346
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.782050853396022e-06,
+      "loss": 0.7835,
+      "step": 14347
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.780774115651269e-06,
+      "loss": 0.8464,
+      "step": 14348
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7794975432064364e-06,
+      "loss": 0.8893,
+      "step": 14349
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7782211360954578e-06,
+      "loss": 0.8986,
+      "step": 14350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7769448943522524e-06,
+      "loss": 0.9034,
+      "step": 14351
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7756688180107436e-06,
+      "loss": 0.9026,
+      "step": 14352
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7743929071048534e-06,
+      "loss": 0.9128,
+      "step": 14353
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.773117161668486e-06,
+      "loss": 0.9124,
+      "step": 14354
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7718415817355515e-06,
+      "loss": 0.8682,
+      "step": 14355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.770566167339955e-06,
+      "loss": 0.8762,
+      "step": 14356
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.769290918515599e-06,
+      "loss": 0.9981,
+      "step": 14357
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.768015835296367e-06,
+      "loss": 0.9743,
+      "step": 14358
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.766740917716163e-06,
+      "loss": 0.8791,
+      "step": 14359
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7654661658088633e-06,
+      "loss": 0.9198,
+      "step": 14360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7641915796083517e-06,
+      "loss": 0.8855,
+      "step": 14361
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.762917159148507e-06,
+      "loss": 0.9356,
+      "step": 14362
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.761642904463204e-06,
+      "loss": 0.9113,
+      "step": 14363
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.760368815586304e-06,
+      "loss": 0.7966,
+      "step": 14364
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7590948925516745e-06,
+      "loss": 0.8956,
+      "step": 14365
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7578211353931782e-06,
+      "loss": 0.8727,
+      "step": 14366
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.756547544144664e-06,
+      "loss": 0.3923,
+      "step": 14367
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7552741188399846e-06,
+      "loss": 0.8963,
+      "step": 14368
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7540008595129875e-06,
+      "loss": 0.9816,
+      "step": 14369
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7527277661975135e-06,
+      "loss": 0.9065,
+      "step": 14370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.751454838927402e-06,
+      "loss": 0.8351,
+      "step": 14371
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.750182077736486e-06,
+      "loss": 0.8889,
+      "step": 14372
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7489094826585894e-06,
+      "loss": 0.8348,
+      "step": 14373
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.74763705372754e-06,
+      "loss": 0.8701,
+      "step": 14374
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.746364790977156e-06,
+      "loss": 0.8668,
+      "step": 14375
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.745092694441257e-06,
+      "loss": 0.8034,
+      "step": 14376
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.743820764153647e-06,
+      "loss": 0.9656,
+      "step": 14377
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7425490001481368e-06,
+      "loss": 0.8588,
+      "step": 14378
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7412774024585295e-06,
+      "loss": 0.922,
+      "step": 14379
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.740005971118619e-06,
+      "loss": 0.9462,
+      "step": 14380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.738734706162198e-06,
+      "loss": 0.9152,
+      "step": 14381
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7374636076230584e-06,
+      "loss": 0.8499,
+      "step": 14382
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.736192675534984e-06,
+      "loss": 0.9278,
+      "step": 14383
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7349219099317546e-06,
+      "loss": 0.91,
+      "step": 14384
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.733651310847145e-06,
+      "loss": 0.8413,
+      "step": 14385
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7323808783149306e-06,
+      "loss": 0.9045,
+      "step": 14386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7311106123688703e-06,
+      "loss": 0.8744,
+      "step": 14387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7298405130427296e-06,
+      "loss": 0.9087,
+      "step": 14388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.728570580370271e-06,
+      "loss": 0.9042,
+      "step": 14389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7273008143852396e-06,
+      "loss": 0.8135,
+      "step": 14390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7260312151213894e-06,
+      "loss": 0.9427,
+      "step": 14391
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.724761782612467e-06,
+      "loss": 0.8628,
+      "step": 14392
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.723492516892205e-06,
+      "loss": 0.9069,
+      "step": 14393
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7222234179943416e-06,
+      "loss": 0.8086,
+      "step": 14394
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.720954485952616e-06,
+      "loss": 0.9131,
+      "step": 14395
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.719685720800745e-06,
+      "loss": 0.7639,
+      "step": 14396
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7184171225724553e-06,
+      "loss": 0.8038,
+      "step": 14397
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.717148691301464e-06,
+      "loss": 0.8311,
+      "step": 14398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7158804270214876e-06,
+      "loss": 0.9501,
+      "step": 14399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7146123297662295e-06,
+      "loss": 0.7885,
+      "step": 14400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7133443995693975e-06,
+      "loss": 0.8499,
+      "step": 14401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712076636464694e-06,
+      "loss": 0.9699,
+      "step": 14402
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.710809040485809e-06,
+      "loss": 0.8905,
+      "step": 14403
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.709541611666437e-06,
+      "loss": 0.9008,
+      "step": 14404
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.708274350040264e-06,
+      "loss": 0.8202,
+      "step": 14405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.707007255640973e-06,
+      "loss": 0.8723,
+      "step": 14406
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.705740328502242e-06,
+      "loss": 0.7932,
+      "step": 14407
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7044735686577483e-06,
+      "loss": 0.8442,
+      "step": 14408
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.703206976141153e-06,
+      "loss": 0.8221,
+      "step": 14409
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.701940550986124e-06,
+      "loss": 0.8949,
+      "step": 14410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7006742932263228e-06,
+      "loss": 0.8633,
+      "step": 14411
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6994082028954083e-06,
+      "loss": 0.8814,
+      "step": 14412
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.698142280027024e-06,
+      "loss": 0.8931,
+      "step": 14413
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.69687652465482e-06,
+      "loss": 0.8393,
+      "step": 14414
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6956109368124428e-06,
+      "loss": 0.8763,
+      "step": 14415
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.694345516533523e-06,
+      "loss": 0.8648,
+      "step": 14416
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.693080263851695e-06,
+      "loss": 0.9214,
+      "step": 14417
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6918151788005962e-06,
+      "loss": 0.8785,
+      "step": 14418
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6905502614138423e-06,
+      "loss": 0.8622,
+      "step": 14419
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.689285511725057e-06,
+      "loss": 0.8283,
+      "step": 14420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6880209297678583e-06,
+      "loss": 0.8641,
+      "step": 14421
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.68675651557585e-06,
+      "loss": 0.9961,
+      "step": 14422
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6854922691826435e-06,
+      "loss": 0.94,
+      "step": 14423
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6842281906218414e-06,
+      "loss": 0.8041,
+      "step": 14424
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.682964279927044e-06,
+      "loss": 0.8755,
+      "step": 14425
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.681700537131837e-06,
+      "loss": 0.8479,
+      "step": 14426
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6804369622698143e-06,
+      "loss": 0.913,
+      "step": 14427
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6791735553745624e-06,
+      "loss": 0.8763,
+      "step": 14428
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6779103164796516e-06,
+      "loss": 0.8092,
+      "step": 14429
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6766472456186676e-06,
+      "loss": 0.8913,
+      "step": 14430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6753843428251814e-06,
+      "loss": 0.7902,
+      "step": 14431
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6741216081327522e-06,
+      "loss": 0.9285,
+      "step": 14432
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6728590415749465e-06,
+      "loss": 0.8865,
+      "step": 14433
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671596643185321e-06,
+      "loss": 0.8738,
+      "step": 14434
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.670334412997432e-06,
+      "loss": 0.9584,
+      "step": 14435
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6690723510448224e-06,
+      "loss": 0.8851,
+      "step": 14436
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6678104573610385e-06,
+      "loss": 0.8156,
+      "step": 14437
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6665487319796235e-06,
+      "loss": 0.8631,
+      "step": 14438
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.665287174934107e-06,
+      "loss": 0.9305,
+      "step": 14439
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.664025786258023e-06,
+      "loss": 0.9112,
+      "step": 14440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6627645659848965e-06,
+      "loss": 0.8845,
+      "step": 14441
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66150351414825e-06,
+      "loss": 0.8778,
+      "step": 14442
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.660242630781601e-06,
+      "loss": 0.96,
+      "step": 14443
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.658981915918466e-06,
+      "loss": 0.872,
+      "step": 14444
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6577213695923454e-06,
+      "loss": 0.9223,
+      "step": 14445
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6564609918367467e-06,
+      "loss": 0.896,
+      "step": 14446
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.655200782685171e-06,
+      "loss": 0.8592,
+      "step": 14447
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6539407421711147e-06,
+      "loss": 0.8436,
+      "step": 14448
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6526808703280627e-06,
+      "loss": 0.8731,
+      "step": 14449
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.651421167189504e-06,
+      "loss": 0.8461,
+      "step": 14450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6501616327889235e-06,
+      "loss": 0.8869,
+      "step": 14451
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6489022671597896e-06,
+      "loss": 0.9653,
+      "step": 14452
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.647643070335578e-06,
+      "loss": 0.8865,
+      "step": 14453
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6463840423497643e-06,
+      "loss": 0.9728,
+      "step": 14454
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.645125183235804e-06,
+      "loss": 0.9052,
+      "step": 14455
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6438664930271574e-06,
+      "loss": 0.9204,
+      "step": 14456
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6426079717572837e-06,
+      "loss": 0.8912,
+      "step": 14457
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6413496194596254e-06,
+      "loss": 0.9055,
+      "step": 14458
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.640091436167632e-06,
+      "loss": 0.8801,
+      "step": 14459
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.638833421914746e-06,
+      "loss": 0.8273,
+      "step": 14460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6375755767344047e-06,
+      "loss": 0.9253,
+      "step": 14461
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6363179006600347e-06,
+      "loss": 0.9382,
+      "step": 14462
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.635060393725066e-06,
+      "loss": 0.4315,
+      "step": 14463
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.633803055962928e-06,
+      "loss": 0.8535,
+      "step": 14464
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6325458874070252e-06,
+      "loss": 0.8693,
+      "step": 14465
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6312888880907848e-06,
+      "loss": 0.9175,
+      "step": 14466
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6300320580476147e-06,
+      "loss": 0.8852,
+      "step": 14467
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.628775397310914e-06,
+      "loss": 0.9275,
+      "step": 14468
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.627518905914087e-06,
+      "loss": 0.8459,
+      "step": 14469
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.626262583890533e-06,
+      "loss": 0.904,
+      "step": 14470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6250064312736367e-06,
+      "loss": 0.8988,
+      "step": 14471
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6237504480967877e-06,
+      "loss": 0.9417,
+      "step": 14472
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.62249463439337e-06,
+      "loss": 0.9227,
+      "step": 14473
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.621238990196765e-06,
+      "loss": 0.9006,
+      "step": 14474
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6199835155403372e-06,
+      "loss": 0.9353,
+      "step": 14475
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6187282104574594e-06,
+      "loss": 0.943,
+      "step": 14476
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6174730749815033e-06,
+      "loss": 0.8293,
+      "step": 14477
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6162181091458203e-06,
+      "loss": 0.8862,
+      "step": 14478
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.614963312983769e-06,
+      "loss": 0.9058,
+      "step": 14479
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.613708686528704e-06,
+      "loss": 0.9115,
+      "step": 14480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6124542298139644e-06,
+      "loss": 0.8455,
+      "step": 14481
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.611199942872896e-06,
+      "loss": 0.3996,
+      "step": 14482
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6099458257388377e-06,
+      "loss": 0.7443,
+      "step": 14483
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.608691878445123e-06,
+      "loss": 0.9056,
+      "step": 14484
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.607438101025076e-06,
+      "loss": 0.42,
+      "step": 14485
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6061844935120236e-06,
+      "loss": 0.9423,
+      "step": 14486
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.604931055939288e-06,
+      "loss": 0.8773,
+      "step": 14487
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.603677788340174e-06,
+      "loss": 0.8424,
+      "step": 14488
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6024246907480034e-06,
+      "loss": 0.8469,
+      "step": 14489
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6011717631960798e-06,
+      "loss": 0.8906,
+      "step": 14490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5999190057177e-06,
+      "loss": 0.8262,
+      "step": 14491
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5986664183461628e-06,
+      "loss": 0.887,
+      "step": 14492
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5974140011147653e-06,
+      "loss": 0.8604,
+      "step": 14493
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.596161754056787e-06,
+      "loss": 0.8496,
+      "step": 14494
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.594909677205517e-06,
+      "loss": 0.9075,
+      "step": 14495
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5936577705942312e-06,
+      "loss": 0.8671,
+      "step": 14496
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.592406034256208e-06,
+      "loss": 0.9417,
+      "step": 14497
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5911544682247124e-06,
+      "loss": 0.9332,
+      "step": 14498
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5899030725330097e-06,
+      "loss": 0.8722,
+      "step": 14499
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5886518472143637e-06,
+      "loss": 0.9664,
+      "step": 14500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5874007923020283e-06,
+      "loss": 0.8961,
+      "step": 14501
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.586149907829256e-06,
+      "loss": 0.8633,
+      "step": 14502
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5848991938292975e-06,
+      "loss": 0.87,
+      "step": 14503
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5836486503353886e-06,
+      "loss": 0.9529,
+      "step": 14504
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5823982773807707e-06,
+      "loss": 0.9048,
+      "step": 14505
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5811480749986794e-06,
+      "loss": 0.8898,
+      "step": 14506
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5798980432223385e-06,
+      "loss": 0.9424,
+      "step": 14507
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.578648182084975e-06,
+      "loss": 0.8624,
+      "step": 14508
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.577398491619808e-06,
+      "loss": 0.8874,
+      "step": 14509
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5761489718600584e-06,
+      "loss": 0.9264,
+      "step": 14510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.574899622838929e-06,
+      "loss": 0.8347,
+      "step": 14511
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5736504445896247e-06,
+      "loss": 0.7984,
+      "step": 14512
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5724014371453597e-06,
+      "loss": 0.9218,
+      "step": 14513
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5711526005393205e-06,
+      "loss": 0.9868,
+      "step": 14514
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5699039348047025e-06,
+      "loss": 0.86,
+      "step": 14515
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5686554399746974e-06,
+      "loss": 0.9155,
+      "step": 14516
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.567407116082482e-06,
+      "loss": 0.8107,
+      "step": 14517
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.566158963161238e-06,
+      "loss": 0.9455,
+      "step": 14518
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5649109812441453e-06,
+      "loss": 0.853,
+      "step": 14519
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5636631703643642e-06,
+      "loss": 0.8831,
+      "step": 14520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.562415530555067e-06,
+      "loss": 0.8542,
+      "step": 14521
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5611680618494105e-06,
+      "loss": 0.9653,
+      "step": 14522
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5599207642805576e-06,
+      "loss": 0.8057,
+      "step": 14523
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.558673637881649e-06,
+      "loss": 0.8511,
+      "step": 14524
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5574266826858405e-06,
+      "loss": 0.8613,
+      "step": 14525
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5561798987262775e-06,
+      "loss": 0.9333,
+      "step": 14526
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5549332860360886e-06,
+      "loss": 0.8783,
+      "step": 14527
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.553686844648413e-06,
+      "loss": 0.9329,
+      "step": 14528
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.552440574596382e-06,
+      "loss": 0.8479,
+      "step": 14529
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5511944759131125e-06,
+      "loss": 1.0209,
+      "step": 14530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.54994854863173e-06,
+      "loss": 0.4151,
+      "step": 14531
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.548702792785347e-06,
+      "loss": 0.8779,
+      "step": 14532
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.547457208407079e-06,
+      "loss": 0.9049,
+      "step": 14533
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5462117955300268e-06,
+      "loss": 0.4131,
+      "step": 14534
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5449665541872933e-06,
+      "loss": 0.9165,
+      "step": 14535
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.543721484411976e-06,
+      "loss": 0.8923,
+      "step": 14536
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.542476586237169e-06,
+      "loss": 0.8445,
+      "step": 14537
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5412318596959584e-06,
+      "loss": 0.8748,
+      "step": 14538
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.539987304821432e-06,
+      "loss": 0.9593,
+      "step": 14539
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5387429216466616e-06,
+      "loss": 0.883,
+      "step": 14540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5374987102047243e-06,
+      "loss": 0.879,
+      "step": 14541
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5362546705286947e-06,
+      "loss": 0.8861,
+      "step": 14542
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5350108026516293e-06,
+      "loss": 0.8901,
+      "step": 14543
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5337671066065936e-06,
+      "loss": 0.9049,
+      "step": 14544
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5325235824266436e-06,
+      "loss": 0.8814,
+      "step": 14545
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5312802301448334e-06,
+      "loss": 0.8551,
+      "step": 14546
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5300370497941994e-06,
+      "loss": 0.9308,
+      "step": 14547
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528794041407798e-06,
+      "loss": 0.9508,
+      "step": 14548
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5275512050186566e-06,
+      "loss": 0.9282,
+      "step": 14549
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.526308540659813e-06,
+      "loss": 0.8795,
+      "step": 14550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.525066048364294e-06,
+      "loss": 0.8452,
+      "step": 14551
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.523823728165128e-06,
+      "loss": 0.9491,
+      "step": 14552
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.522581580095327e-06,
+      "loss": 0.9024,
+      "step": 14553
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.521339604187909e-06,
+      "loss": 0.9503,
+      "step": 14554
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.520097800475889e-06,
+      "loss": 0.8996,
+      "step": 14555
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.518856168992266e-06,
+      "loss": 0.8456,
+      "step": 14556
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.517614709770043e-06,
+      "loss": 0.8319,
+      "step": 14557
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5163734228422174e-06,
+      "loss": 0.8672,
+      "step": 14558
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.515132308241782e-06,
+      "loss": 0.8603,
+      "step": 14559
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.513891366001724e-06,
+      "loss": 0.9082,
+      "step": 14560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5126505961550237e-06,
+      "loss": 0.907,
+      "step": 14561
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.511409998734666e-06,
+      "loss": 0.8818,
+      "step": 14562
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5101695737736163e-06,
+      "loss": 0.9313,
+      "step": 14563
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.508929321304847e-06,
+      "loss": 0.7976,
+      "step": 14564
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507689241361326e-06,
+      "loss": 0.8812,
+      "step": 14565
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5064493339760065e-06,
+      "loss": 0.9667,
+      "step": 14566
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.505209599181848e-06,
+      "loss": 0.8347,
+      "step": 14567
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5039700370118037e-06,
+      "loss": 0.9513,
+      "step": 14568
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5027306474988133e-06,
+      "loss": 0.8811,
+      "step": 14569
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.501491430675822e-06,
+      "loss": 0.9437,
+      "step": 14570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5002523865757665e-06,
+      "loss": 0.9533,
+      "step": 14571
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4990135152315787e-06,
+      "loss": 0.8395,
+      "step": 14572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.497774816676186e-06,
+      "loss": 0.8755,
+      "step": 14573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.496536290942514e-06,
+      "loss": 0.9198,
+      "step": 14574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4952979380634823e-06,
+      "loss": 0.4351,
+      "step": 14575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.494059758071999e-06,
+      "loss": 0.8522,
+      "step": 14576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4928217510009767e-06,
+      "loss": 0.929,
+      "step": 14577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4915839168833243e-06,
+      "loss": 0.7794,
+      "step": 14578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.490346255751934e-06,
+      "loss": 0.8802,
+      "step": 14579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4891087676397063e-06,
+      "loss": 0.8889,
+      "step": 14580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.487871452579531e-06,
+      "loss": 0.89,
+      "step": 14581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.486634310604298e-06,
+      "loss": 0.8603,
+      "step": 14582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.48539734174688e-06,
+      "loss": 0.8981,
+      "step": 14583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.484160546040166e-06,
+      "loss": 0.8797,
+      "step": 14584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.48292392351702e-06,
+      "loss": 0.9157,
+      "step": 14585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4816874742103133e-06,
+      "loss": 0.8745,
+      "step": 14586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4804511981529087e-06,
+      "loss": 0.8677,
+      "step": 14587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.479215095377668e-06,
+      "loss": 0.868,
+      "step": 14588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.47797916591744e-06,
+      "loss": 0.8349,
+      "step": 14589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4767434098050768e-06,
+      "loss": 0.9261,
+      "step": 14590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4755078270734265e-06,
+      "loss": 0.9177,
+      "step": 14591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.474272417755323e-06,
+      "loss": 0.8933,
+      "step": 14592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4730371818836063e-06,
+      "loss": 0.8647,
+      "step": 14593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4718021194911056e-06,
+      "loss": 0.944,
+      "step": 14594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4705672306106495e-06,
+      "loss": 1.0254,
+      "step": 14595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.469332515275059e-06,
+      "loss": 0.907,
+      "step": 14596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4680979735171547e-06,
+      "loss": 0.9371,
+      "step": 14597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4668636053697437e-06,
+      "loss": 0.4362,
+      "step": 14598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.465629410865635e-06,
+      "loss": 0.9104,
+      "step": 14599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4643953900376348e-06,
+      "loss": 0.8663,
+      "step": 14600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.463161542918544e-06,
+      "loss": 0.8589,
+      "step": 14601
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4619278695411495e-06,
+      "loss": 0.8672,
+      "step": 14602
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4606943699382457e-06,
+      "loss": 0.8982,
+      "step": 14603
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.459461044142619e-06,
+      "loss": 0.8464,
+      "step": 14604
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.458227892187046e-06,
+      "loss": 0.9209,
+      "step": 14605
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4569949141043003e-06,
+      "loss": 0.8966,
+      "step": 14606
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4557621099271632e-06,
+      "loss": 0.911,
+      "step": 14607
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.454529479688392e-06,
+      "loss": 0.8793,
+      "step": 14608
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.453297023420752e-06,
+      "loss": 0.8685,
+      "step": 14609
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4520647411569995e-06,
+      "loss": 0.8865,
+      "step": 14610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4508326329298913e-06,
+      "loss": 0.8552,
+      "step": 14611
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4496006987721685e-06,
+      "loss": 0.8743,
+      "step": 14612
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4483689387165777e-06,
+      "loss": 1.0325,
+      "step": 14613
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4471373527958607e-06,
+      "loss": 0.8442,
+      "step": 14614
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4459059410427465e-06,
+      "loss": 0.8766,
+      "step": 14615
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4446747034899663e-06,
+      "loss": 0.8834,
+      "step": 14616
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.443443640170249e-06,
+      "loss": 0.9416,
+      "step": 14617
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.442212751116305e-06,
+      "loss": 0.7566,
+      "step": 14618
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.440982036360859e-06,
+      "loss": 0.8415,
+      "step": 14619
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.439751495936624e-06,
+      "loss": 0.7894,
+      "step": 14620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4385211298762976e-06,
+      "loss": 0.9293,
+      "step": 14621
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.437290938212585e-06,
+      "loss": 0.8939,
+      "step": 14622
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4360609209781847e-06,
+      "loss": 0.9206,
+      "step": 14623
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4348310782057913e-06,
+      "loss": 0.8922,
+      "step": 14624
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4336014099280868e-06,
+      "loss": 0.863,
+      "step": 14625
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4323719161777568e-06,
+      "loss": 0.9506,
+      "step": 14626
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4311425969874846e-06,
+      "loss": 0.8143,
+      "step": 14627
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.429913452389936e-06,
+      "loss": 0.858,
+      "step": 14628
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4286844824177835e-06,
+      "loss": 0.939,
+      "step": 14629
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.427455687103692e-06,
+      "loss": 0.8898,
+      "step": 14630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.426227066480323e-06,
+      "loss": 0.8995,
+      "step": 14631
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.424998620580331e-06,
+      "loss": 0.9056,
+      "step": 14632
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.423770349436368e-06,
+      "loss": 0.8775,
+      "step": 14633
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.422542253081076e-06,
+      "loss": 0.9001,
+      "step": 14634
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4213143315470988e-06,
+      "loss": 0.9604,
+      "step": 14635
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4200865848670726e-06,
+      "loss": 0.8666,
+      "step": 14636
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4188590130736333e-06,
+      "loss": 0.9082,
+      "step": 14637
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4176316161994017e-06,
+      "loss": 0.8293,
+      "step": 14638
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4164043942770044e-06,
+      "loss": 0.8558,
+      "step": 14639
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4151773473390627e-06,
+      "loss": 0.925,
+      "step": 14640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4139504754181817e-06,
+      "loss": 0.9523,
+      "step": 14641
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4127237785469737e-06,
+      "loss": 0.8466,
+      "step": 14642
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4114972567580497e-06,
+      "loss": 0.9099,
+      "step": 14643
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.410270910084e-06,
+      "loss": 0.8772,
+      "step": 14644
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4090447385574245e-06,
+      "loss": 0.9211,
+      "step": 14645
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4078187422109157e-06,
+      "loss": 0.9077,
+      "step": 14646
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.406592921077051e-06,
+      "loss": 0.879,
+      "step": 14647
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.405367275188417e-06,
+      "loss": 0.4568,
+      "step": 14648
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40414180457759e-06,
+      "loss": 0.9117,
+      "step": 14649
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.402916509277142e-06,
+      "loss": 0.876,
+      "step": 14650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4016913893196367e-06,
+      "loss": 0.9562,
+      "step": 14651
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.400466444737638e-06,
+      "loss": 0.9403,
+      "step": 14652
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3992416755637035e-06,
+      "loss": 0.8878,
+      "step": 14653
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3980170818303857e-06,
+      "loss": 0.8249,
+      "step": 14654
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3967926635702353e-06,
+      "loss": 0.9023,
+      "step": 14655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.395568420815797e-06,
+      "loss": 0.835,
+      "step": 14656
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3943443535996033e-06,
+      "loss": 0.8737,
+      "step": 14657
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3931204619541925e-06,
+      "loss": 0.9158,
+      "step": 14658
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3918967459120944e-06,
+      "loss": 0.914,
+      "step": 14659
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3906732055058367e-06,
+      "loss": 0.9997,
+      "step": 14660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3894498407679333e-06,
+      "loss": 0.8728,
+      "step": 14661
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3882266517309037e-06,
+      "loss": 0.9026,
+      "step": 14662
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.387003638427261e-06,
+      "loss": 0.835,
+      "step": 14663
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3857808008895054e-06,
+      "loss": 0.7953,
+      "step": 14664
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.384558139150139e-06,
+      "loss": 0.8744,
+      "step": 14665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.383335653241668e-06,
+      "loss": 0.916,
+      "step": 14666
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3821133431965756e-06,
+      "loss": 0.927,
+      "step": 14667
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3808912090473524e-06,
+      "loss": 0.8474,
+      "step": 14668
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.379669250826484e-06,
+      "loss": 0.7959,
+      "step": 14669
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3784474685664426e-06,
+      "loss": 0.8235,
+      "step": 14670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3772258622997044e-06,
+      "loss": 0.8994,
+      "step": 14671
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.376004432058739e-06,
+      "loss": 0.8134,
+      "step": 14672
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3747831778760143e-06,
+      "loss": 0.9182,
+      "step": 14673
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3735620997839824e-06,
+      "loss": 0.8758,
+      "step": 14674
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3723411978151012e-06,
+      "loss": 0.4151,
+      "step": 14675
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3711204720018253e-06,
+      "loss": 0.8865,
+      "step": 14676
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3698999223765905e-06,
+      "loss": 0.8558,
+      "step": 14677
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.368679548971846e-06,
+      "loss": 0.8928,
+      "step": 14678
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3674593518200282e-06,
+      "loss": 0.8897,
+      "step": 14679
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3662393309535636e-06,
+      "loss": 0.8999,
+      "step": 14680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.36501948640488e-06,
+      "loss": 0.9091,
+      "step": 14681
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.363799818206405e-06,
+      "loss": 0.9202,
+      "step": 14682
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3625803263905463e-06,
+      "loss": 0.8117,
+      "step": 14683
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.361361010989723e-06,
+      "loss": 0.8638,
+      "step": 14684
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3601418720363433e-06,
+      "loss": 0.9204,
+      "step": 14685
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3589229095628108e-06,
+      "loss": 0.847,
+      "step": 14686
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3577041236015195e-06,
+      "loss": 0.905,
+      "step": 14687
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.356485514184867e-06,
+      "loss": 0.3989,
+      "step": 14688
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3552670813452415e-06,
+      "loss": 0.8958,
+      "step": 14689
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3540488251150283e-06,
+      "loss": 0.8191,
+      "step": 14690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3528307455266062e-06,
+      "loss": 0.8705,
+      "step": 14691
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3516128426123553e-06,
+      "loss": 0.9593,
+      "step": 14692
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3503951164046376e-06,
+      "loss": 0.8551,
+      "step": 14693
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.349177566935824e-06,
+      "loss": 0.9322,
+      "step": 14694
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3479601942382778e-06,
+      "loss": 0.8952,
+      "step": 14695
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3467429983443477e-06,
+      "loss": 0.8766,
+      "step": 14696
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3455259792863915e-06,
+      "loss": 0.9273,
+      "step": 14697
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.344309137096754e-06,
+      "loss": 0.8498,
+      "step": 14698
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3430924718077816e-06,
+      "loss": 0.832,
+      "step": 14699
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3418759834518056e-06,
+      "loss": 0.9216,
+      "step": 14700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3406596720611573e-06,
+      "loss": 0.8837,
+      "step": 14701
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3394435376681765e-06,
+      "loss": 0.8872,
+      "step": 14702
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3382275803051756e-06,
+      "loss": 0.908,
+      "step": 14703
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3370118000044783e-06,
+      "loss": 0.8645,
+      "step": 14704
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3357961967983998e-06,
+      "loss": 0.8803,
+      "step": 14705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3345807707192433e-06,
+      "loss": 0.886,
+      "step": 14706
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3333655217993186e-06,
+      "loss": 0.9092,
+      "step": 14707
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.332150450070927e-06,
+      "loss": 0.8353,
+      "step": 14708
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.330935555566358e-06,
+      "loss": 0.8349,
+      "step": 14709
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3297208383179047e-06,
+      "loss": 0.891,
+      "step": 14710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.328506298357853e-06,
+      "loss": 0.8435,
+      "step": 14711
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.327291935718485e-06,
+      "loss": 0.8818,
+      "step": 14712
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3260777504320764e-06,
+      "loss": 0.84,
+      "step": 14713
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3248637425308983e-06,
+      "loss": 0.8319,
+      "step": 14714
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.323649912047222e-06,
+      "loss": 0.9447,
+      "step": 14715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.322436259013303e-06,
+      "loss": 0.887,
+      "step": 14716
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3212227834614007e-06,
+      "loss": 0.8689,
+      "step": 14717
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.320009485423773e-06,
+      "loss": 0.8726,
+      "step": 14718
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3187963649326606e-06,
+      "loss": 0.8185,
+      "step": 14719
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3175834220203097e-06,
+      "loss": 0.9038,
+      "step": 14720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316370656718959e-06,
+      "loss": 0.9051,
+      "step": 14721
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3151580690608465e-06,
+      "loss": 0.8585,
+      "step": 14722
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.313945659078195e-06,
+      "loss": 0.8167,
+      "step": 14723
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3127334268032307e-06,
+      "loss": 0.8504,
+      "step": 14724
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.311521372268175e-06,
+      "loss": 0.911,
+      "step": 14725
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3103094955052417e-06,
+      "loss": 0.8965,
+      "step": 14726
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.309097796546643e-06,
+      "loss": 0.8078,
+      "step": 14727
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3078862754245854e-06,
+      "loss": 0.4019,
+      "step": 14728
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3066749321712643e-06,
+      "loss": 0.8365,
+      "step": 14729
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3054637668188784e-06,
+      "loss": 0.8725,
+      "step": 14730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3042527793996236e-06,
+      "loss": 0.8168,
+      "step": 14731
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3030419699456793e-06,
+      "loss": 0.8628,
+      "step": 14732
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.301831338489231e-06,
+      "loss": 0.9418,
+      "step": 14733
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3006208850624556e-06,
+      "loss": 0.7915,
+      "step": 14734
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2994106096975286e-06,
+      "loss": 0.9191,
+      "step": 14735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.298200512426608e-06,
+      "loss": 0.8711,
+      "step": 14736
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2969905932818714e-06,
+      "loss": 0.7587,
+      "step": 14737
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.295780852295465e-06,
+      "loss": 0.9766,
+      "step": 14738
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2945712894995463e-06,
+      "loss": 0.9806,
+      "step": 14739
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2933619049262644e-06,
+      "loss": 0.9233,
+      "step": 14740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.292152698607768e-06,
+      "loss": 0.8963,
+      "step": 14741
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2909436705761865e-06,
+      "loss": 0.9668,
+      "step": 14742
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2897348208636616e-06,
+      "loss": 0.8701,
+      "step": 14743
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2885261495023234e-06,
+      "loss": 0.9066,
+      "step": 14744
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2873176565242926e-06,
+      "loss": 0.9198,
+      "step": 14745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.286109341961691e-06,
+      "loss": 0.8813,
+      "step": 14746
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.284901205846636e-06,
+      "loss": 0.8376,
+      "step": 14747
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.283693248211236e-06,
+      "loss": 0.8837,
+      "step": 14748
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2824854690875998e-06,
+      "loss": 0.8221,
+      "step": 14749
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.281277868507827e-06,
+      "loss": 0.8939,
+      "step": 14750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.280070446504019e-06,
+      "loss": 0.9108,
+      "step": 14751
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2788632031082602e-06,
+      "loss": 0.8903,
+      "step": 14752
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2776561383526417e-06,
+      "loss": 0.8458,
+      "step": 14753
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.276449252269248e-06,
+      "loss": 0.8732,
+      "step": 14754
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2752425448901505e-06,
+      "loss": 0.9402,
+      "step": 14755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.274036016247426e-06,
+      "loss": 0.8706,
+      "step": 14756
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2728296663731465e-06,
+      "loss": 0.8481,
+      "step": 14757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2716234952993676e-06,
+      "loss": 0.8433,
+      "step": 14758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270417503058151e-06,
+      "loss": 0.842,
+      "step": 14759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.269211689681552e-06,
+      "loss": 0.3764,
+      "step": 14760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.268006055201619e-06,
+      "loss": 0.8301,
+      "step": 14761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266800599650396e-06,
+      "loss": 0.8415,
+      "step": 14762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2655953230599223e-06,
+      "loss": 0.8546,
+      "step": 14763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.264390225462238e-06,
+      "loss": 0.9328,
+      "step": 14764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.263185306889365e-06,
+      "loss": 0.8819,
+      "step": 14765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.261980567373333e-06,
+      "loss": 0.9262,
+      "step": 14766
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2607760069461648e-06,
+      "loss": 0.87,
+      "step": 14767
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.259571625639869e-06,
+      "loss": 0.9449,
+      "step": 14768
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2583674234864626e-06,
+      "loss": 0.8994,
+      "step": 14769
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2571634005179498e-06,
+      "loss": 0.9468,
+      "step": 14770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.255959556766336e-06,
+      "loss": 0.824,
+      "step": 14771
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.254755892263608e-06,
+      "loss": 0.9003,
+      "step": 14772
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2535524070417713e-06,
+      "loss": 0.8262,
+      "step": 14773
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2523491011328036e-06,
+      "loss": 0.8649,
+      "step": 14774
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.25114597456869e-06,
+      "loss": 0.9149,
+      "step": 14775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2499430273814083e-06,
+      "loss": 0.9039,
+      "step": 14776
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.248740259602935e-06,
+      "loss": 0.8876,
+      "step": 14777
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2475376712652316e-06,
+      "loss": 0.8371,
+      "step": 14778
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.246335262400264e-06,
+      "loss": 0.7811,
+      "step": 14779
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2451330330399964e-06,
+      "loss": 0.8987,
+      "step": 14780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2439309832163744e-06,
+      "loss": 0.913,
+      "step": 14781
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2427291129613502e-06,
+      "loss": 0.8528,
+      "step": 14782
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241527422306868e-06,
+      "loss": 0.912,
+      "step": 14783
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2403259112848697e-06,
+      "loss": 0.8909,
+      "step": 14784
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2391245799272876e-06,
+      "loss": 0.8356,
+      "step": 14785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2379234282660554e-06,
+      "loss": 0.8799,
+      "step": 14786
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2367224563330923e-06,
+      "loss": 0.366,
+      "step": 14787
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2355216641603226e-06,
+      "loss": 0.8819,
+      "step": 14788
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2343210517796607e-06,
+      "loss": 0.8425,
+      "step": 14789
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.233120619223021e-06,
+      "loss": 0.8494,
+      "step": 14790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.231920366522304e-06,
+      "loss": 0.8762,
+      "step": 14791
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.230720293709414e-06,
+      "loss": 0.8911,
+      "step": 14792
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2295204008162506e-06,
+      "loss": 0.9063,
+      "step": 14793
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2283206878746986e-06,
+      "loss": 0.8539,
+      "step": 14794
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2271211549166458e-06,
+      "loss": 0.8844,
+      "step": 14795
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2259218019739834e-06,
+      "loss": 0.9257,
+      "step": 14796
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2247226290785806e-06,
+      "loss": 0.9668,
+      "step": 14797
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.223523636262311e-06,
+      "loss": 0.9157,
+      "step": 14798
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2223248235570436e-06,
+      "loss": 0.9066,
+      "step": 14799
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2211261909946457e-06,
+      "loss": 0.9073,
+      "step": 14800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.219927738606967e-06,
+      "loss": 0.8514,
+      "step": 14801
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2187294664258663e-06,
+      "loss": 0.892,
+      "step": 14802
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.217531374483195e-06,
+      "loss": 0.9066,
+      "step": 14803
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.21633346281079e-06,
+      "loss": 0.922,
+      "step": 14804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2151357314404928e-06,
+      "loss": 0.8916,
+      "step": 14805
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.213938180404144e-06,
+      "loss": 0.8699,
+      "step": 14806
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.21274080973356e-06,
+      "loss": 0.8639,
+      "step": 14807
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2115436194605763e-06,
+      "loss": 0.9139,
+      "step": 14808
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2103466096170144e-06,
+      "loss": 0.905,
+      "step": 14809
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2091497802346816e-06,
+      "loss": 0.8532,
+      "step": 14810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2079531313453916e-06,
+      "loss": 0.9096,
+      "step": 14811
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2067566629809498e-06,
+      "loss": 0.8248,
+      "step": 14812
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2055603751731613e-06,
+      "loss": 0.8473,
+      "step": 14813
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2043642679538144e-06,
+      "loss": 0.8117,
+      "step": 14814
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2031683413547034e-06,
+      "loss": 0.9234,
+      "step": 14815
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2019725954076187e-06,
+      "loss": 0.8897,
+      "step": 14816
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.200777030144335e-06,
+      "loss": 0.8517,
+      "step": 14817
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.199581645596633e-06,
+      "loss": 0.8902,
+      "step": 14818
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.198386441796284e-06,
+      "loss": 0.8282,
+      "step": 14819
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1971914187750543e-06,
+      "loss": 0.84,
+      "step": 14820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1959965765647082e-06,
+      "loss": 0.8982,
+      "step": 14821
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1948019151970044e-06,
+      "loss": 0.8708,
+      "step": 14822
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1936074347036907e-06,
+      "loss": 0.9006,
+      "step": 14823
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.192413135116518e-06,
+      "loss": 0.8922,
+      "step": 14824
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1912190164672284e-06,
+      "loss": 0.9296,
+      "step": 14825
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1900250787875654e-06,
+      "loss": 0.8348,
+      "step": 14826
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.188831322109254e-06,
+      "loss": 0.8309,
+      "step": 14827
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1876377464640264e-06,
+      "loss": 0.9224,
+      "step": 14828
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.186444351883611e-06,
+      "loss": 0.897,
+      "step": 14829
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1852511383997197e-06,
+      "loss": 0.9395,
+      "step": 14830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.184058106044067e-06,
+      "loss": 0.9328,
+      "step": 14831
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.182865254848371e-06,
+      "loss": 0.9568,
+      "step": 14832
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1816725848443276e-06,
+      "loss": 0.9403,
+      "step": 14833
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1804800960636394e-06,
+      "loss": 0.9422,
+      "step": 14834
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1792877885380056e-06,
+      "loss": 0.8289,
+      "step": 14835
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1780956622991088e-06,
+      "loss": 0.8303,
+      "step": 14836
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1769037173786376e-06,
+      "loss": 0.8303,
+      "step": 14837
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1757119538082736e-06,
+      "loss": 0.9266,
+      "step": 14838
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.174520371619695e-06,
+      "loss": 0.9006,
+      "step": 14839
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.173328970844566e-06,
+      "loss": 0.8776,
+      "step": 14840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.172137751514556e-06,
+      "loss": 0.885,
+      "step": 14841
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.170946713661326e-06,
+      "loss": 0.8146,
+      "step": 14842
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1697558573165332e-06,
+      "loss": 0.8611,
+      "step": 14843
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.168565182511828e-06,
+      "loss": 0.8866,
+      "step": 14844
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1673746892788627e-06,
+      "loss": 0.8858,
+      "step": 14845
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1661843776492705e-06,
+      "loss": 0.8371,
+      "step": 14846
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1649942476546923e-06,
+      "loss": 0.8482,
+      "step": 14847
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1638042993267613e-06,
+      "loss": 0.7979,
+      "step": 14848
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.162614532697108e-06,
+      "loss": 0.9179,
+      "step": 14849
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1614249477973467e-06,
+      "loss": 0.8508,
+      "step": 14850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1602355446591013e-06,
+      "loss": 0.866,
+      "step": 14851
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1590463233139855e-06,
+      "loss": 0.8847,
+      "step": 14852
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.157857283793603e-06,
+      "loss": 0.8616,
+      "step": 14853
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1566684261295565e-06,
+      "loss": 0.4125,
+      "step": 14854
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.155479750353453e-06,
+      "loss": 0.8902,
+      "step": 14855
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1542912564968776e-06,
+      "loss": 0.8788,
+      "step": 14856
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1531029445914217e-06,
+      "loss": 0.8903,
+      "step": 14857
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.151914814668673e-06,
+      "loss": 0.4145,
+      "step": 14858
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.150726866760204e-06,
+      "loss": 0.9054,
+      "step": 14859
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.149539100897593e-06,
+      "loss": 0.91,
+      "step": 14860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.148351517112408e-06,
+      "loss": 0.9173,
+      "step": 14861
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1471641154362174e-06,
+      "loss": 0.7467,
+      "step": 14862
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1459768959005754e-06,
+      "loss": 0.9785,
+      "step": 14863
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1447898585370386e-06,
+      "loss": 0.9382,
+      "step": 14864
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1436030033771616e-06,
+      "loss": 0.8072,
+      "step": 14865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.142416330452479e-06,
+      "loss": 0.8219,
+      "step": 14866
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1412298397945428e-06,
+      "loss": 0.8651,
+      "step": 14867
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1400435314348867e-06,
+      "loss": 0.9055,
+      "step": 14868
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1388574054050347e-06,
+      "loss": 0.9047,
+      "step": 14869
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1376714617365178e-06,
+      "loss": 0.9115,
+      "step": 14870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1364857004608585e-06,
+      "loss": 0.9233,
+      "step": 14871
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.135300121609567e-06,
+      "loss": 0.8262,
+      "step": 14872
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.134114725214158e-06,
+      "loss": 0.9482,
+      "step": 14873
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1329295113061386e-06,
+      "loss": 0.8692,
+      "step": 14874
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.131744479917013e-06,
+      "loss": 0.9235,
+      "step": 14875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1305596310782717e-06,
+      "loss": 0.8717,
+      "step": 14876
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.129374964821409e-06,
+      "loss": 0.8726,
+      "step": 14877
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.128190481177913e-06,
+      "loss": 0.9501,
+      "step": 14878
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1270061801792663e-06,
+      "loss": 1.0164,
+      "step": 14879
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1258220618569446e-06,
+      "loss": 0.8429,
+      "step": 14880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.124638126242425e-06,
+      "loss": 0.878,
+      "step": 14881
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1234543733671687e-06,
+      "loss": 0.8711,
+      "step": 14882
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1222708032626414e-06,
+      "loss": 0.8328,
+      "step": 14883
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1210874159603044e-06,
+      "loss": 0.8663,
+      "step": 14884
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1199042114916032e-06,
+      "loss": 0.8639,
+      "step": 14885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.118721189887991e-06,
+      "loss": 0.8642,
+      "step": 14886
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117538351180911e-06,
+      "loss": 0.8712,
+      "step": 14887
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.116355695401804e-06,
+      "loss": 0.8672,
+      "step": 14888
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.115173222582096e-06,
+      "loss": 0.8963,
+      "step": 14889
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1139909327532193e-06,
+      "loss": 0.8092,
+      "step": 14890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.112808825946604e-06,
+      "loss": 0.8687,
+      "step": 14891
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1116269021936618e-06,
+      "loss": 0.8275,
+      "step": 14892
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.110445161525809e-06,
+      "loss": 0.9212,
+      "step": 14893
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1092636039744583e-06,
+      "loss": 0.826,
+      "step": 14894
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.108082229571008e-06,
+      "loss": 0.807,
+      "step": 14895
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1069010383468613e-06,
+      "loss": 0.9358,
+      "step": 14896
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.105720030333411e-06,
+      "loss": 0.8561,
+      "step": 14897
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104539205562053e-06,
+      "loss": 0.8473,
+      "step": 14898
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.103358564064163e-06,
+      "loss": 0.9035,
+      "step": 14899
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1021781058711266e-06,
+      "loss": 0.9134,
+      "step": 14900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1009978310143173e-06,
+      "loss": 0.8582,
+      "step": 14901
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.099817739525107e-06,
+      "loss": 0.8832,
+      "step": 14902
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.09863783143486e-06,
+      "loss": 0.8695,
+      "step": 14903
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0974581067749397e-06,
+      "loss": 0.8606,
+      "step": 14904
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.096278565576698e-06,
+      "loss": 0.8451,
+      "step": 14905
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.095099207871486e-06,
+      "loss": 0.9108,
+      "step": 14906
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.093920033690655e-06,
+      "loss": 0.8652,
+      "step": 14907
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.092741043065538e-06,
+      "loss": 0.9515,
+      "step": 14908
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0915622360274756e-06,
+      "loss": 0.8242,
+      "step": 14909
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0903836126077977e-06,
+      "loss": 0.859,
+      "step": 14910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089205172837837e-06,
+      "loss": 0.8742,
+      "step": 14911
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.088026916748905e-06,
+      "loss": 0.7955,
+      "step": 14912
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0868488443723243e-06,
+      "loss": 0.9694,
+      "step": 14913
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.085670955739406e-06,
+      "loss": 0.8098,
+      "step": 14914
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.084493250881456e-06,
+      "loss": 0.8687,
+      "step": 14915
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0833157298297776e-06,
+      "loss": 0.8614,
+      "step": 14916
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.08213839261567e-06,
+      "loss": 0.8541,
+      "step": 14917
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0809612392704203e-06,
+      "loss": 0.8802,
+      "step": 14918
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.079784269825319e-06,
+      "loss": 0.7917,
+      "step": 14919
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.078607484311651e-06,
+      "loss": 0.4076,
+      "step": 14920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.077430882760687e-06,
+      "loss": 0.9411,
+      "step": 14921
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.076254465203705e-06,
+      "loss": 0.898,
+      "step": 14922
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0750782316719706e-06,
+      "loss": 0.8974,
+      "step": 14923
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0739021821967518e-06,
+      "loss": 0.7787,
+      "step": 14924
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.072726316809296e-06,
+      "loss": 0.8343,
+      "step": 14925
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0715506355408662e-06,
+      "loss": 0.9172,
+      "step": 14926
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.070375138422711e-06,
+      "loss": 0.8719,
+      "step": 14927
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0691998254860668e-06,
+      "loss": 0.8652,
+      "step": 14928
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.068024696762175e-06,
+      "loss": 0.8849,
+      "step": 14929
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.066849752282274e-06,
+      "loss": 0.8627,
+      "step": 14930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.065674992077584e-06,
+      "loss": 0.9022,
+      "step": 14931
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0645004161793337e-06,
+      "loss": 0.8521,
+      "step": 14932
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0633260246187435e-06,
+      "loss": 0.9386,
+      "step": 14933
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0621518174270225e-06,
+      "loss": 0.7861,
+      "step": 14934
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0609777946353824e-06,
+      "loss": 0.8981,
+      "step": 14935
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0598039562750272e-06,
+      "loss": 0.8691,
+      "step": 14936
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.058630302377157e-06,
+      "loss": 0.8921,
+      "step": 14937
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0574568329729648e-06,
+      "loss": 0.8782,
+      "step": 14938
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0562835480936402e-06,
+      "loss": 0.9266,
+      "step": 14939
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.055110447770373e-06,
+      "loss": 0.9079,
+      "step": 14940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0539375320343344e-06,
+      "loss": 0.945,
+      "step": 14941
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.052764800916703e-06,
+      "loss": 0.9204,
+      "step": 14942
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.051592254448652e-06,
+      "loss": 0.9972,
+      "step": 14943
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0504198926613402e-06,
+      "loss": 0.8001,
+      "step": 14944
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0492477155859303e-06,
+      "loss": 0.8284,
+      "step": 14945
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0480757232535773e-06,
+      "loss": 0.8349,
+      "step": 14946
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.046903915695435e-06,
+      "loss": 0.8774,
+      "step": 14947
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0457322929426424e-06,
+      "loss": 0.9368,
+      "step": 14948
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0445608550263427e-06,
+      "loss": 0.8598,
+      "step": 14949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.043389601977672e-06,
+      "loss": 0.8282,
+      "step": 14950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.04221853382776e-06,
+      "loss": 0.9233,
+      "step": 14951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0410476506077335e-06,
+      "loss": 0.8537,
+      "step": 14952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0398769523487147e-06,
+      "loss": 0.4235,
+      "step": 14953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0387064390818153e-06,
+      "loss": 0.8256,
+      "step": 14954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.037536110838147e-06,
+      "loss": 0.8958,
+      "step": 14955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.036365967648821e-06,
+      "loss": 0.874,
+      "step": 14956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.035196009544932e-06,
+      "loss": 0.9092,
+      "step": 14957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0340262365575768e-06,
+      "loss": 0.9013,
+      "step": 14958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.032856648717849e-06,
+      "loss": 0.9053,
+      "step": 14959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0316872460568336e-06,
+      "loss": 0.8726,
+      "step": 14960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0305180286056135e-06,
+      "loss": 0.8799,
+      "step": 14961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0293489963952673e-06,
+      "loss": 0.9247,
+      "step": 14962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0281801494568607e-06,
+      "loss": 0.9171,
+      "step": 14963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0270114878214617e-06,
+      "loss": 0.9116,
+      "step": 14964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0258430115201356e-06,
+      "loss": 0.8436,
+      "step": 14965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.024674720583939e-06,
+      "loss": 0.7805,
+      "step": 14966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0235066150439186e-06,
+      "loss": 0.9541,
+      "step": 14967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.022338694931124e-06,
+      "loss": 0.8614,
+      "step": 14968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0211709602766017e-06,
+      "loss": 0.9222,
+      "step": 14969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0200034111113807e-06,
+      "loss": 1.0592,
+      "step": 14970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.018836047466497e-06,
+      "loss": 0.8715,
+      "step": 14971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0176688693729782e-06,
+      "loss": 0.8448,
+      "step": 14972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.016501876861847e-06,
+      "loss": 0.9386,
+      "step": 14973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0153350699641194e-06,
+      "loss": 0.9136,
+      "step": 14974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0141684487108082e-06,
+      "loss": 0.935,
+      "step": 14975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0130020131329242e-06,
+      "loss": 0.8358,
+      "step": 14976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0118357632614638e-06,
+      "loss": 0.8609,
+      "step": 14977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0106696991274278e-06,
+      "loss": 0.8915,
+      "step": 14978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.009503820761812e-06,
+      "loss": 0.8438,
+      "step": 14979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.008338128195597e-06,
+      "loss": 0.4134,
+      "step": 14980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0071726214597698e-06,
+      "loss": 0.8743,
+      "step": 14981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.00600730058531e-06,
+      "loss": 0.9522,
+      "step": 14982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0048421656031857e-06,
+      "loss": 0.8548,
+      "step": 14983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.003677216544364e-06,
+      "loss": 0.9327,
+      "step": 14984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0025124534398187e-06,
+      "loss": 0.8751,
+      "step": 14985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.001347876320496e-06,
+      "loss": 0.8689,
+      "step": 14986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.000183485217354e-06,
+      "loss": 0.8786,
+      "step": 14987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9990192801613404e-06,
+      "loss": 0.7973,
+      "step": 14988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.997855261183401e-06,
+      "loss": 0.8626,
+      "step": 14989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9966914283144697e-06,
+      "loss": 0.9013,
+      "step": 14990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9955277815854813e-06,
+      "loss": 0.9279,
+      "step": 14991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9943643210273665e-06,
+      "loss": 0.8116,
+      "step": 14992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9932010466710447e-06,
+      "loss": 0.8341,
+      "step": 14993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9920379585474367e-06,
+      "loss": 0.8582,
+      "step": 14994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9908750566874555e-06,
+      "loss": 0.8922,
+      "step": 14995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.989712341122011e-06,
+      "loss": 0.8653,
+      "step": 14996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9885498118820046e-06,
+      "loss": 0.8753,
+      "step": 14997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.987387468998341e-06,
+      "loss": 0.7818,
+      "step": 14998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9862253125019045e-06,
+      "loss": 0.8694,
+      "step": 14999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9850633424235896e-06,
+      "loss": 0.8578,
+      "step": 15000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9839015587942787e-06,
+      "loss": 0.706,
+      "step": 15001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.982739961644855e-06,
+      "loss": 0.858,
+      "step": 15002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9815785510061856e-06,
+      "loss": 0.8687,
+      "step": 15003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.980417326909142e-06,
+      "loss": 0.829,
+      "step": 15004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.979256289384592e-06,
+      "loss": 0.8602,
+      "step": 15005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9780954384633885e-06,
+      "loss": 0.4666,
+      "step": 15006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9769347741763887e-06,
+      "loss": 0.8466,
+      "step": 15007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.975774296554441e-06,
+      "loss": 0.8253,
+      "step": 15008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.97461400562839e-06,
+      "loss": 0.8927,
+      "step": 15009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9734539014290755e-06,
+      "loss": 0.8803,
+      "step": 15010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.972293983987333e-06,
+      "loss": 0.4528,
+      "step": 15011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9711342533339883e-06,
+      "loss": 0.8339,
+      "step": 15012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9699747094998667e-06,
+      "loss": 0.8161,
+      "step": 15013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.968815352515788e-06,
+      "loss": 0.9157,
+      "step": 15014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9676561824125716e-06,
+      "loss": 0.8455,
+      "step": 15015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.966497199221018e-06,
+      "loss": 0.8924,
+      "step": 15016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9653384029719356e-06,
+      "loss": 0.9012,
+      "step": 15017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.964179793696127e-06,
+      "loss": 0.3409,
+      "step": 15018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9630213714243783e-06,
+      "loss": 0.7205,
+      "step": 15019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9618631361874874e-06,
+      "loss": 0.9454,
+      "step": 15020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.960705088016238e-06,
+      "loss": 0.875,
+      "step": 15021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.959547226941406e-06,
+      "loss": 0.8091,
+      "step": 15022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.958389552993768e-06,
+      "loss": 0.8958,
+      "step": 15023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.957232066204092e-06,
+      "loss": 0.8942,
+      "step": 15024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9560747666031477e-06,
+      "loss": 0.9437,
+      "step": 15025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9549176542216886e-06,
+      "loss": 0.9489,
+      "step": 15026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9537607290904713e-06,
+      "loss": 0.8269,
+      "step": 15027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9526039912402504e-06,
+      "loss": 0.843,
+      "step": 15028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9514474407017625e-06,
+      "loss": 0.7794,
+      "step": 15029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9502910775057516e-06,
+      "loss": 0.8517,
+      "step": 15030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.949134901682952e-06,
+      "loss": 0.9489,
+      "step": 15031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9479789132640947e-06,
+      "loss": 0.8937,
+      "step": 15032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9468231122799028e-06,
+      "loss": 0.8335,
+      "step": 15033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.945667498761101e-06,
+      "loss": 0.9529,
+      "step": 15034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.944512072738397e-06,
+      "loss": 0.7826,
+      "step": 15035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9433568342425034e-06,
+      "loss": 0.8415,
+      "step": 15036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.942201783304126e-06,
+      "loss": 0.8426,
+      "step": 15037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.941046919953967e-06,
+      "loss": 0.8855,
+      "step": 15038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.939892244222716e-06,
+      "loss": 0.9425,
+      "step": 15039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9387377561410666e-06,
+      "loss": 0.8406,
+      "step": 15040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9375834557397053e-06,
+      "loss": 0.8669,
+      "step": 15041
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9364293430493073e-06,
+      "loss": 0.8902,
+      "step": 15042
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9352754181005463e-06,
+      "loss": 0.9289,
+      "step": 15043
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9341216809241014e-06,
+      "loss": 0.8106,
+      "step": 15044
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9329681315506297e-06,
+      "loss": 0.8113,
+      "step": 15045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9318147700107946e-06,
+      "loss": 0.9697,
+      "step": 15046
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9306615963352524e-06,
+      "loss": 0.9332,
+      "step": 15047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.929508610554648e-06,
+      "loss": 0.9173,
+      "step": 15048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.92835581269963e-06,
+      "loss": 0.9083,
+      "step": 15049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9272032028008367e-06,
+      "loss": 0.8622,
+      "step": 15050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9260507808889094e-06,
+      "loss": 0.9351,
+      "step": 15051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.924898546994469e-06,
+      "loss": 0.963,
+      "step": 15052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9237465011481436e-06,
+      "loss": 0.8534,
+      "step": 15053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9225946433805587e-06,
+      "loss": 0.8671,
+      "step": 15054
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9214429737223192e-06,
+      "loss": 0.9426,
+      "step": 15055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9202914922040425e-06,
+      "loss": 0.8775,
+      "step": 15056
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9191401988563372e-06,
+      "loss": 0.9169,
+      "step": 15057
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9179890937097933e-06,
+      "loss": 0.845,
+      "step": 15058
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9168381767950126e-06,
+      "loss": 0.8166,
+      "step": 15059
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.915687448142586e-06,
+      "loss": 0.9298,
+      "step": 15060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.914536907783092e-06,
+      "loss": 0.8654,
+      "step": 15061
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.913386555747115e-06,
+      "loss": 0.9152,
+      "step": 15062
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9122363920652297e-06,
+      "loss": 0.9029,
+      "step": 15063
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9110864167680097e-06,
+      "loss": 0.8037,
+      "step": 15064
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.909936629886012e-06,
+      "loss": 0.8105,
+      "step": 15065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9087870314498025e-06,
+      "loss": 0.8701,
+      "step": 15066
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.907637621489934e-06,
+      "loss": 0.4442,
+      "step": 15067
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9064884000369576e-06,
+      "loss": 0.8995,
+      "step": 15068
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.905339367121418e-06,
+      "loss": 0.861,
+      "step": 15069
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9041905227738576e-06,
+      "loss": 0.8512,
+      "step": 15070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9030418670248074e-06,
+      "loss": 0.9421,
+      "step": 15071
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901893399904797e-06,
+      "loss": 0.8728,
+      "step": 15072
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.900745121444357e-06,
+      "loss": 0.8936,
+      "step": 15073
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8995970316740007e-06,
+      "loss": 0.8829,
+      "step": 15074
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8984491306242457e-06,
+      "loss": 0.8014,
+      "step": 15075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.897301418325602e-06,
+      "loss": 0.9251,
+      "step": 15076
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.896153894808578e-06,
+      "loss": 0.9267,
+      "step": 15077
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.895006560103667e-06,
+      "loss": 0.908,
+      "step": 15078
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8938594142413636e-06,
+      "loss": 0.9393,
+      "step": 15079
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.892712457252167e-06,
+      "loss": 0.3794,
+      "step": 15080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8915656891665533e-06,
+      "loss": 0.8722,
+      "step": 15081
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8904191100150047e-06,
+      "loss": 0.7022,
+      "step": 15082
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.889272719827999e-06,
+      "loss": 0.7544,
+      "step": 15083
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.888126518636001e-06,
+      "loss": 0.9697,
+      "step": 15084
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8869805064694768e-06,
+      "loss": 0.8552,
+      "step": 15085
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8858346833588867e-06,
+      "loss": 0.8644,
+      "step": 15086
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.884689049334689e-06,
+      "loss": 0.8688,
+      "step": 15087
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8835436044273266e-06,
+      "loss": 0.9279,
+      "step": 15088
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.882398348667248e-06,
+      "loss": 0.9996,
+      "step": 15089
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8812532820848926e-06,
+      "loss": 0.9324,
+      "step": 15090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.880108404710693e-06,
+      "loss": 0.8233,
+      "step": 15091
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8789637165750816e-06,
+      "loss": 0.9577,
+      "step": 15092
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.877819217708484e-06,
+      "loss": 0.3877,
+      "step": 15093
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8766749081413147e-06,
+      "loss": 0.9544,
+      "step": 15094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8755307879039906e-06,
+      "loss": 0.9263,
+      "step": 15095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.874386857026924e-06,
+      "loss": 0.8707,
+      "step": 15096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.873243115540514e-06,
+      "loss": 0.8353,
+      "step": 15097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8720995634751624e-06,
+      "loss": 0.8612,
+      "step": 15098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8709562008612636e-06,
+      "loss": 0.8865,
+      "step": 15099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.869813027729209e-06,
+      "loss": 0.9826,
+      "step": 15100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8686700441093786e-06,
+      "loss": 0.8872,
+      "step": 15101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8675272500321526e-06,
+      "loss": 0.8704,
+      "step": 15102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8663846455279077e-06,
+      "loss": 0.909,
+      "step": 15103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.86524223062701e-06,
+      "loss": 0.8692,
+      "step": 15104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8641000053598265e-06,
+      "loss": 0.8979,
+      "step": 15105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8629579697567166e-06,
+      "loss": 0.8045,
+      "step": 15106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.861816123848031e-06,
+      "loss": 0.9548,
+      "step": 15107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.86067446766412e-06,
+      "loss": 0.9072,
+      "step": 15108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8595330012353295e-06,
+      "loss": 0.8168,
+      "step": 15109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8583917245919944e-06,
+      "loss": 0.9083,
+      "step": 15110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8572506377644503e-06,
+      "loss": 0.9304,
+      "step": 15111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8561097407830276e-06,
+      "loss": 0.9227,
+      "step": 15112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.854969033678051e-06,
+      "loss": 0.9012,
+      "step": 15113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8538285164798317e-06,
+      "loss": 0.8976,
+      "step": 15114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.852688189218691e-06,
+      "loss": 0.9243,
+      "step": 15115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8515480519249384e-06,
+      "loss": 0.8732,
+      "step": 15116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.850408104628871e-06,
+      "loss": 0.8767,
+      "step": 15117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.849268347360792e-06,
+      "loss": 0.9301,
+      "step": 15118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8481287801509947e-06,
+      "loss": 0.9799,
+      "step": 15119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.846989403029764e-06,
+      "loss": 0.8076,
+      "step": 15120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.845850216027385e-06,
+      "loss": 0.8961,
+      "step": 15121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.84471121917414e-06,
+      "loss": 0.8918,
+      "step": 15122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8435724125002963e-06,
+      "loss": 0.8811,
+      "step": 15123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8424337960361224e-06,
+      "loss": 0.8223,
+      "step": 15124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.841295369811885e-06,
+      "loss": 0.8749,
+      "step": 15125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8401571338578404e-06,
+      "loss": 0.3721,
+      "step": 15126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8390190882042413e-06,
+      "loss": 0.8912,
+      "step": 15127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8378812328813366e-06,
+      "loss": 0.9374,
+      "step": 15128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.836743567919372e-06,
+      "loss": 0.8496,
+      "step": 15129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8356060933485794e-06,
+      "loss": 0.8887,
+      "step": 15130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8344688091991933e-06,
+      "loss": 0.8623,
+      "step": 15131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8333317155014473e-06,
+      "loss": 0.972,
+      "step": 15132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8321948122855547e-06,
+      "loss": 0.7982,
+      "step": 15133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.831058099581737e-06,
+      "loss": 0.8261,
+      "step": 15134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8299215774202084e-06,
+      "loss": 0.8384,
+      "step": 15135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8287852458311783e-06,
+      "loss": 0.9431,
+      "step": 15136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.827649104844843e-06,
+      "loss": 0.9217,
+      "step": 15137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.826513154491403e-06,
+      "loss": 0.8778,
+      "step": 15138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8253773948010497e-06,
+      "loss": 0.8156,
+      "step": 15139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8242418258039715e-06,
+      "loss": 0.9355,
+      "step": 15140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8231064475303493e-06,
+      "loss": 0.8025,
+      "step": 15141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8219712600103653e-06,
+      "loss": 0.9055,
+      "step": 15142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8208362632741825e-06,
+      "loss": 0.8591,
+      "step": 15143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8197014573519743e-06,
+      "loss": 0.9172,
+      "step": 15144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8185668422739023e-06,
+      "loss": 0.9744,
+      "step": 15145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8174324180701195e-06,
+      "loss": 0.9788,
+      "step": 15146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8162981847707795e-06,
+      "loss": 0.8688,
+      "step": 15147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.815164142406028e-06,
+      "loss": 0.9124,
+      "step": 15148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8140302910060092e-06,
+      "loss": 0.8207,
+      "step": 15149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8128966306008577e-06,
+      "loss": 0.9708,
+      "step": 15150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8117631612207084e-06,
+      "loss": 0.9159,
+      "step": 15151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8106298828956814e-06,
+      "loss": 0.8804,
+      "step": 15152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8094967956559007e-06,
+      "loss": 0.8903,
+      "step": 15153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.808363899531483e-06,
+      "loss": 0.9809,
+      "step": 15154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.807231194552543e-06,
+      "loss": 0.8134,
+      "step": 15155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8060986807491786e-06,
+      "loss": 0.8602,
+      "step": 15156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8049663581514965e-06,
+      "loss": 0.9397,
+      "step": 15157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8038342267895935e-06,
+      "loss": 0.8717,
+      "step": 15158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.802702286693554e-06,
+      "loss": 0.4317,
+      "step": 15159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8015705378934687e-06,
+      "loss": 0.8222,
+      "step": 15160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8004389804194167e-06,
+      "loss": 0.8284,
+      "step": 15161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.799307614301474e-06,
+      "loss": 0.9705,
+      "step": 15162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.798176439569711e-06,
+      "loss": 0.8939,
+      "step": 15163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7970454562541927e-06,
+      "loss": 0.9071,
+      "step": 15164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7959146643849833e-06,
+      "loss": 0.8929,
+      "step": 15165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7947840639921308e-06,
+      "loss": 0.8445,
+      "step": 15166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7936536551056883e-06,
+      "loss": 0.8787,
+      "step": 15167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.792523437755705e-06,
+      "loss": 0.9359,
+      "step": 15168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791393411972213e-06,
+      "loss": 0.9058,
+      "step": 15169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7902635777852517e-06,
+      "loss": 0.807,
+      "step": 15170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7891339352248534e-06,
+      "loss": 0.9127,
+      "step": 15171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7880044843210365e-06,
+      "loss": 0.9905,
+      "step": 15172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.78687522510382e-06,
+      "loss": 0.8216,
+      "step": 15173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7857461576032284e-06,
+      "loss": 0.8171,
+      "step": 15174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7846172818492614e-06,
+      "loss": 0.8844,
+      "step": 15175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.783488597871926e-06,
+      "loss": 0.361,
+      "step": 15176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7823601057012216e-06,
+      "loss": 0.8696,
+      "step": 15177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7812318053671462e-06,
+      "loss": 0.9665,
+      "step": 15178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.780103696899681e-06,
+      "loss": 0.915,
+      "step": 15179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.778975780328813e-06,
+      "loss": 0.9329,
+      "step": 15180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7778480556845257e-06,
+      "loss": 0.9479,
+      "step": 15181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.776720522996785e-06,
+      "loss": 0.8719,
+      "step": 15182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775593182295563e-06,
+      "loss": 0.8453,
+      "step": 15183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774466033610823e-06,
+      "loss": 0.8899,
+      "step": 15184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7733390769725243e-06,
+      "loss": 0.8676,
+      "step": 15185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7722123124106184e-06,
+      "loss": 0.8909,
+      "step": 15186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7710857399550573e-06,
+      "loss": 0.8755,
+      "step": 15187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.769959359635779e-06,
+      "loss": 0.8413,
+      "step": 15188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.768833171482722e-06,
+      "loss": 0.9789,
+      "step": 15189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.767707175525822e-06,
+      "loss": 0.9252,
+      "step": 15190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7665813717950075e-06,
+      "loss": 0.8631,
+      "step": 15191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.765455760320196e-06,
+      "loss": 0.8036,
+      "step": 15192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7643303411313073e-06,
+      "loss": 0.9025,
+      "step": 15193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.763205114258257e-06,
+      "loss": 0.8892,
+      "step": 15194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7620800797309475e-06,
+      "loss": 0.8434,
+      "step": 15195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7609552375792825e-06,
+      "loss": 0.8645,
+      "step": 15196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7598305878331597e-06,
+      "loss": 0.9322,
+      "step": 15197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7587061305224706e-06,
+      "loss": 0.8507,
+      "step": 15198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7575818656771023e-06,
+      "loss": 0.869,
+      "step": 15199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.756457793326939e-06,
+      "loss": 0.8917,
+      "step": 15200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.755333913501852e-06,
+      "loss": 0.8588,
+      "step": 15201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.754210226231715e-06,
+      "loss": 0.9435,
+      "step": 15202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.753086731546395e-06,
+      "loss": 0.8395,
+      "step": 15203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.751963429475756e-06,
+      "loss": 0.9282,
+      "step": 15204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.750840320049648e-06,
+      "loss": 0.9076,
+      "step": 15205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7497174032979246e-06,
+      "loss": 0.7951,
+      "step": 15206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.748594679250436e-06,
+      "loss": 0.8179,
+      "step": 15207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7474721479370114e-06,
+      "loss": 0.8415,
+      "step": 15208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7463498093874985e-06,
+      "loss": 0.8656,
+      "step": 15209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7452276636317253e-06,
+      "loss": 0.8324,
+      "step": 15210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.744105710699512e-06,
+      "loss": 0.9491,
+      "step": 15211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7429839506206823e-06,
+      "loss": 0.7615,
+      "step": 15212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.74186238342505e-06,
+      "loss": 1.0076,
+      "step": 15213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.74074100914243e-06,
+      "loss": 0.875,
+      "step": 15214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7396198278026196e-06,
+      "loss": 0.8238,
+      "step": 15215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7384988394354226e-06,
+      "loss": 0.8515,
+      "step": 15216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7373780440706356e-06,
+      "loss": 0.8774,
+      "step": 15217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7362574417380427e-06,
+      "loss": 0.8895,
+      "step": 15218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.735137032467431e-06,
+      "loss": 0.806,
+      "step": 15219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7340168162885795e-06,
+      "loss": 0.8325,
+      "step": 15220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7328967932312623e-06,
+      "loss": 0.8707,
+      "step": 15221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7317769633252478e-06,
+      "loss": 0.8076,
+      "step": 15222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7306573266003045e-06,
+      "loss": 0.9061,
+      "step": 15223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.729537883086184e-06,
+      "loss": 0.8537,
+      "step": 15224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.728418632812643e-06,
+      "loss": 1.0635,
+      "step": 15225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.72729957580943e-06,
+      "loss": 0.7817,
+      "step": 15226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.726180712106291e-06,
+      "loss": 0.9001,
+      "step": 15227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7250620417329577e-06,
+      "loss": 0.8244,
+      "step": 15228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.723943564719167e-06,
+      "loss": 0.859,
+      "step": 15229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.722825281094649e-06,
+      "loss": 0.8384,
+      "step": 15230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.72170719088912e-06,
+      "loss": 0.8594,
+      "step": 15231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7205892941323e-06,
+      "loss": 0.8607,
+      "step": 15232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.719471590853907e-06,
+      "loss": 0.967,
+      "step": 15233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.718354081083642e-06,
+      "loss": 0.8466,
+      "step": 15234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7172367648512088e-06,
+      "loss": 0.9267,
+      "step": 15235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7161196421863066e-06,
+      "loss": 0.8891,
+      "step": 15236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7150027131186243e-06,
+      "loss": 0.8257,
+      "step": 15237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.713885977677848e-06,
+      "loss": 0.8827,
+      "step": 15238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.712769435893662e-06,
+      "loss": 0.8278,
+      "step": 15239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.711653087795744e-06,
+      "loss": 0.4434,
+      "step": 15240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.71053693341376e-06,
+      "loss": 0.8301,
+      "step": 15241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.709420972777379e-06,
+      "loss": 0.8323,
+      "step": 15242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7083052059162653e-06,
+      "loss": 0.8675,
+      "step": 15243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7071896328600643e-06,
+      "loss": 0.8579,
+      "step": 15244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7060742536384366e-06,
+      "loss": 0.9003,
+      "step": 15245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.704959068281029e-06,
+      "loss": 0.8764,
+      "step": 15246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7038440768174733e-06,
+      "loss": 0.9311,
+      "step": 15247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7027292792774105e-06,
+      "loss": 0.8534,
+      "step": 15248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7016146756904717e-06,
+      "loss": 0.8576,
+      "step": 15249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7005002660862755e-06,
+      "loss": 0.8851,
+      "step": 15250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6993860504944457e-06,
+      "loss": 0.9299,
+      "step": 15251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.698272028944596e-06,
+      "loss": 0.8695,
+      "step": 15252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6971582014663413e-06,
+      "loss": 0.8702,
+      "step": 15253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6960445680892764e-06,
+      "loss": 0.9793,
+      "step": 15254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.694931128843006e-06,
+      "loss": 0.8885,
+      "step": 15255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.693817883757122e-06,
+      "loss": 0.975,
+      "step": 15256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.692704832861215e-06,
+      "loss": 0.8619,
+      "step": 15257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6915919761848665e-06,
+      "loss": 0.8939,
+      "step": 15258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.69047931375766e-06,
+      "loss": 0.9113,
+      "step": 15259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6893668456091627e-06,
+      "loss": 0.862,
+      "step": 15260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6882545717689444e-06,
+      "loss": 0.8839,
+      "step": 15261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.687142492266569e-06,
+      "loss": 0.8762,
+      "step": 15262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6860306071315967e-06,
+      "loss": 0.8231,
+      "step": 15263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.684918916393575e-06,
+      "loss": 0.8661,
+      "step": 15264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.683807420082054e-06,
+      "loss": 0.8482,
+      "step": 15265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6826961182265786e-06,
+      "loss": 0.8578,
+      "step": 15266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6815850108566776e-06,
+      "loss": 0.4066,
+      "step": 15267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6804740980018904e-06,
+      "loss": 0.9115,
+      "step": 15268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6793633796917464e-06,
+      "loss": 0.8208,
+      "step": 15269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.67825285595576e-06,
+      "loss": 0.8587,
+      "step": 15270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6771425268234486e-06,
+      "loss": 0.9029,
+      "step": 15271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6760323923243294e-06,
+      "loss": 0.8688,
+      "step": 15272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6749224524879013e-06,
+      "loss": 0.936,
+      "step": 15273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6738127073436694e-06,
+      "loss": 0.8671,
+      "step": 15274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.672703156921127e-06,
+      "loss": 0.8755,
+      "step": 15275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.67159380124977e-06,
+      "loss": 0.8469,
+      "step": 15276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6704846403590766e-06,
+      "loss": 0.8884,
+      "step": 15277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6693756742785306e-06,
+      "loss": 0.8728,
+      "step": 15278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6682669030376064e-06,
+      "loss": 0.9149,
+      "step": 15279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.667158326665774e-06,
+      "loss": 0.9124,
+      "step": 15280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6660499451924994e-06,
+      "loss": 0.9112,
+      "step": 15281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6649417586472426e-06,
+      "loss": 0.8415,
+      "step": 15282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.663833767059455e-06,
+      "loss": 0.8445,
+      "step": 15283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.662725970458586e-06,
+      "loss": 0.8771,
+      "step": 15284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6616183688740836e-06,
+      "loss": 0.8288,
+      "step": 15285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6605109623353818e-06,
+      "loss": 0.8698,
+      "step": 15286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.659403750871916e-06,
+      "loss": 0.8606,
+      "step": 15287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6582967345131138e-06,
+      "loss": 0.8374,
+      "step": 15288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.657189913288404e-06,
+      "loss": 0.8429,
+      "step": 15289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6560832872271945e-06,
+      "loss": 0.9052,
+      "step": 15290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.654976856358903e-06,
+      "loss": 0.8753,
+      "step": 15291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.653870620712943e-06,
+      "loss": 0.8486,
+      "step": 15292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6527645803187085e-06,
+      "loss": 0.8891,
+      "step": 15293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6516587352056e-06,
+      "loss": 0.8648,
+      "step": 15294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6505530854030137e-06,
+      "loss": 0.8958,
+      "step": 15295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6494476309403283e-06,
+      "loss": 0.9388,
+      "step": 15296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6483423718469302e-06,
+      "loss": 0.8289,
+      "step": 15297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6472373081521996e-06,
+      "loss": 0.8595,
+      "step": 15298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6461324398854993e-06,
+      "loss": 0.8018,
+      "step": 15299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6450277670762005e-06,
+      "loss": 0.8539,
+      "step": 15300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6439232897536627e-06,
+      "loss": 0.8197,
+      "step": 15301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.642819007947246e-06,
+      "loss": 0.8489,
+      "step": 15302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6417149216862914e-06,
+      "loss": 0.3867,
+      "step": 15303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.640611031000153e-06,
+      "loss": 0.8413,
+      "step": 15304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.639507335918171e-06,
+      "loss": 0.9299,
+      "step": 15305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.638403836469675e-06,
+      "loss": 0.7929,
+      "step": 15306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6373005326839973e-06,
+      "loss": 0.8256,
+      "step": 15307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6361974245904652e-06,
+      "loss": 0.8889,
+      "step": 15308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6350945122183926e-06,
+      "loss": 0.8184,
+      "step": 15309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6339917955970952e-06,
+      "loss": 0.8876,
+      "step": 15310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6328892747558842e-06,
+      "loss": 0.4081,
+      "step": 15311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6317869497240657e-06,
+      "loss": 0.9384,
+      "step": 15312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63068482053093e-06,
+      "loss": 0.9054,
+      "step": 15313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.629582887205776e-06,
+      "loss": 0.8776,
+      "step": 15314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.628481149777892e-06,
+      "loss": 0.8929,
+      "step": 15315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.627379608276559e-06,
+      "loss": 0.9304,
+      "step": 15316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.626278262731056e-06,
+      "loss": 0.8801,
+      "step": 15317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6251771131706583e-06,
+      "loss": 0.8714,
+      "step": 15318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6240761596246267e-06,
+      "loss": 0.8091,
+      "step": 15319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6229754021222265e-06,
+      "loss": 0.4172,
+      "step": 15320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6218748406927184e-06,
+      "loss": 0.8692,
+      "step": 15321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6207744753653463e-06,
+      "loss": 0.8415,
+      "step": 15322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.619674306169362e-06,
+      "loss": 0.8163,
+      "step": 15323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.618574333134005e-06,
+      "loss": 0.8854,
+      "step": 15324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.617474556288515e-06,
+      "loss": 0.8711,
+      "step": 15325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.616374975662114e-06,
+      "loss": 0.7821,
+      "step": 15326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.615275591284038e-06,
+      "loss": 0.9102,
+      "step": 15327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6141764031835003e-06,
+      "loss": 0.8677,
+      "step": 15328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6130774113897185e-06,
+      "loss": 0.8146,
+      "step": 15329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.611978615931903e-06,
+      "loss": 0.9124,
+      "step": 15330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.61088001683926e-06,
+      "loss": 0.8494,
+      "step": 15331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6097816141409835e-06,
+      "loss": 0.9222,
+      "step": 15332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6086834078662726e-06,
+      "loss": 0.8631,
+      "step": 15333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6075853980443176e-06,
+      "loss": 0.8878,
+      "step": 15334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.606487584704297e-06,
+      "loss": 0.93,
+      "step": 15335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6053899678753914e-06,
+      "loss": 0.8763,
+      "step": 15336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6042925475867753e-06,
+      "loss": 0.876,
+      "step": 15337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6031953238676167e-06,
+      "loss": 0.8181,
+      "step": 15338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6020982967470786e-06,
+      "loss": 0.8701,
+      "step": 15339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.601001466254318e-06,
+      "loss": 0.9413,
+      "step": 15340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5999048324184915e-06,
+      "loss": 0.8428,
+      "step": 15341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5988083952687403e-06,
+      "loss": 0.9502,
+      "step": 15342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597712154834209e-06,
+      "loss": 0.8516,
+      "step": 15343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5966161111440358e-06,
+      "loss": 0.8842,
+      "step": 15344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5955202642273494e-06,
+      "loss": 0.8895,
+      "step": 15345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594424614113277e-06,
+      "loss": 0.9034,
+      "step": 15346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.593329160830944e-06,
+      "loss": 0.8199,
+      "step": 15347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5922339044094592e-06,
+      "loss": 0.7527,
+      "step": 15348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.591138844877936e-06,
+      "loss": 0.9025,
+      "step": 15349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.590043982265482e-06,
+      "loss": 0.8038,
+      "step": 15350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.588949316601195e-06,
+      "loss": 0.8624,
+      "step": 15351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.587854847914171e-06,
+      "loss": 0.8693,
+      "step": 15352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5867605762335004e-06,
+      "loss": 0.8828,
+      "step": 15353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5856665015882686e-06,
+      "loss": 0.8465,
+      "step": 15354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5845726240075517e-06,
+      "loss": 0.9356,
+      "step": 15355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5834789435204245e-06,
+      "loss": 0.8842,
+      "step": 15356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5823854601559593e-06,
+      "loss": 0.7803,
+      "step": 15357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.581292173943214e-06,
+      "loss": 0.9535,
+      "step": 15358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5801990849112503e-06,
+      "loss": 0.8833,
+      "step": 15359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5791061930891205e-06,
+      "loss": 0.9167,
+      "step": 15360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5780134985058745e-06,
+      "loss": 0.8883,
+      "step": 15361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5769210011905486e-06,
+      "loss": 0.8975,
+      "step": 15362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5758287011721893e-06,
+      "loss": 0.8633,
+      "step": 15363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5747365984798222e-06,
+      "loss": 0.8185,
+      "step": 15364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5736446931424752e-06,
+      "loss": 0.832,
+      "step": 15365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5725529851891693e-06,
+      "loss": 0.8252,
+      "step": 15366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5714614746489264e-06,
+      "loss": 0.8297,
+      "step": 15367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.57037016155075e-06,
+      "loss": 0.9279,
+      "step": 15368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5692790459236494e-06,
+      "loss": 0.856,
+      "step": 15369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.568188127796628e-06,
+      "loss": 0.8344,
+      "step": 15370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5670974071986754e-06,
+      "loss": 0.8647,
+      "step": 15371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.566006884158785e-06,
+      "loss": 0.9175,
+      "step": 15372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.56491655870594e-06,
+      "loss": 0.8133,
+      "step": 15373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5638264308691217e-06,
+      "loss": 0.8564,
+      "step": 15374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5627365006773044e-06,
+      "loss": 1.0218,
+      "step": 15375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.561646768159459e-06,
+      "loss": 0.8657,
+      "step": 15376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.560557233344545e-06,
+      "loss": 0.8861,
+      "step": 15377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5594678962615225e-06,
+      "loss": 0.9812,
+      "step": 15378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5583787569393447e-06,
+      "loss": 0.8246,
+      "step": 15379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5572898154069647e-06,
+      "loss": 0.9198,
+      "step": 15380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.556201071693317e-06,
+      "loss": 0.8862,
+      "step": 15381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5551125258273436e-06,
+      "loss": 0.8673,
+      "step": 15382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5540241778379794e-06,
+      "loss": 0.8446,
+      "step": 15383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552936027754146e-06,
+      "loss": 0.8385,
+      "step": 15384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.551848075604767e-06,
+      "loss": 0.8887,
+      "step": 15385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5507603214187593e-06,
+      "loss": 0.8754,
+      "step": 15386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.549672765225035e-06,
+      "loss": 0.9615,
+      "step": 15387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5485854070524997e-06,
+      "loss": 0.8626,
+      "step": 15388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5474982469300523e-06,
+      "loss": 0.8839,
+      "step": 15389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.546411284886595e-06,
+      "loss": 0.8416,
+      "step": 15390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5453245209510092e-06,
+      "loss": 0.873,
+      "step": 15391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5442379551521833e-06,
+      "loss": 0.9469,
+      "step": 15392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5431515875190007e-06,
+      "loss": 0.8599,
+      "step": 15393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5420654180803283e-06,
+      "loss": 0.8302,
+      "step": 15394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5409794468650405e-06,
+      "loss": 0.8297,
+      "step": 15395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.539893673902002e-06,
+      "loss": 0.9647,
+      "step": 15396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.538808099220064e-06,
+      "loss": 0.9222,
+      "step": 15397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5377227228480883e-06,
+      "loss": 0.4202,
+      "step": 15398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5366375448149226e-06,
+      "loss": 0.925,
+      "step": 15399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.535552565149404e-06,
+      "loss": 0.8614,
+      "step": 15400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.534467783880373e-06,
+      "loss": 0.8969,
+      "step": 15401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5333832010366622e-06,
+      "loss": 0.8119,
+      "step": 15402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.532298816647101e-06,
+      "loss": 0.8938,
+      "step": 15403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.531214630740506e-06,
+      "loss": 0.8584,
+      "step": 15404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5301306433456953e-06,
+      "loss": 0.4045,
+      "step": 15405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.529046854491484e-06,
+      "loss": 0.8664,
+      "step": 15406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.527963264206671e-06,
+      "loss": 0.8744,
+      "step": 15407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5268798725200606e-06,
+      "loss": 0.8123,
+      "step": 15408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5257966794604483e-06,
+      "loss": 0.9006,
+      "step": 15409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5247136850566234e-06,
+      "loss": 0.8221,
+      "step": 15410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523630889337372e-06,
+      "loss": 0.8558,
+      "step": 15411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.522548292331476e-06,
+      "loss": 0.823,
+      "step": 15412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.521465894067703e-06,
+      "loss": 0.9147,
+      "step": 15413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.520383694574824e-06,
+      "loss": 0.894,
+      "step": 15414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.519301693881605e-06,
+      "loss": 0.8472,
+      "step": 15415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5182198920168064e-06,
+      "loss": 0.8322,
+      "step": 15416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5171382890091745e-06,
+      "loss": 0.8858,
+      "step": 15417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5160568848874612e-06,
+      "loss": 0.9189,
+      "step": 15418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.51497567968041e-06,
+      "loss": 0.8594,
+      "step": 15419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5138946734167557e-06,
+      "loss": 0.7909,
+      "step": 15420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5128138661252266e-06,
+      "loss": 0.8791,
+      "step": 15421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.51173325783456e-06,
+      "loss": 0.8982,
+      "step": 15422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5106528485734684e-06,
+      "loss": 0.8531,
+      "step": 15423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.50957263837067e-06,
+      "loss": 0.989,
+      "step": 15424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5084926272548792e-06,
+      "loss": 0.4082,
+      "step": 15425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507412815254795e-06,
+      "loss": 0.9151,
+      "step": 15426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5063332023991206e-06,
+      "loss": 0.4478,
+      "step": 15427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.505253788716552e-06,
+      "loss": 0.8728,
+      "step": 15428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5041745742357805e-06,
+      "loss": 0.9615,
+      "step": 15429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5030955589854843e-06,
+      "loss": 0.8999,
+      "step": 15430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5020167429943454e-06,
+      "loss": 0.8936,
+      "step": 15431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5009381262910427e-06,
+      "loss": 0.8792,
+      "step": 15432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.499859708904232e-06,
+      "loss": 0.8184,
+      "step": 15433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4987814908625884e-06,
+      "loss": 0.8777,
+      "step": 15434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4977034721947677e-06,
+      "loss": 0.4067,
+      "step": 15435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4966256529294186e-06,
+      "loss": 0.824,
+      "step": 15436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.495548033095189e-06,
+      "loss": 0.8813,
+      "step": 15437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4944706127207252e-06,
+      "loss": 0.8087,
+      "step": 15438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4933933918346565e-06,
+      "loss": 0.3713,
+      "step": 15439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4923163704656197e-06,
+      "loss": 0.8587,
+      "step": 15440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.491239548642238e-06,
+      "loss": 0.8764,
+      "step": 15441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.490162926393136e-06,
+      "loss": 0.8884,
+      "step": 15442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4890865037469237e-06,
+      "loss": 0.8719,
+      "step": 15443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4880102807322136e-06,
+      "loss": 0.8175,
+      "step": 15444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.486934257377611e-06,
+      "loss": 0.9096,
+      "step": 15445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4858584337117166e-06,
+      "loss": 0.9427,
+      "step": 15446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.484782809763121e-06,
+      "loss": 0.8813,
+      "step": 15447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4837073855604186e-06,
+      "loss": 0.8449,
+      "step": 15448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.482632161132187e-06,
+      "loss": 0.8837,
+      "step": 15449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.481557136507007e-06,
+      "loss": 0.4662,
+      "step": 15450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.48048231171345e-06,
+      "loss": 0.8521,
+      "step": 15451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.47940768678009e-06,
+      "loss": 0.8972,
+      "step": 15452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4783332617354795e-06,
+      "loss": 0.8934,
+      "step": 15453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4772590366081816e-06,
+      "loss": 0.8404,
+      "step": 15454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4761850114267492e-06,
+      "loss": 0.9266,
+      "step": 15455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4751111862197207e-06,
+      "loss": 0.461,
+      "step": 15456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4740375610156453e-06,
+      "loss": 0.909,
+      "step": 15457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.472964135843059e-06,
+      "loss": 0.8869,
+      "step": 15458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4718909107304857e-06,
+      "loss": 0.8514,
+      "step": 15459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.470817885706455e-06,
+      "loss": 0.7811,
+      "step": 15460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4697450607994875e-06,
+      "loss": 0.756,
+      "step": 15461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.468672436038094e-06,
+      "loss": 0.8855,
+      "step": 15462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.467600011450786e-06,
+      "loss": 0.9082,
+      "step": 15463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4665277870660653e-06,
+      "loss": 0.8919,
+      "step": 15464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4654557629124364e-06,
+      "loss": 0.8039,
+      "step": 15465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4643839390183844e-06,
+      "loss": 0.8975,
+      "step": 15466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4633123154124005e-06,
+      "loss": 0.925,
+      "step": 15467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.462240892122968e-06,
+      "loss": 0.9517,
+      "step": 15468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4611696691785637e-06,
+      "loss": 0.877,
+      "step": 15469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4600986466076592e-06,
+      "loss": 0.9379,
+      "step": 15470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4590278244387243e-06,
+      "loss": 0.9494,
+      "step": 15471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.457957202700215e-06,
+      "loss": 0.9474,
+      "step": 15472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4568867814205886e-06,
+      "loss": 0.8397,
+      "step": 15473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.455816560628299e-06,
+      "loss": 0.863,
+      "step": 15474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4547465403517866e-06,
+      "loss": 0.3855,
+      "step": 15475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4536767206194946e-06,
+      "loss": 0.844,
+      "step": 15476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4526071014598553e-06,
+      "loss": 0.9123,
+      "step": 15477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4515376829013026e-06,
+      "loss": 0.8911,
+      "step": 15478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4504684649722545e-06,
+      "loss": 0.89,
+      "step": 15479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4493994477011286e-06,
+      "loss": 0.9387,
+      "step": 15480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4483306311163478e-06,
+      "loss": 0.9089,
+      "step": 15481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.447262015246311e-06,
+      "loss": 0.9044,
+      "step": 15482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.446193600119424e-06,
+      "loss": 0.8292,
+      "step": 15483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4451253857640866e-06,
+      "loss": 0.3826,
+      "step": 15484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4440573722086856e-06,
+      "loss": 0.813,
+      "step": 15485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.44298955948161e-06,
+      "loss": 0.8961,
+      "step": 15486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.441921947611243e-06,
+      "loss": 0.8947,
+      "step": 15487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.440854536625956e-06,
+      "loss": 0.8757,
+      "step": 15488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.439787326554123e-06,
+      "loss": 0.847,
+      "step": 15489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.438720317424107e-06,
+      "loss": 0.8751,
+      "step": 15490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4376535092642715e-06,
+      "loss": 0.9119,
+      "step": 15491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4365869021029643e-06,
+      "loss": 0.8558,
+      "step": 15492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.435520495968542e-06,
+      "loss": 0.8078,
+      "step": 15493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4344542908893475e-06,
+      "loss": 0.9143,
+      "step": 15494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.433388286893714e-06,
+      "loss": 0.8139,
+      "step": 15495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.432322484009979e-06,
+      "loss": 0.8294,
+      "step": 15496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.431256882266472e-06,
+      "loss": 0.9438,
+      "step": 15497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4301914816915096e-06,
+      "loss": 0.8667,
+      "step": 15498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.429126282313411e-06,
+      "loss": 0.9143,
+      "step": 15499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.42806128416049e-06,
+      "loss": 0.8125,
+      "step": 15500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.426996487261055e-06,
+      "loss": 0.8158,
+      "step": 15501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4259318916434017e-06,
+      "loss": 0.891,
+      "step": 15502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.424867497335828e-06,
+      "loss": 0.8329,
+      "step": 15503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4238033043666243e-06,
+      "loss": 0.8231,
+      "step": 15504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4227393127640762e-06,
+      "loss": 0.9069,
+      "step": 15505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4216755225564638e-06,
+      "loss": 0.8707,
+      "step": 15506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4206119337720622e-06,
+      "loss": 0.9081,
+      "step": 15507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4195485464391366e-06,
+      "loss": 0.8711,
+      "step": 15508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4184853605859537e-06,
+      "loss": 0.8682,
+      "step": 15509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4174223762407736e-06,
+      "loss": 0.899,
+      "step": 15510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.416359593431843e-06,
+      "loss": 0.7781,
+      "step": 15511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.415297012187413e-06,
+      "loss": 0.8377,
+      "step": 15512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.414234632535726e-06,
+      "loss": 0.9064,
+      "step": 15513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4131724545050216e-06,
+      "loss": 0.9208,
+      "step": 15514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4121104781235227e-06,
+      "loss": 0.8649,
+      "step": 15515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4110487034194673e-06,
+      "loss": 0.3935,
+      "step": 15516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4099871304210663e-06,
+      "loss": 0.8761,
+      "step": 15517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4089257591565396e-06,
+      "loss": 0.9273,
+      "step": 15518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4078645896540962e-06,
+      "loss": 0.8471,
+      "step": 15519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4068036219419433e-06,
+      "loss": 0.8908,
+      "step": 15520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4057428560482755e-06,
+      "loss": 0.8366,
+      "step": 15521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.404682292001289e-06,
+      "loss": 0.8132,
+      "step": 15522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4036219298291754e-06,
+      "loss": 0.8972,
+      "step": 15523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.402561769560112e-06,
+      "loss": 0.7751,
+      "step": 15524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4015018112222798e-06,
+      "loss": 0.8766,
+      "step": 15525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4004420548438513e-06,
+      "loss": 0.9432,
+      "step": 15526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3993825004529927e-06,
+      "loss": 0.8899,
+      "step": 15527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.398323148077867e-06,
+      "loss": 0.8735,
+      "step": 15528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3972639977466304e-06,
+      "loss": 0.39,
+      "step": 15529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3962050494874357e-06,
+      "loss": 0.87,
+      "step": 15530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.395146303328424e-06,
+      "loss": 0.8839,
+      "step": 15531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3940877592977387e-06,
+      "loss": 0.8469,
+      "step": 15532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3930294174235157e-06,
+      "loss": 0.8427,
+      "step": 15533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.39197127773388e-06,
+      "loss": 0.8285,
+      "step": 15534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3909133402569597e-06,
+      "loss": 0.861,
+      "step": 15535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3898556050208743e-06,
+      "loss": 0.8267,
+      "step": 15536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3887980720537316e-06,
+      "loss": 0.9224,
+      "step": 15537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3877407413836453e-06,
+      "loss": 0.8913,
+      "step": 15538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.386683613038715e-06,
+      "loss": 0.9088,
+      "step": 15539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.385626687047039e-06,
+      "loss": 0.8246,
+      "step": 15540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.384569963436709e-06,
+      "loss": 0.8617,
+      "step": 15541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.383513442235812e-06,
+      "loss": 0.8267,
+      "step": 15542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3824571234724314e-06,
+      "loss": 0.8999,
+      "step": 15543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3814010071746384e-06,
+      "loss": 0.8867,
+      "step": 15544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3803450933705052e-06,
+      "loss": 0.8444,
+      "step": 15545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.379289382088099e-06,
+      "loss": 0.9396,
+      "step": 15546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3782338733554742e-06,
+      "loss": 0.8715,
+      "step": 15547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.37717856720069e-06,
+      "loss": 0.8663,
+      "step": 15548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3761234636517915e-06,
+      "loss": 0.9035,
+      "step": 15549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3750685627368276e-06,
+      "loss": 0.9145,
+      "step": 15550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3740138644838273e-06,
+      "loss": 0.879,
+      "step": 15551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.372959368920834e-06,
+      "loss": 0.8283,
+      "step": 15552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3719050760758678e-06,
+      "loss": 0.9034,
+      "step": 15553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3708509859769513e-06,
+      "loss": 0.9334,
+      "step": 15554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3697970986521036e-06,
+      "loss": 0.8833,
+      "step": 15555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3687434141293364e-06,
+      "loss": 0.8719,
+      "step": 15556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.367689932436651e-06,
+      "loss": 0.9226,
+      "step": 15557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.36663665360205e-06,
+      "loss": 0.8001,
+      "step": 15558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.365583577653532e-06,
+      "loss": 0.8095,
+      "step": 15559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3645307046190803e-06,
+      "loss": 0.9008,
+      "step": 15560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3634780345266805e-06,
+      "loss": 0.8155,
+      "step": 15561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.362425567404314e-06,
+      "loss": 0.815,
+      "step": 15562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3613733032799535e-06,
+      "loss": 0.8336,
+      "step": 15563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3603212421815656e-06,
+      "loss": 0.8217,
+      "step": 15564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.359269384137117e-06,
+      "loss": 0.8772,
+      "step": 15565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.358217729174559e-06,
+      "loss": 0.861,
+      "step": 15566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.357166277321845e-06,
+      "loss": 0.8784,
+      "step": 15567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3561150286069233e-06,
+      "loss": 0.8478,
+      "step": 15568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.355063983057736e-06,
+      "loss": 0.863,
+      "step": 15569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.354013140702215e-06,
+      "loss": 0.9068,
+      "step": 15570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.352962501568291e-06,
+      "loss": 0.9106,
+      "step": 15571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3519120656838935e-06,
+      "loss": 0.8798,
+      "step": 15572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.350861833076935e-06,
+      "loss": 0.8845,
+      "step": 15573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3498118037753304e-06,
+      "loss": 0.8716,
+      "step": 15574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.348761977806997e-06,
+      "loss": 0.9225,
+      "step": 15575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3477123551998282e-06,
+      "loss": 0.9306,
+      "step": 15576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.346662935981725e-06,
+      "loss": 0.9016,
+      "step": 15577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3456137201805796e-06,
+      "loss": 0.9221,
+      "step": 15578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344564707824283e-06,
+      "loss": 0.8107,
+      "step": 15579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3435158989407105e-06,
+      "loss": 0.9692,
+      "step": 15580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.342467293557741e-06,
+      "loss": 0.9276,
+      "step": 15581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3414188917032467e-06,
+      "loss": 0.9708,
+      "step": 15582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3403706934050897e-06,
+      "loss": 0.9049,
+      "step": 15583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.33932269869113e-06,
+      "loss": 0.786,
+      "step": 15584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3382749075892273e-06,
+      "loss": 0.8816,
+      "step": 15585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.337227320127222e-06,
+      "loss": 0.8691,
+      "step": 15586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.336179936332965e-06,
+      "loss": 0.8597,
+      "step": 15587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3351327562342962e-06,
+      "loss": 0.865,
+      "step": 15588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.334085779859041e-06,
+      "loss": 0.9135,
+      "step": 15589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3330390072350297e-06,
+      "loss": 0.8392,
+      "step": 15590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.331992438390086e-06,
+      "loss": 0.8127,
+      "step": 15591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3309460733520283e-06,
+      "loss": 0.8815,
+      "step": 15592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.329899912148662e-06,
+      "loss": 0.8248,
+      "step": 15593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.328853954807796e-06,
+      "loss": 0.7581,
+      "step": 15594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3278082013572344e-06,
+      "loss": 0.9111,
+      "step": 15595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3267626518247645e-06,
+      "loss": 0.8818,
+      "step": 15596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.32571730623818e-06,
+      "loss": 0.9063,
+      "step": 15597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3246721646252634e-06,
+      "loss": 0.8495,
+      "step": 15598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.323627227013796e-06,
+      "loss": 0.8993,
+      "step": 15599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.322582493431549e-06,
+      "loss": 0.8224,
+      "step": 15600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3215379639062942e-06,
+      "loss": 0.8476,
+      "step": 15601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3204936384657873e-06,
+      "loss": 0.8256,
+      "step": 15602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3194495171377895e-06,
+      "loss": 0.398,
+      "step": 15603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.318405599950051e-06,
+      "loss": 0.8538,
+      "step": 15604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3173618869303218e-06,
+      "loss": 0.9563,
+      "step": 15605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.316318378106335e-06,
+      "loss": 0.8738,
+      "step": 15606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3152750735058318e-06,
+      "loss": 0.8321,
+      "step": 15607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3142319731565423e-06,
+      "loss": 0.8188,
+      "step": 15608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3131890770861854e-06,
+      "loss": 0.8422,
+      "step": 15609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3121463853224814e-06,
+      "loss": 0.8898,
+      "step": 15610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.311103897893152e-06,
+      "loss": 0.88,
+      "step": 15611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.310061614825897e-06,
+      "loss": 0.8958,
+      "step": 15612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3090195361484213e-06,
+      "loss": 0.8802,
+      "step": 15613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3079776618884244e-06,
+      "loss": 0.8651,
+      "step": 15614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3069359920735945e-06,
+      "loss": 0.9122,
+      "step": 15615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3058945267316193e-06,
+      "loss": 0.7486,
+      "step": 15616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.304853265890179e-06,
+      "loss": 0.8786,
+      "step": 15617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.303812209576953e-06,
+      "loss": 0.8553,
+      "step": 15618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3027713578196077e-06,
+      "loss": 0.9705,
+      "step": 15619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3017307106458077e-06,
+      "loss": 0.8724,
+      "step": 15620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3006902680832156e-06,
+      "loss": 0.909,
+      "step": 15621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2996500301594772e-06,
+      "loss": 0.9287,
+      "step": 15622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.298609996902249e-06,
+      "loss": 0.8601,
+      "step": 15623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2975701683391736e-06,
+      "loss": 0.8286,
+      "step": 15624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.296530544497885e-06,
+      "loss": 0.851,
+      "step": 15625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2954911254060153e-06,
+      "loss": 0.8681,
+      "step": 15626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.294451911091191e-06,
+      "loss": 0.9739,
+      "step": 15627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2934129015810392e-06,
+      "loss": 0.8111,
+      "step": 15628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.292374096903166e-06,
+      "loss": 0.8641,
+      "step": 15629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2913354970851877e-06,
+      "loss": 0.8796,
+      "step": 15630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2902971021547103e-06,
+      "loss": 0.8042,
+      "step": 15631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.289258912139327e-06,
+      "loss": 0.909,
+      "step": 15632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2882209270666333e-06,
+      "loss": 0.8722,
+      "step": 15633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.287183146964226e-06,
+      "loss": 0.8909,
+      "step": 15634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286145571859677e-06,
+      "loss": 0.8996,
+      "step": 15635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2851082017805704e-06,
+      "loss": 0.8493,
+      "step": 15636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2840710367544783e-06,
+      "loss": 1.0081,
+      "step": 15637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.283034076808963e-06,
+      "loss": 0.9432,
+      "step": 15638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2819973219715886e-06,
+      "loss": 0.8424,
+      "step": 15639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2809607722699103e-06,
+      "loss": 0.8597,
+      "step": 15640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2799244277314826e-06,
+      "loss": 0.8578,
+      "step": 15641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2788882883838425e-06,
+      "loss": 0.4539,
+      "step": 15642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.277852354254533e-06,
+      "loss": 0.8635,
+      "step": 15643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2768166253710923e-06,
+      "loss": 0.8715,
+      "step": 15644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2757811017610386e-06,
+      "loss": 0.8829,
+      "step": 15645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2747457834519037e-06,
+      "loss": 0.9215,
+      "step": 15646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2737106704712064e-06,
+      "loss": 0.8498,
+      "step": 15647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2726757628464525e-06,
+      "loss": 0.8912,
+      "step": 15648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2716410606051507e-06,
+      "loss": 0.9116,
+      "step": 15649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2706065637748066e-06,
+      "loss": 0.9273,
+      "step": 15650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2695722723829085e-06,
+      "loss": 0.9014,
+      "step": 15651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.268538186456951e-06,
+      "loss": 0.8322,
+      "step": 15652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2675043060244183e-06,
+      "loss": 0.9648,
+      "step": 15653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2664706311127914e-06,
+      "loss": 0.8554,
+      "step": 15654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.265437161749542e-06,
+      "loss": 0.4612,
+      "step": 15655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2644038979621374e-06,
+      "loss": 0.8701,
+      "step": 15656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.263370839778043e-06,
+      "loss": 0.8498,
+      "step": 15657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2623379872247155e-06,
+      "loss": 0.8967,
+      "step": 15658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.261305340329607e-06,
+      "loss": 0.8704,
+      "step": 15659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2602728991201663e-06,
+      "loss": 0.4014,
+      "step": 15660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2592406636238306e-06,
+      "loss": 0.8186,
+      "step": 15661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2582086338680365e-06,
+      "loss": 0.8371,
+      "step": 15662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2571768098802182e-06,
+      "loss": 0.9308,
+      "step": 15663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.256145191687794e-06,
+      "loss": 0.8742,
+      "step": 15664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2551137793181866e-06,
+      "loss": 0.8279,
+      "step": 15665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2540825727988093e-06,
+      "loss": 0.8651,
+      "step": 15666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2530515721570735e-06,
+      "loss": 0.9089,
+      "step": 15667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.252020777420376e-06,
+      "loss": 0.907,
+      "step": 15668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2509901886161145e-06,
+      "loss": 0.8854,
+      "step": 15669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2499598057716888e-06,
+      "loss": 0.7606,
+      "step": 15670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.248929628914478e-06,
+      "loss": 0.8563,
+      "step": 15671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.247899658071864e-06,
+      "loss": 0.9031,
+      "step": 15672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2468698932712276e-06,
+      "loss": 0.7884,
+      "step": 15673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2458403345399303e-06,
+      "loss": 0.8218,
+      "step": 15674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.244810981905341e-06,
+      "loss": 0.8661,
+      "step": 15675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2437818353948193e-06,
+      "loss": 0.9129,
+      "step": 15676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.24275289503572e-06,
+      "loss": 0.901,
+      "step": 15677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2417241608553873e-06,
+      "loss": 0.9905,
+      "step": 15678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.240695632881165e-06,
+      "loss": 0.8616,
+      "step": 15679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239667311140393e-06,
+      "loss": 0.8555,
+      "step": 15680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.238639195660397e-06,
+      "loss": 0.8467,
+      "step": 15681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2376112864685084e-06,
+      "loss": 0.9078,
+      "step": 15682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.236583583592049e-06,
+      "loss": 0.8451,
+      "step": 15683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2355560870583283e-06,
+      "loss": 0.8731,
+      "step": 15684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2345287968946593e-06,
+      "loss": 0.8413,
+      "step": 15685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.233501713128349e-06,
+      "loss": 0.885,
+      "step": 15686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.23247483578669e-06,
+      "loss": 0.8048,
+      "step": 15687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2314481648969776e-06,
+      "loss": 0.9335,
+      "step": 15688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2304217004865012e-06,
+      "loss": 0.9135,
+      "step": 15689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.229395442582546e-06,
+      "loss": 0.8316,
+      "step": 15690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2283693912123805e-06,
+      "loss": 0.9316,
+      "step": 15691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.227343546403282e-06,
+      "loss": 0.8379,
+      "step": 15692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2263179081825135e-06,
+      "loss": 0.9252,
+      "step": 15693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2252924765773376e-06,
+      "loss": 0.7711,
+      "step": 15694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2242672516150077e-06,
+      "loss": 0.8633,
+      "step": 15695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2232422333227767e-06,
+      "loss": 0.8778,
+      "step": 15696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2222174217278826e-06,
+      "loss": 0.8505,
+      "step": 15697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2211928168575647e-06,
+      "loss": 0.9092,
+      "step": 15698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2201684187390616e-06,
+      "loss": 0.8465,
+      "step": 15699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2191442273995947e-06,
+      "loss": 0.8552,
+      "step": 15700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2181202428663863e-06,
+      "loss": 0.8814,
+      "step": 15701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.217096465166654e-06,
+      "loss": 0.4234,
+      "step": 15702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.216072894327612e-06,
+      "loss": 0.8143,
+      "step": 15703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2150495303764564e-06,
+      "loss": 0.9417,
+      "step": 15704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2140263733403956e-06,
+      "loss": 0.9215,
+      "step": 15705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2130034232466247e-06,
+      "loss": 0.9046,
+      "step": 15706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2119806801223263e-06,
+      "loss": 0.8741,
+      "step": 15707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.210958143994687e-06,
+      "loss": 0.8948,
+      "step": 15708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2099358148908867e-06,
+      "loss": 0.3455,
+      "step": 15709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2089136928380928e-06,
+      "loss": 0.9192,
+      "step": 15710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2078917778634745e-06,
+      "loss": 0.9167,
+      "step": 15711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2068700699941958e-06,
+      "loss": 0.8827,
+      "step": 15712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2058485692574074e-06,
+      "loss": 0.8552,
+      "step": 15713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2048272756802623e-06,
+      "loss": 0.9162,
+      "step": 15714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2038061892899054e-06,
+      "loss": 0.8699,
+      "step": 15715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2027853101134755e-06,
+      "loss": 0.8626,
+      "step": 15716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.201764638178108e-06,
+      "loss": 0.8632,
+      "step": 15717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200744173510928e-06,
+      "loss": 0.9218,
+      "step": 15718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.199723916139065e-06,
+      "loss": 0.9247,
+      "step": 15719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1987038660896276e-06,
+      "loss": 0.9283,
+      "step": 15720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.197684023389731e-06,
+      "loss": 0.8488,
+      "step": 15721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1966643880664862e-06,
+      "loss": 0.8734,
+      "step": 15722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1956449601469863e-06,
+      "loss": 0.8936,
+      "step": 15723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1946257396583305e-06,
+      "loss": 0.9434,
+      "step": 15724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.193606726627606e-06,
+      "loss": 0.9177,
+      "step": 15725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1925879210819035e-06,
+      "loss": 0.9002,
+      "step": 15726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1915693230482937e-06,
+      "loss": 0.4221,
+      "step": 15727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1905509325538542e-06,
+      "loss": 0.975,
+      "step": 15728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1895327496256504e-06,
+      "loss": 0.9036,
+      "step": 15729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1885147742907465e-06,
+      "loss": 0.929,
+      "step": 15730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.187497006576198e-06,
+      "loss": 0.9105,
+      "step": 15731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1864794465090587e-06,
+      "loss": 0.8294,
+      "step": 15732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1854620941163696e-06,
+      "loss": 0.8491,
+      "step": 15733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.184444949425173e-06,
+      "loss": 0.8597,
+      "step": 15734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1834280124625064e-06,
+      "loss": 0.8462,
+      "step": 15735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1824112832553923e-06,
+      "loss": 0.8706,
+      "step": 15736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1813947618308584e-06,
+      "loss": 0.3949,
+      "step": 15737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.180378448215922e-06,
+      "loss": 0.811,
+      "step": 15738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1793623424375986e-06,
+      "loss": 0.8772,
+      "step": 15739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.178346444522885e-06,
+      "loss": 0.9397,
+      "step": 15740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1773307544987976e-06,
+      "loss": 0.8189,
+      "step": 15741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1763152723923196e-06,
+      "loss": 0.924,
+      "step": 15742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175299998230448e-06,
+      "loss": 0.8978,
+      "step": 15743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174284932040164e-06,
+      "loss": 0.9838,
+      "step": 15744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1732700738484523e-06,
+      "loss": 0.8741,
+      "step": 15745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.17225542368228e-06,
+      "loss": 0.8947,
+      "step": 15746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1712409815686187e-06,
+      "loss": 0.8658,
+      "step": 15747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.170226747534434e-06,
+      "loss": 0.8494,
+      "step": 15748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169212721606676e-06,
+      "loss": 0.9271,
+      "step": 15749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.168198903812302e-06,
+      "loss": 0.8588,
+      "step": 15750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1671852941782547e-06,
+      "loss": 0.8932,
+      "step": 15751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1661718927314767e-06,
+      "loss": 0.8062,
+      "step": 15752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1651586994989027e-06,
+      "loss": 0.8677,
+      "step": 15753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1641457145074627e-06,
+      "loss": 0.939,
+      "step": 15754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1631329377840828e-06,
+      "loss": 0.8919,
+      "step": 15755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162120369355676e-06,
+      "loss": 0.8952,
+      "step": 15756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.161108009249158e-06,
+      "loss": 0.9257,
+      "step": 15757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1600958574914388e-06,
+      "loss": 0.9476,
+      "step": 15758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1590839141094156e-06,
+      "loss": 0.8968,
+      "step": 15759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1580721791299862e-06,
+      "loss": 0.8672,
+      "step": 15760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1570606525800453e-06,
+      "loss": 0.8915,
+      "step": 15761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1560493344864707e-06,
+      "loss": 0.8553,
+      "step": 15762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.155038224876145e-06,
+      "loss": 0.8478,
+      "step": 15763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.154027323775948e-06,
+      "loss": 0.864,
+      "step": 15764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.153016631212741e-06,
+      "loss": 0.8558,
+      "step": 15765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1520061472133903e-06,
+      "loss": 0.9271,
+      "step": 15766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150995871804753e-06,
+      "loss": 0.8367,
+      "step": 15767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1499858050136836e-06,
+      "loss": 0.831,
+      "step": 15768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1489759468670224e-06,
+      "loss": 0.9179,
+      "step": 15769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1479662973916147e-06,
+      "loss": 0.8479,
+      "step": 15770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1469568566142972e-06,
+      "loss": 0.8924,
+      "step": 15771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.145947624561895e-06,
+      "loss": 0.8607,
+      "step": 15772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.144938601261235e-06,
+      "loss": 0.8612,
+      "step": 15773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.143929786739135e-06,
+      "loss": 0.8531,
+      "step": 15774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1429211810224094e-06,
+      "loss": 0.8502,
+      "step": 15775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.141912784137864e-06,
+      "loss": 0.9774,
+      "step": 15776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1409045961123067e-06,
+      "loss": 0.9576,
+      "step": 15777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1398966169725243e-06,
+      "loss": 0.8608,
+      "step": 15778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.138888846745314e-06,
+      "loss": 0.9083,
+      "step": 15779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1378812854574606e-06,
+      "loss": 0.8725,
+      "step": 15780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1368739331357447e-06,
+      "loss": 0.8577,
+      "step": 15781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1358667898069364e-06,
+      "loss": 0.7734,
+      "step": 15782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1348598554978083e-06,
+      "loss": 0.9465,
+      "step": 15783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.133853130235124e-06,
+      "loss": 0.8613,
+      "step": 15784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1328466140456373e-06,
+      "loss": 0.8739,
+      "step": 15785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1318403069561024e-06,
+      "loss": 0.8189,
+      "step": 15786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1308342089932656e-06,
+      "loss": 0.7866,
+      "step": 15787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1298283201838677e-06,
+      "loss": 0.8336,
+      "step": 15788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1288226405546455e-06,
+      "loss": 0.8652,
+      "step": 15789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127817170132329e-06,
+      "loss": 0.9269,
+      "step": 15790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.12681190894364e-06,
+      "loss": 0.8954,
+      "step": 15791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.125806857015298e-06,
+      "loss": 0.8348,
+      "step": 15792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1248020143740166e-06,
+      "loss": 0.8708,
+      "step": 15793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.123797381046505e-06,
+      "loss": 0.4346,
+      "step": 15794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1227929570594617e-06,
+      "loss": 0.8556,
+      "step": 15795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1217887424395857e-06,
+      "loss": 0.4494,
+      "step": 15796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1207847372135693e-06,
+      "loss": 0.8278,
+      "step": 15797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1197809414080927e-06,
+      "loss": 0.8471,
+      "step": 15798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1187773550498357e-06,
+      "loss": 0.8358,
+      "step": 15799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1177739781654803e-06,
+      "loss": 0.858,
+      "step": 15800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1167708107816886e-06,
+      "loss": 0.9449,
+      "step": 15801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1157678529251247e-06,
+      "loss": 0.8861,
+      "step": 15802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1147651046224503e-06,
+      "loss": 0.9166,
+      "step": 15803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1137625659003092e-06,
+      "loss": 0.8665,
+      "step": 15804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.112760236785354e-06,
+      "loss": 0.8866,
+      "step": 15805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1117581173042223e-06,
+      "loss": 0.8768,
+      "step": 15806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110756207483554e-06,
+      "loss": 0.8504,
+      "step": 15807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1097545073499715e-06,
+      "loss": 0.8717,
+      "step": 15808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.108753016930104e-06,
+      "loss": 0.9986,
+      "step": 15809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1077517362505708e-06,
+      "loss": 0.8479,
+      "step": 15810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1067506653379775e-06,
+      "loss": 0.9179,
+      "step": 15811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1057498042189384e-06,
+      "loss": 0.8367,
+      "step": 15812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.104749152920057e-06,
+      "loss": 0.3674,
+      "step": 15813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1037487114679235e-06,
+      "loss": 0.8672,
+      "step": 15814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1027484798891296e-06,
+      "loss": 1.033,
+      "step": 15815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.101748458210262e-06,
+      "loss": 0.9737,
+      "step": 15816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.100748646457903e-06,
+      "loss": 0.8332,
+      "step": 15817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.09974904465862e-06,
+      "loss": 0.8545,
+      "step": 15818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.098749652838985e-06,
+      "loss": 0.8682,
+      "step": 15819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0977504710255627e-06,
+      "loss": 0.86,
+      "step": 15820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0967514992449047e-06,
+      "loss": 0.9615,
+      "step": 15821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095752737523562e-06,
+      "loss": 0.8827,
+      "step": 15822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0947541858880903e-06,
+      "loss": 0.4302,
+      "step": 15823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.093755844365021e-06,
+      "loss": 0.8817,
+      "step": 15824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.092757712980891e-06,
+      "loss": 0.8617,
+      "step": 15825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0917597917622335e-06,
+      "loss": 0.7779,
+      "step": 15826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0907620807355656e-06,
+      "loss": 0.9143,
+      "step": 15827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.089764579927408e-06,
+      "loss": 0.9139,
+      "step": 15828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.088767289364274e-06,
+      "loss": 0.8929,
+      "step": 15829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0877702090726724e-06,
+      "loss": 0.9044,
+      "step": 15830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0867733390791e-06,
+      "loss": 0.928,
+      "step": 15831
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0857766794100544e-06,
+      "loss": 0.8496,
+      "step": 15832
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.084780230092027e-06,
+      "loss": 0.8397,
+      "step": 15833
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.083783991151497e-06,
+      "loss": 0.8904,
+      "step": 15834
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0827879626149493e-06,
+      "loss": 0.8417,
+      "step": 15835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0817921445088584e-06,
+      "loss": 0.9112,
+      "step": 15836
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.080796536859686e-06,
+      "loss": 0.988,
+      "step": 15837
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.079801139693898e-06,
+      "loss": 0.8714,
+      "step": 15838
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.078805953037951e-06,
+      "loss": 0.8459,
+      "step": 15839
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0778109769182932e-06,
+      "loss": 0.9648,
+      "step": 15840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.076816211361371e-06,
+      "loss": 0.9575,
+      "step": 15841
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.075821656393625e-06,
+      "loss": 0.8682,
+      "step": 15842
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.074827312041493e-06,
+      "loss": 0.8859,
+      "step": 15843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.073833178331396e-06,
+      "loss": 0.9331,
+      "step": 15844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0728392552897603e-06,
+      "loss": 0.8933,
+      "step": 15845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.071845542943003e-06,
+      "loss": 0.9083,
+      "step": 15846
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.070852041317537e-06,
+      "loss": 0.8909,
+      "step": 15847
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0698587504397684e-06,
+      "loss": 0.8906,
+      "step": 15848
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.068865670336099e-06,
+      "loss": 0.8438,
+      "step": 15849
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0678728010329196e-06,
+      "loss": 0.8543,
+      "step": 15850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0668801425566207e-06,
+      "loss": 0.8943,
+      "step": 15851
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0658876949335917e-06,
+      "loss": 0.8708,
+      "step": 15852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0648954581902014e-06,
+      "loss": 0.8168,
+      "step": 15853
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0639034323528286e-06,
+      "loss": 0.9009,
+      "step": 15854
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.062911617447838e-06,
+      "loss": 0.7924,
+      "step": 15855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.061920013501595e-06,
+      "loss": 0.8499,
+      "step": 15856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.060928620540449e-06,
+      "loss": 0.8864,
+      "step": 15857
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0599374385907488e-06,
+      "loss": 0.8086,
+      "step": 15858
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0589464676788507e-06,
+      "loss": 0.9091,
+      "step": 15859
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.057955707831082e-06,
+      "loss": 0.8647,
+      "step": 15860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0569651590737794e-06,
+      "loss": 0.8546,
+      "step": 15861
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.055974821433274e-06,
+      "loss": 0.9781,
+      "step": 15862
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0549846949358833e-06,
+      "loss": 0.8782,
+      "step": 15863
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.053994779607924e-06,
+      "loss": 0.9072,
+      "step": 15864
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0530050754757093e-06,
+      "loss": 0.8252,
+      "step": 15865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0520155825655453e-06,
+      "loss": 0.8012,
+      "step": 15866
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051026300903728e-06,
+      "loss": 0.789,
+      "step": 15867
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0500372305165516e-06,
+      "loss": 0.9253,
+      "step": 15868
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.04904837143031e-06,
+      "loss": 0.8509,
+      "step": 15869
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0480597236712753e-06,
+      "loss": 0.8575,
+      "step": 15870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.047071287265735e-06,
+      "loss": 0.9157,
+      "step": 15871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0460830622399585e-06,
+      "loss": 0.9197,
+      "step": 15872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.045095048620207e-06,
+      "loss": 0.8991,
+      "step": 15873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0441072464327448e-06,
+      "loss": 0.9063,
+      "step": 15874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.043119655703827e-06,
+      "loss": 0.8817,
+      "step": 15875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0421322764596997e-06,
+      "loss": 0.8251,
+      "step": 15876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0411451087266064e-06,
+      "loss": 0.8492,
+      "step": 15877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.040158152530788e-06,
+      "loss": 0.9421,
+      "step": 15878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.039171407898476e-06,
+      "loss": 0.8746,
+      "step": 15879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.038184874855894e-06,
+      "loss": 0.9749,
+      "step": 15880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.037198553429264e-06,
+      "loss": 0.889,
+      "step": 15881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0362124436448026e-06,
+      "loss": 0.9091,
+      "step": 15882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.035226545528719e-06,
+      "loss": 0.9132,
+      "step": 15883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.034240859107217e-06,
+      "loss": 0.8719,
+      "step": 15884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0332553844064975e-06,
+      "loss": 0.9182,
+      "step": 15885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0322701214527484e-06,
+      "loss": 0.7396,
+      "step": 15886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.03128507027216e-06,
+      "loss": 0.8695,
+      "step": 15887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.030300230890916e-06,
+      "loss": 0.8962,
+      "step": 15888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0293156033351868e-06,
+      "loss": 0.8629,
+      "step": 15889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0283311876311453e-06,
+      "loss": 0.9439,
+      "step": 15890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0273469838049565e-06,
+      "loss": 0.8425,
+      "step": 15891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.026362991882781e-06,
+      "loss": 0.8728,
+      "step": 15892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0253792118907667e-06,
+      "loss": 0.8869,
+      "step": 15893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0243956438550662e-06,
+      "loss": 0.9321,
+      "step": 15894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0234122878018236e-06,
+      "loss": 0.8344,
+      "step": 15895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.02242914375717e-06,
+      "loss": 0.8392,
+      "step": 15896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.021446211747238e-06,
+      "loss": 0.9,
+      "step": 15897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020463491798157e-06,
+      "loss": 0.8567,
+      "step": 15898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0194809839360384e-06,
+      "loss": 0.9114,
+      "step": 15899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.018498688187003e-06,
+      "loss": 0.9296,
+      "step": 15900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0175166045771575e-06,
+      "loss": 0.9071,
+      "step": 15901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016534733132601e-06,
+      "loss": 0.8707,
+      "step": 15902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015553073879435e-06,
+      "loss": 0.7961,
+      "step": 15903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0145716268437477e-06,
+      "loss": 0.8661,
+      "step": 15904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0135903920516266e-06,
+      "loss": 0.8811,
+      "step": 15905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0126093695291517e-06,
+      "loss": 0.7926,
+      "step": 15906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.011628559302398e-06,
+      "loss": 0.917,
+      "step": 15907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.010647961397435e-06,
+      "loss": 0.7422,
+      "step": 15908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.009667575840323e-06,
+      "loss": 0.8967,
+      "step": 15909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0086874026571203e-06,
+      "loss": 0.8863,
+      "step": 15910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.007707441873882e-06,
+      "loss": 0.8895,
+      "step": 15911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.006727693516649e-06,
+      "loss": 0.9197,
+      "step": 15912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0057481576114657e-06,
+      "loss": 0.9451,
+      "step": 15913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0047688341843653e-06,
+      "loss": 0.8642,
+      "step": 15914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0037897232613803e-06,
+      "loss": 0.9173,
+      "step": 15915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.00281082486853e-06,
+      "loss": 0.9467,
+      "step": 15916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001832139031834e-06,
+      "loss": 0.9036,
+      "step": 15917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.000853665777305e-06,
+      "loss": 0.8494,
+      "step": 15918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.999875405130949e-06,
+      "loss": 0.8736,
+      "step": 15919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9988973571187685e-06,
+      "loss": 0.8509,
+      "step": 15920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.99791952176676e-06,
+      "loss": 0.8661,
+      "step": 15921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.996941899100909e-06,
+      "loss": 0.8441,
+      "step": 15922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.995964489147203e-06,
+      "loss": 0.7696,
+      "step": 15923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9949872919316216e-06,
+      "loss": 0.917,
+      "step": 15924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.994010307480132e-06,
+      "loss": 0.8934,
+      "step": 15925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.993033535818705e-06,
+      "loss": 0.8442,
+      "step": 15926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9920569769733024e-06,
+      "loss": 0.8826,
+      "step": 15927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9910806309698804e-06,
+      "loss": 0.9381,
+      "step": 15928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.990104497834384e-06,
+      "loss": 0.8747,
+      "step": 15929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9891285775927684e-06,
+      "loss": 0.8893,
+      "step": 15930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.988152870270962e-06,
+      "loss": 0.8773,
+      "step": 15931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.987177375894901e-06,
+      "loss": 0.9281,
+      "step": 15932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9862020944905146e-06,
+      "loss": 0.8972,
+      "step": 15933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985227026083727e-06,
+      "loss": 0.787,
+      "step": 15934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.984252170700448e-06,
+      "loss": 0.8158,
+      "step": 15935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.983277528366593e-06,
+      "loss": 0.8393,
+      "step": 15936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.982303099108066e-06,
+      "loss": 0.8206,
+      "step": 15937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.981328882950764e-06,
+      "loss": 0.8382,
+      "step": 15938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9803548799205817e-06,
+      "loss": 0.8769,
+      "step": 15939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9793810900434073e-06,
+      "loss": 0.8316,
+      "step": 15940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.978407513345123e-06,
+      "loss": 0.9134,
+      "step": 15941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9774341498516057e-06,
+      "loss": 0.7808,
+      "step": 15942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.976460999588726e-06,
+      "loss": 0.9307,
+      "step": 15943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9754880625823526e-06,
+      "loss": 0.8608,
+      "step": 15944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.974515338858338e-06,
+      "loss": 0.9066,
+      "step": 15945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9735428284425393e-06,
+      "loss": 0.9262,
+      "step": 15946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972570531360809e-06,
+      "loss": 0.8913,
+      "step": 15947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9715984476389825e-06,
+      "loss": 0.8741,
+      "step": 15948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9706265773029e-06,
+      "loss": 0.9782,
+      "step": 15949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9696549203783954e-06,
+      "loss": 0.8153,
+      "step": 15950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968683476891289e-06,
+      "loss": 0.8367,
+      "step": 15951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9677122468674003e-06,
+      "loss": 0.8497,
+      "step": 15952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.966741230332552e-06,
+      "loss": 0.9565,
+      "step": 15953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.965770427312543e-06,
+      "loss": 0.8672,
+      "step": 15954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9647998378331814e-06,
+      "loss": 0.8487,
+      "step": 15955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9638294619202625e-06,
+      "loss": 0.8674,
+      "step": 15956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.962859299599581e-06,
+      "loss": 0.9138,
+      "step": 15957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.961889350896917e-06,
+      "loss": 0.852,
+      "step": 15958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.960919615838054e-06,
+      "loss": 0.9167,
+      "step": 15959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9599500944487694e-06,
+      "loss": 0.9106,
+      "step": 15960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.958980786754825e-06,
+      "loss": 0.8513,
+      "step": 15961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9580116927819882e-06,
+      "loss": 0.8654,
+      "step": 15962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.957042812556016e-06,
+      "loss": 0.9263,
+      "step": 15963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.956074146102659e-06,
+      "loss": 0.8766,
+      "step": 15964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9551056934476653e-06,
+      "loss": 0.8431,
+      "step": 15965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9541374546167757e-06,
+      "loss": 0.9245,
+      "step": 15966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.953169429635721e-06,
+      "loss": 0.9221,
+      "step": 15967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9522016185302327e-06,
+      "loss": 0.8565,
+      "step": 15968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951234021326034e-06,
+      "loss": 0.9497,
+      "step": 15969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9502666380488442e-06,
+      "loss": 0.86,
+      "step": 15970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9492994687243715e-06,
+      "loss": 0.8626,
+      "step": 15971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.948332513378325e-06,
+      "loss": 0.9873,
+      "step": 15972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.947365772036406e-06,
+      "loss": 0.9675,
+      "step": 15973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.946399244724305e-06,
+      "loss": 0.7458,
+      "step": 15974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9454329314677144e-06,
+      "loss": 0.8741,
+      "step": 15975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9444668322923167e-06,
+      "loss": 0.9369,
+      "step": 15976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.943500947223791e-06,
+      "loss": 0.8197,
+      "step": 15977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942535276287809e-06,
+      "loss": 0.9004,
+      "step": 15978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9415698195100384e-06,
+      "loss": 0.8291,
+      "step": 15979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.940604576916136e-06,
+      "loss": 0.8373,
+      "step": 15980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.93963954853176e-06,
+      "loss": 0.8263,
+      "step": 15981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9386747343825584e-06,
+      "loss": 0.9514,
+      "step": 15982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9377101344941783e-06,
+      "loss": 0.8166,
+      "step": 15983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9367457488922526e-06,
+      "loss": 0.8874,
+      "step": 15984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935781577602416e-06,
+      "loss": 0.871,
+      "step": 15985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9348176206502965e-06,
+      "loss": 0.8526,
+      "step": 15986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933853878061511e-06,
+      "loss": 0.8418,
+      "step": 15987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9328903498616745e-06,
+      "loss": 0.8634,
+      "step": 15988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9319270360764044e-06,
+      "loss": 0.8732,
+      "step": 15989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.930963936731296e-06,
+      "loss": 0.7993,
+      "step": 15990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.93000105185195e-06,
+      "loss": 0.8847,
+      "step": 15991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.92903838146396e-06,
+      "loss": 0.8639,
+      "step": 15992
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9280759255929138e-06,
+      "loss": 0.8823,
+      "step": 15993
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.927113684264389e-06,
+      "loss": 0.8661,
+      "step": 15994
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.926151657503962e-06,
+      "loss": 0.9422,
+      "step": 15995
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9251898453372044e-06,
+      "loss": 0.8531,
+      "step": 15996
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9242282477896768e-06,
+      "loss": 0.8835,
+      "step": 15997
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.923266864886938e-06,
+      "loss": 0.8597,
+      "step": 15998
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.922305696654545e-06,
+      "loss": 0.9301,
+      "step": 15999
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9213447431180352e-06,
+      "loss": 0.8886,
+      "step": 16000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9203840043029586e-06,
+      "loss": 0.9164,
+      "step": 16001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.91942348023485e-06,
+      "loss": 0.7753,
+      "step": 16002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9184631709392333e-06,
+      "loss": 0.8301,
+      "step": 16003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9175030764416345e-06,
+      "loss": 0.867,
+      "step": 16004
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.916543196767575e-06,
+      "loss": 0.9611,
+      "step": 16005
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9155835319425664e-06,
+      "loss": 0.8237,
+      "step": 16006
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9146240819921113e-06,
+      "loss": 0.8646,
+      "step": 16007
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9136648469417153e-06,
+      "loss": 0.952,
+      "step": 16008
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.912705826816873e-06,
+      "loss": 0.8364,
+      "step": 16009
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.911747021643072e-06,
+      "loss": 0.8569,
+      "step": 16010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9107884314457948e-06,
+      "loss": 0.962,
+      "step": 16011
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9098300562505266e-06,
+      "loss": 0.9507,
+      "step": 16012
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9088718960827337e-06,
+      "loss": 0.8633,
+      "step": 16013
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9079139509678846e-06,
+      "loss": 0.9465,
+      "step": 16014
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9069562209314429e-06,
+      "loss": 0.9412,
+      "step": 16015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9059987059988595e-06,
+      "loss": 0.9094,
+      "step": 16016
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.905041406195587e-06,
+      "loss": 0.887,
+      "step": 16017
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9040843215470673e-06,
+      "loss": 0.8846,
+      "step": 16018
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9031274520787435e-06,
+      "loss": 0.8768,
+      "step": 16019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9021707978160409e-06,
+      "loss": 0.8247,
+      "step": 16020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9012143587843912e-06,
+      "loss": 0.9396,
+      "step": 16021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9002581350092152e-06,
+      "loss": 0.8993,
+      "step": 16022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8993021265159227e-06,
+      "loss": 0.9091,
+      "step": 16023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8983463333299313e-06,
+      "loss": 0.8653,
+      "step": 16024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8973907554766435e-06,
+      "loss": 0.9226,
+      "step": 16025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8964353929814516e-06,
+      "loss": 0.8629,
+      "step": 16026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.895480245869753e-06,
+      "loss": 0.8408,
+      "step": 16027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8945253141669351e-06,
+      "loss": 0.9108,
+      "step": 16028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8935705978983753e-06,
+      "loss": 0.8787,
+      "step": 16029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8926160970894502e-06,
+      "loss": 0.8429,
+      "step": 16030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8916618117655294e-06,
+      "loss": 0.7741,
+      "step": 16031
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8907077419519792e-06,
+      "loss": 0.8076,
+      "step": 16032
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.889753887674154e-06,
+      "loss": 0.9351,
+      "step": 16033
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8888002489574076e-06,
+      "loss": 0.8663,
+      "step": 16034
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8878468258270854e-06,
+      "loss": 0.9275,
+      "step": 16035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8868936183085307e-06,
+      "loss": 0.8369,
+      "step": 16036
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8859406264270774e-06,
+      "loss": 0.8799,
+      "step": 16037
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8849878502080577e-06,
+      "loss": 0.8491,
+      "step": 16038
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.88403528967679e-06,
+      "loss": 0.8738,
+      "step": 16039
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8830829448585951e-06,
+      "loss": 0.855,
+      "step": 16040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8821308157787854e-06,
+      "loss": 0.9169,
+      "step": 16041
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8811789024626693e-06,
+      "loss": 0.8552,
+      "step": 16042
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.880227204935543e-06,
+      "loss": 0.8402,
+      "step": 16043
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8792757232227043e-06,
+      "loss": 0.8796,
+      "step": 16044
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.878324457349444e-06,
+      "loss": 0.8231,
+      "step": 16045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.877373407341041e-06,
+      "loss": 0.9151,
+      "step": 16046
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8764225732227747e-06,
+      "loss": 0.8116,
+      "step": 16047
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8754719550199219e-06,
+      "loss": 0.8829,
+      "step": 16048
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8745215527577443e-06,
+      "loss": 0.9005,
+      "step": 16049
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.873571366461503e-06,
+      "loss": 0.8535,
+      "step": 16050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8726213961564554e-06,
+      "loss": 0.9286,
+      "step": 16051
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.871671641867847e-06,
+      "loss": 0.8867,
+      "step": 16052
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8707221036209221e-06,
+      "loss": 0.8719,
+      "step": 16053
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8697727814409206e-06,
+      "loss": 0.9303,
+      "step": 16054
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8688236753530743e-06,
+      "loss": 0.8328,
+      "step": 16055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8678747853826063e-06,
+      "loss": 0.8279,
+      "step": 16056
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8669261115547378e-06,
+      "loss": 0.916,
+      "step": 16057
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8659776538946872e-06,
+      "loss": 0.9218,
+      "step": 16058
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8650294124276558e-06,
+      "loss": 0.9481,
+      "step": 16059
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8640813871788544e-06,
+      "loss": 0.7852,
+      "step": 16060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8631335781734794e-06,
+      "loss": 0.8673,
+      "step": 16061
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8621859854367186e-06,
+      "loss": 0.852,
+      "step": 16062
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8612386089937606e-06,
+      "loss": 0.8479,
+      "step": 16063
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8602914488697865e-06,
+      "loss": 0.9405,
+      "step": 16064
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8593445050899673e-06,
+      "loss": 0.8144,
+      "step": 16065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8583977776794737e-06,
+      "loss": 0.8143,
+      "step": 16066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.857451266663468e-06,
+      "loss": 0.8124,
+      "step": 16067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8565049720671112e-06,
+      "loss": 0.8303,
+      "step": 16068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.855558893915549e-06,
+      "loss": 0.9082,
+      "step": 16069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8546130322339262e-06,
+      "loss": 0.9038,
+      "step": 16070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.853667387047392e-06,
+      "loss": 0.8737,
+      "step": 16071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8527219583810718e-06,
+      "loss": 0.8493,
+      "step": 16072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8517767462600966e-06,
+      "loss": 0.841,
+      "step": 16073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8508317507095918e-06,
+      "loss": 0.8548,
+      "step": 16074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8498869717546696e-06,
+      "loss": 0.8637,
+      "step": 16075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8489424094204434e-06,
+      "loss": 0.8344,
+      "step": 16076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8479980637320205e-06,
+      "loss": 0.9417,
+      "step": 16077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8470539347144967e-06,
+      "loss": 0.9327,
+      "step": 16078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8461100223929684e-06,
+      "loss": 0.8745,
+      "step": 16079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8451663267925213e-06,
+      "loss": 0.4088,
+      "step": 16080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.844222847938244e-06,
+      "loss": 0.4152,
+      "step": 16081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8432795858552021e-06,
+      "loss": 0.9716,
+      "step": 16082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8423365405684757e-06,
+      "loss": 0.8137,
+      "step": 16083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.84139371210313e-06,
+      "loss": 0.8962,
+      "step": 16084
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8404511004842185e-06,
+      "loss": 0.8611,
+      "step": 16085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8395087057367989e-06,
+      "loss": 0.8746,
+      "step": 16086
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8385665278859188e-06,
+      "loss": 0.7779,
+      "step": 16087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8376245669566174e-06,
+      "loss": 0.8725,
+      "step": 16088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.836682822973932e-06,
+      "loss": 0.9363,
+      "step": 16089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.835741295962894e-06,
+      "loss": 0.9031,
+      "step": 16090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.834799985948531e-06,
+      "loss": 0.8579,
+      "step": 16091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8338588929558554e-06,
+      "loss": 0.929,
+      "step": 16092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8329180170098848e-06,
+      "loss": 1.0059,
+      "step": 16093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8319773581356248e-06,
+      "loss": 0.8684,
+      "step": 16094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8310369163580777e-06,
+      "loss": 0.8767,
+      "step": 16095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.83009669170224e-06,
+      "loss": 0.7636,
+      "step": 16096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.829156684193104e-06,
+      "loss": 0.872,
+      "step": 16097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8282168938556489e-06,
+      "loss": 0.8606,
+      "step": 16098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8272773207148553e-06,
+      "loss": 0.9332,
+      "step": 16099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8263379647956992e-06,
+      "loss": 0.9956,
+      "step": 16100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8253988261231415e-06,
+      "loss": 0.9964,
+      "step": 16101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8244599047221467e-06,
+      "loss": 0.877,
+      "step": 16102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8235212006176717e-06,
+      "loss": 0.88,
+      "step": 16103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8225827138346663e-06,
+      "loss": 0.8545,
+      "step": 16104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8216444443980696e-06,
+      "loss": 0.8326,
+      "step": 16105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.820706392332824e-06,
+      "loss": 0.86,
+      "step": 16106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8197685576638613e-06,
+      "loss": 0.9508,
+      "step": 16107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8188309404161065e-06,
+      "loss": 0.988,
+      "step": 16108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8178935406144827e-06,
+      "loss": 0.8752,
+      "step": 16109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8169563582839045e-06,
+      "loss": 0.8884,
+      "step": 16110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8160193934492798e-06,
+      "loss": 0.3934,
+      "step": 16111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8150826461355108e-06,
+      "loss": 0.926,
+      "step": 16112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8141461163675013e-06,
+      "loss": 0.9257,
+      "step": 16113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.813209804170135e-06,
+      "loss": 0.8372,
+      "step": 16114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.812273709568303e-06,
+      "loss": 0.947,
+      "step": 16115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8113378325868837e-06,
+      "loss": 0.8382,
+      "step": 16116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8104021732507548e-06,
+      "loss": 0.9311,
+      "step": 16117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8094667315847781e-06,
+      "loss": 0.7662,
+      "step": 16118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8085315076138231e-06,
+      "loss": 0.8735,
+      "step": 16119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8075965013627484e-06,
+      "loss": 0.9174,
+      "step": 16120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8066617128564e-06,
+      "loss": 0.9223,
+      "step": 16121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8057271421196243e-06,
+      "loss": 0.8533,
+      "step": 16122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8047927891772654e-06,
+      "loss": 0.8388,
+      "step": 16123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8038586540541525e-06,
+      "loss": 0.8513,
+      "step": 16124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8029247367751146e-06,
+      "loss": 0.8943,
+      "step": 16125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8019910373649785e-06,
+      "loss": 0.8428,
+      "step": 16126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8010575558485554e-06,
+      "loss": 0.8479,
+      "step": 16127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.800124292250658e-06,
+      "loss": 0.9122,
+      "step": 16128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7991912465960915e-06,
+      "loss": 0.9119,
+      "step": 16129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7982584189096553e-06,
+      "loss": 0.9098,
+      "step": 16130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7973258092161427e-06,
+      "loss": 0.852,
+      "step": 16131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.796393417540343e-06,
+      "loss": 0.8769,
+      "step": 16132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7954612439070384e-06,
+      "loss": 0.8382,
+      "step": 16133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7945292883410016e-06,
+      "loss": 0.8442,
+      "step": 16134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7935975508670046e-06,
+      "loss": 0.8678,
+      "step": 16135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7926660315098142e-06,
+      "loss": 0.8207,
+      "step": 16136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7917347302941846e-06,
+      "loss": 0.8141,
+      "step": 16137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.790803647244871e-06,
+      "loss": 0.8516,
+      "step": 16138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7898727823866214e-06,
+      "loss": 0.8617,
+      "step": 16139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7889421357441784e-06,
+      "loss": 0.881,
+      "step": 16140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7880117073422698e-06,
+      "loss": 0.8227,
+      "step": 16141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7870814972056373e-06,
+      "loss": 0.916,
+      "step": 16142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7861515053589951e-06,
+      "loss": 0.4131,
+      "step": 16143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7852217318270659e-06,
+      "loss": 0.9694,
+      "step": 16144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.78429217663456e-06,
+      "loss": 0.8985,
+      "step": 16145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7833628398061875e-06,
+      "loss": 0.9081,
+      "step": 16146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7824337213666432e-06,
+      "loss": 0.8453,
+      "step": 16147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.781504821340626e-06,
+      "loss": 0.8799,
+      "step": 16148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7805761397528265e-06,
+      "loss": 0.9596,
+      "step": 16149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7796476766279235e-06,
+      "loss": 0.8476,
+      "step": 16150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7787194319905964e-06,
+      "loss": 0.8603,
+      "step": 16151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7777914058655177e-06,
+      "loss": 0.9714,
+      "step": 16152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.776863598277353e-06,
+      "loss": 0.9261,
+      "step": 16153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7759360092507604e-06,
+      "loss": 0.8995,
+      "step": 16154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7750086388104004e-06,
+      "loss": 0.831,
+      "step": 16155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7740814869809131e-06,
+      "loss": 0.9356,
+      "step": 16156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7731545537869455e-06,
+      "loss": 0.9381,
+      "step": 16157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7722278392531346e-06,
+      "loss": 0.8979,
+      "step": 16158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7713013434041126e-06,
+      "loss": 0.8432,
+      "step": 16159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7703750662644992e-06,
+      "loss": 0.9712,
+      "step": 16160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7694490078589189e-06,
+      "loss": 0.8028,
+      "step": 16161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7685231682119864e-06,
+      "loss": 0.8332,
+      "step": 16162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.767597547348303e-06,
+      "loss": 0.8238,
+      "step": 16163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.766672145292475e-06,
+      "loss": 0.9243,
+      "step": 16164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7657469620690992e-06,
+      "loss": 0.8268,
+      "step": 16165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7648219977027636e-06,
+      "loss": 0.9313,
+      "step": 16166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7638972522180542e-06,
+      "loss": 0.787,
+      "step": 16167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7629727256395513e-06,
+      "loss": 0.8901,
+      "step": 16168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.762048417991824e-06,
+      "loss": 0.9098,
+      "step": 16169
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7611243292994406e-06,
+      "loss": 0.8542,
+      "step": 16170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7602004595869627e-06,
+      "loss": 0.8612,
+      "step": 16171
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.759276808878948e-06,
+      "loss": 0.8963,
+      "step": 16172
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.758353377199943e-06,
+      "loss": 0.9122,
+      "step": 16173
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7574301645744906e-06,
+      "loss": 0.8757,
+      "step": 16174
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7565071710271342e-06,
+      "loss": 0.8277,
+      "step": 16175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7555843965823992e-06,
+      "loss": 0.8606,
+      "step": 16176
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.754661841264812e-06,
+      "loss": 0.9478,
+      "step": 16177
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7537395050989014e-06,
+      "loss": 0.8565,
+      "step": 16178
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7528173881091749e-06,
+      "loss": 0.9436,
+      "step": 16179
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7518954903201423e-06,
+      "loss": 0.8236,
+      "step": 16180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7509738117563068e-06,
+      "loss": 0.9013,
+      "step": 16181
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7500523524421698e-06,
+      "loss": 0.9053,
+      "step": 16182
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7491311124022158e-06,
+      "loss": 0.827,
+      "step": 16183
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7482100916609334e-06,
+      "loss": 0.944,
+      "step": 16184
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.747289290242804e-06,
+      "loss": 0.8667,
+      "step": 16185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7463687081722991e-06,
+      "loss": 0.8207,
+      "step": 16186
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7454483454738857e-06,
+      "loss": 0.9544,
+      "step": 16187
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7445282021720288e-06,
+      "loss": 0.8607,
+      "step": 16188
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7436082782911833e-06,
+      "loss": 0.8375,
+      "step": 16189
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7426885738557997e-06,
+      "loss": 0.8995,
+      "step": 16190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.741769088890326e-06,
+      "loss": 0.4143,
+      "step": 16191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.740849823419195e-06,
+      "loss": 0.8625,
+      "step": 16192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.739930777466844e-06,
+      "loss": 0.8009,
+      "step": 16193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7390119510576974e-06,
+      "loss": 0.8823,
+      "step": 16194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7380933442161818e-06,
+      "loss": 0.9399,
+      "step": 16195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7371749569667073e-06,
+      "loss": 0.796,
+      "step": 16196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7362567893336834e-06,
+      "loss": 0.8279,
+      "step": 16197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7353388413415206e-06,
+      "loss": 0.869,
+      "step": 16198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7344211130146083e-06,
+      "loss": 0.8326,
+      "step": 16199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7335036043773402e-06,
+      "loss": 0.8447,
+      "step": 16200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7325863154541113e-06,
+      "loss": 0.8585,
+      "step": 16201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7316692462692918e-06,
+      "loss": 0.9301,
+      "step": 16202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7307523968472605e-06,
+      "loss": 0.8827,
+      "step": 16203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7298357672123899e-06,
+      "loss": 0.9862,
+      "step": 16204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7289193573890362e-06,
+      "loss": 0.8691,
+      "step": 16205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7280031674015586e-06,
+      "loss": 0.8194,
+      "step": 16206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7270871972743097e-06,
+      "loss": 0.9204,
+      "step": 16207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.726171447031637e-06,
+      "loss": 0.8732,
+      "step": 16208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7252559166978755e-06,
+      "loss": 0.8426,
+      "step": 16209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7243406062973612e-06,
+      "loss": 0.8129,
+      "step": 16210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7234255158544244e-06,
+      "loss": 0.7619,
+      "step": 16211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7225106453933792e-06,
+      "loss": 0.9236,
+      "step": 16212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7215959949385496e-06,
+      "loss": 0.8811,
+      "step": 16213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7206815645142461e-06,
+      "loss": 0.9256,
+      "step": 16214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.719767354144769e-06,
+      "loss": 0.9368,
+      "step": 16215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7188533638544181e-06,
+      "loss": 0.7935,
+      "step": 16216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7179395936674891e-06,
+      "loss": 0.8669,
+      "step": 16217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7170260436082653e-06,
+      "loss": 0.8373,
+      "step": 16218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.716112713701028e-06,
+      "loss": 0.833,
+      "step": 16219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7151996039700535e-06,
+      "loss": 0.8589,
+      "step": 16220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7142867144396147e-06,
+      "loss": 0.8492,
+      "step": 16221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7133740451339699e-06,
+      "loss": 0.8914,
+      "step": 16222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7124615960773772e-06,
+      "loss": 0.9415,
+      "step": 16223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7115493672940908e-06,
+      "loss": 0.8945,
+      "step": 16224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7106373588083557e-06,
+      "loss": 0.925,
+      "step": 16225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7097255706444126e-06,
+      "loss": 0.9626,
+      "step": 16226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7088140028264977e-06,
+      "loss": 0.8453,
+      "step": 16227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7079026553788346e-06,
+      "loss": 0.7761,
+      "step": 16228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.706991528325649e-06,
+      "loss": 0.8422,
+      "step": 16229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7060806216911562e-06,
+      "loss": 0.3888,
+      "step": 16230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.705169935499571e-06,
+      "loss": 0.8663,
+      "step": 16231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7042594697750925e-06,
+      "loss": 0.8697,
+      "step": 16232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7033492245419225e-06,
+      "loss": 0.8538,
+      "step": 16233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7024391998242573e-06,
+      "loss": 0.896,
+      "step": 16234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7015293956462787e-06,
+      "loss": 0.8978,
+      "step": 16235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7006198120321694e-06,
+      "loss": 0.8732,
+      "step": 16236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.69971044900611e-06,
+      "loss": 0.8295,
+      "step": 16237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.698801306592266e-06,
+      "loss": 0.9503,
+      "step": 16238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.697892384814802e-06,
+      "loss": 0.7665,
+      "step": 16239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.696983683697878e-06,
+      "loss": 0.8658,
+      "step": 16240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6960752032656436e-06,
+      "loss": 0.9894,
+      "step": 16241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6951669435422459e-06,
+      "loss": 0.9388,
+      "step": 16242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6942589045518255e-06,
+      "loss": 0.8634,
+      "step": 16243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.69335108631852e-06,
+      "loss": 0.842,
+      "step": 16244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6924434888664532e-06,
+      "loss": 0.2959,
+      "step": 16245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6915361122197504e-06,
+      "loss": 0.8683,
+      "step": 16246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.690628956402528e-06,
+      "loss": 0.8835,
+      "step": 16247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6897220214388986e-06,
+      "loss": 0.9079,
+      "step": 16248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6888153073529656e-06,
+      "loss": 0.8628,
+      "step": 16249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6879088141688327e-06,
+      "loss": 0.9019,
+      "step": 16250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6870025419105872e-06,
+      "loss": 0.8787,
+      "step": 16251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6860964906023203e-06,
+      "loss": 0.8814,
+      "step": 16252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6851906602681156e-06,
+      "loss": 0.8978,
+      "step": 16253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.684285050932044e-06,
+      "loss": 0.9099,
+      "step": 16254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6833796626181786e-06,
+      "loss": 0.9029,
+      "step": 16255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6824744953505834e-06,
+      "loss": 0.9414,
+      "step": 16256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.681569549153319e-06,
+      "loss": 0.8563,
+      "step": 16257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.680664824050432e-06,
+      "loss": 0.8802,
+      "step": 16258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6797603200659706e-06,
+      "loss": 0.9511,
+      "step": 16259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6788560372239826e-06,
+      "loss": 0.9123,
+      "step": 16260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6779519755484952e-06,
+      "loss": 0.8185,
+      "step": 16261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.677048135063538e-06,
+      "loss": 0.8554,
+      "step": 16262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.67614451579314e-06,
+      "loss": 0.8232,
+      "step": 16263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6752411177613103e-06,
+      "loss": 0.9728,
+      "step": 16264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6743379409920647e-06,
+      "loss": 0.8982,
+      "step": 16265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6734349855094102e-06,
+      "loss": 0.8869,
+      "step": 16266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.672532251337341e-06,
+      "loss": 0.8677,
+      "step": 16267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6716297384998537e-06,
+      "loss": 0.8868,
+      "step": 16268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6707274470209366e-06,
+      "loss": 0.3943,
+      "step": 16269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6698253769245731e-06,
+      "loss": 0.928,
+      "step": 16270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6689235282347315e-06,
+      "loss": 0.8457,
+      "step": 16271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6680219009753917e-06,
+      "loss": 0.8788,
+      "step": 16272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6671204951705145e-06,
+      "loss": 0.8737,
+      "step": 16273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.666219310844056e-06,
+      "loss": 0.9305,
+      "step": 16274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6653183480199697e-06,
+      "loss": 0.878,
+      "step": 16275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.664417606722205e-06,
+      "loss": 0.9107,
+      "step": 16276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6635170869746986e-06,
+      "loss": 0.8531,
+      "step": 16277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6626167888013856e-06,
+      "loss": 0.8373,
+      "step": 16278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6617167122261967e-06,
+      "loss": 0.8791,
+      "step": 16279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6608168572730564e-06,
+      "loss": 0.8,
+      "step": 16280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.659917223965878e-06,
+      "loss": 0.8746,
+      "step": 16281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.659017812328575e-06,
+      "loss": 0.7973,
+      "step": 16282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6581186223850509e-06,
+      "loss": 0.9075,
+      "step": 16283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6572196541592068e-06,
+      "loss": 0.8779,
+      "step": 16284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6563209076749365e-06,
+      "loss": 0.8488,
+      "step": 16285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6554223829561288e-06,
+      "loss": 0.9047,
+      "step": 16286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6545240800266626e-06,
+      "loss": 0.8927,
+      "step": 16287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.653625998910413e-06,
+      "loss": 0.8849,
+      "step": 16288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6527281396312555e-06,
+      "loss": 0.9227,
+      "step": 16289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6518305022130488e-06,
+      "loss": 0.886,
+      "step": 16290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6509330866796513e-06,
+      "loss": 0.8153,
+      "step": 16291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6500358930549175e-06,
+      "loss": 0.7905,
+      "step": 16292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6491389213626951e-06,
+      "loss": 0.8383,
+      "step": 16293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6482421716268215e-06,
+      "loss": 0.8744,
+      "step": 16294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.647345643871131e-06,
+      "loss": 0.891,
+      "step": 16295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6464493381194535e-06,
+      "loss": 0.8675,
+      "step": 16296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.645553254395613e-06,
+      "loss": 0.3731,
+      "step": 16297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6446573927234243e-06,
+      "loss": 0.8059,
+      "step": 16298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6437617531267025e-06,
+      "loss": 0.8554,
+      "step": 16299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6428663356292474e-06,
+      "loss": 0.8377,
+      "step": 16300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.64197114025486e-06,
+      "loss": 0.7455,
+      "step": 16301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6410761670273355e-06,
+      "loss": 0.8818,
+      "step": 16302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6401814159704588e-06,
+      "loss": 0.864,
+      "step": 16303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6392868871080104e-06,
+      "loss": 0.9861,
+      "step": 16304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.638392580463769e-06,
+      "loss": 0.8202,
+      "step": 16305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6374984960615048e-06,
+      "loss": 0.8821,
+      "step": 16306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6366046339249753e-06,
+      "loss": 0.9184,
+      "step": 16307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6357109940779447e-06,
+      "loss": 0.9188,
+      "step": 16308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6348175765441666e-06,
+      "loss": 0.8854,
+      "step": 16309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6339243813473794e-06,
+      "loss": 0.8233,
+      "step": 16310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6330314085113274e-06,
+      "loss": 0.8985,
+      "step": 16311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6321386580597476e-06,
+      "loss": 0.8578,
+      "step": 16312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.631246130016363e-06,
+      "loss": 0.8833,
+      "step": 16313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6303538244048967e-06,
+      "loss": 0.8999,
+      "step": 16314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.629461741249071e-06,
+      "loss": 0.8712,
+      "step": 16315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.628569880572588e-06,
+      "loss": 0.8585,
+      "step": 16316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6276782423991577e-06,
+      "loss": 0.8344,
+      "step": 16317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6267868267524777e-06,
+      "loss": 0.8281,
+      "step": 16318
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6258956336562404e-06,
+      "loss": 0.9191,
+      "step": 16319
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.625004663134132e-06,
+      "loss": 0.8022,
+      "step": 16320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6241139152098361e-06,
+      "loss": 0.9278,
+      "step": 16321
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6232233899070283e-06,
+      "loss": 0.8562,
+      "step": 16322
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6223330872493726e-06,
+      "loss": 0.738,
+      "step": 16323
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6214430072605347e-06,
+      "loss": 0.9101,
+      "step": 16324
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6205531499641759e-06,
+      "loss": 0.8771,
+      "step": 16325
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6196635153839412e-06,
+      "loss": 0.7241,
+      "step": 16326
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.618774103543479e-06,
+      "loss": 0.8543,
+      "step": 16327
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6178849144664298e-06,
+      "loss": 0.9338,
+      "step": 16328
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6169959481764275e-06,
+      "loss": 0.8077,
+      "step": 16329
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6161072046970939e-06,
+      "loss": 0.9033,
+      "step": 16330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6152186840520612e-06,
+      "loss": 0.8968,
+      "step": 16331
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6143303862649363e-06,
+      "loss": 0.8779,
+      "step": 16332
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6134423113593323e-06,
+      "loss": 0.8296,
+      "step": 16333
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6125544593588539e-06,
+      "loss": 0.8869,
+      "step": 16334
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6116668302871009e-06,
+      "loss": 0.4452,
+      "step": 16335
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.61077942416766e-06,
+      "loss": 0.8465,
+      "step": 16336
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6098922410241213e-06,
+      "loss": 0.8664,
+      "step": 16337
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.609005280880067e-06,
+      "loss": 0.8949,
+      "step": 16338
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6081185437590664e-06,
+      "loss": 0.8702,
+      "step": 16339
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6072320296846898e-06,
+      "loss": 0.8469,
+      "step": 16340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6063457386805004e-06,
+      "loss": 0.8821,
+      "step": 16341
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6054596707700554e-06,
+      "loss": 0.8056,
+      "step": 16342
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.604573825976905e-06,
+      "loss": 0.9311,
+      "step": 16343
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6036882043245961e-06,
+      "loss": 0.9282,
+      "step": 16344
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.602802805836663e-06,
+      "loss": 0.875,
+      "step": 16345
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6019176305366401e-06,
+      "loss": 0.9076,
+      "step": 16346
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6010326784480546e-06,
+      "loss": 0.8814,
+      "step": 16347
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6001479495944318e-06,
+      "loss": 0.834,
+      "step": 16348
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5992634439992794e-06,
+      "loss": 0.9138,
+      "step": 16349
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.598379161686111e-06,
+      "loss": 0.8345,
+      "step": 16350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5974951026784314e-06,
+      "loss": 0.8689,
+      "step": 16351
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5966112669997313e-06,
+      "loss": 0.7697,
+      "step": 16352
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5957276546735078e-06,
+      "loss": 0.8357,
+      "step": 16353
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5948442657232443e-06,
+      "loss": 0.8482,
+      "step": 16354
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5939611001724208e-06,
+      "loss": 0.8094,
+      "step": 16355
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.59307815804451e-06,
+      "loss": 0.8844,
+      "step": 16356
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5921954393629801e-06,
+      "loss": 0.8874,
+      "step": 16357
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5913129441512954e-06,
+      "loss": 0.941,
+      "step": 16358
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5904306724329077e-06,
+      "loss": 0.8909,
+      "step": 16359
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5895486242312662e-06,
+      "loss": 0.9116,
+      "step": 16360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5886667995698201e-06,
+      "loss": 0.9132,
+      "step": 16361
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5877851984720016e-06,
+      "loss": 0.873,
+      "step": 16362
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5869038209612442e-06,
+      "loss": 0.9164,
+      "step": 16363
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.586022667060978e-06,
+      "loss": 0.829,
+      "step": 16364
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5851417367946165e-06,
+      "loss": 0.848,
+      "step": 16365
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5842610301855742e-06,
+      "loss": 0.8672,
+      "step": 16366
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5833805472572672e-06,
+      "loss": 0.8884,
+      "step": 16367
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5825002880330898e-06,
+      "loss": 0.8697,
+      "step": 16368
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.581620252536441e-06,
+      "loss": 0.8563,
+      "step": 16369
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5807404407907124e-06,
+      "loss": 0.8723,
+      "step": 16370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5798608528192882e-06,
+      "loss": 0.8959,
+      "step": 16371
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5789814886455446e-06,
+      "loss": 0.9486,
+      "step": 16372
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5781023482928536e-06,
+      "loss": 0.8167,
+      "step": 16373
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.577223431784587e-06,
+      "loss": 0.8674,
+      "step": 16374
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5763447391441e-06,
+      "loss": 0.8439,
+      "step": 16375
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5754662703947477e-06,
+      "loss": 0.887,
+      "step": 16376
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5745880255598812e-06,
+      "loss": 0.9234,
+      "step": 16377
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5737100046628418e-06,
+      "loss": 0.8617,
+      "step": 16378
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5728322077269676e-06,
+      "loss": 0.903,
+      "step": 16379
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5719546347755899e-06,
+      "loss": 0.9133,
+      "step": 16380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.57107728583203e-06,
+      "loss": 0.9621,
+      "step": 16381
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5702001609196094e-06,
+      "loss": 0.8775,
+      "step": 16382
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5693232600616403e-06,
+      "loss": 0.8799,
+      "step": 16383
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5684465832814332e-06,
+      "loss": 0.8301,
+      "step": 16384
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.567570130602284e-06,
+      "loss": 0.867,
+      "step": 16385
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5666939020474892e-06,
+      "loss": 0.8876,
+      "step": 16386
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5658178976403416e-06,
+      "loss": 0.9115,
+      "step": 16387
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.564942117404119e-06,
+      "loss": 0.8915,
+      "step": 16388
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5640665613620987e-06,
+      "loss": 0.8661,
+      "step": 16389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5631912295375595e-06,
+      "loss": 0.8706,
+      "step": 16390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5623161219537587e-06,
+      "loss": 0.8661,
+      "step": 16391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5614412386339583e-06,
+      "loss": 0.8889,
+      "step": 16392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5605665796014158e-06,
+      "loss": 0.8839,
+      "step": 16393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5596921448793711e-06,
+      "loss": 0.89,
+      "step": 16394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5588179344910704e-06,
+      "loss": 0.8961,
+      "step": 16395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.557943948459748e-06,
+      "loss": 0.8343,
+      "step": 16396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5570701868086358e-06,
+      "loss": 0.8966,
+      "step": 16397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5561966495609538e-06,
+      "loss": 0.8529,
+      "step": 16398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.55532333673992e-06,
+      "loss": 0.9196,
+      "step": 16399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5544502483687495e-06,
+      "loss": 0.876,
+      "step": 16400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5535773844706415e-06,
+      "loss": 0.8312,
+      "step": 16401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5527047450688027e-06,
+      "loss": 0.3777,
+      "step": 16402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5518323301864257e-06,
+      "loss": 0.9161,
+      "step": 16403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5509601398466945e-06,
+      "loss": 0.8836,
+      "step": 16404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5500881740727924e-06,
+      "loss": 0.8512,
+      "step": 16405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5492164328878966e-06,
+      "loss": 0.9073,
+      "step": 16406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5483449163151788e-06,
+      "loss": 0.8925,
+      "step": 16407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5474736243777977e-06,
+      "loss": 0.8835,
+      "step": 16408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.546602557098914e-06,
+      "loss": 0.8932,
+      "step": 16409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5457317145016815e-06,
+      "loss": 0.9281,
+      "step": 16410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5448610966092426e-06,
+      "loss": 0.8235,
+      "step": 16411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5439907034447387e-06,
+      "loss": 0.8848,
+      "step": 16412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.543120535031305e-06,
+      "loss": 0.867,
+      "step": 16413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5422505913920683e-06,
+      "loss": 0.92,
+      "step": 16414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5413808725501511e-06,
+      "loss": 0.7877,
+      "step": 16415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.540511378528672e-06,
+      "loss": 0.8877,
+      "step": 16416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5396421093507362e-06,
+      "loss": 0.8118,
+      "step": 16417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5387730650394506e-06,
+      "loss": 0.9484,
+      "step": 16418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5379042456179139e-06,
+      "loss": 0.8553,
+      "step": 16419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.53703565110922e-06,
+      "loss": 0.9183,
+      "step": 16420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5361672815364493e-06,
+      "loss": 0.8303,
+      "step": 16421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5352991369226865e-06,
+      "loss": 0.8425,
+      "step": 16422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.534431217291008e-06,
+      "loss": 0.8367,
+      "step": 16423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5335635226644752e-06,
+      "loss": 0.8789,
+      "step": 16424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5326960530661538e-06,
+      "loss": 0.834,
+      "step": 16425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5318288085191046e-06,
+      "loss": 0.8172,
+      "step": 16426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5309617890463713e-06,
+      "loss": 0.838,
+      "step": 16427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.530094994671002e-06,
+      "loss": 0.9101,
+      "step": 16428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5292284254160361e-06,
+      "loss": 0.8606,
+      "step": 16429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5283620813045008e-06,
+      "loss": 0.8768,
+      "step": 16430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5274959623594265e-06,
+      "loss": 0.8246,
+      "step": 16431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5266300686038338e-06,
+      "loss": 0.8698,
+      "step": 16432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5257644000607385e-06,
+      "loss": 0.8801,
+      "step": 16433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5248989567531436e-06,
+      "loss": 0.9217,
+      "step": 16434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.524033738704056e-06,
+      "loss": 0.8785,
+      "step": 16435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5231687459364708e-06,
+      "loss": 0.91,
+      "step": 16436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5223039784733784e-06,
+      "loss": 0.8298,
+      "step": 16437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5214394363377638e-06,
+      "loss": 0.8473,
+      "step": 16438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5205751195526087e-06,
+      "loss": 0.8813,
+      "step": 16439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.51971102814088e-06,
+      "loss": 0.8922,
+      "step": 16440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5188471621255463e-06,
+      "loss": 0.8832,
+      "step": 16441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5179835215295713e-06,
+      "loss": 0.8146,
+      "step": 16442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5171201063759045e-06,
+      "loss": 0.917,
+      "step": 16443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5162569166874964e-06,
+      "loss": 0.8577,
+      "step": 16444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5153939524872907e-06,
+      "loss": 0.9533,
+      "step": 16445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.514531213798225e-06,
+      "loss": 0.9426,
+      "step": 16446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.513668700643227e-06,
+      "loss": 0.9135,
+      "step": 16447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.512806413045219e-06,
+      "loss": 0.7931,
+      "step": 16448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5119443510271291e-06,
+      "loss": 0.8647,
+      "step": 16449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5110825146118614e-06,
+      "loss": 0.891,
+      "step": 16450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5102209038223249e-06,
+      "loss": 0.879,
+      "step": 16451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5093595186814236e-06,
+      "loss": 1.0034,
+      "step": 16452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5084983592120461e-06,
+      "loss": 0.8732,
+      "step": 16453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.507637425437084e-06,
+      "loss": 0.9247,
+      "step": 16454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5067767173794202e-06,
+      "loss": 0.9106,
+      "step": 16455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5059162350619327e-06,
+      "loss": 0.8165,
+      "step": 16456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5050559785074881e-06,
+      "loss": 0.8758,
+      "step": 16457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5041959477389545e-06,
+      "loss": 0.8983,
+      "step": 16458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5033361427791914e-06,
+      "loss": 0.8998,
+      "step": 16459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5024765636510442e-06,
+      "loss": 0.8046,
+      "step": 16460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5016172103773685e-06,
+      "loss": 0.9332,
+      "step": 16461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.500758082981002e-06,
+      "loss": 0.8745,
+      "step": 16462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4998991814847774e-06,
+      "loss": 0.8834,
+      "step": 16463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4990405059115244e-06,
+      "loss": 0.8934,
+      "step": 16464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4981820562840676e-06,
+      "loss": 0.9133,
+      "step": 16465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4973238326252193e-06,
+      "loss": 0.822,
+      "step": 16466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4964658349577932e-06,
+      "loss": 0.8705,
+      "step": 16467
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4956080633045933e-06,
+      "loss": 0.988,
+      "step": 16468
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.494750517688419e-06,
+      "loss": 0.845,
+      "step": 16469
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4938931981320614e-06,
+      "loss": 0.7864,
+      "step": 16470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4930361046583075e-06,
+      "loss": 0.9547,
+      "step": 16471
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4921792372899369e-06,
+      "loss": 0.8862,
+      "step": 16472
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4913225960497257e-06,
+      "loss": 0.9624,
+      "step": 16473
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4904661809604425e-06,
+      "loss": 0.8666,
+      "step": 16474
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4896099920448514e-06,
+      "loss": 0.8466,
+      "step": 16475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4887540293257052e-06,
+      "loss": 0.9407,
+      "step": 16476
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4878982928257556e-06,
+      "loss": 0.8392,
+      "step": 16477
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4870427825677502e-06,
+      "loss": 0.8352,
+      "step": 16478
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4861874985744228e-06,
+      "loss": 0.8443,
+      "step": 16479
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4853324408685077e-06,
+      "loss": 0.9006,
+      "step": 16480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.484477609472731e-06,
+      "loss": 0.8933,
+      "step": 16481
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4836230044098164e-06,
+      "loss": 0.8291,
+      "step": 16482
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4827686257024743e-06,
+      "loss": 0.907,
+      "step": 16483
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4819144733734137e-06,
+      "loss": 0.9295,
+      "step": 16484
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4810605474453378e-06,
+      "loss": 0.8415,
+      "step": 16485
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4802068479409437e-06,
+      "loss": 0.8767,
+      "step": 16486
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4793533748829203e-06,
+      "loss": 1.0184,
+      "step": 16487
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.478500128293955e-06,
+      "loss": 0.8502,
+      "step": 16488
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4776471081967213e-06,
+      "loss": 0.8574,
+      "step": 16489
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4767943146138953e-06,
+      "loss": 0.8907,
+      "step": 16490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4759417475681436e-06,
+      "loss": 0.8959,
+      "step": 16491
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.475089407082122e-06,
+      "loss": 0.9948,
+      "step": 16492
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4742372931784877e-06,
+      "loss": 0.8785,
+      "step": 16493
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4733854058798892e-06,
+      "loss": 0.8949,
+      "step": 16494
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4725337452089683e-06,
+      "loss": 0.8621,
+      "step": 16495
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4716823111883606e-06,
+      "loss": 0.9117,
+      "step": 16496
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4708311038406976e-06,
+      "loss": 0.8583,
+      "step": 16497
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4699801231886047e-06,
+      "loss": 0.8641,
+      "step": 16498
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.469129369254695e-06,
+      "loss": 0.8893,
+      "step": 16499
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4682788420615834e-06,
+      "loss": 0.8784,
+      "step": 16500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.467428541631879e-06,
+      "loss": 0.7894,
+      "step": 16501
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4665784679881768e-06,
+      "loss": 0.9246,
+      "step": 16502
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.465728621153072e-06,
+      "loss": 0.9702,
+      "step": 16503
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4648790011491544e-06,
+      "loss": 0.8953,
+      "step": 16504
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4640296079990057e-06,
+      "loss": 0.9434,
+      "step": 16505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4631804417251994e-06,
+      "loss": 0.8783,
+      "step": 16506
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.462331502350307e-06,
+      "loss": 0.8603,
+      "step": 16507
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4614827898968932e-06,
+      "loss": 0.8635,
+      "step": 16508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4606343043875137e-06,
+      "loss": 0.8784,
+      "step": 16509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4597860458447221e-06,
+      "loss": 0.9525,
+      "step": 16510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4589380142910659e-06,
+      "loss": 0.8557,
+      "step": 16511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4580902097490802e-06,
+      "loss": 0.886,
+      "step": 16512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4572426322413003e-06,
+      "loss": 0.3986,
+      "step": 16513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.456395281790257e-06,
+      "loss": 0.9184,
+      "step": 16514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4555481584184672e-06,
+      "loss": 0.8202,
+      "step": 16515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.454701262148449e-06,
+      "loss": 0.8323,
+      "step": 16516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4538545930027115e-06,
+      "loss": 0.8395,
+      "step": 16517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4530081510037597e-06,
+      "loss": 0.8927,
+      "step": 16518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4521619361740856e-06,
+      "loss": 0.883,
+      "step": 16519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4513159485361895e-06,
+      "loss": 0.8925,
+      "step": 16520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4504701881125493e-06,
+      "loss": 0.9005,
+      "step": 16521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4496246549256477e-06,
+      "loss": 0.9033,
+      "step": 16522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4487793489979573e-06,
+      "loss": 0.8465,
+      "step": 16523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4479342703519472e-06,
+      "loss": 0.8338,
+      "step": 16524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4470894190100737e-06,
+      "loss": 0.9404,
+      "step": 16525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4462447949947967e-06,
+      "loss": 0.8389,
+      "step": 16526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4454003983285646e-06,
+      "loss": 0.8103,
+      "step": 16527
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4445562290338178e-06,
+      "loss": 0.8435,
+      "step": 16528
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4437122871329956e-06,
+      "loss": 0.8897,
+      "step": 16529
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4428685726485281e-06,
+      "loss": 0.8706,
+      "step": 16530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4420250856028418e-06,
+      "loss": 0.8492,
+      "step": 16531
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.441181826018353e-06,
+      "loss": 0.9143,
+      "step": 16532
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4403387939174796e-06,
+      "loss": 0.8631,
+      "step": 16533
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4394959893226223e-06,
+      "loss": 0.7904,
+      "step": 16534
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4386534122561846e-06,
+      "loss": 0.8982,
+      "step": 16535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4378110627405606e-06,
+      "loss": 0.9424,
+      "step": 16536
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4369689407981434e-06,
+      "loss": 0.8151,
+      "step": 16537
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4361270464513078e-06,
+      "loss": 0.8428,
+      "step": 16538
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4352853797224353e-06,
+      "loss": 0.8969,
+      "step": 16539
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4344439406338984e-06,
+      "loss": 0.8314,
+      "step": 16540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4336027292080556e-06,
+      "loss": 0.8426,
+      "step": 16541
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4327617454672692e-06,
+      "loss": 0.8492,
+      "step": 16542
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4319209894338903e-06,
+      "loss": 0.8797,
+      "step": 16543
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4310804611302653e-06,
+      "loss": 0.9263,
+      "step": 16544
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4302401605787364e-06,
+      "loss": 0.8475,
+      "step": 16545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4294000878016356e-06,
+      "loss": 0.9605,
+      "step": 16546
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4285602428212941e-06,
+      "loss": 0.8185,
+      "step": 16547
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4277206256600295e-06,
+      "loss": 0.8927,
+      "step": 16548
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4268812363401608e-06,
+      "loss": 0.8437,
+      "step": 16549
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4260420748839987e-06,
+      "loss": 0.8062,
+      "step": 16550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4252031413138445e-06,
+      "loss": 0.8899,
+      "step": 16551
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.424364435651997e-06,
+      "loss": 0.9011,
+      "step": 16552
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4235259579207484e-06,
+      "loss": 0.8018,
+      "step": 16553
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.422687708142385e-06,
+      "loss": 0.9481,
+      "step": 16554
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4218496863391862e-06,
+      "loss": 0.8842,
+      "step": 16555
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.421011892533427e-06,
+      "loss": 0.8705,
+      "step": 16556
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4201743267473721e-06,
+      "loss": 0.9092,
+      "step": 16557
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4193369890032837e-06,
+      "loss": 0.8862,
+      "step": 16558
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4184998793234196e-06,
+      "loss": 0.8284,
+      "step": 16559
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4176629977300282e-06,
+      "loss": 0.88,
+      "step": 16560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4168263442453511e-06,
+      "loss": 0.9522,
+      "step": 16561
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.415989918891626e-06,
+      "loss": 0.8522,
+      "step": 16562
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4151537216910882e-06,
+      "loss": 0.8506,
+      "step": 16563
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4143177526659568e-06,
+      "loss": 0.836,
+      "step": 16564
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.413482011838454e-06,
+      "loss": 0.7991,
+      "step": 16565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412646499230793e-06,
+      "loss": 0.8913,
+      "step": 16566
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4118112148651798e-06,
+      "loss": 0.9,
+      "step": 16567
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.410976158763816e-06,
+      "loss": 0.8947,
+      "step": 16568
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4101413309488976e-06,
+      "loss": 0.9971,
+      "step": 16569
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4093067314426113e-06,
+      "loss": 0.9315,
+      "step": 16570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4084723602671403e-06,
+      "loss": 0.8155,
+      "step": 16571
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40763821744466e-06,
+      "loss": 0.8541,
+      "step": 16572
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4068043029973465e-06,
+      "loss": 0.8735,
+      "step": 16573
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4059706169473575e-06,
+      "loss": 0.9381,
+      "step": 16574
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.405137159316854e-06,
+      "loss": 0.7926,
+      "step": 16575
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4043039301279904e-06,
+      "loss": 0.8857,
+      "step": 16576
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.403470929402909e-06,
+      "loss": 0.9591,
+      "step": 16577
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4026381571637505e-06,
+      "loss": 0.9533,
+      "step": 16578
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4018056134326551e-06,
+      "loss": 0.8727,
+      "step": 16579
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4009732982317437e-06,
+      "loss": 0.8932,
+      "step": 16580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4001412115831414e-06,
+      "loss": 0.8369,
+      "step": 16581
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3993093535089652e-06,
+      "loss": 0.8567,
+      "step": 16582
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.398477724031322e-06,
+      "loss": 0.9779,
+      "step": 16583
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.397646323172317e-06,
+      "loss": 0.9419,
+      "step": 16584
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3968151509540473e-06,
+      "loss": 0.943,
+      "step": 16585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3959842073986085e-06,
+      "loss": 0.3888,
+      "step": 16586
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3951534925280796e-06,
+      "loss": 0.4316,
+      "step": 16587
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.394323006364543e-06,
+      "loss": 0.9116,
+      "step": 16588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3934927489300755e-06,
+      "loss": 0.9125,
+      "step": 16589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3926627202467357e-06,
+      "loss": 0.9162,
+      "step": 16590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3918329203365922e-06,
+      "loss": 0.9315,
+      "step": 16591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3910033492217012e-06,
+      "loss": 0.9306,
+      "step": 16592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3901740069241066e-06,
+      "loss": 0.9046,
+      "step": 16593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3893448934658538e-06,
+      "loss": 0.8117,
+      "step": 16594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3885160088689787e-06,
+      "loss": 0.8354,
+      "step": 16595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3876873531555147e-06,
+      "loss": 0.8143,
+      "step": 16596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.386858926347483e-06,
+      "loss": 0.8252,
+      "step": 16597
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3860307284669039e-06,
+      "loss": 0.9105,
+      "step": 16598
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.385202759535793e-06,
+      "loss": 0.9255,
+      "step": 16599
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3843750195761507e-06,
+      "loss": 0.7652,
+      "step": 16600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3835475086099803e-06,
+      "loss": 0.814,
+      "step": 16601
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3827202266592754e-06,
+      "loss": 0.8917,
+      "step": 16602
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3818931737460261e-06,
+      "loss": 0.8527,
+      "step": 16603
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.381066349892214e-06,
+      "loss": 0.8562,
+      "step": 16604
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3802397551198154e-06,
+      "loss": 0.8172,
+      "step": 16605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3794133894507978e-06,
+      "loss": 0.9169,
+      "step": 16606
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3785872529071265e-06,
+      "loss": 0.8144,
+      "step": 16607
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3777613455107596e-06,
+      "loss": 0.9237,
+      "step": 16608
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.376935667283651e-06,
+      "loss": 0.8899,
+      "step": 16609
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3761102182477415e-06,
+      "loss": 0.8288,
+      "step": 16610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3752849984249727e-06,
+      "loss": 0.8078,
+      "step": 16611
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3744600078372804e-06,
+      "loss": 0.849,
+      "step": 16612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3736352465065884e-06,
+      "loss": 0.4322,
+      "step": 16613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3728107144548164e-06,
+      "loss": 0.8343,
+      "step": 16614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3719864117038862e-06,
+      "loss": 0.8672,
+      "step": 16615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3711623382757023e-06,
+      "loss": 0.887,
+      "step": 16616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3703384941921671e-06,
+      "loss": 0.907,
+      "step": 16617
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.369514879475181e-06,
+      "loss": 0.9329,
+      "step": 16618
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3686914941466311e-06,
+      "loss": 0.8527,
+      "step": 16619
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.367868338228403e-06,
+      "loss": 0.9555,
+      "step": 16620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.367045411742376e-06,
+      "loss": 0.9367,
+      "step": 16621
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3662227147104234e-06,
+      "loss": 0.8782,
+      "step": 16622
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3654002471544093e-06,
+      "loss": 0.9893,
+      "step": 16623
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3645780090961946e-06,
+      "loss": 0.8808,
+      "step": 16624
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3637560005576334e-06,
+      "loss": 0.896,
+      "step": 16625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3629342215605745e-06,
+      "loss": 0.9046,
+      "step": 16626
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3621126721268596e-06,
+      "loss": 0.8343,
+      "step": 16627
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3612913522783255e-06,
+      "loss": 0.8925,
+      "step": 16628
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3604702620368005e-06,
+      "loss": 0.8808,
+      "step": 16629
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.359649401424108e-06,
+      "loss": 0.8841,
+      "step": 16630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3588287704620673e-06,
+      "loss": 0.7056,
+      "step": 16631
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3580083691724876e-06,
+      "loss": 0.8363,
+      "step": 16632
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.357188197577175e-06,
+      "loss": 0.8736,
+      "step": 16633
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3563682556979297e-06,
+      "loss": 0.8948,
+      "step": 16634
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3555485435565464e-06,
+      "loss": 0.9529,
+      "step": 16635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3547290611748064e-06,
+      "loss": 0.8719,
+      "step": 16636
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3539098085744928e-06,
+      "loss": 0.8591,
+      "step": 16637
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3530907857773867e-06,
+      "loss": 0.8629,
+      "step": 16638
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3522719928052486e-06,
+      "loss": 0.8888,
+      "step": 16639
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3514534296798444e-06,
+      "loss": 0.9181,
+      "step": 16640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3506350964229332e-06,
+      "loss": 0.9246,
+      "step": 16641
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3498169930562598e-06,
+      "loss": 0.8968,
+      "step": 16642
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3489991196015717e-06,
+      "loss": 0.8777,
+      "step": 16643
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3481814760806066e-06,
+      "loss": 0.9887,
+      "step": 16644
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3473640625150986e-06,
+      "loss": 0.8487,
+      "step": 16645
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.34654687892677e-06,
+      "loss": 0.8772,
+      "step": 16646
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3457299253373412e-06,
+      "loss": 0.8914,
+      "step": 16647
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3449132017685297e-06,
+      "loss": 0.8731,
+      "step": 16648
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3440967082420353e-06,
+      "loss": 0.865,
+      "step": 16649
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3432804447795667e-06,
+      "loss": 0.8691,
+      "step": 16650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3424644114028197e-06,
+      "loss": 0.8631,
+      "step": 16651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.341648608133478e-06,
+      "loss": 0.8818,
+      "step": 16652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3408330349932276e-06,
+      "loss": 0.9731,
+      "step": 16653
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3400176920037478e-06,
+      "loss": 0.9351,
+      "step": 16654
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3392025791867047e-06,
+      "loss": 0.8814,
+      "step": 16655
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3383876965637655e-06,
+      "loss": 1.0001,
+      "step": 16656
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3375730441565892e-06,
+      "loss": 0.84,
+      "step": 16657
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.33675862198683e-06,
+      "loss": 0.9368,
+      "step": 16658
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3359444300761305e-06,
+      "loss": 0.904,
+      "step": 16659
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.335130468446132e-06,
+      "loss": 0.9103,
+      "step": 16660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3343167371184695e-06,
+      "loss": 0.8786,
+      "step": 16661
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3335032361147716e-06,
+      "loss": 0.8646,
+      "step": 16662
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3326899654566582e-06,
+      "loss": 0.381,
+      "step": 16663
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3318769251657493e-06,
+      "loss": 0.9076,
+      "step": 16664
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3310641152636504e-06,
+      "loss": 0.8438,
+      "step": 16665
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3302515357719658e-06,
+      "loss": 0.8941,
+      "step": 16666
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3294391867122958e-06,
+      "loss": 0.9538,
+      "step": 16667
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3286270681062275e-06,
+      "loss": 0.8791,
+      "step": 16668
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.327815179975348e-06,
+      "loss": 0.8754,
+      "step": 16669
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.327003522341237e-06,
+      "loss": 0.8726,
+      "step": 16670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.326192095225468e-06,
+      "loss": 0.8091,
+      "step": 16671
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3253808986496052e-06,
+      "loss": 0.9385,
+      "step": 16672
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.324569932635208e-06,
+      "loss": 0.8661,
+      "step": 16673
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3237591972038387e-06,
+      "loss": 0.8287,
+      "step": 16674
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3229486923770373e-06,
+      "loss": 0.8837,
+      "step": 16675
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3221384181763508e-06,
+      "loss": 0.9957,
+      "step": 16676
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3213283746233163e-06,
+      "loss": 0.9018,
+      "step": 16677
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.320518561739459e-06,
+      "loss": 0.8505,
+      "step": 16678
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3197089795463048e-06,
+      "loss": 0.8739,
+      "step": 16679
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3188996280653754e-06,
+      "loss": 0.898,
+      "step": 16680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.318090507318176e-06,
+      "loss": 0.9672,
+      "step": 16681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3172816173262159e-06,
+      "loss": 0.8595,
+      "step": 16682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3164729581109937e-06,
+      "loss": 0.9258,
+      "step": 16683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.315664529694003e-06,
+      "loss": 0.7967,
+      "step": 16684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.31485633209673e-06,
+      "loss": 0.9423,
+      "step": 16685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3140483653406566e-06,
+      "loss": 0.8028,
+      "step": 16686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3132406294472589e-06,
+      "loss": 0.8806,
+      "step": 16687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3124331244380029e-06,
+      "loss": 0.8325,
+      "step": 16688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3116258503343515e-06,
+      "loss": 0.8646,
+      "step": 16689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3108188071577643e-06,
+      "loss": 0.8474,
+      "step": 16690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3100119949296875e-06,
+      "loss": 0.8787,
+      "step": 16691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.309205413671566e-06,
+      "loss": 0.8268,
+      "step": 16692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3083990634048393e-06,
+      "loss": 0.9152,
+      "step": 16693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3075929441509417e-06,
+      "loss": 0.8634,
+      "step": 16694
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3067870559312933e-06,
+      "loss": 0.9224,
+      "step": 16695
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3059813987673154e-06,
+      "loss": 0.9903,
+      "step": 16696
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3051759726804236e-06,
+      "loss": 0.8264,
+      "step": 16697
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3043707776920233e-06,
+      "loss": 0.839,
+      "step": 16698
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3035658138235164e-06,
+      "loss": 0.7922,
+      "step": 16699
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3027610810963009e-06,
+      "loss": 0.9158,
+      "step": 16700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3019565795317602e-06,
+      "loss": 0.9167,
+      "step": 16701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.301152309151279e-06,
+      "loss": 0.785,
+      "step": 16702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.300348269976236e-06,
+      "loss": 0.9229,
+      "step": 16703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2995444620279984e-06,
+      "loss": 0.9072,
+      "step": 16704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2987408853279325e-06,
+      "loss": 0.894,
+      "step": 16705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2979375398973958e-06,
+      "loss": 0.8243,
+      "step": 16706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2971344257577435e-06,
+      "loss": 0.8582,
+      "step": 16707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2963315429303137e-06,
+      "loss": 0.9209,
+      "step": 16708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295528891436455e-06,
+      "loss": 0.86,
+      "step": 16709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.294726471297495e-06,
+      "loss": 0.9206,
+      "step": 16710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.293924282534762e-06,
+      "loss": 0.8603,
+      "step": 16711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2931223251695802e-06,
+      "loss": 0.8488,
+      "step": 16712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2923205992232635e-06,
+      "loss": 0.8799,
+      "step": 16713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2915191047171183e-06,
+      "loss": 0.8291,
+      "step": 16714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2907178416724497e-06,
+      "loss": 0.8374,
+      "step": 16715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2899168101105553e-06,
+      "loss": 0.8319,
+      "step": 16716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2891160100527222e-06,
+      "loss": 0.8446,
+      "step": 16717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2883154415202381e-06,
+      "loss": 0.4157,
+      "step": 16718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2875151045343782e-06,
+      "loss": 0.902,
+      "step": 16719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2867149991164175e-06,
+      "loss": 0.803,
+      "step": 16720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2859151252876189e-06,
+      "loss": 0.9014,
+      "step": 16721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.285115483069246e-06,
+      "loss": 0.86,
+      "step": 16722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.284316072482551e-06,
+      "loss": 0.8196,
+      "step": 16723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2835168935487785e-06,
+      "loss": 0.835,
+      "step": 16724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2827179462891714e-06,
+      "loss": 0.8964,
+      "step": 16725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2819192307249683e-06,
+      "loss": 0.885,
+      "step": 16726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2811207468773923e-06,
+      "loss": 0.8996,
+      "step": 16727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2803224947676695e-06,
+      "loss": 0.8319,
+      "step": 16728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2795244744170177e-06,
+      "loss": 0.8491,
+      "step": 16729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.278726685846643e-06,
+      "loss": 0.9801,
+      "step": 16730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.277929129077753e-06,
+      "loss": 0.8521,
+      "step": 16731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.277131804131545e-06,
+      "loss": 0.8605,
+      "step": 16732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2763347110292112e-06,
+      "loss": 0.8936,
+      "step": 16733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.275537849791938e-06,
+      "loss": 0.9529,
+      "step": 16734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2747412204409038e-06,
+      "loss": 0.862,
+      "step": 16735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2739448229972851e-06,
+      "loss": 0.8902,
+      "step": 16736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2731486574822437e-06,
+      "loss": 0.37,
+      "step": 16737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2723527239169454e-06,
+      "loss": 0.8962,
+      "step": 16738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2715570223225459e-06,
+      "loss": 0.8687,
+      "step": 16739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.270761552720189e-06,
+      "loss": 0.8571,
+      "step": 16740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2699663151310203e-06,
+      "loss": 0.909,
+      "step": 16741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.269171309576177e-06,
+      "loss": 0.9073,
+      "step": 16742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2683765360767874e-06,
+      "loss": 0.8459,
+      "step": 16743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.267581994653978e-06,
+      "loss": 0.9298,
+      "step": 16744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2667876853288686e-06,
+      "loss": 0.8379,
+      "step": 16745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2659936081225654e-06,
+      "loss": 0.8692,
+      "step": 16746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2651997630561764e-06,
+      "loss": 0.8429,
+      "step": 16747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2644061501508021e-06,
+      "loss": 0.905,
+      "step": 16748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2636127694275368e-06,
+      "loss": 0.9,
+      "step": 16749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.262819620907465e-06,
+      "loss": 0.8808,
+      "step": 16750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2620267046116686e-06,
+      "loss": 0.8815,
+      "step": 16751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.261234020561225e-06,
+      "loss": 0.9199,
+      "step": 16752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2604415687771977e-06,
+      "loss": 0.8276,
+      "step": 16753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2596493492806527e-06,
+      "loss": 0.8286,
+      "step": 16754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2588573620926447e-06,
+      "loss": 0.7685,
+      "step": 16755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2580656072342257e-06,
+      "loss": 0.8774,
+      "step": 16756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2572740847264375e-06,
+      "loss": 0.936,
+      "step": 16757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2564827945903225e-06,
+      "loss": 0.8481,
+      "step": 16758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2556917368469057e-06,
+      "loss": 0.793,
+      "step": 16759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2549009115172162e-06,
+      "loss": 0.9432,
+      "step": 16760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2541103186222714e-06,
+      "loss": 0.8835,
+      "step": 16761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.253319958183089e-06,
+      "loss": 0.3894,
+      "step": 16762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2525298302206689e-06,
+      "loss": 0.9192,
+      "step": 16763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2517399347560165e-06,
+      "loss": 0.8337,
+      "step": 16764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2509502718101274e-06,
+      "loss": 1.0035,
+      "step": 16765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2501608414039846e-06,
+      "loss": 0.8293,
+      "step": 16766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2493716435585723e-06,
+      "loss": 0.8447,
+      "step": 16767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2485826782948717e-06,
+      "loss": 0.8387,
+      "step": 16768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.247793945633845e-06,
+      "loss": 0.889,
+      "step": 16769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.247005445596461e-06,
+      "loss": 0.9129,
+      "step": 16770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2462171782036747e-06,
+      "loss": 0.8909,
+      "step": 16771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2454291434764399e-06,
+      "loss": 0.9042,
+      "step": 16772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2446413414356973e-06,
+      "loss": 0.3709,
+      "step": 16773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.243853772102389e-06,
+      "loss": 0.8635,
+      "step": 16774
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2430664354974486e-06,
+      "loss": 0.8944,
+      "step": 16775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2422793316417992e-06,
+      "loss": 0.9535,
+      "step": 16776
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2414924605563626e-06,
+      "loss": 0.9521,
+      "step": 16777
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.240705822262056e-06,
+      "loss": 0.8403,
+      "step": 16778
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.239919416779779e-06,
+      "loss": 0.8395,
+      "step": 16779
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2391332441304427e-06,
+      "loss": 0.8944,
+      "step": 16780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2383473043349392e-06,
+      "loss": 0.857,
+      "step": 16781
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2375615974141575e-06,
+      "loss": 0.8789,
+      "step": 16782
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2367761233889796e-06,
+      "loss": 0.8152,
+      "step": 16783
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2359908822802835e-06,
+      "loss": 0.8913,
+      "step": 16784
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2352058741089434e-06,
+      "loss": 0.7877,
+      "step": 16785
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2344210988958183e-06,
+      "loss": 0.8656,
+      "step": 16786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2336365566617692e-06,
+      "loss": 0.8872,
+      "step": 16787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2328522474276493e-06,
+      "loss": 0.8642,
+      "step": 16788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2320681712143034e-06,
+      "loss": 0.8641,
+      "step": 16789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.23128432804257e-06,
+      "loss": 0.8921,
+      "step": 16790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2305007179332851e-06,
+      "loss": 0.8026,
+      "step": 16791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2297173409072761e-06,
+      "loss": 0.8898,
+      "step": 16792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2289341969853619e-06,
+      "loss": 0.8618,
+      "step": 16793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2281512861883626e-06,
+      "loss": 0.8617,
+      "step": 16794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2273686085370807e-06,
+      "loss": 0.7895,
+      "step": 16795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2265861640523225e-06,
+      "loss": 0.9114,
+      "step": 16796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2258039527548837e-06,
+      "loss": 0.8462,
+      "step": 16797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2250219746655568e-06,
+      "loss": 0.8876,
+      "step": 16798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2242402298051215e-06,
+      "loss": 0.8857,
+      "step": 16799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.223458718194359e-06,
+      "loss": 0.8998,
+      "step": 16800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2226774398540408e-06,
+      "loss": 0.9538,
+      "step": 16801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2218963948049277e-06,
+      "loss": 0.8858,
+      "step": 16802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2211155830677856e-06,
+      "loss": 0.8534,
+      "step": 16803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2203350046633667e-06,
+      "loss": 0.8088,
+      "step": 16804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2195546596124142e-06,
+      "loss": 0.8359,
+      "step": 16805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2187745479356705e-06,
+      "loss": 0.8775,
+      "step": 16806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2179946696538714e-06,
+      "loss": 0.862,
+      "step": 16807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2172150247877424e-06,
+      "loss": 0.8643,
+      "step": 16808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.216435613358008e-06,
+      "loss": 0.9097,
+      "step": 16809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2156564353853817e-06,
+      "loss": 0.8365,
+      "step": 16810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2148774908905782e-06,
+      "loss": 0.9115,
+      "step": 16811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.214098779894295e-06,
+      "loss": 0.8123,
+      "step": 16812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2133203024172313e-06,
+      "loss": 0.8058,
+      "step": 16813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2125420584800785e-06,
+      "loss": 0.8973,
+      "step": 16814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.211764048103522e-06,
+      "loss": 0.9308,
+      "step": 16815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.21098627130824e-06,
+      "loss": 0.8992,
+      "step": 16816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2102087281149066e-06,
+      "loss": 0.9577,
+      "step": 16817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2094314185441836e-06,
+      "loss": 0.8911,
+      "step": 16818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2086543426167352e-06,
+      "loss": 0.8237,
+      "step": 16819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2078775003532128e-06,
+      "loss": 0.9531,
+      "step": 16820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2071008917742667e-06,
+      "loss": 0.9266,
+      "step": 16821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2063245169005356e-06,
+      "loss": 0.4251,
+      "step": 16822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2055483757526542e-06,
+      "loss": 0.985,
+      "step": 16823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2047724683512552e-06,
+      "loss": 0.8918,
+      "step": 16824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2039967947169573e-06,
+      "loss": 0.8991,
+      "step": 16825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2032213548703752e-06,
+      "loss": 0.9102,
+      "step": 16826
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2024461488321282e-06,
+      "loss": 0.3658,
+      "step": 16827
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2016711766228129e-06,
+      "loss": 0.902,
+      "step": 16828
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2008964382630272e-06,
+      "loss": 0.8574,
+      "step": 16829
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2001219337733682e-06,
+      "loss": 0.8897,
+      "step": 16830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.199347663174416e-06,
+      "loss": 0.8224,
+      "step": 16831
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.198573626486751e-06,
+      "loss": 0.9225,
+      "step": 16832
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1977998237309463e-06,
+      "loss": 0.809,
+      "step": 16833
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1970262549275723e-06,
+      "loss": 0.9076,
+      "step": 16834
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1962529200971839e-06,
+      "loss": 0.3249,
+      "step": 16835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1954798192603379e-06,
+      "loss": 0.936,
+      "step": 16836
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1947069524375842e-06,
+      "loss": 0.941,
+      "step": 16837
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1939343196494602e-06,
+      "loss": 0.4229,
+      "step": 16838
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1931619209165057e-06,
+      "loss": 0.7214,
+      "step": 16839
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1923897562592513e-06,
+      "loss": 0.8748,
+      "step": 16840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1916178256982157e-06,
+      "loss": 0.8486,
+      "step": 16841
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1908461292539187e-06,
+      "loss": 0.8559,
+      "step": 16842
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1900746669468721e-06,
+      "loss": 0.8424,
+      "step": 16843
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1893034387975766e-06,
+      "loss": 0.917,
+      "step": 16844
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1885324448265334e-06,
+      "loss": 0.9337,
+      "step": 16845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.187761685054234e-06,
+      "loss": 0.8504,
+      "step": 16846
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1869911595011674e-06,
+      "loss": 0.9364,
+      "step": 16847
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1862208681878074e-06,
+      "loss": 0.9101,
+      "step": 16848
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1854508111346307e-06,
+      "loss": 0.7915,
+      "step": 16849
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1846809883621035e-06,
+      "loss": 0.8735,
+      "step": 16850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1839113998906881e-06,
+      "loss": 0.8832,
+      "step": 16851
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1831420457408393e-06,
+      "loss": 0.8967,
+      "step": 16852
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1823729259330064e-06,
+      "loss": 0.7951,
+      "step": 16853
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1816040404876283e-06,
+      "loss": 0.938,
+      "step": 16854
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1808353894251446e-06,
+      "loss": 0.9142,
+      "step": 16855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1800669727659842e-06,
+      "loss": 0.8173,
+      "step": 16856
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1792987905305697e-06,
+      "loss": 0.8613,
+      "step": 16857
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1785308427393182e-06,
+      "loss": 0.8202,
+      "step": 16858
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1777631294126412e-06,
+      "loss": 0.9123,
+      "step": 16859
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1769956505709479e-06,
+      "loss": 0.7766,
+      "step": 16860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1762284062346285e-06,
+      "loss": 0.8272,
+      "step": 16861
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1754613964240825e-06,
+      "loss": 0.8756,
+      "step": 16862
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1746946211596954e-06,
+      "loss": 0.8327,
+      "step": 16863
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1739280804618446e-06,
+      "loss": 0.8005,
+      "step": 16864
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1731617743509049e-06,
+      "loss": 0.8831,
+      "step": 16865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1723957028472454e-06,
+      "loss": 0.9329,
+      "step": 16866
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1716298659712244e-06,
+      "loss": 0.9334,
+      "step": 16867
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.170864263743199e-06,
+      "loss": 0.8654,
+      "step": 16868
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.170098896183517e-06,
+      "loss": 0.8363,
+      "step": 16869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1693337633125246e-06,
+      "loss": 0.9238,
+      "step": 16870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1685688651505522e-06,
+      "loss": 0.9348,
+      "step": 16871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1678042017179336e-06,
+      "loss": 0.857,
+      "step": 16872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1670397730349913e-06,
+      "loss": 0.9369,
+      "step": 16873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1662755791220437e-06,
+      "loss": 0.8704,
+      "step": 16874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1655116199994032e-06,
+      "loss": 0.8683,
+      "step": 16875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1647478956873747e-06,
+      "loss": 0.9519,
+      "step": 16876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1639844062062544e-06,
+      "loss": 0.7915,
+      "step": 16877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.163221151576338e-06,
+      "loss": 0.8778,
+      "step": 16878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1624581318179117e-06,
+      "loss": 0.8601,
+      "step": 16879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1616953469512537e-06,
+      "loss": 0.7622,
+      "step": 16880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1609327969966377e-06,
+      "loss": 0.8557,
+      "step": 16881
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1601704819743343e-06,
+      "loss": 0.8252,
+      "step": 16882
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1594084019046049e-06,
+      "loss": 0.8621,
+      "step": 16883
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.158646556807701e-06,
+      "loss": 0.8926,
+      "step": 16884
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1578849467038734e-06,
+      "loss": 0.825,
+      "step": 16885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1571235716133655e-06,
+      "loss": 0.9509,
+      "step": 16886
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1563624315564137e-06,
+      "loss": 0.8174,
+      "step": 16887
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1556015265532483e-06,
+      "loss": 0.8944,
+      "step": 16888
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1548408566240955e-06,
+      "loss": 0.8845,
+      "step": 16889
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1540804217891676e-06,
+      "loss": 0.9028,
+      "step": 16890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.153320222068679e-06,
+      "loss": 0.4362,
+      "step": 16891
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1525602574828377e-06,
+      "loss": 0.8864,
+      "step": 16892
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1518005280518386e-06,
+      "loss": 0.84,
+      "step": 16893
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1510410337958744e-06,
+      "loss": 0.9281,
+      "step": 16894
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1502817747351335e-06,
+      "loss": 0.8407,
+      "step": 16895
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.149522750889799e-06,
+      "loss": 0.8669,
+      "step": 16896
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1487639622800361e-06,
+      "loss": 0.9562,
+      "step": 16897
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1480054089260217e-06,
+      "loss": 0.8889,
+      "step": 16898
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1472470908479127e-06,
+      "loss": 0.8531,
+      "step": 16899
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1464890080658652e-06,
+      "loss": 0.8665,
+      "step": 16900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1457311606000276e-06,
+      "loss": 0.8798,
+      "step": 16901
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1449735484705448e-06,
+      "loss": 0.8788,
+      "step": 16902
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1442161716975508e-06,
+      "loss": 0.8951,
+      "step": 16903
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.143459030301176e-06,
+      "loss": 0.8882,
+      "step": 16904
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1427021243015469e-06,
+      "loss": 0.8645,
+      "step": 16905
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.141945453718778e-06,
+      "loss": 0.9136,
+      "step": 16906
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1411890185729813e-06,
+      "loss": 0.8488,
+      "step": 16907
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1404328188842628e-06,
+      "loss": 0.893,
+      "step": 16908
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1396768546727222e-06,
+      "loss": 0.8812,
+      "step": 16909
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.138921125958451e-06,
+      "loss": 0.8843,
+      "step": 16910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1381656327615354e-06,
+      "loss": 0.8753,
+      "step": 16911
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1374103751020594e-06,
+      "loss": 0.9751,
+      "step": 16912
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1366553530000902e-06,
+      "loss": 0.8186,
+      "step": 16913
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1359005664756994e-06,
+      "loss": 0.8788,
+      "step": 16914
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.135146015548949e-06,
+      "loss": 0.44,
+      "step": 16915
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1343917002398919e-06,
+      "loss": 1.0044,
+      "step": 16916
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1336376205685774e-06,
+      "loss": 0.8329,
+      "step": 16917
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1328837765550493e-06,
+      "loss": 0.9815,
+      "step": 16918
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1321301682193441e-06,
+      "loss": 0.8276,
+      "step": 16919
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1313767955814903e-06,
+      "loss": 0.8949,
+      "step": 16920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1306236586615104e-06,
+      "loss": 0.8583,
+      "step": 16921
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1298707574794254e-06,
+      "loss": 0.8034,
+      "step": 16922
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1291180920552447e-06,
+      "loss": 0.8894,
+      "step": 16923
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1283656624089734e-06,
+      "loss": 0.7902,
+      "step": 16924
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.127613468560612e-06,
+      "loss": 0.9126,
+      "step": 16925
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1268615105301506e-06,
+      "loss": 0.8983,
+      "step": 16926
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1261097883375749e-06,
+      "loss": 0.8644,
+      "step": 16927
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1253583020028691e-06,
+      "loss": 0.8663,
+      "step": 16928
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1246070515460017e-06,
+      "loss": 0.8998,
+      "step": 16929
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1238560369869411e-06,
+      "loss": 0.9334,
+      "step": 16930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1231052583456503e-06,
+      "loss": 0.8533,
+      "step": 16931
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1223547156420834e-06,
+      "loss": 0.8653,
+      "step": 16932
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1216044088961887e-06,
+      "loss": 0.8931,
+      "step": 16933
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1208543381279114e-06,
+      "loss": 0.8736,
+      "step": 16934
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1201045033571822e-06,
+      "loss": 0.9487,
+      "step": 16935
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1193549046039331e-06,
+      "loss": 0.9113,
+      "step": 16936
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.118605541888088e-06,
+      "loss": 0.8585,
+      "step": 16937
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1178564152295668e-06,
+      "loss": 0.9646,
+      "step": 16938
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1171075246482754e-06,
+      "loss": 0.8532,
+      "step": 16939
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1163588701641214e-06,
+      "loss": 0.8616,
+      "step": 16940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1156104517970034e-06,
+      "loss": 0.8569,
+      "step": 16941
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1148622695668109e-06,
+      "loss": 0.8802,
+      "step": 16942
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1141143234934327e-06,
+      "loss": 0.9073,
+      "step": 16943
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1133666135967458e-06,
+      "loss": 0.8818,
+      "step": 16944
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1126191398966246e-06,
+      "loss": 0.8549,
+      "step": 16945
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1118719024129377e-06,
+      "loss": 0.8472,
+      "step": 16946
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1111249011655457e-06,
+      "loss": 0.7735,
+      "step": 16947
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1103781361742994e-06,
+      "loss": 0.8154,
+      "step": 16948
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1096316074590496e-06,
+      "loss": 0.924,
+      "step": 16949
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1088853150396372e-06,
+      "loss": 0.8129,
+      "step": 16950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1081392589359019e-06,
+      "loss": 0.8856,
+      "step": 16951
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1073934391676666e-06,
+      "loss": 0.42,
+      "step": 16952
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1066478557547578e-06,
+      "loss": 0.9201,
+      "step": 16953
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1059025087169927e-06,
+      "loss": 0.846,
+      "step": 16954
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1051573980741802e-06,
+      "loss": 0.9412,
+      "step": 16955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1044125238461234e-06,
+      "loss": 0.8736,
+      "step": 16956
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1036678860526252e-06,
+      "loss": 0.7596,
+      "step": 16957
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1029234847134719e-06,
+      "loss": 0.9137,
+      "step": 16958
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.102179319848451e-06,
+      "loss": 0.889,
+      "step": 16959
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1014353914773424e-06,
+      "loss": 0.8788,
+      "step": 16960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.100691699619919e-06,
+      "loss": 0.8646,
+      "step": 16961
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0999482442959441e-06,
+      "loss": 0.901,
+      "step": 16962
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0992050255251807e-06,
+      "loss": 0.8463,
+      "step": 16963
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.098462043327384e-06,
+      "loss": 0.8921,
+      "step": 16964
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0977192977222973e-06,
+      "loss": 0.3576,
+      "step": 16965
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0969767887296646e-06,
+      "loss": 0.8788,
+      "step": 16966
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.09623451636922e-06,
+      "loss": 0.8351,
+      "step": 16967
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0954924806606937e-06,
+      "loss": 0.8926,
+      "step": 16968
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0947506816238074e-06,
+      "loss": 0.7966,
+      "step": 16969
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0940091192782787e-06,
+      "loss": 0.8114,
+      "step": 16970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.093267793643814e-06,
+      "loss": 0.9304,
+      "step": 16971
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.092526704740119e-06,
+      "loss": 0.8873,
+      "step": 16972
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0917858525868907e-06,
+      "loss": 0.8528,
+      "step": 16973
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0910452372038227e-06,
+      "loss": 0.9491,
+      "step": 16974
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0903048586105947e-06,
+      "loss": 0.3821,
+      "step": 16975
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0895647168268886e-06,
+      "loss": 0.8562,
+      "step": 16976
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0888248118723766e-06,
+      "loss": 0.8629,
+      "step": 16977
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0880851437667228e-06,
+      "loss": 0.4371,
+      "step": 16978
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0873457125295861e-06,
+      "loss": 0.8401,
+      "step": 16979
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0866065181806218e-06,
+      "loss": 0.8609,
+      "step": 16980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0858675607394764e-06,
+      "loss": 0.9266,
+      "step": 16981
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0851288402257898e-06,
+      "loss": 0.8383,
+      "step": 16982
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0843903566591995e-06,
+      "loss": 0.9102,
+      "step": 16983
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.083652110059329e-06,
+      "loss": 0.8471,
+      "step": 16984
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.082914100445801e-06,
+      "loss": 0.8685,
+      "step": 16985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0821763278382325e-06,
+      "loss": 0.8972,
+      "step": 16986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.081438792256234e-06,
+      "loss": 0.8633,
+      "step": 16987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0807014937194049e-06,
+      "loss": 0.8992,
+      "step": 16988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0799644322473423e-06,
+      "loss": 0.8477,
+      "step": 16989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0792276078596409e-06,
+      "loss": 0.9553,
+      "step": 16990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0784910205758758e-06,
+      "loss": 0.8135,
+      "step": 16991
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0777546704156327e-06,
+      "loss": 0.8363,
+      "step": 16992
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0770185573984815e-06,
+      "loss": 0.8294,
+      "step": 16993
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0762826815439852e-06,
+      "loss": 0.8037,
+      "step": 16994
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0755470428717029e-06,
+      "loss": 0.8481,
+      "step": 16995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.074811641401189e-06,
+      "loss": 0.8621,
+      "step": 16996
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0740764771519862e-06,
+      "loss": 0.8761,
+      "step": 16997
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0733415501436372e-06,
+      "loss": 0.8732,
+      "step": 16998
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.072606860395674e-06,
+      "loss": 0.8767,
+      "step": 16999
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0718724079276267e-06,
+      "loss": 0.9216,
+      "step": 17000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0711381927590115e-06,
+      "loss": 0.8918,
+      "step": 17001
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0704042149093451e-06,
+      "loss": 0.8964,
+      "step": 17002
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0696704743981368e-06,
+      "loss": 0.8559,
+      "step": 17003
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0689369712448872e-06,
+      "loss": 0.9286,
+      "step": 17004
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.068203705469093e-06,
+      "loss": 0.9096,
+      "step": 17005
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0674706770902454e-06,
+      "loss": 0.8665,
+      "step": 17006
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0667378861278233e-06,
+      "loss": 0.884,
+      "step": 17007
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0660053326013041e-06,
+      "loss": 0.8456,
+      "step": 17008
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0652730165301617e-06,
+      "loss": 0.8516,
+      "step": 17009
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.064540937933859e-06,
+      "loss": 0.8942,
+      "step": 17010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0638090968318504e-06,
+      "loss": 0.9189,
+      "step": 17011
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0630774932435906e-06,
+      "loss": 0.9304,
+      "step": 17012
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.062346127188526e-06,
+      "loss": 0.9253,
+      "step": 17013
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0616149986860913e-06,
+      "loss": 0.8728,
+      "step": 17014
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0608841077557198e-06,
+      "loss": 0.8665,
+      "step": 17015
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0601534544168424e-06,
+      "loss": 0.8194,
+      "step": 17016
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0594230386888749e-06,
+      "loss": 0.8948,
+      "step": 17017
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0586928605912316e-06,
+      "loss": 0.8665,
+      "step": 17018
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0579629201433216e-06,
+      "loss": 0.7948,
+      "step": 17019
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0572332173645428e-06,
+      "loss": 0.9117,
+      "step": 17020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0565037522742915e-06,
+      "loss": 0.9355,
+      "step": 17021
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.055774524891956e-06,
+      "loss": 0.8124,
+      "step": 17022
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0550455352369204e-06,
+      "loss": 0.91,
+      "step": 17023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.054316783328556e-06,
+      "loss": 0.9415,
+      "step": 17024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0535882691862353e-06,
+      "loss": 0.8499,
+      "step": 17025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0528599928293226e-06,
+      "loss": 0.8529,
+      "step": 17026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.052131954277168e-06,
+      "loss": 0.8391,
+      "step": 17027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0514041535491293e-06,
+      "loss": 0.8116,
+      "step": 17028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0506765906645488e-06,
+      "loss": 0.8776,
+      "step": 17029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0499492656427623e-06,
+      "loss": 0.9203,
+      "step": 17030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.049222178503102e-06,
+      "loss": 0.8499,
+      "step": 17031
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0484953292648959e-06,
+      "loss": 0.8464,
+      "step": 17032
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0477687179474583e-06,
+      "loss": 0.8201,
+      "step": 17033
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0470423445701027e-06,
+      "loss": 0.9652,
+      "step": 17034
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0463162091521374e-06,
+      "loss": 0.8909,
+      "step": 17035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.045590311712863e-06,
+      "loss": 0.9155,
+      "step": 17036
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0448646522715688e-06,
+      "loss": 0.8111,
+      "step": 17037
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.044139230847545e-06,
+      "loss": 0.679,
+      "step": 17038
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0434140474600717e-06,
+      "loss": 0.826,
+      "step": 17039
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0426891021284236e-06,
+      "loss": 0.8713,
+      "step": 17040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0419643948718683e-06,
+      "loss": 0.9609,
+      "step": 17041
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0412399257096717e-06,
+      "loss": 0.8671,
+      "step": 17042
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0405156946610827e-06,
+      "loss": 0.9152,
+      "step": 17043
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.039791701745354e-06,
+      "loss": 0.8927,
+      "step": 17044
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.039067946981731e-06,
+      "loss": 0.8861,
+      "step": 17045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0383444303894453e-06,
+      "loss": 0.9384,
+      "step": 17046
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.037621151987729e-06,
+      "loss": 0.8687,
+      "step": 17047
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0368981117958065e-06,
+      "loss": 0.9499,
+      "step": 17048
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0361753098328985e-06,
+      "loss": 0.9143,
+      "step": 17049
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0354527461182085e-06,
+      "loss": 0.9141,
+      "step": 17050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0347304206709473e-06,
+      "loss": 0.9228,
+      "step": 17051
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0340083335103145e-06,
+      "loss": 0.8488,
+      "step": 17052
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.033286484655498e-06,
+      "loss": 0.8752,
+      "step": 17053
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0325648741256867e-06,
+      "loss": 0.8983,
+      "step": 17054
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0318435019400608e-06,
+      "loss": 0.8851,
+      "step": 17055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0311223681177895e-06,
+      "loss": 0.9087,
+      "step": 17056
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.030401472678042e-06,
+      "loss": 0.9068,
+      "step": 17057
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0296808156399796e-06,
+      "loss": 0.9188,
+      "step": 17058
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.028960397022758e-06,
+      "loss": 0.9335,
+      "step": 17059
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.028240216845522e-06,
+      "loss": 0.837,
+      "step": 17060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0275202751274127e-06,
+      "loss": 0.9601,
+      "step": 17061
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0268005718875674e-06,
+      "loss": 0.8459,
+      "step": 17062
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.026081107145115e-06,
+      "loss": 0.8223,
+      "step": 17063
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0253618809191779e-06,
+      "loss": 0.9183,
+      "step": 17064
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0246428932288731e-06,
+      "loss": 0.9355,
+      "step": 17065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0239241440933078e-06,
+      "loss": 0.9432,
+      "step": 17066
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0232056335315877e-06,
+      "loss": 0.8495,
+      "step": 17067
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0224873615628106e-06,
+      "loss": 0.8841,
+      "step": 17068
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0217693282060637e-06,
+      "loss": 0.8638,
+      "step": 17069
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0210515334804349e-06,
+      "loss": 0.399,
+      "step": 17070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0203339774050002e-06,
+      "loss": 0.8725,
+      "step": 17071
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0196166599988356e-06,
+      "loss": 0.8145,
+      "step": 17072
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.018899581281002e-06,
+      "loss": 0.9056,
+      "step": 17073
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.018182741270559e-06,
+      "loss": 0.9246,
+      "step": 17074
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0174661399865616e-06,
+      "loss": 0.8565,
+      "step": 17075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.016749777448054e-06,
+      "loss": 0.7938,
+      "step": 17076
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0160336536740778e-06,
+      "loss": 0.3962,
+      "step": 17077
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.015317768683669e-06,
+      "loss": 0.9095,
+      "step": 17078
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0146021224958502e-06,
+      "loss": 0.9685,
+      "step": 17079
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0138867151296449e-06,
+      "loss": 0.8896,
+      "step": 17080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0131715466040692e-06,
+      "loss": 0.9936,
+      "step": 17081
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0124566169381277e-06,
+      "loss": 0.9208,
+      "step": 17082
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0117419261508244e-06,
+      "loss": 0.7824,
+      "step": 17083
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0110274742611558e-06,
+      "loss": 0.8648,
+      "step": 17084
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0103132612881117e-06,
+      "loss": 0.8689,
+      "step": 17085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.009599287250671e-06,
+      "loss": 0.8347,
+      "step": 17086
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0088855521678143e-06,
+      "loss": 0.874,
+      "step": 17087
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0081720560585139e-06,
+      "loss": 0.9304,
+      "step": 17088
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0074587989417273e-06,
+      "loss": 0.8677,
+      "step": 17089
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0067457808364167e-06,
+      "loss": 0.8911,
+      "step": 17090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0060330017615327e-06,
+      "loss": 0.9017,
+      "step": 17091
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.005320461736018e-06,
+      "loss": 0.944,
+      "step": 17092
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0046081607788127e-06,
+      "loss": 0.9171,
+      "step": 17093
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0038960989088497e-06,
+      "loss": 0.8478,
+      "step": 17094
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0031842761450528e-06,
+      "loss": 0.828,
+      "step": 17095
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.002472692506342e-06,
+      "loss": 0.8887,
+      "step": 17096
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0017613480116317e-06,
+      "loss": 0.9793,
+      "step": 17097
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0010502426798273e-06,
+      "loss": 0.8268,
+      "step": 17098
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0003393765298298e-06,
+      "loss": 0.8437,
+      "step": 17099
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.996287495805334e-07,
+      "loss": 0.8732,
+      "step": 17100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.98918361850827e-07,
+      "loss": 0.8209,
+      "step": 17101
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.982082133595882e-07,
+      "loss": 0.8117,
+      "step": 17102
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.974983041256935e-07,
+      "loss": 0.8806,
+      "step": 17103
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.967886341680155e-07,
+      "loss": 0.8944,
+      "step": 17104
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.9607920350541e-07,
+      "loss": 0.8583,
+      "step": 17105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.953700121567355e-07,
+      "loss": 0.8607,
+      "step": 17106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.946610601408425e-07,
+      "loss": 0.8703,
+      "step": 17107
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.939523474765756e-07,
+      "loss": 0.9924,
+      "step": 17108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.932438741827643e-07,
+      "loss": 0.8117,
+      "step": 17109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.925356402782494e-07,
+      "loss": 0.9315,
+      "step": 17110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.918276457818466e-07,
+      "loss": 0.9059,
+      "step": 17111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.911198907123777e-07,
+      "loss": 0.8534,
+      "step": 17112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.90412375088653e-07,
+      "loss": 0.874,
+      "step": 17113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.897050989294788e-07,
+      "loss": 0.8285,
+      "step": 17114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.889980622536499e-07,
+      "loss": 0.8924,
+      "step": 17115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.882912650799615e-07,
+      "loss": 0.9133,
+      "step": 17116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.875847074272004e-07,
+      "loss": 0.7611,
+      "step": 17117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.868783893141432e-07,
+      "loss": 0.8238,
+      "step": 17118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.861723107595644e-07,
+      "loss": 0.862,
+      "step": 17119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.854664717822315e-07,
+      "loss": 0.804,
+      "step": 17120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.847608724009028e-07,
+      "loss": 0.7546,
+      "step": 17121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.840555126343343e-07,
+      "loss": 0.9052,
+      "step": 17122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.833503925012756e-07,
+      "loss": 0.9248,
+      "step": 17123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.826455120204625e-07,
+      "loss": 0.8121,
+      "step": 17124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.819408712106338e-07,
+      "loss": 0.8243,
+      "step": 17125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.812364700905165e-07,
+      "loss": 0.9032,
+      "step": 17126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.805323086788354e-07,
+      "loss": 0.7857,
+      "step": 17127
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.798283869943025e-07,
+      "loss": 0.9,
+      "step": 17128
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.791247050556295e-07,
+      "loss": 0.8988,
+      "step": 17129
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.784212628815204e-07,
+      "loss": 0.8839,
+      "step": 17130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.777180604906689e-07,
+      "loss": 0.7471,
+      "step": 17131
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.770150979017667e-07,
+      "loss": 0.9405,
+      "step": 17132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.763123751334979e-07,
+      "loss": 0.8884,
+      "step": 17133
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.75609892204541e-07,
+      "loss": 0.9052,
+      "step": 17134
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.749076491335651e-07,
+      "loss": 0.838,
+      "step": 17135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.74205645939237e-07,
+      "loss": 0.9147,
+      "step": 17136
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.735038826402165e-07,
+      "loss": 0.7999,
+      "step": 17137
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.728023592551505e-07,
+      "loss": 0.8916,
+      "step": 17138
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.721010758026884e-07,
+      "loss": 0.9119,
+      "step": 17139
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.714000323014705e-07,
+      "loss": 0.8124,
+      "step": 17140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.706992287701266e-07,
+      "loss": 0.9095,
+      "step": 17141
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.699986652272851e-07,
+      "loss": 0.3862,
+      "step": 17142
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.692983416915668e-07,
+      "loss": 0.9208,
+      "step": 17143
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.685982581815833e-07,
+      "loss": 0.8754,
+      "step": 17144
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.678984147159409e-07,
+      "loss": 0.8512,
+      "step": 17145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.67198811313248e-07,
+      "loss": 0.9436,
+      "step": 17146
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.66499447992091e-07,
+      "loss": 0.8403,
+      "step": 17147
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.658003247710611e-07,
+      "loss": 0.8389,
+      "step": 17148
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.651014416687422e-07,
+      "loss": 0.8701,
+      "step": 17149
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.644027987037086e-07,
+      "loss": 0.885,
+      "step": 17150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.63704395894528e-07,
+      "loss": 0.8902,
+      "step": 17151
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.630062332597634e-07,
+      "loss": 0.8847,
+      "step": 17152
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.623083108179742e-07,
+      "loss": 0.3933,
+      "step": 17153
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.616106285877058e-07,
+      "loss": 0.9299,
+      "step": 17154
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.609131865875043e-07,
+      "loss": 0.8701,
+      "step": 17155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.60215984835906e-07,
+      "loss": 0.9672,
+      "step": 17156
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.595190233514439e-07,
+      "loss": 0.9725,
+      "step": 17157
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.5882230215264e-07,
+      "loss": 0.8954,
+      "step": 17158
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.581258212580147e-07,
+      "loss": 0.8667,
+      "step": 17159
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.574295806860767e-07,
+      "loss": 0.806,
+      "step": 17160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.567335804553335e-07,
+      "loss": 0.7923,
+      "step": 17161
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.560378205842835e-07,
+      "loss": 0.9312,
+      "step": 17162
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.553423010914199e-07,
+      "loss": 0.8189,
+      "step": 17163
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.546470219952264e-07,
+      "loss": 0.8444,
+      "step": 17164
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.53951983314184e-07,
+      "loss": 0.9199,
+      "step": 17165
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.532571850667693e-07,
+      "loss": 0.8797,
+      "step": 17166
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.525626272714438e-07,
+      "loss": 0.9351,
+      "step": 17167
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.518683099466686e-07,
+      "loss": 0.9314,
+      "step": 17168
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.511742331109031e-07,
+      "loss": 0.8888,
+      "step": 17169
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.504803967825904e-07,
+      "loss": 0.8637,
+      "step": 17170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.497868009801725e-07,
+      "loss": 0.7833,
+      "step": 17171
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.49093445722088e-07,
+      "loss": 0.8506,
+      "step": 17172
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.484003310267598e-07,
+      "loss": 0.8043,
+      "step": 17173
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.477074569126132e-07,
+      "loss": 0.9194,
+      "step": 17174
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.470148233980636e-07,
+      "loss": 0.882,
+      "step": 17175
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.463224305015228e-07,
+      "loss": 0.8714,
+      "step": 17176
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.456302782413884e-07,
+      "loss": 0.8374,
+      "step": 17177
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.449383666360612e-07,
+      "loss": 0.8681,
+      "step": 17178
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.44246695703932e-07,
+      "loss": 0.8009,
+      "step": 17179
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.435552654633795e-07,
+      "loss": 0.8732,
+      "step": 17180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.428640759327867e-07,
+      "loss": 0.8502,
+      "step": 17181
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.421731271305234e-07,
+      "loss": 0.81,
+      "step": 17182
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.414824190749528e-07,
+      "loss": 0.8792,
+      "step": 17183
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.407919517844322e-07,
+      "loss": 0.9442,
+      "step": 17184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.40101725277316e-07,
+      "loss": 0.9334,
+      "step": 17185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.394117395719504e-07,
+      "loss": 0.754,
+      "step": 17186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.387219946866699e-07,
+      "loss": 0.8752,
+      "step": 17187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.380324906398097e-07,
+      "loss": 0.8129,
+      "step": 17188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.373432274496996e-07,
+      "loss": 0.8425,
+      "step": 17189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.366542051346528e-07,
+      "loss": 0.8553,
+      "step": 17190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.359654237129856e-07,
+      "loss": 0.8363,
+      "step": 17191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.352768832030068e-07,
+      "loss": 0.8039,
+      "step": 17192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.34588583623014e-07,
+      "loss": 0.8933,
+      "step": 17193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.339005249913036e-07,
+      "loss": 0.8484,
+      "step": 17194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.332127073261643e-07,
+      "loss": 0.8796,
+      "step": 17195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.325251306458749e-07,
+      "loss": 0.8886,
+      "step": 17196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.318377949687107e-07,
+      "loss": 0.9154,
+      "step": 17197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.311507003129416e-07,
+      "loss": 0.9057,
+      "step": 17198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.304638466968308e-07,
+      "loss": 0.8139,
+      "step": 17199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.297772341386313e-07,
+      "loss": 0.9,
+      "step": 17200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.290908626565931e-07,
+      "loss": 0.8879,
+      "step": 17201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.284047322689615e-07,
+      "loss": 0.8944,
+      "step": 17202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.277188429939699e-07,
+      "loss": 0.9299,
+      "step": 17203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.270331948498479e-07,
+      "loss": 0.8412,
+      "step": 17204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.263477878548255e-07,
+      "loss": 0.8708,
+      "step": 17205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.256626220271137e-07,
+      "loss": 0.9135,
+      "step": 17206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.249776973849245e-07,
+      "loss": 0.8698,
+      "step": 17207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.242930139464667e-07,
+      "loss": 0.8114,
+      "step": 17208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.236085717299325e-07,
+      "loss": 0.9508,
+      "step": 17209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.229243707535163e-07,
+      "loss": 0.849,
+      "step": 17210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.222404110354033e-07,
+      "loss": 0.9038,
+      "step": 17211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.215566925937747e-07,
+      "loss": 0.8762,
+      "step": 17212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.208732154467981e-07,
+      "loss": 0.7932,
+      "step": 17213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.201899796126424e-07,
+      "loss": 0.9089,
+      "step": 17214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.195069851094685e-07,
+      "loss": 0.8714,
+      "step": 17215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.188242319554252e-07,
+      "loss": 0.8901,
+      "step": 17216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.181417201686627e-07,
+      "loss": 0.8642,
+      "step": 17217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.174594497673239e-07,
+      "loss": 0.876,
+      "step": 17218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.167774207695368e-07,
+      "loss": 0.8734,
+      "step": 17219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.160956331934323e-07,
+      "loss": 0.867,
+      "step": 17220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.154140870571326e-07,
+      "loss": 0.9025,
+      "step": 17221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.1473278237875e-07,
+      "loss": 0.8556,
+      "step": 17222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.140517191763931e-07,
+      "loss": 0.8455,
+      "step": 17223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.133708974681643e-07,
+      "loss": 0.348,
+      "step": 17224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.126903172721601e-07,
+      "loss": 0.8975,
+      "step": 17225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.120099786064673e-07,
+      "loss": 0.829,
+      "step": 17226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.113298814891702e-07,
+      "loss": 0.7756,
+      "step": 17227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.106500259383444e-07,
+      "loss": 0.8628,
+      "step": 17228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.099704119720598e-07,
+      "loss": 0.8454,
+      "step": 17229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.092910396083798e-07,
+      "loss": 0.87,
+      "step": 17230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.086119088653644e-07,
+      "loss": 0.8949,
+      "step": 17231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.079330197610581e-07,
+      "loss": 0.8413,
+      "step": 17232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.072543723135074e-07,
+      "loss": 0.9381,
+      "step": 17233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.065759665407514e-07,
+      "loss": 0.839,
+      "step": 17234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.058978024608223e-07,
+      "loss": 0.8736,
+      "step": 17235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.052198800917411e-07,
+      "loss": 0.8257,
+      "step": 17236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.045421994515291e-07,
+      "loss": 0.9229,
+      "step": 17237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.038647605581985e-07,
+      "loss": 0.8466,
+      "step": 17238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.031875634297504e-07,
+      "loss": 0.8251,
+      "step": 17239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.025106080841883e-07,
+      "loss": 0.8969,
+      "step": 17240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.018338945395066e-07,
+      "loss": 0.7935,
+      "step": 17241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.011574228136866e-07,
+      "loss": 0.8847,
+      "step": 17242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.004811929247104e-07,
+      "loss": 0.8437,
+      "step": 17243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.998052048905537e-07,
+      "loss": 0.8883,
+      "step": 17244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.991294587291799e-07,
+      "loss": 0.3969,
+      "step": 17245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.984539544585502e-07,
+      "loss": 0.8559,
+      "step": 17246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.977786920966191e-07,
+      "loss": 0.9329,
+      "step": 17247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.971036716613368e-07,
+      "loss": 0.8742,
+      "step": 17248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.96428893170641e-07,
+      "loss": 0.8658,
+      "step": 17249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.957543566424676e-07,
+      "loss": 0.7531,
+      "step": 17250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.950800620947453e-07,
+      "loss": 0.8547,
+      "step": 17251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.944060095453955e-07,
+      "loss": 0.8176,
+      "step": 17252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.937321990123348e-07,
+      "loss": 0.9501,
+      "step": 17253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.930586305134747e-07,
+      "loss": 0.9198,
+      "step": 17254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.923853040667119e-07,
+      "loss": 0.8417,
+      "step": 17255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.917122196899464e-07,
+      "loss": 0.876,
+      "step": 17256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.910393774010695e-07,
+      "loss": 0.8343,
+      "step": 17257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.903667772179603e-07,
+      "loss": 0.8978,
+      "step": 17258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.896944191584988e-07,
+      "loss": 0.8932,
+      "step": 17259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.890223032405543e-07,
+      "loss": 0.8482,
+      "step": 17260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.883504294819922e-07,
+      "loss": 0.829,
+      "step": 17261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.876787979006685e-07,
+      "loss": 0.8799,
+      "step": 17262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.870074085144343e-07,
+      "loss": 0.8555,
+      "step": 17263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.863362613411352e-07,
+      "loss": 0.8904,
+      "step": 17264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.856653563986095e-07,
+      "loss": 0.9695,
+      "step": 17265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.849946937046894e-07,
+      "loss": 0.8992,
+      "step": 17266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.843242732772017e-07,
+      "loss": 0.9227,
+      "step": 17267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.836540951339623e-07,
+      "loss": 0.7991,
+      "step": 17268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.829841592927845e-07,
+      "loss": 0.8349,
+      "step": 17269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.823144657714766e-07,
+      "loss": 0.8749,
+      "step": 17270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.816450145878364e-07,
+      "loss": 0.8513,
+      "step": 17271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.809758057596562e-07,
+      "loss": 0.9363,
+      "step": 17272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.803068393047242e-07,
+      "loss": 0.8997,
+      "step": 17273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.79638115240824e-07,
+      "loss": 0.8376,
+      "step": 17274
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.789696335857212e-07,
+      "loss": 0.8904,
+      "step": 17275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.783013943571905e-07,
+      "loss": 0.9224,
+      "step": 17276
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.776333975729923e-07,
+      "loss": 0.9151,
+      "step": 17277
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.769656432508789e-07,
+      "loss": 0.898,
+      "step": 17278
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.762981314085983e-07,
+      "loss": 0.7926,
+      "step": 17279
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.756308620638954e-07,
+      "loss": 0.9538,
+      "step": 17280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.749638352345002e-07,
+      "loss": 0.923,
+      "step": 17281
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.742970509381455e-07,
+      "loss": 0.8358,
+      "step": 17282
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.736305091925523e-07,
+      "loss": 0.9405,
+      "step": 17283
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.72964210015439e-07,
+      "loss": 0.7969,
+      "step": 17284
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.722981534245112e-07,
+      "loss": 0.8657,
+      "step": 17285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.716323394374737e-07,
+      "loss": 0.9639,
+      "step": 17286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.709667680720224e-07,
+      "loss": 0.8164,
+      "step": 17287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.703014393458486e-07,
+      "loss": 0.7536,
+      "step": 17288
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.69636353276635e-07,
+      "loss": 0.9004,
+      "step": 17289
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.689715098820617e-07,
+      "loss": 0.9463,
+      "step": 17290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.683069091797947e-07,
+      "loss": 0.7846,
+      "step": 17291
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.676425511874997e-07,
+      "loss": 0.8021,
+      "step": 17292
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.669784359228395e-07,
+      "loss": 0.3924,
+      "step": 17293
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.663145634034587e-07,
+      "loss": 0.9547,
+      "step": 17294
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.656509336470042e-07,
+      "loss": 0.8679,
+      "step": 17295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.649875466711166e-07,
+      "loss": 0.9068,
+      "step": 17296
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.643244024934283e-07,
+      "loss": 0.868,
+      "step": 17297
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.636615011315597e-07,
+      "loss": 0.8493,
+      "step": 17298
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.629988426031377e-07,
+      "loss": 0.8252,
+      "step": 17299
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.623364269257695e-07,
+      "loss": 0.8541,
+      "step": 17300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.61674254117062e-07,
+      "loss": 0.8837,
+      "step": 17301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.610123241946166e-07,
+      "loss": 0.813,
+      "step": 17302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.603506371760272e-07,
+      "loss": 0.8983,
+      "step": 17303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.596891930788776e-07,
+      "loss": 0.8446,
+      "step": 17304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.590279919207501e-07,
+      "loss": 0.8507,
+      "step": 17305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.583670337192207e-07,
+      "loss": 0.8789,
+      "step": 17306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.577063184918521e-07,
+      "loss": 0.8515,
+      "step": 17307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.570458462562092e-07,
+      "loss": 0.8962,
+      "step": 17308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.563856170298446e-07,
+      "loss": 0.8591,
+      "step": 17309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.557256308303075e-07,
+      "loss": 0.8345,
+      "step": 17310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.550658876751395e-07,
+      "loss": 0.8855,
+      "step": 17311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.544063875818776e-07,
+      "loss": 0.8775,
+      "step": 17312
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.537471305680467e-07,
+      "loss": 0.7867,
+      "step": 17313
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.530881166511717e-07,
+      "loss": 0.8199,
+      "step": 17314
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.524293458487664e-07,
+      "loss": 0.7895,
+      "step": 17315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.517708181783447e-07,
+      "loss": 0.8386,
+      "step": 17316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.511125336574033e-07,
+      "loss": 0.858,
+      "step": 17317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.50454492303443e-07,
+      "loss": 0.9247,
+      "step": 17318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.497966941339542e-07,
+      "loss": 0.8501,
+      "step": 17319
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.491391391664161e-07,
+      "loss": 0.7909,
+      "step": 17320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.484818274183082e-07,
+      "loss": 0.8471,
+      "step": 17321
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.478247589071021e-07,
+      "loss": 0.8942,
+      "step": 17322
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.471679336502614e-07,
+      "loss": 0.8692,
+      "step": 17323
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.465113516652424e-07,
+      "loss": 0.8666,
+      "step": 17324
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.458550129694987e-07,
+      "loss": 0.8289,
+      "step": 17325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.451989175804742e-07,
+      "loss": 0.9034,
+      "step": 17326
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.445430655156062e-07,
+      "loss": 0.9168,
+      "step": 17327
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.438874567923261e-07,
+      "loss": 0.8924,
+      "step": 17328
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.432320914280623e-07,
+      "loss": 0.9026,
+      "step": 17329
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.425769694402286e-07,
+      "loss": 0.9468,
+      "step": 17330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.419220908462411e-07,
+      "loss": 0.8922,
+      "step": 17331
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.412674556635047e-07,
+      "loss": 0.888,
+      "step": 17332
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.406130639094212e-07,
+      "loss": 0.819,
+      "step": 17333
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.399589156013787e-07,
+      "loss": 0.8896,
+      "step": 17334
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.393050107567691e-07,
+      "loss": 0.8569,
+      "step": 17335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.386513493929671e-07,
+      "loss": 0.8877,
+      "step": 17336
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.379979315273501e-07,
+      "loss": 0.9132,
+      "step": 17337
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.373447571772842e-07,
+      "loss": 0.8458,
+      "step": 17338
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.36691826360131e-07,
+      "loss": 0.8825,
+      "step": 17339
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.36039139093242e-07,
+      "loss": 0.7602,
+      "step": 17340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.353866953939671e-07,
+      "loss": 0.8197,
+      "step": 17341
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.347344952796476e-07,
+      "loss": 0.9397,
+      "step": 17342
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.340825387676166e-07,
+      "loss": 0.8516,
+      "step": 17343
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.334308258752022e-07,
+      "loss": 0.9085,
+      "step": 17344
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.327793566197284e-07,
+      "loss": 0.7886,
+      "step": 17345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.321281310185081e-07,
+      "loss": 0.9142,
+      "step": 17346
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.314771490888518e-07,
+      "loss": 0.8608,
+      "step": 17347
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.308264108480635e-07,
+      "loss": 0.7803,
+      "step": 17348
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.30175916313436e-07,
+      "loss": 0.7895,
+      "step": 17349
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.295256655022588e-07,
+      "loss": 0.8783,
+      "step": 17350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.288756584318158e-07,
+      "loss": 0.8722,
+      "step": 17351
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.282258951193844e-07,
+      "loss": 0.8894,
+      "step": 17352
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.275763755822331e-07,
+      "loss": 0.9463,
+      "step": 17353
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.269270998376244e-07,
+      "loss": 0.9105,
+      "step": 17354
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.262780679028193e-07,
+      "loss": 0.8211,
+      "step": 17355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.256292797950649e-07,
+      "loss": 0.842,
+      "step": 17356
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.249807355316031e-07,
+      "loss": 0.9041,
+      "step": 17357
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.243324351296789e-07,
+      "loss": 0.8077,
+      "step": 17358
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.236843786065163e-07,
+      "loss": 0.916,
+      "step": 17359
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.230365659793426e-07,
+      "loss": 0.8783,
+      "step": 17360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.223889972653776e-07,
+      "loss": 0.9101,
+      "step": 17361
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.217416724818305e-07,
+      "loss": 0.8714,
+      "step": 17362
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.210945916459068e-07,
+      "loss": 0.9275,
+      "step": 17363
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.204477547748047e-07,
+      "loss": 0.8924,
+      "step": 17364
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.198011618857193e-07,
+      "loss": 0.9273,
+      "step": 17365
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.191548129958326e-07,
+      "loss": 0.8716,
+      "step": 17366
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.185087081223253e-07,
+      "loss": 0.9048,
+      "step": 17367
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.178628472823725e-07,
+      "loss": 0.8905,
+      "step": 17368
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.172172304931348e-07,
+      "loss": 0.9168,
+      "step": 17369
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.165718577717763e-07,
+      "loss": 0.8295,
+      "step": 17370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.159267291354522e-07,
+      "loss": 0.8118,
+      "step": 17371
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.152818446013033e-07,
+      "loss": 0.9247,
+      "step": 17372
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.146372041864736e-07,
+      "loss": 0.8691,
+      "step": 17373
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.139928079080972e-07,
+      "loss": 0.8613,
+      "step": 17374
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.133486557833015e-07,
+      "loss": 0.8462,
+      "step": 17375
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.12704747829205e-07,
+      "loss": 0.9654,
+      "step": 17376
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.120610840629228e-07,
+      "loss": 0.8875,
+      "step": 17377
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.114176645015648e-07,
+      "loss": 0.8639,
+      "step": 17378
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.107744891622305e-07,
+      "loss": 0.8439,
+      "step": 17379
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.101315580620139e-07,
+      "loss": 0.8284,
+      "step": 17380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.09488871218006e-07,
+      "loss": 0.8721,
+      "step": 17381
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.088464286472863e-07,
+      "loss": 0.8243,
+      "step": 17382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.082042303669313e-07,
+      "loss": 0.8561,
+      "step": 17383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.075622763940116e-07,
+      "loss": 0.9357,
+      "step": 17384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.069205667455859e-07,
+      "loss": 0.9078,
+      "step": 17385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.062791014387128e-07,
+      "loss": 0.8496,
+      "step": 17386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.056378804904408e-07,
+      "loss": 0.8909,
+      "step": 17387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.04996903917814e-07,
+      "loss": 0.7855,
+      "step": 17388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.043561717378667e-07,
+      "loss": 0.8431,
+      "step": 17389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.037156839676296e-07,
+      "loss": 0.8981,
+      "step": 17390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.030754406241281e-07,
+      "loss": 0.9055,
+      "step": 17391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.024354417243763e-07,
+      "loss": 0.8961,
+      "step": 17392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.017956872853827e-07,
+      "loss": 0.8889,
+      "step": 17393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.011561773241594e-07,
+      "loss": 0.8837,
+      "step": 17394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.005169118576961e-07,
+      "loss": 0.8459,
+      "step": 17395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.99877890902987e-07,
+      "loss": 0.8008,
+      "step": 17396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.992391144770184e-07,
+      "loss": 0.9293,
+      "step": 17397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.986005825967636e-07,
+      "loss": 0.8708,
+      "step": 17398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.979622952791954e-07,
+      "loss": 0.8913,
+      "step": 17399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.973242525412806e-07,
+      "loss": 0.7923,
+      "step": 17400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.966864543999786e-07,
+      "loss": 0.8349,
+      "step": 17401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.960489008722372e-07,
+      "loss": 0.8417,
+      "step": 17402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.954115919750038e-07,
+      "loss": 0.9205,
+      "step": 17403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.947745277252195e-07,
+      "loss": 0.912,
+      "step": 17404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.941377081398127e-07,
+      "loss": 0.905,
+      "step": 17405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.935011332357113e-07,
+      "loss": 0.892,
+      "step": 17406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.928648030298381e-07,
+      "loss": 0.8046,
+      "step": 17407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.922287175391008e-07,
+      "loss": 0.9502,
+      "step": 17408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.915928767804082e-07,
+      "loss": 0.8243,
+      "step": 17409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.909572807706611e-07,
+      "loss": 0.8916,
+      "step": 17410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.903219295267495e-07,
+      "loss": 0.8581,
+      "step": 17411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.89686823065563e-07,
+      "loss": 0.8645,
+      "step": 17412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.890519614039826e-07,
+      "loss": 0.7931,
+      "step": 17413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.884173445588816e-07,
+      "loss": 0.8228,
+      "step": 17414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.877829725471253e-07,
+      "loss": 0.9111,
+      "step": 17415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.871488453855757e-07,
+      "loss": 0.7901,
+      "step": 17416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.865149630910906e-07,
+      "loss": 0.8652,
+      "step": 17417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.85881325680512e-07,
+      "loss": 0.9179,
+      "step": 17418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.852479331706852e-07,
+      "loss": 0.9038,
+      "step": 17419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.846147855784458e-07,
+      "loss": 0.8826,
+      "step": 17420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.839818829206192e-07,
+      "loss": 0.8374,
+      "step": 17421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.833492252140284e-07,
+      "loss": 0.9266,
+      "step": 17422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.82716812475488e-07,
+      "loss": 0.8793,
+      "step": 17423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.8208464472181e-07,
+      "loss": 0.8954,
+      "step": 17424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.81452721969792e-07,
+      "loss": 0.9169,
+      "step": 17425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.808210442362319e-07,
+      "loss": 0.842,
+      "step": 17426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.801896115379215e-07,
+      "loss": 0.7914,
+      "step": 17427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.795584238916376e-07,
+      "loss": 0.8712,
+      "step": 17428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.789274813141612e-07,
+      "loss": 0.8583,
+      "step": 17429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.782967838222633e-07,
+      "loss": 0.7879,
+      "step": 17430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.776663314327027e-07,
+      "loss": 0.9282,
+      "step": 17431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.770361241622381e-07,
+      "loss": 0.8417,
+      "step": 17432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.76406162027622e-07,
+      "loss": 0.9306,
+      "step": 17433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.757764450455929e-07,
+      "loss": 0.9281,
+      "step": 17434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.751469732328909e-07,
+      "loss": 0.9035,
+      "step": 17435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.745177466062481e-07,
+      "loss": 0.8784,
+      "step": 17436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.738887651823868e-07,
+      "loss": 0.7849,
+      "step": 17437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.732600289780245e-07,
+      "loss": 0.8845,
+      "step": 17438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.726315380098726e-07,
+      "loss": 0.9024,
+      "step": 17439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.720032922946352e-07,
+      "loss": 0.4299,
+      "step": 17440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.713752918490114e-07,
+      "loss": 0.8316,
+      "step": 17441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.707475366896922e-07,
+      "loss": 0.8968,
+      "step": 17442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.701200268333642e-07,
+      "loss": 0.8862,
+      "step": 17443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.694927622967019e-07,
+      "loss": 0.9324,
+      "step": 17444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.688657430963808e-07,
+      "loss": 0.4115,
+      "step": 17445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.682389692490666e-07,
+      "loss": 0.7686,
+      "step": 17446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.676124407714147e-07,
+      "loss": 0.8813,
+      "step": 17447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.669861576800807e-07,
+      "loss": 0.9315,
+      "step": 17448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.663601199917092e-07,
+      "loss": 0.9197,
+      "step": 17449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.657343277229412e-07,
+      "loss": 0.9105,
+      "step": 17450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.651087808904067e-07,
+      "loss": 0.8913,
+      "step": 17451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.644834795107325e-07,
+      "loss": 0.8626,
+      "step": 17452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.63858423600542e-07,
+      "loss": 0.8819,
+      "step": 17453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.63233613176445e-07,
+      "loss": 0.823,
+      "step": 17454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.626090482550485e-07,
+      "loss": 0.945,
+      "step": 17455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.619847288529547e-07,
+      "loss": 0.8605,
+      "step": 17456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.613606549867558e-07,
+      "loss": 0.8639,
+      "step": 17457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.607368266730375e-07,
+      "loss": 0.4012,
+      "step": 17458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.601132439283842e-07,
+      "loss": 0.9196,
+      "step": 17459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.594899067693673e-07,
+      "loss": 0.8506,
+      "step": 17460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.588668152125544e-07,
+      "loss": 0.8736,
+      "step": 17461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.582439692745069e-07,
+      "loss": 0.8239,
+      "step": 17462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.576213689717815e-07,
+      "loss": 0.9097,
+      "step": 17463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.569990143209205e-07,
+      "loss": 0.8715,
+      "step": 17464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.563769053384706e-07,
+      "loss": 0.9285,
+      "step": 17465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.557550420409676e-07,
+      "loss": 0.9296,
+      "step": 17466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.55133424444936e-07,
+      "loss": 0.9077,
+      "step": 17467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.545120525669003e-07,
+      "loss": 0.8534,
+      "step": 17468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.538909264233751e-07,
+      "loss": 0.7889,
+      "step": 17469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.532700460308673e-07,
+      "loss": 0.925,
+      "step": 17470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.526494114058813e-07,
+      "loss": 0.9198,
+      "step": 17471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.520290225649129e-07,
+      "loss": 0.8175,
+      "step": 17472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.514088795244523e-07,
+      "loss": 0.8362,
+      "step": 17473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.507889823009795e-07,
+      "loss": 0.8762,
+      "step": 17474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.501693309109714e-07,
+      "loss": 0.8991,
+      "step": 17475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.495499253708982e-07,
+      "loss": 0.9086,
+      "step": 17476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.489307656972222e-07,
+      "loss": 0.9106,
+      "step": 17477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.483118519064014e-07,
+      "loss": 0.897,
+      "step": 17478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.476931840148871e-07,
+      "loss": 0.8369,
+      "step": 17479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.470747620391194e-07,
+      "loss": 0.7991,
+      "step": 17480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.464565859955364e-07,
+      "loss": 0.8235,
+      "step": 17481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.458386559005704e-07,
+      "loss": 0.915,
+      "step": 17482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.452209717706416e-07,
+      "loss": 0.9676,
+      "step": 17483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.44603533622169e-07,
+      "loss": 0.8635,
+      "step": 17484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.439863414715654e-07,
+      "loss": 0.9611,
+      "step": 17485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.433693953352339e-07,
+      "loss": 0.9407,
+      "step": 17486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.427526952295683e-07,
+      "loss": 0.8808,
+      "step": 17487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.421362411709676e-07,
+      "loss": 0.9347,
+      "step": 17488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.415200331758099e-07,
+      "loss": 0.8236,
+      "step": 17489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.409040712604754e-07,
+      "loss": 0.8765,
+      "step": 17490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.402883554413365e-07,
+      "loss": 0.8592,
+      "step": 17491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.39672885734758e-07,
+      "loss": 0.8148,
+      "step": 17492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.390576621570967e-07,
+      "loss": 0.8674,
+      "step": 17493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.384426847247062e-07,
+      "loss": 0.9264,
+      "step": 17494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.378279534539323e-07,
+      "loss": 0.8866,
+      "step": 17495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.372134683611121e-07,
+      "loss": 0.9023,
+      "step": 17496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.36599229462579e-07,
+      "loss": 0.8965,
+      "step": 17497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.359852367746578e-07,
+      "loss": 0.7763,
+      "step": 17498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.353714903136688e-07,
+      "loss": 0.4005,
+      "step": 17499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.347579900959234e-07,
+      "loss": 0.8654,
+      "step": 17500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.341447361377296e-07,
+      "loss": 0.9658,
+      "step": 17501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.335317284553877e-07,
+      "loss": 0.8995,
+      "step": 17502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.329189670651882e-07,
+      "loss": 0.9333,
+      "step": 17503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.323064519834178e-07,
+      "loss": 0.8888,
+      "step": 17504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.316941832263591e-07,
+      "loss": 0.8733,
+      "step": 17505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.310821608102814e-07,
+      "loss": 0.7874,
+      "step": 17506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.304703847514538e-07,
+      "loss": 0.8026,
+      "step": 17507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.29858855066139e-07,
+      "loss": 0.8683,
+      "step": 17508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.29247571770586e-07,
+      "loss": 0.8887,
+      "step": 17509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.286365348810443e-07,
+      "loss": 0.8762,
+      "step": 17510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.280257444137551e-07,
+      "loss": 0.3952,
+      "step": 17511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.274152003849522e-07,
+      "loss": 0.8596,
+      "step": 17512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.268049028108614e-07,
+      "loss": 0.9123,
+      "step": 17513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.261948517077066e-07,
+      "loss": 0.3599,
+      "step": 17514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.255850470917025e-07,
+      "loss": 0.8382,
+      "step": 17515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.249754889790539e-07,
+      "loss": 0.9088,
+      "step": 17516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.243661773859634e-07,
+      "loss": 0.8097,
+      "step": 17517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.23757112328628e-07,
+      "loss": 0.8578,
+      "step": 17518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.231482938232315e-07,
+      "loss": 0.9122,
+      "step": 17519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.225397218859598e-07,
+      "loss": 0.9167,
+      "step": 17520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.219313965329855e-07,
+      "loss": 0.8728,
+      "step": 17521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.21323317780479e-07,
+      "loss": 0.8896,
+      "step": 17522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.207154856445997e-07,
+      "loss": 0.8921,
+      "step": 17523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.201079001415068e-07,
+      "loss": 0.7816,
+      "step": 17524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.195005612873463e-07,
+      "loss": 0.8732,
+      "step": 17525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.188934690982619e-07,
+      "loss": 0.8719,
+      "step": 17526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.182866235903885e-07,
+      "loss": 0.7917,
+      "step": 17527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.176800247798576e-07,
+      "loss": 0.9033,
+      "step": 17528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.170736726827887e-07,
+      "loss": 0.8555,
+      "step": 17529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.164675673152987e-07,
+      "loss": 0.8724,
+      "step": 17530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.158617086935005e-07,
+      "loss": 0.7591,
+      "step": 17531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.152560968334932e-07,
+      "loss": 0.9108,
+      "step": 17532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.14650731751374e-07,
+      "loss": 0.8948,
+      "step": 17533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.140456134632335e-07,
+      "loss": 0.4143,
+      "step": 17534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.134407419851552e-07,
+      "loss": 0.8619,
+      "step": 17535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.128361173332155e-07,
+      "loss": 0.9333,
+      "step": 17536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.122317395234868e-07,
+      "loss": 0.9115,
+      "step": 17537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.116276085720286e-07,
+      "loss": 0.8057,
+      "step": 17538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.110237244949003e-07,
+      "loss": 0.8972,
+      "step": 17539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.104200873081535e-07,
+      "loss": 0.8529,
+      "step": 17540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.09816697027832e-07,
+      "loss": 0.3409,
+      "step": 17541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.092135536699707e-07,
+      "loss": 0.3808,
+      "step": 17542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.086106572506013e-07,
+      "loss": 0.9091,
+      "step": 17543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.080080077857509e-07,
+      "loss": 0.9016,
+      "step": 17544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.074056052914335e-07,
+      "loss": 0.8531,
+      "step": 17545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.068034497836607e-07,
+      "loss": 0.7587,
+      "step": 17546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.062015412784417e-07,
+      "loss": 0.899,
+      "step": 17547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.055998797917696e-07,
+      "loss": 0.8521,
+      "step": 17548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.049984653396369e-07,
+      "loss": 0.8251,
+      "step": 17549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.043972979380298e-07,
+      "loss": 0.8454,
+      "step": 17550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.037963776029267e-07,
+      "loss": 0.7992,
+      "step": 17551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.03195704350298e-07,
+      "loss": 0.9109,
+      "step": 17552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.025952781961087e-07,
+      "loss": 0.8892,
+      "step": 17553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.019950991563207e-07,
+      "loss": 0.886,
+      "step": 17554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.013951672468811e-07,
+      "loss": 0.8229,
+      "step": 17555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.007954824837382e-07,
+      "loss": 0.8808,
+      "step": 17556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.001960448828326e-07,
+      "loss": 0.8722,
+      "step": 17557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.995968544600906e-07,
+      "loss": 0.86,
+      "step": 17558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.989979112314449e-07,
+      "loss": 0.9167,
+      "step": 17559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.983992152128128e-07,
+      "loss": 0.8988,
+      "step": 17560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.978007664201048e-07,
+      "loss": 0.8392,
+      "step": 17561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.972025648692282e-07,
+      "loss": 0.8995,
+      "step": 17562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.966046105760826e-07,
+      "loss": 0.8486,
+      "step": 17563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.96006903556563e-07,
+      "loss": 0.7179,
+      "step": 17564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.954094438265513e-07,
+      "loss": 0.8656,
+      "step": 17565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.948122314019301e-07,
+      "loss": 0.8516,
+      "step": 17566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.942152662985746e-07,
+      "loss": 0.8371,
+      "step": 17567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.936185485323465e-07,
+      "loss": 0.8494,
+      "step": 17568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.930220781191088e-07,
+      "loss": 0.874,
+      "step": 17569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.924258550747154e-07,
+      "loss": 0.8606,
+      "step": 17570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.918298794150114e-07,
+      "loss": 0.8817,
+      "step": 17571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.912341511558384e-07,
+      "loss": 0.8404,
+      "step": 17572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.906386703130319e-07,
+      "loss": 0.3525,
+      "step": 17573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.900434369024156e-07,
+      "loss": 0.8476,
+      "step": 17574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.894484509398125e-07,
+      "loss": 0.8094,
+      "step": 17575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.888537124410344e-07,
+      "loss": 0.8141,
+      "step": 17576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.882592214218931e-07,
+      "loss": 0.879,
+      "step": 17577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.876649778981847e-07,
+      "loss": 0.8088,
+      "step": 17578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.870709818857058e-07,
+      "loss": 0.8418,
+      "step": 17579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.864772334002446e-07,
+      "loss": 0.8404,
+      "step": 17580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.858837324575807e-07,
+      "loss": 0.8515,
+      "step": 17581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.852904790734871e-07,
+      "loss": 0.8519,
+      "step": 17582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.846974732637379e-07,
+      "loss": 0.8366,
+      "step": 17583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.841047150440883e-07,
+      "loss": 0.7864,
+      "step": 17584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.835122044302967e-07,
+      "loss": 0.7897,
+      "step": 17585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.829199414381116e-07,
+      "loss": 0.8383,
+      "step": 17586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.823279260832716e-07,
+      "loss": 0.7415,
+      "step": 17587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.81736158381513e-07,
+      "loss": 0.8745,
+      "step": 17588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.811446383485653e-07,
+      "loss": 0.8306,
+      "step": 17589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.805533660001517e-07,
+      "loss": 0.9335,
+      "step": 17590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.79962341351984e-07,
+      "loss": 0.7986,
+      "step": 17591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.793715644197729e-07,
+      "loss": 0.8812,
+      "step": 17592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.787810352192226e-07,
+      "loss": 0.9356,
+      "step": 17593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.781907537660227e-07,
+      "loss": 0.7987,
+      "step": 17594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.776007200758672e-07,
+      "loss": 0.8881,
+      "step": 17595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.770109341644393e-07,
+      "loss": 0.832,
+      "step": 17596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.764213960474119e-07,
+      "loss": 0.9085,
+      "step": 17597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.758321057404549e-07,
+      "loss": 0.8923,
+      "step": 17598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.752430632592311e-07,
+      "loss": 0.896,
+      "step": 17599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.74654268619398e-07,
+      "loss": 0.904,
+      "step": 17600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.740657218366031e-07,
+      "loss": 0.8485,
+      "step": 17601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.734774229264895e-07,
+      "loss": 0.9262,
+      "step": 17602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.728893719046958e-07,
+      "loss": 0.8922,
+      "step": 17603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.723015687868484e-07,
+      "loss": 0.8431,
+      "step": 17604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.717140135885714e-07,
+      "loss": 0.82,
+      "step": 17605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.711267063254845e-07,
+      "loss": 0.844,
+      "step": 17606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.705396470131931e-07,
+      "loss": 0.8743,
+      "step": 17607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.699528356673025e-07,
+      "loss": 0.8235,
+      "step": 17608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.693662723034122e-07,
+      "loss": 0.8924,
+      "step": 17609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.687799569371079e-07,
+      "loss": 0.8923,
+      "step": 17610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.681938895839746e-07,
+      "loss": 0.8535,
+      "step": 17611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.6760807025959e-07,
+      "loss": 0.8796,
+      "step": 17612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.67022498979526e-07,
+      "loss": 0.895,
+      "step": 17613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.664371757593424e-07,
+      "loss": 0.8388,
+      "step": 17614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.65852100614599e-07,
+      "loss": 0.7978,
+      "step": 17615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.65267273560849e-07,
+      "loss": 0.8905,
+      "step": 17616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.646826946136286e-07,
+      "loss": 0.8528,
+      "step": 17617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.640983637884823e-07,
+      "loss": 0.841,
+      "step": 17618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.635142811009409e-07,
+      "loss": 0.9369,
+      "step": 17619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.629304465665254e-07,
+      "loss": 0.8727,
+      "step": 17620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.623468602007544e-07,
+      "loss": 0.8207,
+      "step": 17621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.617635220191399e-07,
+      "loss": 0.8791,
+      "step": 17622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.61180432037185e-07,
+      "loss": 0.4014,
+      "step": 17623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.605975902703887e-07,
+      "loss": 0.9406,
+      "step": 17624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.600149967342407e-07,
+      "loss": 0.4414,
+      "step": 17625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.594326514442296e-07,
+      "loss": 0.8026,
+      "step": 17626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.588505544158275e-07,
+      "loss": 0.8977,
+      "step": 17627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.582687056645098e-07,
+      "loss": 0.9155,
+      "step": 17628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.57687105205741e-07,
+      "loss": 0.8615,
+      "step": 17629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.571057530549796e-07,
+      "loss": 0.8523,
+      "step": 17630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.565246492276756e-07,
+      "loss": 0.8443,
+      "step": 17631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.559437937392766e-07,
+      "loss": 0.9599,
+      "step": 17632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.55363186605219e-07,
+      "loss": 0.7914,
+      "step": 17633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.54782827840934e-07,
+      "loss": 0.8647,
+      "step": 17634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.542027174618504e-07,
+      "loss": 0.886,
+      "step": 17635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.536228554833835e-07,
+      "loss": 0.8578,
+      "step": 17636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.530432419209465e-07,
+      "loss": 0.9308,
+      "step": 17637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.524638767899438e-07,
+      "loss": 0.8678,
+      "step": 17638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.518847601057787e-07,
+      "loss": 0.7385,
+      "step": 17639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.513058918838366e-07,
+      "loss": 0.4122,
+      "step": 17640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.507272721395063e-07,
+      "loss": 0.8777,
+      "step": 17641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.5014890088817e-07,
+      "loss": 0.8876,
+      "step": 17642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.495707781451966e-07,
+      "loss": 0.8579,
+      "step": 17643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.489929039259513e-07,
+      "loss": 0.8991,
+      "step": 17644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.484152782457976e-07,
+      "loss": 0.8755,
+      "step": 17645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.47837901120083e-07,
+      "loss": 0.8341,
+      "step": 17646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.472607725641555e-07,
+      "loss": 0.8577,
+      "step": 17647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.466838925933549e-07,
+      "loss": 0.9296,
+      "step": 17648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.461072612230158e-07,
+      "loss": 0.8728,
+      "step": 17649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.455308784684611e-07,
+      "loss": 0.9401,
+      "step": 17650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.449547443450121e-07,
+      "loss": 0.8455,
+      "step": 17651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.443788588679823e-07,
+      "loss": 0.8668,
+      "step": 17652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.438032220526747e-07,
+      "loss": 0.8735,
+      "step": 17653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.432278339143927e-07,
+      "loss": 0.8673,
+      "step": 17654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.426526944684309e-07,
+      "loss": 0.376,
+      "step": 17655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.420778037300713e-07,
+      "loss": 0.8075,
+      "step": 17656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.415031617145951e-07,
+      "loss": 0.8889,
+      "step": 17657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.409287684372789e-07,
+      "loss": 0.8689,
+      "step": 17658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.403546239133851e-07,
+      "loss": 0.8479,
+      "step": 17659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.397807281581759e-07,
+      "loss": 0.8806,
+      "step": 17660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.392070811869044e-07,
+      "loss": 0.9397,
+      "step": 17661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.386336830148187e-07,
+      "loss": 0.855,
+      "step": 17662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.380605336571577e-07,
+      "loss": 0.8223,
+      "step": 17663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.374876331291546e-07,
+      "loss": 0.9046,
+      "step": 17664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.369149814460374e-07,
+      "loss": 0.8457,
+      "step": 17665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.36342578623026e-07,
+      "loss": 0.8868,
+      "step": 17666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.357704246753337e-07,
+      "loss": 0.9349,
+      "step": 17667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.351985196181721e-07,
+      "loss": 0.897,
+      "step": 17668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.346268634667352e-07,
+      "loss": 0.9317,
+      "step": 17669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.3405545623622e-07,
+      "loss": 0.6973,
+      "step": 17670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.334842979418165e-07,
+      "loss": 0.9156,
+      "step": 17671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.329133885987004e-07,
+      "loss": 0.8973,
+      "step": 17672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.323427282220485e-07,
+      "loss": 0.8835,
+      "step": 17673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.317723168270273e-07,
+      "loss": 0.8476,
+      "step": 17674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.312021544288005e-07,
+      "loss": 0.85,
+      "step": 17675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.306322410425169e-07,
+      "loss": 0.8683,
+      "step": 17676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.300625766833312e-07,
+      "loss": 0.9016,
+      "step": 17677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.294931613663791e-07,
+      "loss": 0.8868,
+      "step": 17678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.289239951067961e-07,
+      "loss": 0.3848,
+      "step": 17679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.2835507791971e-07,
+      "loss": 0.7547,
+      "step": 17680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.277864098202457e-07,
+      "loss": 0.9713,
+      "step": 17681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.27217990823512e-07,
+      "loss": 0.9259,
+      "step": 17682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.266498209446192e-07,
+      "loss": 0.9322,
+      "step": 17683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.260819001986706e-07,
+      "loss": 0.8548,
+      "step": 17684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.255142286007576e-07,
+      "loss": 0.8228,
+      "step": 17685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.249468061659691e-07,
+      "loss": 0.9025,
+      "step": 17686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.243796329093876e-07,
+      "loss": 1.0357,
+      "step": 17687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.238127088460877e-07,
+      "loss": 0.868,
+      "step": 17688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.232460339911372e-07,
+      "loss": 0.8679,
+      "step": 17689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.226796083595976e-07,
+      "loss": 0.8721,
+      "step": 17690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.221134319665256e-07,
+      "loss": 0.9106,
+      "step": 17691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.21547504826967e-07,
+      "loss": 0.9883,
+      "step": 17692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.209818269559642e-07,
+      "loss": 0.8995,
+      "step": 17693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.204163983685541e-07,
+      "loss": 0.8617,
+      "step": 17694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.198512190797612e-07,
+      "loss": 0.8955,
+      "step": 17695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.192862891046114e-07,
+      "loss": 0.9767,
+      "step": 17696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.187216084581172e-07,
+      "loss": 0.7992,
+      "step": 17697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.181571771552897e-07,
+      "loss": 0.7924,
+      "step": 17698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.175929952111282e-07,
+      "loss": 0.764,
+      "step": 17699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.170290626406295e-07,
+      "loss": 0.9949,
+      "step": 17700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.164653794587805e-07,
+      "loss": 0.8988,
+      "step": 17701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.15901945680566e-07,
+      "loss": 0.8886,
+      "step": 17702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.153387613209594e-07,
+      "loss": 0.8962,
+      "step": 17703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.147758263949322e-07,
+      "loss": 0.9344,
+      "step": 17704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.142131409174434e-07,
+      "loss": 0.9445,
+      "step": 17705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.13650704903449e-07,
+      "loss": 0.8942,
+      "step": 17706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.130885183679003e-07,
+      "loss": 0.8246,
+      "step": 17707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.125265813257364e-07,
+      "loss": 0.9402,
+      "step": 17708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.119648937918954e-07,
+      "loss": 0.8526,
+      "step": 17709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.114034557813043e-07,
+      "loss": 0.8675,
+      "step": 17710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.108422673088877e-07,
+      "loss": 0.8492,
+      "step": 17711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.102813283895581e-07,
+      "loss": 0.8654,
+      "step": 17712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.097206390382293e-07,
+      "loss": 0.8793,
+      "step": 17713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.091601992698004e-07,
+      "loss": 0.8405,
+      "step": 17714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.086000090991672e-07,
+      "loss": 0.7996,
+      "step": 17715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.080400685412202e-07,
+      "loss": 0.8884,
+      "step": 17716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.074803776108418e-07,
+      "loss": 0.8583,
+      "step": 17717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.069209363229078e-07,
+      "loss": 0.8822,
+      "step": 17718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.063617446922864e-07,
+      "loss": 0.8252,
+      "step": 17719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.058028027338436e-07,
+      "loss": 0.8872,
+      "step": 17720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.052441104624307e-07,
+      "loss": 0.9467,
+      "step": 17721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.046856678929003e-07,
+      "loss": 0.8648,
+      "step": 17722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.041274750400939e-07,
+      "loss": 0.8218,
+      "step": 17723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.035695319188484e-07,
+      "loss": 0.8815,
+      "step": 17724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.03011838543992e-07,
+      "loss": 0.9337,
+      "step": 17725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.024543949303507e-07,
+      "loss": 0.835,
+      "step": 17726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.018972010927382e-07,
+      "loss": 0.8928,
+      "step": 17727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.013402570459625e-07,
+      "loss": 0.8699,
+      "step": 17728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.007835628048298e-07,
+      "loss": 0.9123,
+      "step": 17729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.002271183841357e-07,
+      "loss": 0.92,
+      "step": 17730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.996709237986687e-07,
+      "loss": 0.9032,
+      "step": 17731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.991149790632123e-07,
+      "loss": 0.847,
+      "step": 17732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.985592841925436e-07,
+      "loss": 0.9292,
+      "step": 17733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.980038392014309e-07,
+      "loss": 0.9102,
+      "step": 17734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.974486441046356e-07,
+      "loss": 0.8815,
+      "step": 17735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.968936989169205e-07,
+      "loss": 0.9389,
+      "step": 17736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.963390036530292e-07,
+      "loss": 0.9431,
+      "step": 17737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.957845583277078e-07,
+      "loss": 0.8731,
+      "step": 17738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.952303629556922e-07,
+      "loss": 0.9038,
+      "step": 17739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.946764175517128e-07,
+      "loss": 0.8342,
+      "step": 17740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.941227221304924e-07,
+      "loss": 0.9572,
+      "step": 17741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.935692767067469e-07,
+      "loss": 0.8508,
+      "step": 17742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.93016081295188e-07,
+      "loss": 0.7942,
+      "step": 17743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.92463135910516e-07,
+      "loss": 0.8293,
+      "step": 17744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.919104405674303e-07,
+      "loss": 0.9169,
+      "step": 17745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.913579952806192e-07,
+      "loss": 0.8057,
+      "step": 17746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.908058000647676e-07,
+      "loss": 0.9393,
+      "step": 17747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.902538549345516e-07,
+      "loss": 0.8673,
+      "step": 17748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.897021599046415e-07,
+      "loss": 0.7863,
+      "step": 17749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.891507149897002e-07,
+      "loss": 0.8262,
+      "step": 17750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.885995202043848e-07,
+      "loss": 0.8871,
+      "step": 17751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.880485755633447e-07,
+      "loss": 0.9027,
+      "step": 17752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.874978810812249e-07,
+      "loss": 0.8788,
+      "step": 17753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.869474367726602e-07,
+      "loss": 0.9467,
+      "step": 17754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.863972426522823e-07,
+      "loss": 0.8412,
+      "step": 17755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.858472987347152e-07,
+      "loss": 0.8577,
+      "step": 17756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.852976050345749e-07,
+      "loss": 0.8198,
+      "step": 17757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.847481615664696e-07,
+      "loss": 0.8661,
+      "step": 17758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.841989683450067e-07,
+      "loss": 0.8815,
+      "step": 17759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.836500253847799e-07,
+      "loss": 0.8586,
+      "step": 17760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.83101332700382e-07,
+      "loss": 0.8367,
+      "step": 17761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.825528903063971e-07,
+      "loss": 0.9327,
+      "step": 17762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.820046982173977e-07,
+      "loss": 0.9143,
+      "step": 17763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.814567564479578e-07,
+      "loss": 0.8797,
+      "step": 17764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.809090650126414e-07,
+      "loss": 0.9223,
+      "step": 17765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.803616239260046e-07,
+      "loss": 0.8047,
+      "step": 17766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.798144332025968e-07,
+      "loss": 0.8546,
+      "step": 17767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.792674928569619e-07,
+      "loss": 0.4159,
+      "step": 17768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.787208029036396e-07,
+      "loss": 0.8208,
+      "step": 17769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.781743633571557e-07,
+      "loss": 0.8442,
+      "step": 17770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.776281742320366e-07,
+      "loss": 0.8478,
+      "step": 17771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.770822355428007e-07,
+      "loss": 0.9125,
+      "step": 17772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.765365473039563e-07,
+      "loss": 0.8019,
+      "step": 17773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.759911095300075e-07,
+      "loss": 0.8872,
+      "step": 17774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.754459222354536e-07,
+      "loss": 0.7932,
+      "step": 17775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.74900985434782e-07,
+      "loss": 0.8164,
+      "step": 17776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.743562991424767e-07,
+      "loss": 0.871,
+      "step": 17777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.738118633730161e-07,
+      "loss": 0.9141,
+      "step": 17778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.732676781408731e-07,
+      "loss": 0.9469,
+      "step": 17779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.72723743460506e-07,
+      "loss": 0.8107,
+      "step": 17780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.721800593463756e-07,
+      "loss": 0.771,
+      "step": 17781
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.716366258129313e-07,
+      "loss": 0.8448,
+      "step": 17782
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.710934428746173e-07,
+      "loss": 0.8701,
+      "step": 17783
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.705505105458708e-07,
+      "loss": 0.8525,
+      "step": 17784
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.700078288411237e-07,
+      "loss": 0.9242,
+      "step": 17785
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.694653977747966e-07,
+      "loss": 0.8974,
+      "step": 17786
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.689232173613091e-07,
+      "loss": 0.869,
+      "step": 17787
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.683812876150708e-07,
+      "loss": 0.4297,
+      "step": 17788
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.678396085504878e-07,
+      "loss": 0.7623,
+      "step": 17789
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.672981801819533e-07,
+      "loss": 0.9171,
+      "step": 17790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.667570025238611e-07,
+      "loss": 0.8529,
+      "step": 17791
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.662160755905943e-07,
+      "loss": 0.8361,
+      "step": 17792
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.656753993965291e-07,
+      "loss": 0.885,
+      "step": 17793
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.651349739560352e-07,
+      "loss": 0.8759,
+      "step": 17794
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.6459479928348e-07,
+      "loss": 0.8631,
+      "step": 17795
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.640548753932184e-07,
+      "loss": 0.8792,
+      "step": 17796
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.635152022996016e-07,
+      "loss": 0.8831,
+      "step": 17797
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.629757800169732e-07,
+      "loss": 0.8711,
+      "step": 17798
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.6243660855967e-07,
+      "loss": 0.8578,
+      "step": 17799
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.618976879420235e-07,
+      "loss": 0.9305,
+      "step": 17800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.613590181783557e-07,
+      "loss": 0.8962,
+      "step": 17801
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.608205992829874e-07,
+      "loss": 0.8773,
+      "step": 17802
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.602824312702259e-07,
+      "loss": 0.8173,
+      "step": 17803
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.597445141543756e-07,
+      "loss": 0.8242,
+      "step": 17804
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.59206847949737e-07,
+      "loss": 0.8621,
+      "step": 17805
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.586694326705932e-07,
+      "loss": 0.855,
+      "step": 17806
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.581322683312362e-07,
+      "loss": 0.8485,
+      "step": 17807
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.575953549459401e-07,
+      "loss": 0.9384,
+      "step": 17808
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.570586925289734e-07,
+      "loss": 0.9276,
+      "step": 17809
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.565222810946025e-07,
+      "loss": 0.8471,
+      "step": 17810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.55986120657086e-07,
+      "loss": 0.865,
+      "step": 17811
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.554502112306692e-07,
+      "loss": 0.8557,
+      "step": 17812
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.549145528296007e-07,
+      "loss": 0.8252,
+      "step": 17813
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.543791454681147e-07,
+      "loss": 0.8551,
+      "step": 17814
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.538439891604452e-07,
+      "loss": 0.9033,
+      "step": 17815
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.533090839208133e-07,
+      "loss": 0.8187,
+      "step": 17816
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.527744297634352e-07,
+      "loss": 0.851,
+      "step": 17817
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.522400267025241e-07,
+      "loss": 0.8908,
+      "step": 17818
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.517058747522819e-07,
+      "loss": 0.8598,
+      "step": 17819
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.511719739269073e-07,
+      "loss": 0.879,
+      "step": 17820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.506383242405911e-07,
+      "loss": 0.9251,
+      "step": 17821
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.501049257075153e-07,
+      "loss": 0.905,
+      "step": 17822
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.495717783418564e-07,
+      "loss": 0.8658,
+      "step": 17823
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.490388821577885e-07,
+      "loss": 0.8644,
+      "step": 17824
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.485062371694716e-07,
+      "loss": 0.9093,
+      "step": 17825
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.479738433910642e-07,
+      "loss": 0.8511,
+      "step": 17826
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.474417008367173e-07,
+      "loss": 0.9573,
+      "step": 17827
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.469098095205749e-07,
+      "loss": 0.9162,
+      "step": 17828
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.463781694567716e-07,
+      "loss": 0.9125,
+      "step": 17829
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.458467806594392e-07,
+      "loss": 0.8653,
+      "step": 17830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.453156431427031e-07,
+      "loss": 0.818,
+      "step": 17831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.447847569206788e-07,
+      "loss": 0.9039,
+      "step": 17832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.44254122007476e-07,
+      "loss": 0.8997,
+      "step": 17833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.437237384172012e-07,
+      "loss": 0.9441,
+      "step": 17834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.431936061639476e-07,
+      "loss": 0.8454,
+      "step": 17835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.426637252618072e-07,
+      "loss": 0.919,
+      "step": 17836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.421340957248634e-07,
+      "loss": 0.8808,
+      "step": 17837
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.416047175671945e-07,
+      "loss": 0.8264,
+      "step": 17838
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.410755908028698e-07,
+      "loss": 0.8697,
+      "step": 17839
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.40546715445951e-07,
+      "loss": 0.9892,
+      "step": 17840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.40018091510498e-07,
+      "loss": 0.9434,
+      "step": 17841
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.394897190105597e-07,
+      "loss": 0.8812,
+      "step": 17842
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.389615979601793e-07,
+      "loss": 0.8927,
+      "step": 17843
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.384337283733954e-07,
+      "loss": 0.8434,
+      "step": 17844
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.379061102642357e-07,
+      "loss": 0.9389,
+      "step": 17845
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.373787436467248e-07,
+      "loss": 0.8302,
+      "step": 17846
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.368516285348802e-07,
+      "loss": 0.8487,
+      "step": 17847
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.363247649427106e-07,
+      "loss": 0.8134,
+      "step": 17848
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.357981528842204e-07,
+      "loss": 0.9021,
+      "step": 17849
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.352717923734064e-07,
+      "loss": 0.9098,
+      "step": 17850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.347456834242593e-07,
+      "loss": 0.94,
+      "step": 17851
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.342198260507603e-07,
+      "loss": 0.8777,
+      "step": 17852
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.336942202668871e-07,
+      "loss": 0.8369,
+      "step": 17853
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.331688660866108e-07,
+      "loss": 0.882,
+      "step": 17854
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.326437635238935e-07,
+      "loss": 0.9258,
+      "step": 17855
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.321189125926918e-07,
+      "loss": 0.8742,
+      "step": 17856
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.315943133069579e-07,
+      "loss": 0.9012,
+      "step": 17857
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.310699656806317e-07,
+      "loss": 0.8427,
+      "step": 17858
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.305458697276511e-07,
+      "loss": 0.3826,
+      "step": 17859
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.300220254619471e-07,
+      "loss": 0.8401,
+      "step": 17860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.294984328974406e-07,
+      "loss": 0.8613,
+      "step": 17861
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.289750920480508e-07,
+      "loss": 0.8911,
+      "step": 17862
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.284520029276852e-07,
+      "loss": 0.9202,
+      "step": 17863
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.279291655502484e-07,
+      "loss": 0.9171,
+      "step": 17864
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.274065799296346e-07,
+      "loss": 0.9012,
+      "step": 17865
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.268842460797363e-07,
+      "loss": 0.8798,
+      "step": 17866
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.263621640144367e-07,
+      "loss": 0.9092,
+      "step": 17867
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.258403337476093e-07,
+      "loss": 0.7981,
+      "step": 17868
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253187552931249e-07,
+      "loss": 0.9141,
+      "step": 17869
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.247974286648494e-07,
+      "loss": 0.917,
+      "step": 17870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.242763538766338e-07,
+      "loss": 0.8349,
+      "step": 17871
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.237555309423315e-07,
+      "loss": 0.9115,
+      "step": 17872
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.232349598757846e-07,
+      "loss": 0.9069,
+      "step": 17873
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.22714640690829e-07,
+      "loss": 0.8858,
+      "step": 17874
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.221945734012935e-07,
+      "loss": 0.8731,
+      "step": 17875
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.216747580210013e-07,
+      "loss": 0.8051,
+      "step": 17876
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.211551945637683e-07,
+      "loss": 0.8089,
+      "step": 17877
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.206358830434055e-07,
+      "loss": 0.8866,
+      "step": 17878
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.201168234737131e-07,
+      "loss": 0.9312,
+      "step": 17879
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.19598015868491e-07,
+      "loss": 0.3438,
+      "step": 17880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.190794602415239e-07,
+      "loss": 0.8616,
+      "step": 17881
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.185611566065974e-07,
+      "loss": 0.9076,
+      "step": 17882
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.18043104977487e-07,
+      "loss": 0.8388,
+      "step": 17883
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.175253053679608e-07,
+      "loss": 0.9408,
+      "step": 17884
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.170077577917809e-07,
+      "loss": 0.8586,
+      "step": 17885
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.164904622627053e-07,
+      "loss": 0.9168,
+      "step": 17886
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.159734187944832e-07,
+      "loss": 0.8508,
+      "step": 17887
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.154566274008543e-07,
+      "loss": 0.2938,
+      "step": 17888
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.149400880955558e-07,
+      "loss": 0.8043,
+      "step": 17889
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.144238008923163e-07,
+      "loss": 0.3842,
+      "step": 17890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.139077658048586e-07,
+      "loss": 0.8257,
+      "step": 17891
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.133919828468992e-07,
+      "loss": 0.8939,
+      "step": 17892
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.128764520321472e-07,
+      "loss": 0.8384,
+      "step": 17893
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.123611733743017e-07,
+      "loss": 0.891,
+      "step": 17894
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.118461468870617e-07,
+      "loss": 0.839,
+      "step": 17895
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.113313725841152e-07,
+      "loss": 0.885,
+      "step": 17896
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.108168504791422e-07,
+      "loss": 0.8488,
+      "step": 17897
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.103025805858197e-07,
+      "loss": 0.8214,
+      "step": 17898
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.097885629178157e-07,
+      "loss": 0.8136,
+      "step": 17899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.092747974887957e-07,
+      "loss": 0.8305,
+      "step": 17900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.087612843124079e-07,
+      "loss": 0.849,
+      "step": 17901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.082480234023091e-07,
+      "loss": 0.8713,
+      "step": 17902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.077350147721349e-07,
+      "loss": 0.8625,
+      "step": 17903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.072222584355235e-07,
+      "loss": 0.8995,
+      "step": 17904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.067097544061017e-07,
+      "loss": 0.9561,
+      "step": 17905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.061975026974953e-07,
+      "loss": 0.8236,
+      "step": 17906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.056855033233132e-07,
+      "loss": 0.9038,
+      "step": 17907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.051737562971681e-07,
+      "loss": 0.8591,
+      "step": 17908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.046622616326613e-07,
+      "loss": 0.8543,
+      "step": 17909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.041510193433852e-07,
+      "loss": 0.9089,
+      "step": 17910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0364002944293e-07,
+      "loss": 0.8665,
+      "step": 17911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.031292919448772e-07,
+      "loss": 0.8908,
+      "step": 17912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.026188068628013e-07,
+      "loss": 0.8703,
+      "step": 17913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.021085742102705e-07,
+      "loss": 0.345,
+      "step": 17914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.015985940008461e-07,
+      "loss": 0.8579,
+      "step": 17915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.01088866248084e-07,
+      "loss": 0.969,
+      "step": 17916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.005793909655299e-07,
+      "loss": 0.8772,
+      "step": 17917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.000701681667253e-07,
+      "loss": 0.9301,
+      "step": 17918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.995611978652082e-07,
+      "loss": 0.9682,
+      "step": 17919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.990524800745023e-07,
+      "loss": 0.889,
+      "step": 17920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9854401480813e-07,
+      "loss": 0.8645,
+      "step": 17921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.980358020796072e-07,
+      "loss": 0.7904,
+      "step": 17922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.975278419024399e-07,
+      "loss": 0.806,
+      "step": 17923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.970201342901282e-07,
+      "loss": 0.3954,
+      "step": 17924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.965126792561703e-07,
+      "loss": 0.4168,
+      "step": 17925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.960054768140499e-07,
+      "loss": 0.8772,
+      "step": 17926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.954985269772494e-07,
+      "loss": 0.9746,
+      "step": 17927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.949918297592437e-07,
+      "loss": 0.8801,
+      "step": 17928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.944853851734999e-07,
+      "loss": 0.91,
+      "step": 17929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.93979193233477e-07,
+      "loss": 0.8624,
+      "step": 17930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.93473253952631e-07,
+      "loss": 0.8144,
+      "step": 17931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.92967567344409e-07,
+      "loss": 0.8279,
+      "step": 17932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.92462133422249e-07,
+      "loss": 0.7748,
+      "step": 17933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919569521995882e-07,
+      "loss": 0.9044,
+      "step": 17934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.914520236898513e-07,
+      "loss": 0.8536,
+      "step": 17935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.909473479064608e-07,
+      "loss": 0.8104,
+      "step": 17936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.904429248628284e-07,
+      "loss": 0.8715,
+      "step": 17937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.899387545723644e-07,
+      "loss": 0.86,
+      "step": 17938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.894348370484648e-07,
+      "loss": 0.8526,
+      "step": 17939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.889311723045254e-07,
+      "loss": 0.805,
+      "step": 17940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.884277603539323e-07,
+      "loss": 0.8055,
+      "step": 17941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.87924601210068e-07,
+      "loss": 0.8179,
+      "step": 17942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.874216948863019e-07,
+      "loss": 0.8586,
+      "step": 17943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.869190413960035e-07,
+      "loss": 0.8976,
+      "step": 17944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.864166407525328e-07,
+      "loss": 0.8935,
+      "step": 17945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.859144929692416e-07,
+      "loss": 0.9626,
+      "step": 17946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.854125980594771e-07,
+      "loss": 0.8028,
+      "step": 17947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.849109560365783e-07,
+      "loss": 0.8615,
+      "step": 17948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.844095669138794e-07,
+      "loss": 0.8343,
+      "step": 17949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.839084307047059e-07,
+      "loss": 0.9048,
+      "step": 17950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.834075474223798e-07,
+      "loss": 0.8959,
+      "step": 17951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.829069170802103e-07,
+      "loss": 0.8713,
+      "step": 17952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.824065396915057e-07,
+      "loss": 0.8269,
+      "step": 17953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.819064152695651e-07,
+      "loss": 0.8483,
+      "step": 17954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.814065438276827e-07,
+      "loss": 0.8806,
+      "step": 17955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.80906925379141e-07,
+      "loss": 0.7787,
+      "step": 17956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.804075599372227e-07,
+      "loss": 0.8431,
+      "step": 17957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.799084475151994e-07,
+      "loss": 0.9186,
+      "step": 17958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.794095881263361e-07,
+      "loss": 0.8901,
+      "step": 17959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.789109817838899e-07,
+      "loss": 0.8355,
+      "step": 17960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.784126285011192e-07,
+      "loss": 0.4146,
+      "step": 17961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.779145282912645e-07,
+      "loss": 0.9108,
+      "step": 17962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.774166811675662e-07,
+      "loss": 0.8455,
+      "step": 17963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.769190871432572e-07,
+      "loss": 0.9034,
+      "step": 17964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7642174623156346e-07,
+      "loss": 0.9078,
+      "step": 17965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.759246584457011e-07,
+      "loss": 0.8674,
+      "step": 17966
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7542782379888405e-07,
+      "loss": 0.8883,
+      "step": 17967
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7493124230431955e-07,
+      "loss": 0.8301,
+      "step": 17968
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7443491397520134e-07,
+      "loss": 0.8984,
+      "step": 17969
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.739388388247246e-07,
+      "loss": 0.8183,
+      "step": 17970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.734430168660731e-07,
+      "loss": 0.863,
+      "step": 17971
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7294744811242633e-07,
+      "loss": 0.8569,
+      "step": 17972
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.72452132576956e-07,
+      "loss": 0.8067,
+      "step": 17973
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.719570702728271e-07,
+      "loss": 0.9074,
+      "step": 17974
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.714622612131958e-07,
+      "loss": 0.8429,
+      "step": 17975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7096770541121494e-07,
+      "loss": 0.8156,
+      "step": 17976
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.704734028800295e-07,
+      "loss": 0.9158,
+      "step": 17977
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.6997935363277905e-07,
+      "loss": 0.8386,
+      "step": 17978
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6948555768259075e-07,
+      "loss": 0.9009,
+      "step": 17979
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6899201504259196e-07,
+      "loss": 0.8391,
+      "step": 17980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6849872572590107e-07,
+      "loss": 0.8817,
+      "step": 17981
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.680056897456264e-07,
+      "loss": 0.8437,
+      "step": 17982
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6751290711487317e-07,
+      "loss": 0.9197,
+      "step": 17983
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6702037784674305e-07,
+      "loss": 0.7281,
+      "step": 17984
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.665281019543211e-07,
+      "loss": 0.8629,
+      "step": 17985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.660360794506946e-07,
+      "loss": 0.834,
+      "step": 17986
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.65544310348941e-07,
+      "loss": 0.9024,
+      "step": 17987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.650527946621297e-07,
+      "loss": 0.9465,
+      "step": 17988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.645615324033237e-07,
+      "loss": 1.0361,
+      "step": 17989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6407052358558357e-07,
+      "loss": 0.9106,
+      "step": 17990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.635797682219578e-07,
+      "loss": 0.8777,
+      "step": 17991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6308926632548933e-07,
+      "loss": 0.8656,
+      "step": 17992
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6259901790921544e-07,
+      "loss": 0.8682,
+      "step": 17993
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.621090229861691e-07,
+      "loss": 0.8112,
+      "step": 17994
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6161928156936873e-07,
+      "loss": 0.8486,
+      "step": 17995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.611297936718351e-07,
+      "loss": 0.8056,
+      "step": 17996
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.606405593065788e-07,
+      "loss": 0.9259,
+      "step": 17997
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.601515784865995e-07,
+      "loss": 0.9082,
+      "step": 17998
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.596628512248957e-07,
+      "loss": 0.9053,
+      "step": 17999
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5917437753445926e-07,
+      "loss": 0.8769,
+      "step": 18000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5868615742826973e-07,
+      "loss": 0.8417,
+      "step": 18001
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.581981909193045e-07,
+      "loss": 0.8809,
+      "step": 18002
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5771047802053327e-07,
+      "loss": 0.8434,
+      "step": 18003
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.572230187449211e-07,
+      "loss": 0.8804,
+      "step": 18004
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5673581310542114e-07,
+      "loss": 0.8439,
+      "step": 18005
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.562488611149829e-07,
+      "loss": 0.8143,
+      "step": 18006
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.557621627865505e-07,
+      "loss": 0.8787,
+      "step": 18007
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5527571813305915e-07,
+      "loss": 0.8489,
+      "step": 18008
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.547895271674385e-07,
+      "loss": 0.9064,
+      "step": 18009
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.543035899026116e-07,
+      "loss": 0.8708,
+      "step": 18010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5381790635149135e-07,
+      "loss": 0.8154,
+      "step": 18011
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.533324765269886e-07,
+      "loss": 0.7711,
+      "step": 18012
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5284730044200533e-07,
+      "loss": 0.8422,
+      "step": 18013
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.523623781094377e-07,
+      "loss": 0.3682,
+      "step": 18014
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5187770954217334e-07,
+      "loss": 0.8517,
+      "step": 18015
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5139329475309414e-07,
+      "loss": 0.9125,
+      "step": 18016
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.509091337550764e-07,
+      "loss": 0.877,
+      "step": 18017
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.504252265609865e-07,
+      "loss": 0.7952,
+      "step": 18018
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4994157318368646e-07,
+      "loss": 0.8601,
+      "step": 18019
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.494581736360348e-07,
+      "loss": 0.911,
+      "step": 18020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4897502793087576e-07,
+      "loss": 0.9434,
+      "step": 18021
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4849213608105235e-07,
+      "loss": 0.8738,
+      "step": 18022
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.480094980993999e-07,
+      "loss": 0.8737,
+      "step": 18023
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.475271139987447e-07,
+      "loss": 0.8547,
+      "step": 18024
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4704498379190887e-07,
+      "loss": 0.8806,
+      "step": 18025
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.465631074917065e-07,
+      "loss": 0.364,
+      "step": 18026
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.460814851109474e-07,
+      "loss": 0.8176,
+      "step": 18027
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4560011666242907e-07,
+      "loss": 0.9638,
+      "step": 18028
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.45119002158948e-07,
+      "loss": 0.8449,
+      "step": 18029
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.446381416132906e-07,
+      "loss": 0.9417,
+      "step": 18030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4415753503823875e-07,
+      "loss": 0.8989,
+      "step": 18031
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4367718244656467e-07,
+      "loss": 0.7983,
+      "step": 18032
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.431970838510391e-07,
+      "loss": 0.8546,
+      "step": 18033
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4271723926441857e-07,
+      "loss": 0.8279,
+      "step": 18034
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.422376486994584e-07,
+      "loss": 0.8576,
+      "step": 18035
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.417583121689062e-07,
+      "loss": 0.8928,
+      "step": 18036
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4127922968550064e-07,
+      "loss": 0.8723,
+      "step": 18037
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4080040126197596e-07,
+      "loss": 0.9434,
+      "step": 18038
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4032182691105987e-07,
+      "loss": 0.9295,
+      "step": 18039
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3984350664547205e-07,
+      "loss": 0.8935,
+      "step": 18040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.393654404779235e-07,
+      "loss": 0.8618,
+      "step": 18041
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3888762842112186e-07,
+      "loss": 0.9172,
+      "step": 18042
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3841007048776806e-07,
+      "loss": 0.8842,
+      "step": 18043
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.379327666905542e-07,
+      "loss": 0.895,
+      "step": 18044
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.374557170421656e-07,
+      "loss": 0.8971,
+      "step": 18045
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3697892155528446e-07,
+      "loss": 0.9162,
+      "step": 18046
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.365023802425805e-07,
+      "loss": 0.9018,
+      "step": 18047
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.360260931167204e-07,
+      "loss": 0.866,
+      "step": 18048
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3555006019036505e-07,
+      "loss": 0.885,
+      "step": 18049
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.350742814761644e-07,
+      "loss": 0.85,
+      "step": 18050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.345987569867649e-07,
+      "loss": 0.8668,
+      "step": 18051
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3412348673480544e-07,
+      "loss": 0.8771,
+      "step": 18052
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.336484707329203e-07,
+      "loss": 0.8499,
+      "step": 18053
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3317370899373044e-07,
+      "loss": 0.8477,
+      "step": 18054
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.326992015298581e-07,
+      "loss": 0.8274,
+      "step": 18055
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.322249483539154e-07,
+      "loss": 0.8515,
+      "step": 18056
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3175094947850436e-07,
+      "loss": 0.7361,
+      "step": 18057
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3127720491622506e-07,
+      "loss": 0.9568,
+      "step": 18058
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.308037146796706e-07,
+      "loss": 0.9014,
+      "step": 18059
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.303304787814233e-07,
+      "loss": 0.8829,
+      "step": 18060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.29857497234063e-07,
+      "loss": 0.8344,
+      "step": 18061
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2938477005015853e-07,
+      "loss": 0.8822,
+      "step": 18062
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.289122972422788e-07,
+      "loss": 0.9287,
+      "step": 18063
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.28440078822977e-07,
+      "loss": 0.8391,
+      "step": 18064
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.279681148048065e-07,
+      "loss": 0.814,
+      "step": 18065
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2749640520031166e-07,
+      "loss": 0.8591,
+      "step": 18066
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2702495002202913e-07,
+      "loss": 0.9151,
+      "step": 18067
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2655374928249005e-07,
+      "loss": 0.9037,
+      "step": 18068
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.260828029942188e-07,
+      "loss": 0.8979,
+      "step": 18069
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.25612111169732e-07,
+      "loss": 0.9014,
+      "step": 18070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.251416738215397e-07,
+      "loss": 0.8916,
+      "step": 18071
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2467149096214743e-07,
+      "loss": 0.9309,
+      "step": 18072
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2420156260405075e-07,
+      "loss": 0.3765,
+      "step": 18073
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.237318887597386e-07,
+      "loss": 0.9402,
+      "step": 18074
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2326246944169645e-07,
+      "loss": 0.8696,
+      "step": 18075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.227933046624011e-07,
+      "loss": 0.8422,
+      "step": 18076
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2232439443432026e-07,
+      "loss": 0.8174,
+      "step": 18077
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2185573876991737e-07,
+      "loss": 0.891,
+      "step": 18078
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.213873376816513e-07,
+      "loss": 0.8247,
+      "step": 18079
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.209191911819688e-07,
+      "loss": 0.9215,
+      "step": 18080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.204512992833143e-07,
+      "loss": 0.878,
+      "step": 18081
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.199836619981246e-07,
+      "loss": 0.8746,
+      "step": 18082
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1951627933882635e-07,
+      "loss": 0.8197,
+      "step": 18083
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1904915131784406e-07,
+      "loss": 0.8102,
+      "step": 18084
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.185822779475934e-07,
+      "loss": 0.9391,
+      "step": 18085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1811565924048114e-07,
+      "loss": 0.9594,
+      "step": 18086
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1764929520891175e-07,
+      "loss": 0.8567,
+      "step": 18087
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.171831858652797e-07,
+      "loss": 0.814,
+      "step": 18088
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.167173312219752e-07,
+      "loss": 0.8091,
+      "step": 18089
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.162517312913772e-07,
+      "loss": 0.8619,
+      "step": 18090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1578638608586464e-07,
+      "loss": 0.803,
+      "step": 18091
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.153212956178021e-07,
+      "loss": 0.8686,
+      "step": 18092
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1485645989955303e-07,
+      "loss": 0.9312,
+      "step": 18093
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1439187894347197e-07,
+      "loss": 0.9006,
+      "step": 18094
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.13927552761908e-07,
+      "loss": 0.8765,
+      "step": 18095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1346348136720003e-07,
+      "loss": 0.7561,
+      "step": 18096
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1299966477168387e-07,
+      "loss": 0.8314,
+      "step": 18097
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1253610298768844e-07,
+      "loss": 0.8468,
+      "step": 18098
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.120727960275328e-07,
+      "loss": 0.8851,
+      "step": 18099
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1160974390353045e-07,
+      "loss": 0.8061,
+      "step": 18100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.111469466279916e-07,
+      "loss": 0.3955,
+      "step": 18101
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.106844042132141e-07,
+      "loss": 0.7878,
+      "step": 18102
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.102221166714937e-07,
+      "loss": 0.8265,
+      "step": 18103
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0976008401511723e-07,
+      "loss": 0.8877,
+      "step": 18104
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.09298306256366e-07,
+      "loss": 0.8902,
+      "step": 18105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0883678340751023e-07,
+      "loss": 0.8029,
+      "step": 18106
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.083755154808189e-07,
+      "loss": 0.9164,
+      "step": 18107
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.079145024885522e-07,
+      "loss": 0.8877,
+      "step": 18108
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.074537444429627e-07,
+      "loss": 0.8553,
+      "step": 18109
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.06993241356296e-07,
+      "loss": 0.9228,
+      "step": 18110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0653299324079355e-07,
+      "loss": 0.8254,
+      "step": 18111
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.060730001086888e-07,
+      "loss": 0.8684,
+      "step": 18112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.056132619722031e-07,
+      "loss": 0.4263,
+      "step": 18113
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.051537788435611e-07,
+      "loss": 0.9192,
+      "step": 18114
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.046945507349731e-07,
+      "loss": 0.8424,
+      "step": 18115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0423557765864485e-07,
+      "loss": 0.9117,
+      "step": 18116
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.037768596267755e-07,
+      "loss": 0.8236,
+      "step": 18117
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.033183966515575e-07,
+      "loss": 0.8682,
+      "step": 18118
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0286018874517664e-07,
+      "loss": 0.4624,
+      "step": 18119
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0240223591980987e-07,
+      "loss": 0.8463,
+      "step": 18120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.019445381876308e-07,
+      "loss": 0.8491,
+      "step": 18121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0148709556080413e-07,
+      "loss": 0.8622,
+      "step": 18122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0102990805148676e-07,
+      "loss": 0.4497,
+      "step": 18123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.005729756718313e-07,
+      "loss": 0.8976,
+      "step": 18124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0011629843398345e-07,
+      "loss": 0.9376,
+      "step": 18125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.996598763500803e-07,
+      "loss": 0.8556,
+      "step": 18126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.992037094322532e-07,
+      "loss": 0.8978,
+      "step": 18127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.987477976926257e-07,
+      "loss": 0.8904,
+      "step": 18128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9829214114331605e-07,
+      "loss": 0.8998,
+      "step": 18129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9783673979643554e-07,
+      "loss": 0.4699,
+      "step": 18130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.973815936640879e-07,
+      "loss": 0.9414,
+      "step": 18131
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.969267027583701e-07,
+      "loss": 0.8665,
+      "step": 18132
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9647206709137244e-07,
+      "loss": 0.4072,
+      "step": 18133
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.960176866751797e-07,
+      "loss": 0.8592,
+      "step": 18134
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9556356152186783e-07,
+      "loss": 0.8366,
+      "step": 18135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.95109691643506e-07,
+      "loss": 0.8564,
+      "step": 18136
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9465607705215904e-07,
+      "loss": 0.9074,
+      "step": 18137
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9420271775988393e-07,
+      "loss": 0.7854,
+      "step": 18138
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.937496137787289e-07,
+      "loss": 0.9896,
+      "step": 18139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9329676512073975e-07,
+      "loss": 0.9505,
+      "step": 18140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.928441717979492e-07,
+      "loss": 0.9136,
+      "step": 18141
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9239183382238753e-07,
+      "loss": 0.9167,
+      "step": 18142
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.919397512060785e-07,
+      "loss": 0.9101,
+      "step": 18143
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.914879239610392e-07,
+      "loss": 0.9409,
+      "step": 18144
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9103635209927437e-07,
+      "loss": 0.825,
+      "step": 18145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9058503563279004e-07,
+      "loss": 0.9343,
+      "step": 18146
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9013397457358104e-07,
+      "loss": 0.8348,
+      "step": 18147
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.896831689336322e-07,
+      "loss": 0.8003,
+      "step": 18148
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.892326187249318e-07,
+      "loss": 0.8673,
+      "step": 18149
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8878232395945124e-07,
+      "loss": 0.8374,
+      "step": 18150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8833228464915997e-07,
+      "loss": 0.8895,
+      "step": 18151
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.878825008060172e-07,
+      "loss": 0.8192,
+      "step": 18152
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.874329724419812e-07,
+      "loss": 0.8772,
+      "step": 18153
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8698369956899794e-07,
+      "loss": 0.4045,
+      "step": 18154
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.86534682199009e-07,
+      "loss": 0.8385,
+      "step": 18155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.860859203439471e-07,
+      "loss": 0.9194,
+      "step": 18156
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8563741401574374e-07,
+      "loss": 0.8644,
+      "step": 18157
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8518916322631604e-07,
+      "loss": 0.864,
+      "step": 18158
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.847411679875779e-07,
+      "loss": 0.8733,
+      "step": 18159
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8429342831143856e-07,
+      "loss": 0.9142,
+      "step": 18160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.838459442097986e-07,
+      "loss": 0.8592,
+      "step": 18161
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8339871569454955e-07,
+      "loss": 0.8862,
+      "step": 18162
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8295174277758084e-07,
+      "loss": 0.8549,
+      "step": 18163
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8250502547077076e-07,
+      "loss": 0.849,
+      "step": 18164
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8205856378599194e-07,
+      "loss": 0.8854,
+      "step": 18165
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8161235773511164e-07,
+      "loss": 0.8808,
+      "step": 18166
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8116640732999144e-07,
+      "loss": 0.8497,
+      "step": 18167
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8072071258248076e-07,
+      "loss": 0.8369,
+      "step": 18168
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8027527350442794e-07,
+      "loss": 0.9287,
+      "step": 18169
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7983009010767233e-07,
+      "loss": 0.7863,
+      "step": 18170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7938516240404455e-07,
+      "loss": 0.8924,
+      "step": 18171
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.789404904053695e-07,
+      "loss": 0.8904,
+      "step": 18172
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.784960741234711e-07,
+      "loss": 0.9281,
+      "step": 18173
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.780519135701566e-07,
+      "loss": 0.9115,
+      "step": 18174
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7760800875723314e-07,
+      "loss": 0.8583,
+      "step": 18175
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.771643596965002e-07,
+      "loss": 0.3607,
+      "step": 18176
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.767209663997473e-07,
+      "loss": 0.8954,
+      "step": 18177
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.762778288787594e-07,
+      "loss": 0.8716,
+      "step": 18178
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7583494714531596e-07,
+      "loss": 0.9094,
+      "step": 18179
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7539232121118874e-07,
+      "loss": 0.8994,
+      "step": 18180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.749499510881405e-07,
+      "loss": 0.8592,
+      "step": 18181
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7450783678792955e-07,
+      "loss": 0.8826,
+      "step": 18182
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7406597832230883e-07,
+      "loss": 0.8996,
+      "step": 18183
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7362437570301667e-07,
+      "loss": 0.8165,
+      "step": 18184
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.73183028941797e-07,
+      "loss": 0.9335,
+      "step": 18185
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7274193805037716e-07,
+      "loss": 0.8769,
+      "step": 18186
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.723011030404811e-07,
+      "loss": 0.8605,
+      "step": 18187
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.71860523923826e-07,
+      "loss": 0.8271,
+      "step": 18188
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.714202007121226e-07,
+      "loss": 0.8861,
+      "step": 18189
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.709801334170715e-07,
+      "loss": 0.9199,
+      "step": 18190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.705403220503723e-07,
+      "loss": 0.9367,
+      "step": 18191
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.701007666237122e-07,
+      "loss": 0.7965,
+      "step": 18192
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.696614671487775e-07,
+      "loss": 0.903,
+      "step": 18193
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.69222423637241e-07,
+      "loss": 0.8748,
+      "step": 18194
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.687836361007724e-07,
+      "loss": 0.8904,
+      "step": 18195
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6834510455103444e-07,
+      "loss": 0.7933,
+      "step": 18196
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6790682899968454e-07,
+      "loss": 0.8552,
+      "step": 18197
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6746880945836893e-07,
+      "loss": 0.4274,
+      "step": 18198
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6703104593873274e-07,
+      "loss": 0.4591,
+      "step": 18199
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6659353845240885e-07,
+      "loss": 0.9431,
+      "step": 18200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.661562870110258e-07,
+      "loss": 0.866,
+      "step": 18201
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6571929162620536e-07,
+      "loss": 0.9558,
+      "step": 18202
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.652825523095638e-07,
+      "loss": 0.8682,
+      "step": 18203
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6484606907270737e-07,
+      "loss": 0.9022,
+      "step": 18204
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6440984192723794e-07,
+      "loss": 0.8999,
+      "step": 18205
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6397387088475175e-07,
+      "loss": 0.9356,
+      "step": 18206
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6353815595683294e-07,
+      "loss": 0.9019,
+      "step": 18207
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.631026971550633e-07,
+      "loss": 0.8823,
+      "step": 18208
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6266749449101914e-07,
+      "loss": 0.8352,
+      "step": 18209
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.622325479762656e-07,
+      "loss": 0.9226,
+      "step": 18210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.617978576223635e-07,
+      "loss": 0.93,
+      "step": 18211
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6136342344086697e-07,
+      "loss": 0.8994,
+      "step": 18212
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6092924544332106e-07,
+      "loss": 0.8804,
+      "step": 18213
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6049532364126785e-07,
+      "loss": 0.8658,
+      "step": 18214
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6006165804623904e-07,
+      "loss": 0.8181,
+      "step": 18215
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.596282486697633e-07,
+      "loss": 0.9229,
+      "step": 18216
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5919509552335586e-07,
+      "loss": 0.8663,
+      "step": 18217
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.58762198618533e-07,
+      "loss": 0.9522,
+      "step": 18218
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.583295579667989e-07,
+      "loss": 0.9169,
+      "step": 18219
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5789717357965436e-07,
+      "loss": 0.8928,
+      "step": 18220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.574650454685902e-07,
+      "loss": 0.8892,
+      "step": 18221
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.570331736450927e-07,
+      "loss": 0.9122,
+      "step": 18222
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5660155812063945e-07,
+      "loss": 0.7618,
+      "step": 18223
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.561701989067035e-07,
+      "loss": 0.9271,
+      "step": 18224
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5573909601475e-07,
+      "loss": 0.8682,
+      "step": 18225
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.553082494562354e-07,
+      "loss": 0.8594,
+      "step": 18226
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5487765924261177e-07,
+      "loss": 0.8734,
+      "step": 18227
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.544473253853242e-07,
+      "loss": 0.8619,
+      "step": 18228
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.540172478958126e-07,
+      "loss": 0.8424,
+      "step": 18229
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5358742678550327e-07,
+      "loss": 0.887,
+      "step": 18230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.531578620658227e-07,
+      "loss": 0.8578,
+      "step": 18231
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.527285537481906e-07,
+      "loss": 0.8938,
+      "step": 18232
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5229950184401453e-07,
+      "loss": 0.7687,
+      "step": 18233
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5187070636469867e-07,
+      "loss": 0.8462,
+      "step": 18234
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.514421673216406e-07,
+      "loss": 0.9402,
+      "step": 18235
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.51013884726229e-07,
+      "loss": 0.9185,
+      "step": 18236
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5058585858984916e-07,
+      "loss": 0.8392,
+      "step": 18237
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.501580889238776e-07,
+      "loss": 0.888,
+      "step": 18238
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.497305757396807e-07,
+      "loss": 0.4036,
+      "step": 18239
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4930331904862393e-07,
+      "loss": 0.7502,
+      "step": 18240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.488763188620625e-07,
+      "loss": 0.9021,
+      "step": 18241
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4844957519134746e-07,
+      "loss": 0.9451,
+      "step": 18242
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.480230880478175e-07,
+      "loss": 0.8364,
+      "step": 18243
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4759685744281125e-07,
+      "loss": 0.9192,
+      "step": 18244
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4717088338765747e-07,
+      "loss": 0.873,
+      "step": 18245
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.467451658936749e-07,
+      "loss": 0.9133,
+      "step": 18246
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4631970497218226e-07,
+      "loss": 0.841,
+      "step": 18247
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4589450063448713e-07,
+      "loss": 0.9132,
+      "step": 18248
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.454695528918883e-07,
+      "loss": 0.7689,
+      "step": 18249
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.450448617556823e-07,
+      "loss": 0.8431,
+      "step": 18250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4462042723715783e-07,
+      "loss": 0.8708,
+      "step": 18251
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4419624934759476e-07,
+      "loss": 0.924,
+      "step": 18252
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4377232809826746e-07,
+      "loss": 0.8897,
+      "step": 18253
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.433486635004424e-07,
+      "loss": 0.4459,
+      "step": 18254
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.429252555653806e-07,
+      "loss": 0.7733,
+      "step": 18255
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.425021043043375e-07,
+      "loss": 0.8066,
+      "step": 18256
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4207920972855744e-07,
+      "loss": 0.9039,
+      "step": 18257
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4165657184928257e-07,
+      "loss": 0.7874,
+      "step": 18258
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.412341906777439e-07,
+      "loss": 0.8449,
+      "step": 18259
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.408120662251702e-07,
+      "loss": 0.8692,
+      "step": 18260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.403901985027802e-07,
+      "loss": 0.8604,
+      "step": 18261
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3996858752178506e-07,
+      "loss": 0.9289,
+      "step": 18262
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3954723329339243e-07,
+      "loss": 0.8295,
+      "step": 18263
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3912613582880115e-07,
+      "loss": 0.8701,
+      "step": 18264
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.387052951392045e-07,
+      "loss": 0.971,
+      "step": 18265
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.382847112357857e-07,
+      "loss": 0.891,
+      "step": 18266
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.378643841297247e-07,
+      "loss": 0.8862,
+      "step": 18267
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.374443138321937e-07,
+      "loss": 0.957,
+      "step": 18268
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3702450035435597e-07,
+      "loss": 0.9257,
+      "step": 18269
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3660494370737153e-07,
+      "loss": 0.8737,
+      "step": 18270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.361856439023914e-07,
+      "loss": 0.9219,
+      "step": 18271
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.357666009505589e-07,
+      "loss": 0.8772,
+      "step": 18272
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.353478148630129e-07,
+      "loss": 0.8924,
+      "step": 18273
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.349292856508857e-07,
+      "loss": 0.9352,
+      "step": 18274
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.345110133252971e-07,
+      "loss": 0.9109,
+      "step": 18275
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3409299789736726e-07,
+      "loss": 0.9529,
+      "step": 18276
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3367523937820614e-07,
+      "loss": 0.8102,
+      "step": 18277
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.33257737778917e-07,
+      "loss": 0.9274,
+      "step": 18278
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.328404931105966e-07,
+      "loss": 0.3838,
+      "step": 18279
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3242350538433497e-07,
+      "loss": 0.8959,
+      "step": 18280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3200677461121656e-07,
+      "loss": 0.9204,
+      "step": 18281
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.315903008023147e-07,
+      "loss": 0.9376,
+      "step": 18282
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3117408396870056e-07,
+      "loss": 0.9458,
+      "step": 18283
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.307581241214375e-07,
+      "loss": 0.8208,
+      "step": 18284
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.303424212715778e-07,
+      "loss": 0.8851,
+      "step": 18285
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.299269754301748e-07,
+      "loss": 0.8232,
+      "step": 18286
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2951178660826866e-07,
+      "loss": 0.9345,
+      "step": 18287
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.290968548168927e-07,
+      "loss": 0.879,
+      "step": 18288
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.286821800670781e-07,
+      "loss": 0.8964,
+      "step": 18289
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.282677623698449e-07,
+      "loss": 0.8262,
+      "step": 18290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2785360173620885e-07,
+      "loss": 0.8826,
+      "step": 18291
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2743969817717657e-07,
+      "loss": 0.7918,
+      "step": 18292
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2702605170375156e-07,
+      "loss": 0.9397,
+      "step": 18293
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2661266232692614e-07,
+      "loss": 0.852,
+      "step": 18294
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.26199530057687e-07,
+      "loss": 0.7801,
+      "step": 18295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2578665490701653e-07,
+      "loss": 0.8633,
+      "step": 18296
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.253740368858893e-07,
+      "loss": 0.8565,
+      "step": 18297
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.249616760052687e-07,
+      "loss": 0.9083,
+      "step": 18298
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.245495722761172e-07,
+      "loss": 0.8428,
+      "step": 18299
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2413772570938697e-07,
+      "loss": 0.8436,
+      "step": 18300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.237261363160271e-07,
+      "loss": 0.8396,
+      "step": 18301
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.233148041069722e-07,
+      "loss": 0.8813,
+      "step": 18302
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.229037290931614e-07,
+      "loss": 0.9063,
+      "step": 18303
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2249291128551463e-07,
+      "loss": 0.9329,
+      "step": 18304
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.220823506949544e-07,
+      "loss": 0.8762,
+      "step": 18305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.21672047332392e-07,
+      "loss": 0.8117,
+      "step": 18306
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2126200120873307e-07,
+      "loss": 0.8554,
+      "step": 18307
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2085221233487564e-07,
+      "loss": 0.8523,
+      "step": 18308
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2044268072171205e-07,
+      "loss": 0.8589,
+      "step": 18309
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.200334063801269e-07,
+      "loss": 0.9037,
+      "step": 18310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1962438932099826e-07,
+      "loss": 0.9009,
+      "step": 18311
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1921562955519626e-07,
+      "loss": 0.9034,
+      "step": 18312
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1880712709358776e-07,
+      "loss": 0.9091,
+      "step": 18313
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.183988819470274e-07,
+      "loss": 0.8734,
+      "step": 18314
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1799089412636765e-07,
+      "loss": 0.9098,
+      "step": 18315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.175831636424542e-07,
+      "loss": 0.8565,
+      "step": 18316
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.171756905061196e-07,
+      "loss": 0.9261,
+      "step": 18317
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1676847472819626e-07,
+      "loss": 0.8037,
+      "step": 18318
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.163615163195088e-07,
+      "loss": 0.8334,
+      "step": 18319
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1595481529087203e-07,
+      "loss": 0.8693,
+      "step": 18320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1554837165309603e-07,
+      "loss": 0.8764,
+      "step": 18321
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.151421854169834e-07,
+      "loss": 0.847,
+      "step": 18322
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1473625659333096e-07,
+      "loss": 0.9228,
+      "step": 18323
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1433058519292683e-07,
+      "loss": 0.9152,
+      "step": 18324
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.139251712265523e-07,
+      "loss": 0.8779,
+      "step": 18325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.135200147049855e-07,
+      "loss": 0.9308,
+      "step": 18326
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1311511563899335e-07,
+      "loss": 0.8583,
+      "step": 18327
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.127104740393372e-07,
+      "loss": 0.8856,
+      "step": 18328
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.12306089916774e-07,
+      "loss": 0.9094,
+      "step": 18329
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.119019632820508e-07,
+      "loss": 0.9002,
+      "step": 18330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.114980941459078e-07,
+      "loss": 0.8684,
+      "step": 18331
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.110944825190798e-07,
+      "loss": 0.827,
+      "step": 18332
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.10691128412296e-07,
+      "loss": 0.7855,
+      "step": 18333
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.102880318362744e-07,
+      "loss": 0.9267,
+      "step": 18334
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.098851928017299e-07,
+      "loss": 0.9259,
+      "step": 18335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0948261131937164e-07,
+      "loss": 0.8606,
+      "step": 18336
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0908028739989547e-07,
+      "loss": 0.9141,
+      "step": 18337
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.086782210539985e-07,
+      "loss": 0.889,
+      "step": 18338
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.082764122923665e-07,
+      "loss": 0.8577,
+      "step": 18339
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0787486112567763e-07,
+      "loss": 0.7835,
+      "step": 18340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0747356756460565e-07,
+      "loss": 0.859,
+      "step": 18341
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0707253161981533e-07,
+      "loss": 0.9159,
+      "step": 18342
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.066717533019692e-07,
+      "loss": 0.8211,
+      "step": 18343
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0627123262171543e-07,
+      "loss": 0.8896,
+      "step": 18344
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.058709695897011e-07,
+      "loss": 0.8811,
+      "step": 18345
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.054709642165654e-07,
+      "loss": 0.8,
+      "step": 18346
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0507121651293767e-07,
+      "loss": 0.8202,
+      "step": 18347
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0467172648944497e-07,
+      "loss": 0.8554,
+      "step": 18348
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0427249415670544e-07,
+      "loss": 0.8892,
+      "step": 18349
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.038735195253284e-07,
+      "loss": 0.9356,
+      "step": 18350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0347480260591867e-07,
+      "loss": 0.8814,
+      "step": 18351
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0307634340907663e-07,
+      "loss": 0.9385,
+      "step": 18352
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0267814194538834e-07,
+      "loss": 0.4117,
+      "step": 18353
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0228019822544084e-07,
+      "loss": 0.8546,
+      "step": 18354
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0188251225980905e-07,
+      "loss": 0.9495,
+      "step": 18355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.014850840590644e-07,
+      "loss": 0.8941,
+      "step": 18356
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0108791363376853e-07,
+      "loss": 0.8939,
+      "step": 18357
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.006910009944797e-07,
+      "loss": 0.8643,
+      "step": 18358
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0029434615174605e-07,
+      "loss": 0.8434,
+      "step": 18359
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.998979491161103e-07,
+      "loss": 0.9221,
+      "step": 18360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.995018098981073e-07,
+      "loss": 0.8293,
+      "step": 18361
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9910592850826983e-07,
+      "loss": 0.8651,
+      "step": 18362
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9871030495711493e-07,
+      "loss": 0.8434,
+      "step": 18363
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.98314939255161e-07,
+      "loss": 0.8691,
+      "step": 18364
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9791983141291727e-07,
+      "loss": 0.8817,
+      "step": 18365
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.975249814408809e-07,
+      "loss": 0.8956,
+      "step": 18366
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9713038934955027e-07,
+      "loss": 0.8258,
+      "step": 18367
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.967360551494125e-07,
+      "loss": 0.9234,
+      "step": 18368
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.963419788509492e-07,
+      "loss": 0.8902,
+      "step": 18369
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.959481604646319e-07,
+      "loss": 0.918,
+      "step": 18370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9555460000092907e-07,
+      "loss": 0.8562,
+      "step": 18371
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.951612974703033e-07,
+      "loss": 0.9818,
+      "step": 18372
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.947682528832041e-07,
+      "loss": 0.8693,
+      "step": 18373
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9437546625008085e-07,
+      "loss": 0.8194,
+      "step": 18374
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.939829375813741e-07,
+      "loss": 0.871,
+      "step": 18375
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9359066688751435e-07,
+      "loss": 0.8779,
+      "step": 18376
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.931986541789289e-07,
+      "loss": 0.8531,
+      "step": 18377
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.928068994660371e-07,
+      "loss": 0.8871,
+      "step": 18378
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9241540275925293e-07,
+      "loss": 0.9369,
+      "step": 18379
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9202416406897803e-07,
+      "loss": 0.8613,
+      "step": 18380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.916331834056141e-07,
+      "loss": 0.8338,
+      "step": 18381
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.912424607795528e-07,
+      "loss": 0.8328,
+      "step": 18382
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9085199620117686e-07,
+      "loss": 0.8338,
+      "step": 18383
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.90461789680867e-07,
+      "loss": 0.8958,
+      "step": 18384
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9007184122899267e-07,
+      "loss": 0.9615,
+      "step": 18385
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8968215085592e-07,
+      "loss": 0.8126,
+      "step": 18386
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8929271857200403e-07,
+      "loss": 0.9494,
+      "step": 18387
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.889035443875987e-07,
+      "loss": 0.8381,
+      "step": 18388
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.885146283130458e-07,
+      "loss": 0.8938,
+      "step": 18389
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.881259703586814e-07,
+      "loss": 0.8086,
+      "step": 18390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.877375705348373e-07,
+      "loss": 0.8532,
+      "step": 18391
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8734942885183634e-07,
+      "loss": 0.8385,
+      "step": 18392
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8696154531999477e-07,
+      "loss": 0.8781,
+      "step": 18393
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.865739199496209e-07,
+      "loss": 0.85,
+      "step": 18394
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.861865527510199e-07,
+      "loss": 0.8633,
+      "step": 18395
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8579944373448463e-07,
+      "loss": 0.7946,
+      "step": 18396
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8541259291030577e-07,
+      "loss": 0.9075,
+      "step": 18397
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8502600028876613e-07,
+      "loss": 0.8641,
+      "step": 18398
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.846396658801387e-07,
+      "loss": 0.9261,
+      "step": 18399
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.84253589694693e-07,
+      "loss": 0.864,
+      "step": 18400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8386777174269186e-07,
+      "loss": 0.7095,
+      "step": 18401
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.834822120343872e-07,
+      "loss": 0.8692,
+      "step": 18402
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.830969105800285e-07,
+      "loss": 0.8484,
+      "step": 18403
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8271186738985546e-07,
+      "loss": 0.8712,
+      "step": 18404
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.823270824741042e-07,
+      "loss": 0.4379,
+      "step": 18405
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.819425558429989e-07,
+      "loss": 0.8918,
+      "step": 18406
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.815582875067624e-07,
+      "loss": 0.8738,
+      "step": 18407
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.811742774756054e-07,
+      "loss": 0.8265,
+      "step": 18408
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8079052575973764e-07,
+      "loss": 0.8261,
+      "step": 18409
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.804070323693564e-07,
+      "loss": 0.7994,
+      "step": 18410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.800237973146569e-07,
+      "loss": 0.9034,
+      "step": 18411
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.796408206058232e-07,
+      "loss": 0.8936,
+      "step": 18412
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7925810225303386e-07,
+      "loss": 0.9402,
+      "step": 18413
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.788756422664629e-07,
+      "loss": 0.9253,
+      "step": 18414
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.784934406562745e-07,
+      "loss": 0.822,
+      "step": 18415
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7811149743262713e-07,
+      "loss": 0.8601,
+      "step": 18416
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7772981260567265e-07,
+      "loss": 0.9619,
+      "step": 18417
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.773483861855575e-07,
+      "loss": 0.829,
+      "step": 18418
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.769672181824157e-07,
+      "loss": 0.8395,
+      "step": 18419
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7658630860638026e-07,
+      "loss": 0.9046,
+      "step": 18420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7620565746757757e-07,
+      "loss": 0.822,
+      "step": 18421
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.758252647761206e-07,
+      "loss": 0.9046,
+      "step": 18422
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7544513054212354e-07,
+      "loss": 0.3088,
+      "step": 18423
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7506525477568935e-07,
+      "loss": 0.8019,
+      "step": 18424
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.746856374869122e-07,
+      "loss": 0.8673,
+      "step": 18425
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.74306278685883e-07,
+      "loss": 0.8539,
+      "step": 18426
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.739271783826858e-07,
+      "loss": 0.8028,
+      "step": 18427
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.73548336587397e-07,
+      "loss": 0.8978,
+      "step": 18428
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.73169753310083e-07,
+      "loss": 0.9093,
+      "step": 18429
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7279142856080907e-07,
+      "loss": 0.8584,
+      "step": 18430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7241336234962943e-07,
+      "loss": 0.822,
+      "step": 18431
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7203555468659047e-07,
+      "loss": 0.8758,
+      "step": 18432
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7165800558173863e-07,
+      "loss": 0.8411,
+      "step": 18433
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.712807150451058e-07,
+      "loss": 0.8792,
+      "step": 18434
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7090368308671954e-07,
+      "loss": 0.8518,
+      "step": 18435
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.705269097166019e-07,
+      "loss": 0.8539,
+      "step": 18436
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.70150394944767e-07,
+      "loss": 0.9332,
+      "step": 18437
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.697741387812225e-07,
+      "loss": 0.9152,
+      "step": 18438
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6939814123596696e-07,
+      "loss": 0.8694,
+      "step": 18439
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.690224023189969e-07,
+      "loss": 0.7984,
+      "step": 18440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.686469220402976e-07,
+      "loss": 0.9074,
+      "step": 18441
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6827170040984783e-07,
+      "loss": 0.8532,
+      "step": 18442
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.678967374376218e-07,
+      "loss": 0.8891,
+      "step": 18443
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6752203313358596e-07,
+      "loss": 0.8463,
+      "step": 18444
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.671475875076979e-07,
+      "loss": 0.8938,
+      "step": 18445
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.667734005699119e-07,
+      "loss": 0.8569,
+      "step": 18446
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6639947233017217e-07,
+      "loss": 0.9243,
+      "step": 18447
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.660258027984175e-07,
+      "loss": 0.8813,
+      "step": 18448
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.656523919845799e-07,
+      "loss": 0.8493,
+      "step": 18449
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6527923989858485e-07,
+      "loss": 0.8994,
+      "step": 18450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6490634655034874e-07,
+      "loss": 0.8228,
+      "step": 18451
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.645337119497826e-07,
+      "loss": 0.8632,
+      "step": 18452
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6416133610679183e-07,
+      "loss": 0.8796,
+      "step": 18453
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.637892190312752e-07,
+      "loss": 0.8106,
+      "step": 18454
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.634173607331181e-07,
+      "loss": 0.8569,
+      "step": 18455
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6304576122221035e-07,
+      "loss": 0.8396,
+      "step": 18456
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6267442050842417e-07,
+      "loss": 0.9166,
+      "step": 18457
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.623033386016305e-07,
+      "loss": 0.9008,
+      "step": 18458
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6193251551169253e-07,
+      "loss": 0.9286,
+      "step": 18459
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6156195124846796e-07,
+      "loss": 0.8011,
+      "step": 18460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6119164582180223e-07,
+      "loss": 0.9022,
+      "step": 18461
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.608215992415408e-07,
+      "loss": 0.8907,
+      "step": 18462
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6045181151751917e-07,
+      "loss": 0.8366,
+      "step": 18463
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.600822826595628e-07,
+      "loss": 0.8989,
+      "step": 18464
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5971301267749603e-07,
+      "loss": 0.8741,
+      "step": 18465
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5934400158113326e-07,
+      "loss": 0.8744,
+      "step": 18466
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.589752493802811e-07,
+      "loss": 0.8539,
+      "step": 18467
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.586067560847416e-07,
+      "loss": 0.8452,
+      "step": 18468
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.582385217043093e-07,
+      "loss": 0.8904,
+      "step": 18469
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.578705462487718e-07,
+      "loss": 0.874,
+      "step": 18470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.575028297279081e-07,
+      "loss": 0.8416,
+      "step": 18471
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.571353721514913e-07,
+      "loss": 0.9426,
+      "step": 18472
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.567681735292904e-07,
+      "loss": 0.8133,
+      "step": 18473
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5640123387106197e-07,
+      "loss": 0.8582,
+      "step": 18474
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5603455318656154e-07,
+      "loss": 0.9008,
+      "step": 18475
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.556681314855336e-07,
+      "loss": 0.8249,
+      "step": 18476
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.553019687777181e-07,
+      "loss": 0.8368,
+      "step": 18477
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5493606507284605e-07,
+      "loss": 0.8311,
+      "step": 18478
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.545704203806432e-07,
+      "loss": 0.8938,
+      "step": 18479
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5420503471082934e-07,
+      "loss": 0.939,
+      "step": 18480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5383990807311354e-07,
+      "loss": 0.8559,
+      "step": 18481
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.534750404772024e-07,
+      "loss": 0.884,
+      "step": 18482
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.531104319327937e-07,
+      "loss": 0.8055,
+      "step": 18483
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5274608244957757e-07,
+      "loss": 0.8453,
+      "step": 18484
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.523819920372372e-07,
+      "loss": 0.8864,
+      "step": 18485
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.520181607054528e-07,
+      "loss": 0.8558,
+      "step": 18486
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.516545884638899e-07,
+      "loss": 0.9018,
+      "step": 18487
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.512912753222152e-07,
+      "loss": 0.8776,
+      "step": 18488
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.509282212900843e-07,
+      "loss": 0.8542,
+      "step": 18489
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5056542637714844e-07,
+      "loss": 0.8765,
+      "step": 18490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5020289059304536e-07,
+      "loss": 0.8741,
+      "step": 18491
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4984061394741744e-07,
+      "loss": 0.9259,
+      "step": 18492
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.49478596449888e-07,
+      "loss": 0.8346,
+      "step": 18493
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.491168381100817e-07,
+      "loss": 0.9503,
+      "step": 18494
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4875533893761406e-07,
+      "loss": 0.8321,
+      "step": 18495
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4839409894209297e-07,
+      "loss": 0.8835,
+      "step": 18496
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.480331181331175e-07,
+      "loss": 0.843,
+      "step": 18497
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.476723965202854e-07,
+      "loss": 0.7929,
+      "step": 18498
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4731193411318243e-07,
+      "loss": 0.845,
+      "step": 18499
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.469517309213898e-07,
+      "loss": 0.8433,
+      "step": 18500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.465917869544798e-07,
+      "loss": 0.9107,
+      "step": 18501
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4623210222202153e-07,
+      "loss": 0.921,
+      "step": 18502
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.45872676733574e-07,
+      "loss": 0.7422,
+      "step": 18503
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4551351049869055e-07,
+      "loss": 0.8928,
+      "step": 18504
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.451546035269181e-07,
+      "loss": 0.8942,
+      "step": 18505
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4479595582779457e-07,
+      "loss": 0.8854,
+      "step": 18506
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.444375674108523e-07,
+      "loss": 0.826,
+      "step": 18507
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.440794382856182e-07,
+      "loss": 0.8591,
+      "step": 18508
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4372156846161124e-07,
+      "loss": 0.8186,
+      "step": 18509
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4336395794834045e-07,
+      "loss": 0.8661,
+      "step": 18510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4300660675531386e-07,
+      "loss": 0.8668,
+      "step": 18511
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4264951489202824e-07,
+      "loss": 0.9127,
+      "step": 18512
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.422926823679728e-07,
+      "loss": 0.8594,
+      "step": 18513
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4193610919263533e-07,
+      "loss": 0.8783,
+      "step": 18514
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4157979537548946e-07,
+      "loss": 0.9533,
+      "step": 18515
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4122374092600876e-07,
+      "loss": 0.9279,
+      "step": 18516
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.408679458536556e-07,
+      "loss": 0.8654,
+      "step": 18517
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.405124101678857e-07,
+      "loss": 0.8527,
+      "step": 18518
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4015713387815053e-07,
+      "loss": 0.8582,
+      "step": 18519
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.398021169938913e-07,
+      "loss": 0.8873,
+      "step": 18520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.394473595245439e-07,
+      "loss": 0.8867,
+      "step": 18521
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3909286147953956e-07,
+      "loss": 0.8782,
+      "step": 18522
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.387386228682986e-07,
+      "loss": 0.8559,
+      "step": 18523
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3838464370023573e-07,
+      "loss": 0.3543,
+      "step": 18524
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3803092398476e-07,
+      "loss": 0.8569,
+      "step": 18525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.37677463731274e-07,
+      "loss": 0.848,
+      "step": 18526
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.373242629491701e-07,
+      "loss": 0.7939,
+      "step": 18527
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3697132164783977e-07,
+      "loss": 0.9458,
+      "step": 18528
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3661863983665878e-07,
+      "loss": 0.877,
+      "step": 18529
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3626621752500523e-07,
+      "loss": 0.8708,
+      "step": 18530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3591405472224272e-07,
+      "loss": 0.8326,
+      "step": 18531
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3556215143773487e-07,
+      "loss": 0.3615,
+      "step": 18532
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3521050768083197e-07,
+      "loss": 0.9356,
+      "step": 18533
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3485912346088102e-07,
+      "loss": 0.9142,
+      "step": 18534
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.345079987872223e-07,
+      "loss": 0.9103,
+      "step": 18535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3415713366918724e-07,
+      "loss": 0.8413,
+      "step": 18536
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3380652811610173e-07,
+      "loss": 0.8604,
+      "step": 18537
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3345618213728492e-07,
+      "loss": 0.8967,
+      "step": 18538
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3310609574204835e-07,
+      "loss": 0.9267,
+      "step": 18539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3275626893969673e-07,
+      "loss": 0.8971,
+      "step": 18540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3240670173952927e-07,
+      "loss": 0.8757,
+      "step": 18541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3205739415083528e-07,
+      "loss": 0.8886,
+      "step": 18542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.317083461828995e-07,
+      "loss": 0.8696,
+      "step": 18543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3135955784499897e-07,
+      "loss": 0.9373,
+      "step": 18544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3101102914640517e-07,
+      "loss": 0.8659,
+      "step": 18545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3066276009638066e-07,
+      "loss": 0.8622,
+      "step": 18546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.303147507041825e-07,
+      "loss": 0.8363,
+      "step": 18547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.299670009790611e-07,
+      "loss": 0.911,
+      "step": 18548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2961951093025791e-07,
+      "loss": 0.844,
+      "step": 18549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.292722805670078e-07,
+      "loss": 0.3896,
+      "step": 18550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2892530989854221e-07,
+      "loss": 0.8623,
+      "step": 18551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2857859893408273e-07,
+      "loss": 0.8295,
+      "step": 18552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2823214768284308e-07,
+      "loss": 0.7634,
+      "step": 18553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2788595615403475e-07,
+      "loss": 0.7794,
+      "step": 18554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2754002435685484e-07,
+      "loss": 0.8969,
+      "step": 18555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2719435230049936e-07,
+      "loss": 0.8312,
+      "step": 18556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2684893999415758e-07,
+      "loss": 0.7994,
+      "step": 18557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2650378744700885e-07,
+      "loss": 0.868,
+      "step": 18558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2615889466822693e-07,
+      "loss": 0.851,
+      "step": 18559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2581426166697895e-07,
+      "loss": 0.8546,
+      "step": 18560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2546988845242535e-07,
+      "loss": 0.9103,
+      "step": 18561
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2512577503371657e-07,
+      "loss": 0.9192,
+      "step": 18562
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.24781921420002e-07,
+      "loss": 0.8547,
+      "step": 18563
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2443832762041983e-07,
+      "loss": 0.8702,
+      "step": 18564
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.240949936441017e-07,
+      "loss": 0.8863,
+      "step": 18565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2375191950017362e-07,
+      "loss": 0.7881,
+      "step": 18566
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2340910519775273e-07,
+      "loss": 0.8841,
+      "step": 18567
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2306655074595397e-07,
+      "loss": 0.7538,
+      "step": 18568
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2272425615387895e-07,
+      "loss": 0.9059,
+      "step": 18569
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2238222143062704e-07,
+      "loss": 0.3791,
+      "step": 18570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2204044658528878e-07,
+      "loss": 0.8641,
+      "step": 18571
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.216989316269469e-07,
+      "loss": 0.8726,
+      "step": 18572
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2135767656467965e-07,
+      "loss": 0.4633,
+      "step": 18573
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.210166814075565e-07,
+      "loss": 0.9281,
+      "step": 18574
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2067594616464237e-07,
+      "loss": 0.8767,
+      "step": 18575
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.203354708449923e-07,
+      "loss": 0.8796,
+      "step": 18576
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1999525545765678e-07,
+      "loss": 0.8764,
+      "step": 18577
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1965530001167634e-07,
+      "loss": 0.8413,
+      "step": 18578
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1931560451608825e-07,
+      "loss": 0.7696,
+      "step": 18579
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.189761689799208e-07,
+      "loss": 0.8221,
+      "step": 18580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.186369934121957e-07,
+      "loss": 0.8512,
+      "step": 18581
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1829807782192792e-07,
+      "loss": 0.8226,
+      "step": 18582
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1795942221812583e-07,
+      "loss": 0.9583,
+      "step": 18583
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1762102660979e-07,
+      "loss": 0.8901,
+      "step": 18584
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1728289100591327e-07,
+      "loss": 0.8676,
+      "step": 18585
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1694501541548508e-07,
+      "loss": 0.8309,
+      "step": 18586
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1660739984748714e-07,
+      "loss": 0.8675,
+      "step": 18587
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1627004431088893e-07,
+      "loss": 0.8578,
+      "step": 18588
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1593294881465887e-07,
+      "loss": 0.8732,
+      "step": 18589
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1559611336775755e-07,
+      "loss": 0.9026,
+      "step": 18590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1525953797913557e-07,
+      "loss": 0.8429,
+      "step": 18591
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1492322265774023e-07,
+      "loss": 0.9065,
+      "step": 18592
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1458716741250995e-07,
+      "loss": 0.9534,
+      "step": 18593
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1425137225237757e-07,
+      "loss": 0.8735,
+      "step": 18594
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1391583718626597e-07,
+      "loss": 0.9454,
+      "step": 18595
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1358056222309576e-07,
+      "loss": 0.8737,
+      "step": 18596
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.132455473717765e-07,
+      "loss": 0.8936,
+      "step": 18597
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1291079264121218e-07,
+      "loss": 0.8247,
+      "step": 18598
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1257629804030234e-07,
+      "loss": 0.8292,
+      "step": 18599
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1224206357793653e-07,
+      "loss": 0.8422,
+      "step": 18600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1190808926299656e-07,
+      "loss": 0.9025,
+      "step": 18601
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1157437510436086e-07,
+      "loss": 0.3724,
+      "step": 18602
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.112409211109001e-07,
+      "loss": 0.8478,
+      "step": 18603
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.10907727291475e-07,
+      "loss": 0.869,
+      "step": 18604
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1057479365494182e-07,
+      "loss": 0.8913,
+      "step": 18605
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1024212021015011e-07,
+      "loss": 0.8929,
+      "step": 18606
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0990970696594282e-07,
+      "loss": 0.8234,
+      "step": 18607
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0957755393115398e-07,
+      "loss": 0.8459,
+      "step": 18608
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0924566111460987e-07,
+      "loss": 0.9029,
+      "step": 18609
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0891402852513675e-07,
+      "loss": 0.9406,
+      "step": 18610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0858265617154427e-07,
+      "loss": 0.8832,
+      "step": 18611
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0825154406264313e-07,
+      "loss": 0.8367,
+      "step": 18612
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0792069220723298e-07,
+      "loss": 0.8816,
+      "step": 18613
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.075901006141068e-07,
+      "loss": 0.8584,
+      "step": 18614
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0725976929205082e-07,
+      "loss": 0.3644,
+      "step": 18615
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0692969824984698e-07,
+      "loss": 0.9379,
+      "step": 18616
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0659988749626713e-07,
+      "loss": 0.8886,
+      "step": 18617
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0627033704007648e-07,
+      "loss": 0.8419,
+      "step": 18618
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0594104689003468e-07,
+      "loss": 0.8646,
+      "step": 18619
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0561201705489476e-07,
+      "loss": 0.9173,
+      "step": 18620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0528324754339858e-07,
+      "loss": 0.847,
+      "step": 18621
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0495473836428915e-07,
+      "loss": 0.8734,
+      "step": 18622
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0462648952629616e-07,
+      "loss": 0.8798,
+      "step": 18623
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0429850103814376e-07,
+      "loss": 0.458,
+      "step": 18624
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0397077290854828e-07,
+      "loss": 0.8897,
+      "step": 18625
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0364330514622278e-07,
+      "loss": 0.8211,
+      "step": 18626
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0331609775986915e-07,
+      "loss": 0.8897,
+      "step": 18627
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0298915075818383e-07,
+      "loss": 0.9238,
+      "step": 18628
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.026624641498587e-07,
+      "loss": 0.94,
+      "step": 18629
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.023360379435757e-07,
+      "loss": 0.978,
+      "step": 18630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0200987214801015e-07,
+      "loss": 0.9129,
+      "step": 18631
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0168396677183178e-07,
+      "loss": 0.8025,
+      "step": 18632
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0135832182370362e-07,
+      "loss": 0.8259,
+      "step": 18633
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.010329373122788e-07,
+      "loss": 0.7907,
+      "step": 18634
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0070781324620814e-07,
+      "loss": 0.8911,
+      "step": 18635
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0038294963413251e-07,
+      "loss": 0.7728,
+      "step": 18636
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0005834648468503e-07,
+      "loss": 0.8749,
+      "step": 18637
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9973400380649323e-07,
+      "loss": 0.8692,
+      "step": 18638
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9940992160818018e-07,
+      "loss": 0.8492,
+      "step": 18639
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9908609989835793e-07,
+      "loss": 0.8419,
+      "step": 18640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9876253868563178e-07,
+      "loss": 0.9117,
+      "step": 18641
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9843923797860486e-07,
+      "loss": 0.8043,
+      "step": 18642
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9811619778586811e-07,
+      "loss": 0.815,
+      "step": 18643
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9779341811600795e-07,
+      "loss": 0.7957,
+      "step": 18644
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9747089897760309e-07,
+      "loss": 0.8741,
+      "step": 18645
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9714864037922666e-07,
+      "loss": 0.8764,
+      "step": 18646
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9682664232944403e-07,
+      "loss": 0.8282,
+      "step": 18647
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9650490483681172e-07,
+      "loss": 0.8284,
+      "step": 18648
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9618342790988397e-07,
+      "loss": 0.3596,
+      "step": 18649
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9586221155720176e-07,
+      "loss": 0.8838,
+      "step": 18650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.95541255787306e-07,
+      "loss": 0.8444,
+      "step": 18651
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.952205606087254e-07,
+      "loss": 0.8269,
+      "step": 18652
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9490012602998432e-07,
+      "loss": 0.8791,
+      "step": 18653
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9457995205959922e-07,
+      "loss": 0.8828,
+      "step": 18654
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9426003870608002e-07,
+      "loss": 0.8523,
+      "step": 18655
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9394038597792987e-07,
+      "loss": 0.9011,
+      "step": 18656
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9362099388364418e-07,
+      "loss": 0.9339,
+      "step": 18657
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.933018624317129e-07,
+      "loss": 0.8442,
+      "step": 18658
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9298299163061917e-07,
+      "loss": 0.8555,
+      "step": 18659
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9266438148883514e-07,
+      "loss": 0.9165,
+      "step": 18660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.923460320148307e-07,
+      "loss": 0.8813,
+      "step": 18661
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.920279432170691e-07,
+      "loss": 0.8121,
+      "step": 18662
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9171011510400128e-07,
+      "loss": 0.8742,
+      "step": 18663
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9139254768407722e-07,
+      "loss": 0.9554,
+      "step": 18664
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.910752409657357e-07,
+      "loss": 0.882,
+      "step": 18665
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9075819495741331e-07,
+      "loss": 0.9037,
+      "step": 18666
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9044140966753333e-07,
+      "loss": 0.8682,
+      "step": 18667
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9012488510451788e-07,
+      "loss": 0.9163,
+      "step": 18668
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8980862127677803e-07,
+      "loss": 0.8574,
+      "step": 18669
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8949261819272147e-07,
+      "loss": 0.9471,
+      "step": 18670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8917687586074596e-07,
+      "loss": 0.8905,
+      "step": 18671
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8886139428924477e-07,
+      "loss": 0.9183,
+      "step": 18672
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8854617348660231e-07,
+      "loss": 0.8161,
+      "step": 18673
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8823121346119633e-07,
+      "loss": 0.878,
+      "step": 18674
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.87916514221399e-07,
+      "loss": 0.8282,
+      "step": 18675
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8760207577557364e-07,
+      "loss": 0.7751,
+      "step": 18676
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8728789813207804e-07,
+      "loss": 0.8293,
+      "step": 18677
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8697398129926436e-07,
+      "loss": 0.8834,
+      "step": 18678
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8666032528547373e-07,
+      "loss": 0.9145,
+      "step": 18679
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.863469300990428e-07,
+      "loss": 0.48,
+      "step": 18680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8603379574830383e-07,
+      "loss": 0.9394,
+      "step": 18681
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.857209222415779e-07,
+      "loss": 0.8331,
+      "step": 18682
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.854083095871795e-07,
+      "loss": 0.9014,
+      "step": 18683
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8509595779341971e-07,
+      "loss": 0.9629,
+      "step": 18684
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8478386686860083e-07,
+      "loss": 0.937,
+      "step": 18685
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8447203682101512e-07,
+      "loss": 0.8754,
+      "step": 18686
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8416046765895257e-07,
+      "loss": 0.8542,
+      "step": 18687
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8384915939069547e-07,
+      "loss": 0.929,
+      "step": 18688
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8353811202451498e-07,
+      "loss": 0.9056,
+      "step": 18689
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8322732556868007e-07,
+      "loss": 0.9118,
+      "step": 18690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8291680003145074e-07,
+      "loss": 0.9287,
+      "step": 18691
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8260653542108154e-07,
+      "loss": 0.4063,
+      "step": 18692
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8229653174581696e-07,
+      "loss": 0.818,
+      "step": 18693
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8198678901389823e-07,
+      "loss": 0.7892,
+      "step": 18694
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.816773072335587e-07,
+      "loss": 0.9817,
+      "step": 18695
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8136808641302185e-07,
+      "loss": 0.9283,
+      "step": 18696
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8105912656050662e-07,
+      "loss": 0.8667,
+      "step": 18697
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8075042768422645e-07,
+      "loss": 0.9025,
+      "step": 18698
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.804419897923848e-07,
+      "loss": 0.9366,
+      "step": 18699
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8013381289318065e-07,
+      "loss": 0.899,
+      "step": 18700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7982589699480412e-07,
+      "loss": 0.8761,
+      "step": 18701
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7951824210543867e-07,
+      "loss": 0.8612,
+      "step": 18702
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.792108482332622e-07,
+      "loss": 0.8666,
+      "step": 18703
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7890371538644591e-07,
+      "loss": 0.8766,
+      "step": 18704
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.785968435731522e-07,
+      "loss": 0.9038,
+      "step": 18705
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7829023280153678e-07,
+      "loss": 0.372,
+      "step": 18706
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7798388307974978e-07,
+      "loss": 0.8691,
+      "step": 18707
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7767779441593358e-07,
+      "loss": 0.9531,
+      "step": 18708
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7737196681822278e-07,
+      "loss": 0.7921,
+      "step": 18709
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7706640029474642e-07,
+      "loss": 0.9469,
+      "step": 18710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7676109485362803e-07,
+      "loss": 0.9286,
+      "step": 18711
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7645605050297888e-07,
+      "loss": 0.859,
+      "step": 18712
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7615126725090804e-07,
+      "loss": 0.9377,
+      "step": 18713
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.758467451055179e-07,
+      "loss": 0.8178,
+      "step": 18714
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7554248407489982e-07,
+      "loss": 0.932,
+      "step": 18715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7523848416714174e-07,
+      "loss": 0.9135,
+      "step": 18716
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7493474539032607e-07,
+      "loss": 0.4768,
+      "step": 18717
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7463126775252192e-07,
+      "loss": 0.8697,
+      "step": 18718
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.743280512617962e-07,
+      "loss": 0.7793,
+      "step": 18719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.740250959262102e-07,
+      "loss": 0.8061,
+      "step": 18720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7372240175381527e-07,
+      "loss": 0.8579,
+      "step": 18721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.73419968752655e-07,
+      "loss": 0.7801,
+      "step": 18722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7311779693076846e-07,
+      "loss": 0.8844,
+      "step": 18723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7281588629618928e-07,
+      "loss": 0.9068,
+      "step": 18724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.725142368569388e-07,
+      "loss": 0.8693,
+      "step": 18725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7221284862103615e-07,
+      "loss": 0.9465,
+      "step": 18726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7191172159649162e-07,
+      "loss": 0.8388,
+      "step": 18727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7161085579130876e-07,
+      "loss": 0.865,
+      "step": 18728
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7131025121348454e-07,
+      "loss": 0.8362,
+      "step": 18729
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7100990787100814e-07,
+      "loss": 0.8453,
+      "step": 18730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.707098257718631e-07,
+      "loss": 0.9069,
+      "step": 18731
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7041000492402426e-07,
+      "loss": 0.8653,
+      "step": 18732
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7011044533546073e-07,
+      "loss": 0.8933,
+      "step": 18733
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6981114701413615e-07,
+      "loss": 0.9168,
+      "step": 18734
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6951210996800306e-07,
+      "loss": 0.8611,
+      "step": 18735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6921333420500952e-07,
+      "loss": 0.8381,
+      "step": 18736
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.689148197331003e-07,
+      "loss": 0.7831,
+      "step": 18737
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6861656656020464e-07,
+      "loss": 0.8974,
+      "step": 18738
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6831857469425282e-07,
+      "loss": 0.8937,
+      "step": 18739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6802084414316523e-07,
+      "loss": 0.8483,
+      "step": 18740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6772337491485324e-07,
+      "loss": 0.8335,
+      "step": 18741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6742616701722503e-07,
+      "loss": 0.9099,
+      "step": 18742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.671292204581787e-07,
+      "loss": 0.9327,
+      "step": 18743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6683253524560904e-07,
+      "loss": 0.8372,
+      "step": 18744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6653611138739868e-07,
+      "loss": 0.8152,
+      "step": 18745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.662399488914279e-07,
+      "loss": 0.8871,
+      "step": 18746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6594404776556937e-07,
+      "loss": 0.8929,
+      "step": 18747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6564840801768457e-07,
+      "loss": 0.8043,
+      "step": 18748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6535302965563382e-07,
+      "loss": 0.8512,
+      "step": 18749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6505791268726868e-07,
+      "loss": 0.8837,
+      "step": 18750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.647630571204295e-07,
+      "loss": 0.8305,
+      "step": 18751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.644684629629556e-07,
+      "loss": 0.916,
+      "step": 18752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6417413022267847e-07,
+      "loss": 0.9043,
+      "step": 18753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6388005890741743e-07,
+      "loss": 0.8634,
+      "step": 18754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6358624902499176e-07,
+      "loss": 0.9085,
+      "step": 18755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6329270058320855e-07,
+      "loss": 0.8457,
+      "step": 18756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6299941358987158e-07,
+      "loss": 0.8371,
+      "step": 18757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.627063880527746e-07,
+      "loss": 0.796,
+      "step": 18758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6241362397970583e-07,
+      "loss": 0.8845,
+      "step": 18759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6212112137844904e-07,
+      "loss": 0.9158,
+      "step": 18760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.618288802567758e-07,
+      "loss": 0.8785,
+      "step": 18761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6153690062245432e-07,
+      "loss": 0.4623,
+      "step": 18762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6124518248324726e-07,
+      "loss": 0.8911,
+      "step": 18763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6095372584690405e-07,
+      "loss": 0.8388,
+      "step": 18764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.606625307211751e-07,
+      "loss": 0.8943,
+      "step": 18765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.603715971137987e-07,
+      "loss": 0.9753,
+      "step": 18766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6008092503250638e-07,
+      "loss": 0.8956,
+      "step": 18767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5979051448502426e-07,
+      "loss": 0.9113,
+      "step": 18768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5950036547907278e-07,
+      "loss": 0.8458,
+      "step": 18769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5921047802236246e-07,
+      "loss": 0.8337,
+      "step": 18770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5892085212259822e-07,
+      "loss": 0.8382,
+      "step": 18771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5863148778747838e-07,
+      "loss": 0.8893,
+      "step": 18772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5834238502469347e-07,
+      "loss": 0.8973,
+      "step": 18773
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5805354384192727e-07,
+      "loss": 0.7808,
+      "step": 18774
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5776496424685704e-07,
+      "loss": 0.8848,
+      "step": 18775
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.574766462471544e-07,
+      "loss": 0.892,
+      "step": 18776
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5718858985048102e-07,
+      "loss": 0.9979,
+      "step": 18777
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5690079506449296e-07,
+      "loss": 0.9085,
+      "step": 18778
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5661326189684078e-07,
+      "loss": 0.7907,
+      "step": 18779
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5632599035516505e-07,
+      "loss": 0.8796,
+      "step": 18780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5603898044710186e-07,
+      "loss": 0.8062,
+      "step": 18781
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5575223218027958e-07,
+      "loss": 0.8596,
+      "step": 18782
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5546574556231987e-07,
+      "loss": 0.8965,
+      "step": 18783
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5517952060083774e-07,
+      "loss": 0.8333,
+      "step": 18784
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5489355730343935e-07,
+      "loss": 0.8886,
+      "step": 18785
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5460785567772528e-07,
+      "loss": 0.8641,
+      "step": 18786
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5432241573129057e-07,
+      "loss": 0.8441,
+      "step": 18787
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.540372374717214e-07,
+      "loss": 0.831,
+      "step": 18788
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.537523209065983e-07,
+      "loss": 0.9087,
+      "step": 18789
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5346766604349195e-07,
+      "loss": 0.8305,
+      "step": 18790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5318327288996959e-07,
+      "loss": 0.8597,
+      "step": 18791
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5289914145358853e-07,
+      "loss": 0.8199,
+      "step": 18792
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5261527174190382e-07,
+      "loss": 0.9056,
+      "step": 18793
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5233166376245722e-07,
+      "loss": 0.8353,
+      "step": 18794
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5204831752278714e-07,
+      "loss": 0.7774,
+      "step": 18795
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5176523303042755e-07,
+      "loss": 0.9413,
+      "step": 18796
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.51482410292898e-07,
+      "loss": 0.8955,
+      "step": 18797
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5119984931771802e-07,
+      "loss": 0.8713,
+      "step": 18798
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.509175501123994e-07,
+      "loss": 0.7929,
+      "step": 18799
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5063551268444275e-07,
+      "loss": 0.8374,
+      "step": 18800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5035373704134436e-07,
+      "loss": 0.9353,
+      "step": 18801
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.50072223190596e-07,
+      "loss": 0.911,
+      "step": 18802
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.497909711396772e-07,
+      "loss": 0.8405,
+      "step": 18803
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4950998089606427e-07,
+      "loss": 0.8361,
+      "step": 18804
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4922925246722563e-07,
+      "loss": 0.8413,
+      "step": 18805
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4894878586062312e-07,
+      "loss": 0.8399,
+      "step": 18806
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4866858108371074e-07,
+      "loss": 0.8949,
+      "step": 18807
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.483886381439359e-07,
+      "loss": 0.8716,
+      "step": 18808
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.481089570487404e-07,
+      "loss": 0.8671,
+      "step": 18809
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4782953780555497e-07,
+      "loss": 0.9286,
+      "step": 18810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4755038042180925e-07,
+      "loss": 0.9324,
+      "step": 18811
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4727148490492284e-07,
+      "loss": 0.9462,
+      "step": 18812
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4699285126230645e-07,
+      "loss": 0.4382,
+      "step": 18813
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4671447950136641e-07,
+      "loss": 0.8673,
+      "step": 18814
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4643636962950347e-07,
+      "loss": 0.8218,
+      "step": 18815
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4615852165410726e-07,
+      "loss": 0.8558,
+      "step": 18816
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.45880935582563e-07,
+      "loss": 0.8957,
+      "step": 18817
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.456036114222492e-07,
+      "loss": 0.8529,
+      "step": 18818
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4532654918053668e-07,
+      "loss": 0.8457,
+      "step": 18819
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4504974886478951e-07,
+      "loss": 0.8029,
+      "step": 18820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4477321048236402e-07,
+      "loss": 0.7834,
+      "step": 18821
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4449693404061105e-07,
+      "loss": 0.8988,
+      "step": 18822
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4422091954687356e-07,
+      "loss": 0.92,
+      "step": 18823
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4394516700848794e-07,
+      "loss": 0.7573,
+      "step": 18824
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4366967643278273e-07,
+      "loss": 0.9026,
+      "step": 18825
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.43394447827081e-07,
+      "loss": 0.8962,
+      "step": 18826
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.431194811986969e-07,
+      "loss": 0.7784,
+      "step": 18827
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.428447765549401e-07,
+      "loss": 0.7569,
+      "step": 18828
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4257033390311148e-07,
+      "loss": 0.8631,
+      "step": 18829
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4229615325050406e-07,
+      "loss": 0.827,
+      "step": 18830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4202223460440645e-07,
+      "loss": 0.9424,
+      "step": 18831
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4174857797209951e-07,
+      "loss": 0.8745,
+      "step": 18832
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.414751833608552e-07,
+      "loss": 0.8148,
+      "step": 18833
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4120205077794102e-07,
+      "loss": 0.8694,
+      "step": 18834
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4092918023061784e-07,
+      "loss": 0.8231,
+      "step": 18835
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4065657172613656e-07,
+      "loss": 0.9067,
+      "step": 18836
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4038422527174245e-07,
+      "loss": 0.9176,
+      "step": 18837
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.401121408746753e-07,
+      "loss": 0.8184,
+      "step": 18838
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3984031854216597e-07,
+      "loss": 0.404,
+      "step": 18839
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.395687582814398e-07,
+      "loss": 0.826,
+      "step": 18840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3929746009971434e-07,
+      "loss": 0.932,
+      "step": 18841
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3902642400420162e-07,
+      "loss": 0.7473,
+      "step": 18842
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3875565000210255e-07,
+      "loss": 0.8401,
+      "step": 18843
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3848513810061582e-07,
+      "loss": 0.8399,
+      "step": 18844
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.382148883069323e-07,
+      "loss": 0.8097,
+      "step": 18845
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3794490062823297e-07,
+      "loss": 0.8952,
+      "step": 18846
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3767517507169426e-07,
+      "loss": 0.9153,
+      "step": 18847
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3740571164448714e-07,
+      "loss": 0.8802,
+      "step": 18848
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.371365103537703e-07,
+      "loss": 0.8048,
+      "step": 18849
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686757120670135e-07,
+      "loss": 0.9217,
+      "step": 18850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3659889421042792e-07,
+      "loss": 0.8932,
+      "step": 18851
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3633047937208987e-07,
+      "loss": 0.8336,
+      "step": 18852
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3606232669882257e-07,
+      "loss": 0.9536,
+      "step": 18853
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3579443619775368e-07,
+      "loss": 0.8955,
+      "step": 18854
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3552680787600193e-07,
+      "loss": 0.9149,
+      "step": 18855
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3525944174068163e-07,
+      "loss": 0.8984,
+      "step": 18856
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.349923377988982e-07,
+      "loss": 0.8598,
+      "step": 18857
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3472549605775154e-07,
+      "loss": 0.9189,
+      "step": 18858
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3445891652433485e-07,
+      "loss": 0.8682,
+      "step": 18859
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3419259920573246e-07,
+      "loss": 0.7751,
+      "step": 18860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3392654410902316e-07,
+      "loss": 0.8921,
+      "step": 18861
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3366075124127688e-07,
+      "loss": 0.8917,
+      "step": 18862
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3339522060956013e-07,
+      "loss": 0.9428,
+      "step": 18863
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.331299522209306e-07,
+      "loss": 0.8735,
+      "step": 18864
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3286494608243717e-07,
+      "loss": 0.9186,
+      "step": 18865
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3260020220112302e-07,
+      "loss": 0.8664,
+      "step": 18866
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3233572058402699e-07,
+      "loss": 0.9183,
+      "step": 18867
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3207150123817792e-07,
+      "loss": 0.8182,
+      "step": 18868
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3180754417059683e-07,
+      "loss": 0.8088,
+      "step": 18869
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3154384938830146e-07,
+      "loss": 0.8935,
+      "step": 18870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.312804168982995e-07,
+      "loss": 0.8243,
+      "step": 18871
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.310172467075932e-07,
+      "loss": 0.8383,
+      "step": 18872
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.307543388231769e-07,
+      "loss": 0.8649,
+      "step": 18873
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3049169325203947e-07,
+      "loss": 0.8546,
+      "step": 18874
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3022931000115867e-07,
+      "loss": 0.8953,
+      "step": 18875
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2996718907751226e-07,
+      "loss": 0.9357,
+      "step": 18876
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2970533048806466e-07,
+      "loss": 0.9398,
+      "step": 18877
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2944373423977697e-07,
+      "loss": 0.9535,
+      "step": 18878
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2918240033960028e-07,
+      "loss": 0.889,
+      "step": 18879
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2892132879448238e-07,
+      "loss": 0.9009,
+      "step": 18880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2866051961136217e-07,
+      "loss": 0.8389,
+      "step": 18881
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2839997279717075e-07,
+      "loss": 0.8664,
+      "step": 18882
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2813968835883372e-07,
+      "loss": 0.8385,
+      "step": 18883
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2787966630326888e-07,
+      "loss": 0.8384,
+      "step": 18884
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2761990663738732e-07,
+      "loss": 0.863,
+      "step": 18885
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2736040936809357e-07,
+      "loss": 0.8838,
+      "step": 18886
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2710117450228542e-07,
+      "loss": 0.8295,
+      "step": 18887
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2684220204685073e-07,
+      "loss": 0.8829,
+      "step": 18888
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2658349200867505e-07,
+      "loss": 0.3787,
+      "step": 18889
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2632504439463288e-07,
+      "loss": 0.8837,
+      "step": 18890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2606685921159545e-07,
+      "loss": 0.8224,
+      "step": 18891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2580893646642167e-07,
+      "loss": 0.8033,
+      "step": 18892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2555127616597053e-07,
+      "loss": 0.847,
+      "step": 18893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2529387831708873e-07,
+      "loss": 0.8229,
+      "step": 18894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.250367429266175e-07,
+      "loss": 0.8552,
+      "step": 18895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2477987000139023e-07,
+      "loss": 0.7906,
+      "step": 18896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2452325954823707e-07,
+      "loss": 0.8899,
+      "step": 18897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2426691157397587e-07,
+      "loss": 0.9465,
+      "step": 18898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2401082608542114e-07,
+      "loss": 0.8574,
+      "step": 18899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.237550030893797e-07,
+      "loss": 0.8891,
+      "step": 18900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2349944259265056e-07,
+      "loss": 0.9957,
+      "step": 18901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2324414460202494e-07,
+      "loss": 0.901,
+      "step": 18902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2298910912429074e-07,
+      "loss": 0.8181,
+      "step": 18903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2273433616622476e-07,
+      "loss": 0.8405,
+      "step": 18904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2247982573459938e-07,
+      "loss": 0.9108,
+      "step": 18905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2222557783618028e-07,
+      "loss": 0.8652,
+      "step": 18906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2197159247772206e-07,
+      "loss": 0.9063,
+      "step": 18907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.217178696659782e-07,
+      "loss": 0.9783,
+      "step": 18908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2146440940769e-07,
+      "loss": 0.9632,
+      "step": 18909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.212112117095976e-07,
+      "loss": 0.843,
+      "step": 18910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2095827657842674e-07,
+      "loss": 0.7672,
+      "step": 18911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2070560402090204e-07,
+      "loss": 0.9142,
+      "step": 18912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2045319404374035e-07,
+      "loss": 0.8989,
+      "step": 18913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.202010466536485e-07,
+      "loss": 0.8346,
+      "step": 18914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1994916185732896e-07,
+      "loss": 0.8667,
+      "step": 18915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1969753966147634e-07,
+      "loss": 0.8531,
+      "step": 18916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1944618007277865e-07,
+      "loss": 0.8329,
+      "step": 18917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1919508309791717e-07,
+      "loss": 0.8428,
+      "step": 18918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1894424874356547e-07,
+      "loss": 0.9348,
+      "step": 18919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1869367701639046e-07,
+      "loss": 0.8071,
+      "step": 18920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1844336792305122e-07,
+      "loss": 0.918,
+      "step": 18921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1819332147020246e-07,
+      "loss": 0.922,
+      "step": 18922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1794353766448885e-07,
+      "loss": 0.8203,
+      "step": 18923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1769401651254952e-07,
+      "loss": 0.8945,
+      "step": 18924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1744475802101695e-07,
+      "loss": 0.851,
+      "step": 18925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1719576219651585e-07,
+      "loss": 0.8438,
+      "step": 18926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1694702904566313e-07,
+      "loss": 0.8241,
+      "step": 18927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1669855857507018e-07,
+      "loss": 0.8564,
+      "step": 18928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1645035079134392e-07,
+      "loss": 0.9557,
+      "step": 18929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.16202405701078e-07,
+      "loss": 0.9607,
+      "step": 18930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1595472331086377e-07,
+      "loss": 0.9317,
+      "step": 18931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1570730362728488e-07,
+      "loss": 0.8865,
+      "step": 18932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1546014665691719e-07,
+      "loss": 0.8096,
+      "step": 18933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1521325240632875e-07,
+      "loss": 0.8174,
+      "step": 18934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.149666208820821e-07,
+      "loss": 0.809,
+      "step": 18935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1472025209073423e-07,
+      "loss": 0.815,
+      "step": 18936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.14474146038831e-07,
+      "loss": 0.9364,
+      "step": 18937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1422830273291497e-07,
+      "loss": 0.811,
+      "step": 18938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1398272217951977e-07,
+      "loss": 0.8953,
+      "step": 18939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1373740438517356e-07,
+      "loss": 0.9132,
+      "step": 18940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1349234935639441e-07,
+      "loss": 0.8355,
+      "step": 18941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1324755709969937e-07,
+      "loss": 0.843,
+      "step": 18942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1300302762159099e-07,
+      "loss": 0.8328,
+      "step": 18943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1275876092857075e-07,
+      "loss": 0.9082,
+      "step": 18944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1251475702713011e-07,
+      "loss": 0.9097,
+      "step": 18945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.12271015923755e-07,
+      "loss": 0.8699,
+      "step": 18946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1202753762492246e-07,
+      "loss": 0.3802,
+      "step": 18947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.117843221371051e-07,
+      "loss": 0.8492,
+      "step": 18948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1154136946676774e-07,
+      "loss": 0.93,
+      "step": 18949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1129867962036744e-07,
+      "loss": 0.8371,
+      "step": 18950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1105625260435238e-07,
+      "loss": 0.3696,
+      "step": 18951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1081408842516961e-07,
+      "loss": 0.4026,
+      "step": 18952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1057218708925288e-07,
+      "loss": 0.828,
+      "step": 18953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1033054860303261e-07,
+      "loss": 0.8722,
+      "step": 18954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1008917297293254e-07,
+      "loss": 0.386,
+      "step": 18955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0984806020536532e-07,
+      "loss": 0.8436,
+      "step": 18956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0960721030674138e-07,
+      "loss": 0.8846,
+      "step": 18957
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0936662328346226e-07,
+      "loss": 0.8764,
+      "step": 18958
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0912629914192285e-07,
+      "loss": 0.8387,
+      "step": 18959
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0888623788850916e-07,
+      "loss": 0.9448,
+      "step": 18960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0864643952960163e-07,
+      "loss": 0.8983,
+      "step": 18961
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0840690407157628e-07,
+      "loss": 0.9307,
+      "step": 18962
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.081676315207969e-07,
+      "loss": 0.9259,
+      "step": 18963
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0792862188362396e-07,
+      "loss": 0.871,
+      "step": 18964
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0768987516641238e-07,
+      "loss": 0.9223,
+      "step": 18965
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0745139137550486e-07,
+      "loss": 0.9153,
+      "step": 18966
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0721317051724078e-07,
+      "loss": 0.8023,
+      "step": 18967
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0697521259795174e-07,
+      "loss": 0.3828,
+      "step": 18968
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0673751762396378e-07,
+      "loss": 0.8287,
+      "step": 18969
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0650008560159186e-07,
+      "loss": 0.7836,
+      "step": 18970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0626291653714871e-07,
+      "loss": 0.8752,
+      "step": 18971
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0602601043693817e-07,
+      "loss": 0.3516,
+      "step": 18972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0578936730725631e-07,
+      "loss": 0.3993,
+      "step": 18973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0555298715439254e-07,
+      "loss": 0.8066,
+      "step": 18974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0531686998462853e-07,
+      "loss": 0.8775,
+      "step": 18975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0508101580424256e-07,
+      "loss": 0.8009,
+      "step": 18976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0484542461950187e-07,
+      "loss": 0.943,
+      "step": 18977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.046100964366692e-07,
+      "loss": 0.8472,
+      "step": 18978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0437503126199732e-07,
+      "loss": 0.8319,
+      "step": 18979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0414022910173571e-07,
+      "loss": 0.9306,
+      "step": 18980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0390568996212491e-07,
+      "loss": 0.8423,
+      "step": 18981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0367141384939882e-07,
+      "loss": 0.8517,
+      "step": 18982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0343740076978359e-07,
+      "loss": 0.8149,
+      "step": 18983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0320365072949867e-07,
+      "loss": 0.8791,
+      "step": 18984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0297016373475799e-07,
+      "loss": 0.8081,
+      "step": 18985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0273693979176658e-07,
+      "loss": 0.8824,
+      "step": 18986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.025039789067228e-07,
+      "loss": 0.8372,
+      "step": 18987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0227128108582063e-07,
+      "loss": 0.8907,
+      "step": 18988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0203884633524286e-07,
+      "loss": 0.8592,
+      "step": 18989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0180667466116789e-07,
+      "loss": 0.8853,
+      "step": 18990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0157476606976746e-07,
+      "loss": 0.9223,
+      "step": 18991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0134312056720441e-07,
+      "loss": 0.9036,
+      "step": 18992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0111173815963604e-07,
+      "loss": 1.0147,
+      "step": 18993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0088061885321076e-07,
+      "loss": 0.8508,
+      "step": 18994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0064976265407478e-07,
+      "loss": 0.8782,
+      "step": 18995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0041916956835985e-07,
+      "loss": 0.7965,
+      "step": 18996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0018883960219882e-07,
+      "loss": 0.8679,
+      "step": 18997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.995877276171129e-08,
+      "loss": 0.9182,
+      "step": 18998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.972896905301121e-08,
+      "loss": 0.7915,
+      "step": 18999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.949942848220928e-08,
+      "loss": 0.8545,
+      "step": 19000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.927015105540505e-08,
+      "loss": 0.9121,
+      "step": 19001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.904113677869143e-08,
+      "loss": 0.8892,
+      "step": 19002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.881238565815799e-08,
+      "loss": 0.9118,
+      "step": 19003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.858389769988208e-08,
+      "loss": 0.8106,
+      "step": 19004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.835567290993776e-08,
+      "loss": 0.8814,
+      "step": 19005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.812771129439125e-08,
+      "loss": 0.9812,
+      "step": 19006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.790001285930107e-08,
+      "loss": 0.923,
+      "step": 19007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.767257761071902e-08,
+      "loss": 0.9483,
+      "step": 19008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.744540555469028e-08,
+      "loss": 0.8385,
+      "step": 19009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.721849669725225e-08,
+      "loss": 0.8812,
+      "step": 19010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.699185104443786e-08,
+      "loss": 0.8943,
+      "step": 19011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.676546860226898e-08,
+      "loss": 0.9065,
+      "step": 19012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.653934937676413e-08,
+      "loss": 0.863,
+      "step": 19013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.631349337393181e-08,
+      "loss": 0.8682,
+      "step": 19014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.608790059977613e-08,
+      "loss": 0.8607,
+      "step": 19015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.586257106029229e-08,
+      "loss": 0.8991,
+      "step": 19016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.563750476147104e-08,
+      "loss": 0.8708,
+      "step": 19017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.541270170929096e-08,
+      "loss": 0.8094,
+      "step": 19018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.518816190973057e-08,
+      "loss": 0.8271,
+      "step": 19019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.496388536875623e-08,
+      "loss": 0.8699,
+      "step": 19020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.473987209232871e-08,
+      "loss": 0.8455,
+      "step": 19021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.451612208640215e-08,
+      "loss": 0.8308,
+      "step": 19022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.429263535692401e-08,
+      "loss": 0.9343,
+      "step": 19023
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.406941190983621e-08,
+      "loss": 0.8912,
+      "step": 19024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.384645175106733e-08,
+      "loss": 0.8143,
+      "step": 19025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.362375488654707e-08,
+      "loss": 0.8503,
+      "step": 19026
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.340132132219293e-08,
+      "loss": 0.8439,
+      "step": 19027
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.317915106391795e-08,
+      "loss": 0.4249,
+      "step": 19028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.295724411762519e-08,
+      "loss": 0.8191,
+      "step": 19029
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.273560048921437e-08,
+      "loss": 0.91,
+      "step": 19030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.251422018457745e-08,
+      "loss": 0.9478,
+      "step": 19031
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.229310320959528e-08,
+      "loss": 0.8505,
+      "step": 19032
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.20722495701476e-08,
+      "loss": 0.8999,
+      "step": 19033
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.185165927210305e-08,
+      "loss": 0.958,
+      "step": 19034
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.163133232132582e-08,
+      "loss": 0.8494,
+      "step": 19035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.141126872367123e-08,
+      "loss": 0.3482,
+      "step": 19036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.119146848498906e-08,
+      "loss": 0.9103,
+      "step": 19037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.097193161111905e-08,
+      "loss": 0.8283,
+      "step": 19038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.07526581078988e-08,
+      "loss": 0.8849,
+      "step": 19039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.053364798115582e-08,
+      "loss": 0.8367,
+      "step": 19040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.031490123671106e-08,
+      "loss": 0.9048,
+      "step": 19041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.00964178803776e-08,
+      "loss": 0.915,
+      "step": 19042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.987819791796304e-08,
+      "loss": 0.9165,
+      "step": 19043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.96602413552683e-08,
+      "loss": 0.3876,
+      "step": 19044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.944254819808428e-08,
+      "loss": 0.8428,
+      "step": 19045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.922511845219972e-08,
+      "loss": 0.8991,
+      "step": 19046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.900795212339108e-08,
+      "loss": 0.9025,
+      "step": 19047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.879104921743264e-08,
+      "loss": 0.8891,
+      "step": 19048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.857440974008757e-08,
+      "loss": 0.9381,
+      "step": 19049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.83580336971157e-08,
+      "loss": 0.8983,
+      "step": 19050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.814192109426689e-08,
+      "loss": 0.9135,
+      "step": 19051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.792607193728542e-08,
+      "loss": 0.8813,
+      "step": 19052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.771048623190781e-08,
+      "loss": 0.9567,
+      "step": 19053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.749516398386504e-08,
+      "loss": 0.858,
+      "step": 19054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.728010519887919e-08,
+      "loss": 0.8019,
+      "step": 19055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.706530988266682e-08,
+      "loss": 0.8793,
+      "step": 19056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.685077804093778e-08,
+      "loss": 0.8583,
+      "step": 19057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.663650967939196e-08,
+      "loss": 0.8602,
+      "step": 19058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.642250480372593e-08,
+      "loss": 0.8245,
+      "step": 19059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.620876341962736e-08,
+      "loss": 0.9101,
+      "step": 19060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.599528553277725e-08,
+      "loss": 0.9115,
+      "step": 19061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.578207114884996e-08,
+      "loss": 0.8397,
+      "step": 19062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.556912027351205e-08,
+      "loss": 0.7953,
+      "step": 19063
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.535643291242346e-08,
+      "loss": 0.9112,
+      "step": 19064
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.514400907123632e-08,
+      "loss": 0.9149,
+      "step": 19065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.493184875559946e-08,
+      "loss": 0.8704,
+      "step": 19066
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.471995197114836e-08,
+      "loss": 0.8902,
+      "step": 19067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.450831872351739e-08,
+      "loss": 0.8779,
+      "step": 19068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.429694901832985e-08,
+      "loss": 0.862,
+      "step": 19069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.408584286120457e-08,
+      "loss": 0.9438,
+      "step": 19070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.387500025775374e-08,
+      "loss": 0.8369,
+      "step": 19071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.366442121357954e-08,
+      "loss": 0.4097,
+      "step": 19072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.34541057342797e-08,
+      "loss": 0.8258,
+      "step": 19073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.324405382544309e-08,
+      "loss": 0.8585,
+      "step": 19074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.303426549265414e-08,
+      "loss": 0.7848,
+      "step": 19075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.282474074148838e-08,
+      "loss": 0.8896,
+      "step": 19076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.26154795775147e-08,
+      "loss": 0.8622,
+      "step": 19077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.24064820062942e-08,
+      "loss": 0.9644,
+      "step": 19078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.219774803338243e-08,
+      "loss": 0.8873,
+      "step": 19079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.198927766432829e-08,
+      "loss": 0.9427,
+      "step": 19080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.178107090466958e-08,
+      "loss": 0.9219,
+      "step": 19081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.157312775994409e-08,
+      "loss": 0.844,
+      "step": 19082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.136544823567627e-08,
+      "loss": 0.9047,
+      "step": 19083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.115803233738728e-08,
+      "loss": 0.8208,
+      "step": 19084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.095088007058827e-08,
+      "loss": 0.8518,
+      "step": 19085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.074399144078816e-08,
+      "loss": 0.8292,
+      "step": 19086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.053736645348253e-08,
+      "loss": 0.9266,
+      "step": 19087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.03310051141637e-08,
+      "loss": 0.8745,
+      "step": 19088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.012490742831946e-08,
+      "loss": 0.9309,
+      "step": 19089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.991907340142435e-08,
+      "loss": 0.8914,
+      "step": 19090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.971350303895064e-08,
+      "loss": 0.843,
+      "step": 19091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.950819634636176e-08,
+      "loss": 0.8161,
+      "step": 19092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.930315332911442e-08,
+      "loss": 0.9094,
+      "step": 19093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.909837399265986e-08,
+      "loss": 0.9289,
+      "step": 19094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.889385834243923e-08,
+      "loss": 0.8641,
+      "step": 19095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.868960638388933e-08,
+      "loss": 0.8086,
+      "step": 19096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.848561812243804e-08,
+      "loss": 0.8496,
+      "step": 19097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.828189356350657e-08,
+      "loss": 0.4243,
+      "step": 19098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.807843271251281e-08,
+      "loss": 0.907,
+      "step": 19099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.787523557486021e-08,
+      "loss": 0.8462,
+      "step": 19100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.767230215595334e-08,
+      "loss": 0.9004,
+      "step": 19101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.746963246118344e-08,
+      "loss": 0.8727,
+      "step": 19102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.726722649593842e-08,
+      "loss": 0.8319,
+      "step": 19103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.706508426559733e-08,
+      "loss": 0.8816,
+      "step": 19104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.686320577553363e-08,
+      "loss": 0.9368,
+      "step": 19105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.666159103111193e-08,
+      "loss": 0.8032,
+      "step": 19106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.646024003769237e-08,
+      "loss": 0.9328,
+      "step": 19107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.625915280062513e-08,
+      "loss": 0.9222,
+      "step": 19108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.605832932525703e-08,
+      "loss": 0.9396,
+      "step": 19109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.585776961692271e-08,
+      "loss": 0.8605,
+      "step": 19110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.565747368095566e-08,
+      "loss": 0.3953,
+      "step": 19111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.54574415226772e-08,
+      "loss": 0.8337,
+      "step": 19112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.525767314740528e-08,
+      "loss": 0.9165,
+      "step": 19113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.505816856045012e-08,
+      "loss": 1.0685,
+      "step": 19114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.485892776711302e-08,
+      "loss": 0.8318,
+      "step": 19115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.465995077268972e-08,
+      "loss": 0.9588,
+      "step": 19116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.446123758246826e-08,
+      "loss": 0.8914,
+      "step": 19117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.426278820173328e-08,
+      "loss": 0.8551,
+      "step": 19118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.406460263575499e-08,
+      "loss": 0.8975,
+      "step": 19119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.386668088980364e-08,
+      "loss": 0.8924,
+      "step": 19120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.366902296913947e-08,
+      "loss": 0.8838,
+      "step": 19121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.347162887901605e-08,
+      "loss": 0.9117,
+      "step": 19122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.327449862467917e-08,
+      "loss": 0.8557,
+      "step": 19123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.30776322113691e-08,
+      "loss": 0.8932,
+      "step": 19124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.288102964431831e-08,
+      "loss": 0.8652,
+      "step": 19125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.268469092875042e-08,
+      "loss": 0.9402,
+      "step": 19126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.248861606988677e-08,
+      "loss": 0.938,
+      "step": 19127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.229280507293657e-08,
+      "loss": 0.7937,
+      "step": 19128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.209725794310452e-08,
+      "loss": 0.9102,
+      "step": 19129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.19019746855898e-08,
+      "loss": 0.8187,
+      "step": 19130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.170695530558158e-08,
+      "loss": 0.8692,
+      "step": 19131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.15121998082624e-08,
+      "loss": 0.8337,
+      "step": 19132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.131770819881035e-08,
+      "loss": 0.8608,
+      "step": 19133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.112348048239348e-08,
+      "loss": 0.8747,
+      "step": 19134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.092951666417435e-08,
+      "loss": 0.8101,
+      "step": 19135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.073581674930996e-08,
+      "loss": 0.4359,
+      "step": 19136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.054238074294618e-08,
+      "loss": 0.9362,
+      "step": 19137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.034920865022555e-08,
+      "loss": 0.858,
+      "step": 19138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.015630047628174e-08,
+      "loss": 0.8642,
+      "step": 19139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.996365622624291e-08,
+      "loss": 0.8143,
+      "step": 19140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.977127590522937e-08,
+      "loss": 0.866,
+      "step": 19141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.95791595183526e-08,
+      "loss": 0.8615,
+      "step": 19142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.938730707072184e-08,
+      "loss": 0.8673,
+      "step": 19143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.919571856743412e-08,
+      "loss": 0.918,
+      "step": 19144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.900439401358094e-08,
+      "loss": 0.8968,
+      "step": 19145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.881333341425045e-08,
+      "loss": 0.8485,
+      "step": 19146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.862253677451747e-08,
+      "loss": 0.8013,
+      "step": 19147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.843200409945571e-08,
+      "loss": 0.8365,
+      "step": 19148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.824173539412782e-08,
+      "loss": 0.8231,
+      "step": 19149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.805173066359194e-08,
+      "loss": 0.8373,
+      "step": 19150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.786198991289738e-08,
+      "loss": 0.8506,
+      "step": 19151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.767251314708789e-08,
+      "loss": 0.9502,
+      "step": 19152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.748330037119832e-08,
+      "loss": 0.8619,
+      "step": 19153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.729435159026021e-08,
+      "loss": 0.8236,
+      "step": 19154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.710566680929286e-08,
+      "loss": 0.9328,
+      "step": 19155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.691724603331229e-08,
+      "loss": 0.8806,
+      "step": 19156
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.672908926732669e-08,
+      "loss": 0.9522,
+      "step": 19157
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.654119651633761e-08,
+      "loss": 0.7622,
+      "step": 19158
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.635356778533774e-08,
+      "loss": 0.8981,
+      "step": 19159
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.616620307931531e-08,
+      "loss": 0.7778,
+      "step": 19160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.597910240324967e-08,
+      "loss": 0.8858,
+      "step": 19161
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.57922657621135e-08,
+      "loss": 0.8299,
+      "step": 19162
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.560569316087284e-08,
+      "loss": 0.9432,
+      "step": 19163
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.541938460448705e-08,
+      "loss": 0.8295,
+      "step": 19164
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.523334009790661e-08,
+      "loss": 0.8785,
+      "step": 19165
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.504755964607867e-08,
+      "loss": 0.334,
+      "step": 19166
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.48620432539393e-08,
+      "loss": 0.8222,
+      "step": 19167
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.46767909264201e-08,
+      "loss": 0.8043,
+      "step": 19168
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.44918026684438e-08,
+      "loss": 0.8591,
+      "step": 19169
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.430707848492868e-08,
+      "loss": 0.8675,
+      "step": 19170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.412261838078415e-08,
+      "loss": 0.8432,
+      "step": 19171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.393842236091185e-08,
+      "loss": 0.8973,
+      "step": 19172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.375449043020787e-08,
+      "loss": 0.8041,
+      "step": 19173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.357082259356273e-08,
+      "loss": 0.8584,
+      "step": 19174
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.338741885585476e-08,
+      "loss": 0.3699,
+      "step": 19175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.320427922196115e-08,
+      "loss": 0.8368,
+      "step": 19176
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.302140369674914e-08,
+      "loss": 0.8406,
+      "step": 19177
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.283879228507927e-08,
+      "loss": 0.9163,
+      "step": 19178
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.265644499180435e-08,
+      "loss": 0.9171,
+      "step": 19179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.247436182177268e-08,
+      "loss": 0.9338,
+      "step": 19180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.229254277982155e-08,
+      "loss": 0.8772,
+      "step": 19181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.211098787078596e-08,
+      "loss": 0.8906,
+      "step": 19182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.19296970994887e-08,
+      "loss": 0.89,
+      "step": 19183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.17486704707515e-08,
+      "loss": 0.8478,
+      "step": 19184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.156790798938273e-08,
+      "loss": 0.8436,
+      "step": 19185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.138740966018852e-08,
+      "loss": 0.9109,
+      "step": 19186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.120717548796618e-08,
+      "loss": 0.8671,
+      "step": 19187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.102720547750519e-08,
+      "loss": 0.8731,
+      "step": 19188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.084749963358948e-08,
+      "loss": 0.8397,
+      "step": 19189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.066805796099639e-08,
+      "loss": 0.8246,
+      "step": 19190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.048888046449431e-08,
+      "loss": 0.8675,
+      "step": 19191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.030996714884607e-08,
+      "loss": 0.9257,
+      "step": 19192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.013131801880678e-08,
+      "loss": 0.8601,
+      "step": 19193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.995293307912487e-08,
+      "loss": 0.8879,
+      "step": 19194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.977481233454097e-08,
+      "loss": 0.8664,
+      "step": 19195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.959695578978908e-08,
+      "loss": 0.8801,
+      "step": 19196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.941936344959765e-08,
+      "loss": 0.8858,
+      "step": 19197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.924203531868622e-08,
+      "loss": 0.901,
+      "step": 19198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.90649714017677e-08,
+      "loss": 0.9417,
+      "step": 19199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.888817170354943e-08,
+      "loss": 0.9556,
+      "step": 19200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8711636228729885e-08,
+      "loss": 0.9155,
+      "step": 19201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8535364981999744e-08,
+      "loss": 0.8417,
+      "step": 19202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.835935796804637e-08,
+      "loss": 0.9474,
+      "step": 19203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.818361519154714e-08,
+      "loss": 0.8683,
+      "step": 19204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8008136657171646e-08,
+      "loss": 0.8356,
+      "step": 19205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.783292236958504e-08,
+      "loss": 0.9111,
+      "step": 19206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.765797233344583e-08,
+      "loss": 0.965,
+      "step": 19207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.748328655340141e-08,
+      "loss": 0.8309,
+      "step": 19208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.730886503409583e-08,
+      "loss": 0.8691,
+      "step": 19209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.713470778016539e-08,
+      "loss": 0.8324,
+      "step": 19210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6960814796237496e-08,
+      "loss": 0.8041,
+      "step": 19211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.678718608693512e-08,
+      "loss": 0.9071,
+      "step": 19212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6613821656874566e-08,
+      "loss": 0.8662,
+      "step": 19213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6440721510661046e-08,
+      "loss": 0.8483,
+      "step": 19214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.626788565289753e-08,
+      "loss": 0.7854,
+      "step": 19215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6095314088175926e-08,
+      "loss": 0.8199,
+      "step": 19216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.592300682108365e-08,
+      "loss": 0.9256,
+      "step": 19217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.575096385620149e-08,
+      "loss": 0.9316,
+      "step": 19218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.557918519810024e-08,
+      "loss": 0.8818,
+      "step": 19219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5407670851348464e-08,
+      "loss": 0.83,
+      "step": 19220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.52364208205014e-08,
+      "loss": 0.9421,
+      "step": 19221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5065435110112086e-08,
+      "loss": 0.9085,
+      "step": 19222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4894713724725765e-08,
+      "loss": 0.8343,
+      "step": 19223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4724256668879925e-08,
+      "loss": 0.8733,
+      "step": 19224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4554063947104275e-08,
+      "loss": 0.918,
+      "step": 19225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.438413556392297e-08,
+      "loss": 0.8406,
+      "step": 19226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4214471523852396e-08,
+      "loss": 0.8662,
+      "step": 19227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.404507183140228e-08,
+      "loss": 0.8883,
+      "step": 19228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.387593649107459e-08,
+      "loss": 0.8186,
+      "step": 19229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.37070655073646e-08,
+      "loss": 0.8458,
+      "step": 19230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.353845888476206e-08,
+      "loss": 0.3833,
+      "step": 19231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3370116627745605e-08,
+      "loss": 0.3803,
+      "step": 19232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.320203874079277e-08,
+      "loss": 0.8387,
+      "step": 19233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3034225228368876e-08,
+      "loss": 0.8254,
+      "step": 19234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2866676094934786e-08,
+      "loss": 0.8837,
+      "step": 19235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2699391344943616e-08,
+      "loss": 0.8732,
+      "step": 19236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2532370982841805e-08,
+      "loss": 0.853,
+      "step": 19237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2365615013069136e-08,
+      "loss": 0.9292,
+      "step": 19238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.219912344005651e-08,
+      "loss": 0.7996,
+      "step": 19239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.203289626822927e-08,
+      "loss": 0.8583,
+      "step": 19240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1866933502007224e-08,
+      "loss": 0.7923,
+      "step": 19241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.170123514579905e-08,
+      "loss": 0.9238,
+      "step": 19242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.153580120401014e-08,
+      "loss": 0.8652,
+      "step": 19243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.137063168103806e-08,
+      "loss": 0.8778,
+      "step": 19244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.120572658127154e-08,
+      "loss": 0.8133,
+      "step": 19245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1041085909094844e-08,
+      "loss": 0.8405,
+      "step": 19246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0876709668882254e-08,
+      "loss": 0.857,
+      "step": 19247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.071259786500471e-08,
+      "loss": 0.8094,
+      "step": 19248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.054875050182317e-08,
+      "loss": 0.8719,
+      "step": 19249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.038516758369194e-08,
+      "loss": 0.8977,
+      "step": 19250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.022184911495864e-08,
+      "loss": 0.9022,
+      "step": 19251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0058795099966474e-08,
+      "loss": 0.7858,
+      "step": 19252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9896005543047524e-08,
+      "loss": 0.9451,
+      "step": 19253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.973348044852722e-08,
+      "loss": 0.8761,
+      "step": 19254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.957121982072766e-08,
+      "loss": 0.8602,
+      "step": 19255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.940922366396095e-08,
+      "loss": 0.8358,
+      "step": 19256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9247491982531423e-08,
+      "loss": 0.7464,
+      "step": 19257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.908602478073898e-08,
+      "loss": 0.9267,
+      "step": 19258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8924822062875745e-08,
+      "loss": 0.8525,
+      "step": 19259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8763883833224945e-08,
+      "loss": 0.8403,
+      "step": 19260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.860321009606428e-08,
+      "loss": 0.8938,
+      "step": 19261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.844280085566588e-08,
+      "loss": 0.8745,
+      "step": 19262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8282656116290794e-08,
+      "loss": 0.9081,
+      "step": 19263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.812277588219782e-08,
+      "loss": 0.9315,
+      "step": 19264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.796316015763469e-08,
+      "loss": 0.8589,
+      "step": 19265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7803808946844664e-08,
+      "loss": 0.8521,
+      "step": 19266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.764472225406325e-08,
+      "loss": 0.8153,
+      "step": 19267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.748590008351706e-08,
+      "loss": 0.9207,
+      "step": 19268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.732734243943049e-08,
+      "loss": 0.8209,
+      "step": 19269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7169049326014626e-08,
+      "loss": 0.88,
+      "step": 19270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.701102074747943e-08,
+      "loss": 0.8556,
+      "step": 19271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.685325670802376e-08,
+      "loss": 0.8028,
+      "step": 19272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.669575721183983e-08,
+      "loss": 0.9718,
+      "step": 19273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.65385222631165e-08,
+      "loss": 0.8638,
+      "step": 19274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.638155186603044e-08,
+      "loss": 0.8047,
+      "step": 19275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6224846024754964e-08,
+      "loss": 0.8908,
+      "step": 19276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.606840474345342e-08,
+      "loss": 0.8169,
+      "step": 19277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.591222802628692e-08,
+      "loss": 0.9312,
+      "step": 19278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5756315877404366e-08,
+      "loss": 0.9099,
+      "step": 19279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.560066830094911e-08,
+      "loss": 0.9159,
+      "step": 19280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.544528530106007e-08,
+      "loss": 0.8318,
+      "step": 19281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.529016688186616e-08,
+      "loss": 0.9553,
+      "step": 19282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5135313047489634e-08,
+      "loss": 0.805,
+      "step": 19283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.498072380204721e-08,
+      "loss": 0.873,
+      "step": 19284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.482639914964782e-08,
+      "loss": 0.8478,
+      "step": 19285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4672339094392614e-08,
+      "loss": 0.9131,
+      "step": 19286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.451854364037611e-08,
+      "loss": 0.8852,
+      "step": 19287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4365012791687256e-08,
+      "loss": 0.8245,
+      "step": 19288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.421174655240501e-08,
+      "loss": 0.8945,
+      "step": 19289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4058744926605005e-08,
+      "loss": 0.8787,
+      "step": 19290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3906007918352865e-08,
+      "loss": 0.915,
+      "step": 19291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.375353553170647e-08,
+      "loss": 0.8833,
+      "step": 19292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3601327770721456e-08,
+      "loss": 0.8695,
+      "step": 19293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.344938463944126e-08,
+      "loss": 0.9219,
+      "step": 19294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.329770614190487e-08,
+      "loss": 0.8469,
+      "step": 19295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.314629228214351e-08,
+      "loss": 0.8184,
+      "step": 19296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.299514306418284e-08,
+      "loss": 0.8879,
+      "step": 19297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2844258492038546e-08,
+      "loss": 0.8511,
+      "step": 19298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.269363856972186e-08,
+      "loss": 0.8365,
+      "step": 19299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.254328330123625e-08,
+      "loss": 0.886,
+      "step": 19300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.23931926905774e-08,
+      "loss": 0.8151,
+      "step": 19301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2243366741735457e-08,
+      "loss": 0.872,
+      "step": 19302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2093805458690576e-08,
+      "loss": 0.9185,
+      "step": 19303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.194450884542067e-08,
+      "loss": 0.9245,
+      "step": 19304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.179547690589258e-08,
+      "loss": 0.8282,
+      "step": 19305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.164670964406536e-08,
+      "loss": 0.8017,
+      "step": 19306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.149820706389696e-08,
+      "loss": 0.8824,
+      "step": 19307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.134996916933198e-08,
+      "loss": 0.8364,
+      "step": 19308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1201995964310623e-08,
+      "loss": 0.8574,
+      "step": 19309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.105428745276752e-08,
+      "loss": 0.8435,
+      "step": 19310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.090684363862618e-08,
+      "loss": 0.9028,
+      "step": 19311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.075966452580682e-08,
+      "loss": 0.8673,
+      "step": 19312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0612750118220747e-08,
+      "loss": 0.855,
+      "step": 19313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0466100419773725e-08,
+      "loss": 0.8848,
+      "step": 19314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.031971543436153e-08,
+      "loss": 0.9251,
+      "step": 19315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.017359516587771e-08,
+      "loss": 0.9187,
+      "step": 19316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.002773961820361e-08,
+      "loss": 0.8446,
+      "step": 19317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9882148795217236e-08,
+      "loss": 0.8272,
+      "step": 19318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9736822700786604e-08,
+      "loss": 0.8392,
+      "step": 19319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.959176133877751e-08,
+      "loss": 0.8492,
+      "step": 19320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.944696471304244e-08,
+      "loss": 0.8517,
+      "step": 19321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9302432827430516e-08,
+      "loss": 0.9482,
+      "step": 19322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9158165685785345e-08,
+      "loss": 0.8784,
+      "step": 19323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.90141632919383e-08,
+      "loss": 0.8746,
+      "step": 19324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8870425649718544e-08,
+      "loss": 0.8235,
+      "step": 19325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8726952762946357e-08,
+      "loss": 0.8529,
+      "step": 19326
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.858374463543535e-08,
+      "loss": 0.8727,
+      "step": 19327
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.844080127099137e-08,
+      "loss": 0.9272,
+      "step": 19328
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8298122673412484e-08,
+      "loss": 0.9196,
+      "step": 19329
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.815570884649345e-08,
+      "loss": 0.8936,
+      "step": 19330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8013559794017884e-08,
+      "loss": 0.8634,
+      "step": 19331
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7871675519763897e-08,
+      "loss": 0.8558,
+      "step": 19332
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.77300560275029e-08,
+      "loss": 0.8982,
+      "step": 19333
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.758870132099857e-08,
+      "loss": 0.8333,
+      "step": 19334
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7447611404008985e-08,
+      "loss": 0.8597,
+      "step": 19335
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.730678628028339e-08,
+      "loss": 0.8751,
+      "step": 19336
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.716622595356545e-08,
+      "loss": 0.7617,
+      "step": 19337
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.702593042758995e-08,
+      "loss": 0.8193,
+      "step": 19338
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.688589970608725e-08,
+      "loss": 0.858,
+      "step": 19339
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.674613379277769e-08,
+      "loss": 0.918,
+      "step": 19340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.660663269137721e-08,
+      "loss": 0.9093,
+      "step": 19341
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6467396405591713e-08,
+      "loss": 0.825,
+      "step": 19342
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.632842493912492e-08,
+      "loss": 0.8314,
+      "step": 19343
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.618971829566831e-08,
+      "loss": 0.8896,
+      "step": 19344
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6051276478910044e-08,
+      "loss": 0.869,
+      "step": 19345
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5913099492528305e-08,
+      "loss": 0.4084,
+      "step": 19346
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5775187340196805e-08,
+      "loss": 0.8811,
+      "step": 19347
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.563754002557929e-08,
+      "loss": 0.8116,
+      "step": 19348
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.550015755233727e-08,
+      "loss": 0.8921,
+      "step": 19349
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5363039924118934e-08,
+      "loss": 0.9114,
+      "step": 19350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5226187144571376e-08,
+      "loss": 0.8477,
+      "step": 19351
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5089599217330575e-08,
+      "loss": 0.4078,
+      "step": 19352
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.495327614602695e-08,
+      "loss": 0.8771,
+      "step": 19353
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4817217934284276e-08,
+      "loss": 0.8793,
+      "step": 19354
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4681424585718546e-08,
+      "loss": 0.8583,
+      "step": 19355
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4545896103939106e-08,
+      "loss": 0.8686,
+      "step": 19356
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.441063249254861e-08,
+      "loss": 0.3939,
+      "step": 19357
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.427563375514087e-08,
+      "loss": 0.9263,
+      "step": 19358
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.414089989530522e-08,
+      "loss": 0.8758,
+      "step": 19359
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4006430916623253e-08,
+      "loss": 0.8658,
+      "step": 19360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3872226822666555e-08,
+      "loss": 0.8134,
+      "step": 19361
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3738287617004484e-08,
+      "loss": 0.8782,
+      "step": 19362
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.360461330319642e-08,
+      "loss": 0.7887,
+      "step": 19363
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3471203884793965e-08,
+      "loss": 0.8789,
+      "step": 19364
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.333805936534429e-08,
+      "loss": 0.7895,
+      "step": 19365
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3205179748386775e-08,
+      "loss": 0.9056,
+      "step": 19366
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.307256503745193e-08,
+      "loss": 0.8431,
+      "step": 19367
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.294021523606472e-08,
+      "loss": 0.851,
+      "step": 19368
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2808130347744546e-08,
+      "loss": 0.7977,
+      "step": 19369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.267631037599972e-08,
+      "loss": 1.0205,
+      "step": 19370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.25447553243341e-08,
+      "loss": 0.7786,
+      "step": 19371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.241346519624489e-08,
+      "loss": 0.8677,
+      "step": 19372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.228243999522263e-08,
+      "loss": 0.8633,
+      "step": 19373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2151679724748974e-08,
+      "loss": 0.8861,
+      "step": 19374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.202118438829893e-08,
+      "loss": 0.8923,
+      "step": 19375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1890953989340834e-08,
+      "loss": 0.9224,
+      "step": 19376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1760988531337466e-08,
+      "loss": 0.8797,
+      "step": 19377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.163128801774162e-08,
+      "loss": 0.901,
+      "step": 19378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.150185245200055e-08,
+      "loss": 0.8481,
+      "step": 19379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.137268183755593e-08,
+      "loss": 0.8872,
+      "step": 19380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.124377617784058e-08,
+      "loss": 0.7682,
+      "step": 19381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1115135476279537e-08,
+      "loss": 0.8373,
+      "step": 19382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.098675973629228e-08,
+      "loss": 0.9089,
+      "step": 19383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.085864896129165e-08,
+      "loss": 0.8597,
+      "step": 19384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.073080315468269e-08,
+      "loss": 0.846,
+      "step": 19385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.060322231986268e-08,
+      "loss": 0.8632,
+      "step": 19386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.047590646022336e-08,
+      "loss": 0.8651,
+      "step": 19387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0348855579147574e-08,
+      "loss": 0.4096,
+      "step": 19388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0222069680012625e-08,
+      "loss": 0.8336,
+      "step": 19389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.009554876618914e-08,
+      "loss": 0.9109,
+      "step": 19390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.996929284103889e-08,
+      "loss": 0.8516,
+      "step": 19391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9843301907918064e-08,
+      "loss": 0.8549,
+      "step": 19392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9717575970175105e-08,
+      "loss": 0.8513,
+      "step": 19393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.959211503115178e-08,
+      "loss": 0.9402,
+      "step": 19394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.946691909418209e-08,
+      "loss": 0.8554,
+      "step": 19395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9341988162595593e-08,
+      "loss": 0.8816,
+      "step": 19396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9217322239710744e-08,
+      "loss": 0.8026,
+      "step": 19397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9092921328841562e-08,
+      "loss": 0.8143,
+      "step": 19398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.89687854332954e-08,
+      "loss": 0.8828,
+      "step": 19399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.884491455637073e-08,
+      "loss": 0.8534,
+      "step": 19400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8721308701359363e-08,
+      "loss": 0.8663,
+      "step": 19401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.859796787154756e-08,
+      "loss": 0.8575,
+      "step": 19402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.847489207021381e-08,
+      "loss": 0.8991,
+      "step": 19403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8352081300628832e-08,
+      "loss": 0.7854,
+      "step": 19404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8229535566056676e-08,
+      "loss": 0.8634,
+      "step": 19405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.810725486975474e-08,
+      "loss": 0.7977,
+      "step": 19406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.798523921497265e-08,
+      "loss": 0.8594,
+      "step": 19407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.786348860495336e-08,
+      "loss": 0.8605,
+      "step": 19408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7742003042934286e-08,
+      "loss": 0.9618,
+      "step": 19409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7620782532142844e-08,
+      "loss": 0.8931,
+      "step": 19410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.74998270758009e-08,
+      "loss": 0.8831,
+      "step": 19411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7379136677124772e-08,
+      "loss": 0.9367,
+      "step": 19412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7258711339320787e-08,
+      "loss": 0.8693,
+      "step": 19413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7138551065590825e-08,
+      "loss": 0.8624,
+      "step": 19414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7018655859127884e-08,
+      "loss": 0.9288,
+      "step": 19415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.689902572311831e-08,
+      "loss": 0.8548,
+      "step": 19416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.677966066074289e-08,
+      "loss": 0.836,
+      "step": 19417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.666056067517353e-08,
+      "loss": 0.8197,
+      "step": 19418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6541725769576587e-08,
+      "loss": 0.8574,
+      "step": 19419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6423155947108425e-08,
+      "loss": 0.9255,
+      "step": 19420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6304851210922078e-08,
+      "loss": 0.8595,
+      "step": 19421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6186811564162805e-08,
+      "loss": 0.9401,
+      "step": 19422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.606903700996588e-08,
+      "loss": 0.9504,
+      "step": 19423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.595152755146213e-08,
+      "loss": 0.9403,
+      "step": 19424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.583428319177683e-08,
+      "loss": 0.9542,
+      "step": 19425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.571730393402305e-08,
+      "loss": 0.8983,
+      "step": 19426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5600589781312746e-08,
+      "loss": 0.8845,
+      "step": 19427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5484140736746766e-08,
+      "loss": 0.8819,
+      "step": 19428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5367956803420414e-08,
+      "loss": 0.8003,
+      "step": 19429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5252037984421217e-08,
+      "loss": 0.8167,
+      "step": 19430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.513638428283116e-08,
+      "loss": 0.8515,
+      "step": 19431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5020995701723338e-08,
+      "loss": 0.9402,
+      "step": 19432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4905872244164186e-08,
+      "loss": 0.8617,
+      "step": 19433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.47910139132157e-08,
+      "loss": 0.9508,
+      "step": 19434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4676420711928774e-08,
+      "loss": 0.9115,
+      "step": 19435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4562092643348746e-08,
+      "loss": 0.8756,
+      "step": 19436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4448029710515407e-08,
+      "loss": 0.8994,
+      "step": 19437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.433423191646189e-08,
+      "loss": 0.8892,
+      "step": 19438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4220699264210225e-08,
+      "loss": 0.8882,
+      "step": 19439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4107431756777987e-08,
+      "loss": 0.8033,
+      "step": 19440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3994429397178332e-08,
+      "loss": 0.9189,
+      "step": 19441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3881692188412185e-08,
+      "loss": 0.8988,
+      "step": 19442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3769220133477156e-08,
+      "loss": 0.8987,
+      "step": 19443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.365701323536196e-08,
+      "loss": 0.882,
+      "step": 19444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3545071497048654e-08,
+      "loss": 0.8504,
+      "step": 19445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3433394921514862e-08,
+      "loss": 0.8084,
+      "step": 19446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3321983511725988e-08,
+      "loss": 0.7681,
+      "step": 19447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3210837270644106e-08,
+      "loss": 0.889,
+      "step": 19448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3099956201223517e-08,
+      "loss": 0.864,
+      "step": 19449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2989340306411867e-08,
+      "loss": 0.8779,
+      "step": 19450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2878989589149024e-08,
+      "loss": 0.8831,
+      "step": 19451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.276890405236709e-08,
+      "loss": 0.9183,
+      "step": 19452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2659083698992613e-08,
+      "loss": 0.856,
+      "step": 19453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2549528531945474e-08,
+      "loss": 0.8928,
+      "step": 19454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2440238554135572e-08,
+      "loss": 0.8685,
+      "step": 19455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2331213768468363e-08,
+      "loss": 0.9348,
+      "step": 19456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2222454177842635e-08,
+      "loss": 0.8452,
+      "step": 19457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.21139597851483e-08,
+      "loss": 0.8904,
+      "step": 19458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2005730593268605e-08,
+      "loss": 0.8878,
+      "step": 19459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.189776660508125e-08,
+      "loss": 0.8665,
+      "step": 19460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1790067823453942e-08,
+      "loss": 0.844,
+      "step": 19461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1682634251252166e-08,
+      "loss": 0.8417,
+      "step": 19462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1575465891328084e-08,
+      "loss": 0.8556,
+      "step": 19463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.146856274653275e-08,
+      "loss": 0.88,
+      "step": 19464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1361924819706114e-08,
+      "loss": 0.845,
+      "step": 19465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1255552113682576e-08,
+      "loss": 0.8972,
+      "step": 19466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1149444631289873e-08,
+      "loss": 0.877,
+      "step": 19467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.104360237534797e-08,
+      "loss": 0.8491,
+      "step": 19468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0938025348670177e-08,
+      "loss": 0.8759,
+      "step": 19469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0832713554062022e-08,
+      "loss": 0.8934,
+      "step": 19470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.072766699432349e-08,
+      "loss": 0.4167,
+      "step": 19471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0622885672245685e-08,
+      "loss": 0.9444,
+      "step": 19472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.051836959061415e-08,
+      "loss": 0.9621,
+      "step": 19473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.041411875220778e-08,
+      "loss": 0.9085,
+      "step": 19474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0310133159795463e-08,
+      "loss": 0.8861,
+      "step": 19475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0206412816141664e-08,
+      "loss": 0.7921,
+      "step": 19476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.010295772400417e-08,
+      "loss": 0.8361,
+      "step": 19477
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9999767886130783e-08,
+      "loss": 0.9228,
+      "step": 19478
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9896843305265978e-08,
+      "loss": 0.849,
+      "step": 19479
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9794183984145344e-08,
+      "loss": 0.8771,
+      "step": 19480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9691789925495585e-08,
+      "loss": 0.8515,
+      "step": 19481
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.958966113204008e-08,
+      "loss": 0.8242,
+      "step": 19482
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.948779760649222e-08,
+      "loss": 0.8361,
+      "step": 19483
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.938619935156094e-08,
+      "loss": 0.9185,
+      "step": 19484
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.928486636994409e-08,
+      "loss": 0.8768,
+      "step": 19485
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9183798664337284e-08,
+      "loss": 0.8604,
+      "step": 19486
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9082996237426156e-08,
+      "loss": 0.8526,
+      "step": 19487
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8982459091889673e-08,
+      "loss": 0.922,
+      "step": 19488
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.888218723039903e-08,
+      "loss": 0.4442,
+      "step": 19489
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8782180655622096e-08,
+      "loss": 0.8873,
+      "step": 19490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.868243937021341e-08,
+      "loss": 0.8665,
+      "step": 19491
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8582963376826413e-08,
+      "loss": 0.8255,
+      "step": 19492
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.848375267810454e-08,
+      "loss": 0.8518,
+      "step": 19493
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.838480727668457e-08,
+      "loss": 0.7896,
+      "step": 19494
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.828612717519551e-08,
+      "loss": 0.9196,
+      "step": 19495
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8187712376261936e-08,
+      "loss": 0.8709,
+      "step": 19496
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8089562882498413e-08,
+      "loss": 0.947,
+      "step": 19497
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.799167869651286e-08,
+      "loss": 0.9065,
+      "step": 19498
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7894059820908747e-08,
+      "loss": 0.9097,
+      "step": 19499
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.779670625828067e-08,
+      "loss": 0.8911,
+      "step": 19500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7699618011214336e-08,
+      "loss": 0.8993,
+      "step": 19501
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7602795082292122e-08,
+      "loss": 0.8482,
+      "step": 19502
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.750623747408642e-08,
+      "loss": 0.8529,
+      "step": 19503
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7409945189162945e-08,
+      "loss": 0.8773,
+      "step": 19504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7313918230081885e-08,
+      "loss": 0.9894,
+      "step": 19505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.721815659939674e-08,
+      "loss": 0.9167,
+      "step": 19506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.712266029965104e-08,
+      "loss": 0.8329,
+      "step": 19507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.702742933338275e-08,
+      "loss": 0.8288,
+      "step": 19508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6932463703125402e-08,
+      "loss": 0.8973,
+      "step": 19509
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6837763411400308e-08,
+      "loss": 0.8216,
+      "step": 19510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.674332846072546e-08,
+      "loss": 0.8348,
+      "step": 19511
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6649158853611068e-08,
+      "loss": 0.8945,
+      "step": 19512
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.655525459256069e-08,
+      "loss": 0.8243,
+      "step": 19513
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6461615680067876e-08,
+      "loss": 0.8542,
+      "step": 19514
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.636824211862398e-08,
+      "loss": 0.9606,
+      "step": 19515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6275133910710338e-08,
+      "loss": 0.9018,
+      "step": 19516
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6182291058800537e-08,
+      "loss": 0.7836,
+      "step": 19517
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6089713565362598e-08,
+      "loss": 0.8446,
+      "step": 19518
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5997401432857884e-08,
+      "loss": 0.9329,
+      "step": 19519
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.590535466373999e-08,
+      "loss": 0.9121,
+      "step": 19520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.581357326045363e-08,
+      "loss": 0.8514,
+      "step": 19521
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.572205722544018e-08,
+      "loss": 0.9337,
+      "step": 19522
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.563080656113103e-08,
+      "loss": 1.0259,
+      "step": 19523
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5539821269952016e-08,
+      "loss": 0.8483,
+      "step": 19524
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5449101354320096e-08,
+      "loss": 0.8861,
+      "step": 19525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.535864681664889e-08,
+      "loss": 0.8531,
+      "step": 19526
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.526845765934093e-08,
+      "loss": 0.9067,
+      "step": 19527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5178533884794288e-08,
+      "loss": 0.8758,
+      "step": 19528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.508887549539706e-08,
+      "loss": 0.7998,
+      "step": 19529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.499948249353511e-08,
+      "loss": 0.902,
+      "step": 19530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.49103548815821e-08,
+      "loss": 0.8421,
+      "step": 19531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4821492661908355e-08,
+      "loss": 0.8023,
+      "step": 19532
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4732895836874206e-08,
+      "loss": 0.8873,
+      "step": 19533
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.464456440883666e-08,
+      "loss": 0.9104,
+      "step": 19534
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4556498380141615e-08,
+      "loss": 0.8714,
+      "step": 19535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4468697753130534e-08,
+      "loss": 0.7866,
+      "step": 19536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4381162530135995e-08,
+      "loss": 0.7639,
+      "step": 19537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4293892713486135e-08,
+      "loss": 0.8048,
+      "step": 19538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.420688830550021e-08,
+      "loss": 0.8561,
+      "step": 19539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4120149308489705e-08,
+      "loss": 0.8773,
+      "step": 19540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4033675724761664e-08,
+      "loss": 0.8423,
+      "step": 19541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3947467556612027e-08,
+      "loss": 0.9426,
+      "step": 19542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3861524806334514e-08,
+      "loss": 0.7761,
+      "step": 19543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3775847476212856e-08,
+      "loss": 0.7449,
+      "step": 19544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3690435568523009e-08,
+      "loss": 0.9022,
+      "step": 19545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3605289085536488e-08,
+      "loss": 0.8387,
+      "step": 19546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.352040802951482e-08,
+      "loss": 0.8559,
+      "step": 19547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3435792402715087e-08,
+      "loss": 0.9155,
+      "step": 19548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.335144220738771e-08,
+      "loss": 0.8748,
+      "step": 19549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3267357445772011e-08,
+      "loss": 0.9622,
+      "step": 19550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3183538120103979e-08,
+      "loss": 0.9118,
+      "step": 19551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3099984232611829e-08,
+      "loss": 0.9054,
+      "step": 19552
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3016695785516009e-08,
+      "loss": 0.8712,
+      "step": 19553
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2933672781030304e-08,
+      "loss": 0.8879,
+      "step": 19554
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2850915221360727e-08,
+      "loss": 0.8568,
+      "step": 19555
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.276842310870774e-08,
+      "loss": 0.8817,
+      "step": 19556
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2686196445264031e-08,
+      "loss": 0.9003,
+      "step": 19557
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2604235233213413e-08,
+      "loss": 0.8397,
+      "step": 19558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2522539474736362e-08,
+      "loss": 0.8216,
+      "step": 19559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2441109172003363e-08,
+      "loss": 0.8546,
+      "step": 19560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2359944327178242e-08,
+      "loss": 0.839,
+      "step": 19561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2279044942420382e-08,
+      "loss": 0.7984,
+      "step": 19562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2198411019876955e-08,
+      "loss": 0.8852,
+      "step": 19563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.211804256169291e-08,
+      "loss": 0.8104,
+      "step": 19564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.203793957000432e-08,
+      "loss": 0.8227,
+      "step": 19565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1958102046939479e-08,
+      "loss": 0.8801,
+      "step": 19566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1878529994621135e-08,
+      "loss": 0.8678,
+      "step": 19567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1799223415164262e-08,
+      "loss": 0.8234,
+      "step": 19568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1720182310676065e-08,
+      "loss": 0.8394,
+      "step": 19569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1641406683257084e-08,
+      "loss": 0.82,
+      "step": 19570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.156289653500231e-08,
+      "loss": 0.9053,
+      "step": 19571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1484651867997854e-08,
+      "loss": 0.3987,
+      "step": 19572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1406672684324271e-08,
+      "loss": 0.9595,
+      "step": 19573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.132895898605213e-08,
+      "loss": 0.8506,
+      "step": 19574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1251510775248663e-08,
+      "loss": 0.9264,
+      "step": 19575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1174328053971118e-08,
+      "loss": 0.7696,
+      "step": 19576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1097410824272292e-08,
+      "loss": 0.908,
+      "step": 19577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1020759088194999e-08,
+      "loss": 0.8442,
+      "step": 19578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0944372847778716e-08,
+      "loss": 0.8579,
+      "step": 19579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0868252105050715e-08,
+      "loss": 0.8274,
+      "step": 19580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0792396862037147e-08,
+      "loss": 0.898,
+      "step": 19581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.071680712075307e-08,
+      "loss": 0.8777,
+      "step": 19582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0641482883206878e-08,
+      "loss": 0.8401,
+      "step": 19583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0566424151401412e-08,
+      "loss": 0.9448,
+      "step": 19584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0491630927330632e-08,
+      "loss": 0.8589,
+      "step": 19585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0417103212984059e-08,
+      "loss": 0.8694,
+      "step": 19586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.034284101034122e-08,
+      "loss": 0.8231,
+      "step": 19587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0268844321377203e-08,
+      "loss": 0.9018,
+      "step": 19588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.01951131480571e-08,
+      "loss": 0.886,
+      "step": 19589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0121647492341568e-08,
+      "loss": 0.8685,
+      "step": 19590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0048447356182379e-08,
+      "loss": 0.9608,
+      "step": 19591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.97551274152686e-09,
+      "loss": 0.826,
+      "step": 19592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.902843650312355e-09,
+      "loss": 0.821,
+      "step": 19593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.830440084470649e-09,
+      "loss": 0.8787,
+      "step": 19594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.75830204592576e-09,
+      "loss": 0.8862,
+      "step": 19595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.686429536595044e-09,
+      "loss": 0.8751,
+      "step": 19596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.614822558389192e-09,
+      "loss": 0.925,
+      "step": 19597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.543481113211128e-09,
+      "loss": 0.8199,
+      "step": 19598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.472405202957113e-09,
+      "loss": 0.8708,
+      "step": 19599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.401594829516748e-09,
+      "loss": 0.8693,
+      "step": 19600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.331049994770747e-09,
+      "loss": 0.8439,
+      "step": 19601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.2607707005965e-09,
+      "loss": 0.8189,
+      "step": 19602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.190756948859181e-09,
+      "loss": 0.8616,
+      "step": 19603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.121008741421745e-09,
+      "loss": 0.8544,
+      "step": 19604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.051526080136042e-09,
+      "loss": 0.8871,
+      "step": 19605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.982308966850595e-09,
+      "loss": 0.9805,
+      "step": 19606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.913357403405043e-09,
+      "loss": 0.848,
+      "step": 19607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.844671391631254e-09,
+      "loss": 0.8656,
+      "step": 19608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.776250933354436e-09,
+      "loss": 0.8597,
+      "step": 19609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.708096030395352e-09,
+      "loss": 0.8133,
+      "step": 19610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.640206684562558e-09,
+      "loss": 0.9565,
+      "step": 19611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.572582897662385e-09,
+      "loss": 0.7902,
+      "step": 19612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.505224671491175e-09,
+      "loss": 0.863,
+      "step": 19613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.438132007839717e-09,
+      "loss": 0.8534,
+      "step": 19614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.37130490849214e-09,
+      "loss": 0.8971,
+      "step": 19615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.304743375222579e-09,
+      "loss": 0.8054,
+      "step": 19616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.238447409802952e-09,
+      "loss": 0.9359,
+      "step": 19617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.172417013991852e-09,
+      "loss": 0.8399,
+      "step": 19618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.10665218954787e-09,
+      "loss": 0.8916,
+      "step": 19619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.041152938216278e-09,
+      "loss": 0.8492,
+      "step": 19620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.975919261739018e-09,
+      "loss": 0.8829,
+      "step": 19621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.910951161849145e-09,
+      "loss": 0.9387,
+      "step": 19622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.846248640275278e-09,
+      "loss": 0.8628,
+      "step": 19623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.781811698736041e-09,
+      "loss": 0.7567,
+      "step": 19624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.7176403389434e-09,
+      "loss": 0.8539,
+      "step": 19625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.653734562603765e-09,
+      "loss": 0.8874,
+      "step": 19626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.590094371415779e-09,
+      "loss": 0.9227,
+      "step": 19627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.526719767070312e-09,
+      "loss": 0.8696,
+      "step": 19628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.463610751251571e-09,
+      "loss": 0.8867,
+      "step": 19629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.400767325638214e-09,
+      "loss": 0.7349,
+      "step": 19630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.338189491900016e-09,
+      "loss": 0.9061,
+      "step": 19631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.275877251700092e-09,
+      "loss": 0.8448,
+      "step": 19632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.213830606693784e-09,
+      "loss": 0.8554,
+      "step": 19633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.152049558531993e-09,
+      "loss": 0.8417,
+      "step": 19634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.0905341088545185e-09,
+      "loss": 0.8855,
+      "step": 19635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.029284259298941e-09,
+      "loss": 0.8563,
+      "step": 19636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.968300011490625e-09,
+      "loss": 0.8727,
+      "step": 19637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.907581367052718e-09,
+      "loss": 0.8437,
+      "step": 19638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.847128327598374e-09,
+      "loss": 0.9886,
+      "step": 19639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.786940894732974e-09,
+      "loss": 0.8488,
+      "step": 19640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.727019070058571e-09,
+      "loss": 0.7415,
+      "step": 19641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.667362855165005e-09,
+      "loss": 0.87,
+      "step": 19642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.607972251641004e-09,
+      "loss": 0.8685,
+      "step": 19643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.548847261063085e-09,
+      "loss": 0.8484,
+      "step": 19644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.489987885003324e-09,
+      "loss": 0.8045,
+      "step": 19645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.431394125024915e-09,
+      "loss": 0.9462,
+      "step": 19646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.373065982687721e-09,
+      "loss": 0.8367,
+      "step": 19647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.315003459539393e-09,
+      "loss": 0.8894,
+      "step": 19648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.257206557125362e-09,
+      "loss": 0.8615,
+      "step": 19649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.199675276981066e-09,
+      "loss": 0.8434,
+      "step": 19650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.142409620634171e-09,
+      "loss": 0.875,
+      "step": 19651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.0854095896090145e-09,
+      "loss": 0.9492,
+      "step": 19652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.028675185418831e-09,
+      "loss": 0.8783,
+      "step": 19653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.972206409572412e-09,
+      "loss": 0.8061,
+      "step": 19654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.91600326357078e-09,
+      "loss": 0.8147,
+      "step": 19655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.860065748908295e-09,
+      "loss": 0.8669,
+      "step": 19656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.8043938670693244e-09,
+      "loss": 0.8607,
+      "step": 19657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.748987619536018e-09,
+      "loss": 0.8759,
+      "step": 19658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.693847007779418e-09,
+      "loss": 0.8407,
+      "step": 19659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.638972033266132e-09,
+      "loss": 0.8786,
+      "step": 19660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.584362697453882e-09,
+      "loss": 0.8522,
+      "step": 19661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.53001900179484e-09,
+      "loss": 0.9512,
+      "step": 19662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.475940947733405e-09,
+      "loss": 0.827,
+      "step": 19663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.422128536706206e-09,
+      "loss": 0.8375,
+      "step": 19664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.368581770144321e-09,
+      "loss": 0.8967,
+      "step": 19665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.315300649469946e-09,
+      "loss": 0.9491,
+      "step": 19666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.262285176100834e-09,
+      "loss": 0.7641,
+      "step": 19667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.20953535144475e-09,
+      "loss": 0.8871,
+      "step": 19668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.157051176903905e-09,
+      "loss": 0.8991,
+      "step": 19669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.104832653872738e-09,
+      "loss": 0.9781,
+      "step": 19670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.052879783741249e-09,
+      "loss": 0.882,
+      "step": 19671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0011925678883355e-09,
+      "loss": 0.3547,
+      "step": 19672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.949771007687343e-09,
+      "loss": 0.8649,
+      "step": 19673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.898615104507176e-09,
+      "loss": 0.893,
+      "step": 19674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.847724859705638e-09,
+      "loss": 0.8607,
+      "step": 19675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.797100274637201e-09,
+      "loss": 0.8421,
+      "step": 19676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7467413506452345e-09,
+      "loss": 0.8788,
+      "step": 19677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.696648089068667e-09,
+      "loss": 0.8593,
+      "step": 19678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.646820491240878e-09,
+      "loss": 0.8068,
+      "step": 19679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.597258558484141e-09,
+      "loss": 0.8254,
+      "step": 19680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.547962292116292e-09,
+      "loss": 0.866,
+      "step": 19681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4989316934473945e-09,
+      "loss": 0.904,
+      "step": 19682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.45016676378085e-09,
+      "loss": 0.9031,
+      "step": 19683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.401667504413398e-09,
+      "loss": 0.841,
+      "step": 19684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3534339166340085e-09,
+      "loss": 0.9332,
+      "step": 19685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.305466001723879e-09,
+      "loss": 0.868,
+      "step": 19686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.257763760957545e-09,
+      "loss": 0.4294,
+      "step": 19687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.210327195605102e-09,
+      "loss": 0.8812,
+      "step": 19688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1631563069244315e-09,
+      "loss": 0.7877,
+      "step": 19689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.116251096171198e-09,
+      "loss": 0.871,
+      "step": 19690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0696115645921795e-09,
+      "loss": 0.9347,
+      "step": 19691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0232377134252765e-09,
+      "loss": 0.8639,
+      "step": 19692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.977129543905056e-09,
+      "loss": 0.9415,
+      "step": 19693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.931287057256095e-09,
+      "loss": 0.8661,
+      "step": 19694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.885710254697417e-09,
+      "loss": 0.8658,
+      "step": 19695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8403991374391654e-09,
+      "loss": 0.8339,
+      "step": 19696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7953537066859335e-09,
+      "loss": 0.8791,
+      "step": 19697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.750573963636761e-09,
+      "loss": 0.8922,
+      "step": 19698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.706059909479587e-09,
+      "loss": 0.4151,
+      "step": 19699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6618115453990186e-09,
+      "loss": 0.8708,
+      "step": 19700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6178288725696732e-09,
+      "loss": 0.873,
+      "step": 19701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5741118921628346e-09,
+      "loss": 0.8807,
+      "step": 19702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.530660605337577e-09,
+      "loss": 0.8156,
+      "step": 19703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.487475013251862e-09,
+      "loss": 0.8518,
+      "step": 19704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4445551170514404e-09,
+      "loss": 0.8534,
+      "step": 19705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4019009178765107e-09,
+      "loss": 0.8665,
+      "step": 19706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3595124168628312e-09,
+      "loss": 0.8815,
+      "step": 19707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3173896151361683e-09,
+      "loss": 0.4472,
+      "step": 19708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.275532513815627e-09,
+      "loss": 0.7672,
+      "step": 19709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.23394111401365e-09,
+      "loss": 0.865,
+      "step": 19710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1926154168371305e-09,
+      "loss": 0.8281,
+      "step": 19711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1515554233829682e-09,
+      "loss": 0.8773,
+      "step": 19712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1107611347425126e-09,
+      "loss": 0.8128,
+      "step": 19713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.070232552001562e-09,
+      "loss": 0.8584,
+      "step": 19714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.029969676234812e-09,
+      "loss": 0.8432,
+      "step": 19715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.989972508514738e-09,
+      "loss": 0.872,
+      "step": 19716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9502410499038236e-09,
+      "loss": 0.8255,
+      "step": 19717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9107753014567808e-09,
+      "loss": 0.8602,
+      "step": 19718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8715752642238804e-09,
+      "loss": 0.9187,
+      "step": 19719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8326409392476216e-09,
+      "loss": 0.8628,
+      "step": 19720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7939723275616228e-09,
+      "loss": 0.7807,
+      "step": 19721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7555694301928393e-09,
+      "loss": 0.9117,
+      "step": 19722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7174322481637873e-09,
+      "loss": 0.8226,
+      "step": 19723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6795607824869896e-09,
+      "loss": 0.8397,
+      "step": 19724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.641955034170529e-09,
+      "loss": 0.9148,
+      "step": 19725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6046150042124964e-09,
+      "loss": 0.8687,
+      "step": 19726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.56754069360432e-09,
+      "loss": 0.8839,
+      "step": 19727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5307321033340994e-09,
+      "loss": 0.9015,
+      "step": 19728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.49418923437883e-09,
+      "loss": 0.8851,
+      "step": 19729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.457912087708847e-09,
+      "loss": 0.9712,
+      "step": 19730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4219006642900446e-09,
+      "loss": 0.8586,
+      "step": 19731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3861549650783243e-09,
+      "loss": 0.823,
+      "step": 19732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.350674991025148e-09,
+      "loss": 0.8124,
+      "step": 19733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3154607430719844e-09,
+      "loss": 0.8609,
+      "step": 19734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2805122221547517e-09,
+      "loss": 0.8361,
+      "step": 19735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2458294292038164e-09,
+      "loss": 0.9068,
+      "step": 19736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2114123651406637e-09,
+      "loss": 0.8806,
+      "step": 19737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.177261030879008e-09,
+      "loss": 0.4101,
+      "step": 19738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1433754273270103e-09,
+      "loss": 0.754,
+      "step": 19739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.109755555385062e-09,
+      "loss": 0.9299,
+      "step": 19740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0764014159480038e-09,
+      "loss": 0.8637,
+      "step": 19741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0433130099006825e-09,
+      "loss": 0.8005,
+      "step": 19742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0104903381235052e-09,
+      "loss": 0.8609,
+      "step": 19743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9779334014891074e-09,
+      "loss": 0.8928,
+      "step": 19744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.945642200862352e-09,
+      "loss": 0.8807,
+      "step": 19745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9136167371014423e-09,
+      "loss": 0.8917,
+      "step": 19746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8818570110579194e-09,
+      "loss": 0.8527,
+      "step": 19747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8503630235755521e-09,
+      "loss": 0.8042,
+      "step": 19748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.819134775491449e-09,
+      "loss": 0.8515,
+      "step": 19749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7881722676360568e-09,
+      "loss": 0.8453,
+      "step": 19750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7574755008309408e-09,
+      "loss": 0.8096,
+      "step": 19751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.727044475894335e-09,
+      "loss": 0.9337,
+      "step": 19752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6968791936333718e-09,
+      "loss": 0.8917,
+      "step": 19753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.666979654849632e-09,
+      "loss": 0.8487,
+      "step": 19754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6373458603391456e-09,
+      "loss": 0.8688,
+      "step": 19755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6079778108879507e-09,
+      "loss": 0.8714,
+      "step": 19756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5788755072787543e-09,
+      "loss": 0.8824,
+      "step": 19757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5500389502820513e-09,
+      "loss": 0.8573,
+      "step": 19758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5214681406672261e-09,
+      "loss": 0.8024,
+      "step": 19759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.493163079192561e-09,
+      "loss": 0.8447,
+      "step": 19760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4651237666096775e-09,
+      "loss": 0.8118,
+      "step": 19761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4373502036646447e-09,
+      "loss": 0.8575,
+      "step": 19762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4098423910946512e-09,
+      "loss": 0.8047,
+      "step": 19763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3826003296324441e-09,
+      "loss": 0.8776,
+      "step": 19764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3556240200007786e-09,
+      "loss": 0.8501,
+      "step": 19765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3289134629168587e-09,
+      "loss": 0.9008,
+      "step": 19766
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3024686590912273e-09,
+      "loss": 0.8545,
+      "step": 19767
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2762896092266553e-09,
+      "loss": 0.8684,
+      "step": 19768
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2503763140181424e-09,
+      "loss": 0.8758,
+      "step": 19769
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2247287741551372e-09,
+      "loss": 0.8455,
+      "step": 19770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1993469903193168e-09,
+      "loss": 0.4417,
+      "step": 19771
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1742309631845861e-09,
+      "loss": 0.8357,
+      "step": 19772
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1493806934192997e-09,
+      "loss": 0.8082,
+      "step": 19773
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1247961816840403e-09,
+      "loss": 0.8857,
+      "step": 19774
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.100477428631619e-09,
+      "loss": 0.8997,
+      "step": 19775
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0764244349092956e-09,
+      "loss": 0.8601,
+      "step": 19776
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0526372011554486e-09,
+      "loss": 0.8734,
+      "step": 19777
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.029115728001795e-09,
+      "loss": 0.8629,
+      "step": 19778
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0058600160745002e-09,
+      "loss": 0.8994,
+      "step": 19779
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.828700659919588e-10,
+      "loss": 0.8636,
+      "step": 19780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.601458783647933e-10,
+      "loss": 0.9448,
+      "step": 19781
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.376874537969649e-10,
+      "loss": 0.8772,
+      "step": 19782
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.154947928857738e-10,
+      "loss": 0.8494,
+      "step": 19783
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.935678962196381e-10,
+      "loss": 0.9212,
+      "step": 19784
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.719067643836454e-10,
+      "loss": 0.8814,
+      "step": 19785
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.505113979506707e-10,
+      "loss": 0.9022,
+      "step": 19786
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.293817974924789e-10,
+      "loss": 0.753,
+      "step": 19787
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.085179635686224e-10,
+      "loss": 0.8404,
+      "step": 19788
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.879198967342127e-10,
+      "loss": 0.859,
+      "step": 19789
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.675875975365899e-10,
+      "loss": 0.9086,
+      "step": 19790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.475210665164323e-10,
+      "loss": 0.9068,
+      "step": 19791
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.277203042066472e-10,
+      "loss": 0.8999,
+      "step": 19792
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.081853111334802e-10,
+      "loss": 0.897,
+      "step": 19793
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.88916087817626e-10,
+      "loss": 0.853,
+      "step": 19794
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.699126347686768e-10,
+      "loss": 0.9058,
+      "step": 19795
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.511749524940047e-10,
+      "loss": 0.8589,
+      "step": 19796
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.327030414909896e-10,
+      "loss": 0.8909,
+      "step": 19797
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.144969022492398e-10,
+      "loss": 0.909,
+      "step": 19798
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.965565352550328e-10,
+      "loss": 0.8753,
+      "step": 19799
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.788819409835445e-10,
+      "loss": 0.8849,
+      "step": 19800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.614731199043988e-10,
+      "loss": 0.823,
+      "step": 19801
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.44330072480559e-10,
+      "loss": 0.8505,
+      "step": 19802
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.274527991694367e-10,
+      "loss": 0.8875,
+      "step": 19803
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.108413004173418e-10,
+      "loss": 0.8789,
+      "step": 19804
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.94495576666143e-10,
+      "loss": 0.8714,
+      "step": 19805
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.784156283510477e-10,
+      "loss": 0.894,
+      "step": 19806
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.626014558994918e-10,
+      "loss": 0.8777,
+      "step": 19807
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.4705305973113953e-10,
+      "loss": 0.8676,
+      "step": 19808
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3177044025899397e-10,
+      "loss": 0.8857,
+      "step": 19809
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1675359789050683e-10,
+      "loss": 0.8994,
+      "step": 19810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0200253302424827e-10,
+      "loss": 0.8507,
+      "step": 19811
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8751724605101683e-10,
+      "loss": 0.9426,
+      "step": 19812
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7329773735828023e-10,
+      "loss": 0.8381,
+      "step": 19813
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.593440073212939e-10,
+      "loss": 0.8479,
+      "step": 19814
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4565605631309285e-10,
+      "loss": 0.9181,
+      "step": 19815
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.322338846967199e-10,
+      "loss": 0.8529,
+      "step": 19816
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.190774928285567e-10,
+      "loss": 0.8255,
+      "step": 19817
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.061868810594337e-10,
+      "loss": 0.858,
+      "step": 19818
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.935620497301894e-10,
+      "loss": 0.7613,
+      "step": 19819
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.812029991772214e-10,
+      "loss": 0.9419,
+      "step": 19820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6910972972915563e-10,
+      "loss": 0.9692,
+      "step": 19821
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.572822417079568e-10,
+      "loss": 0.8311,
+      "step": 19822
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.457205354267078e-10,
+      "loss": 0.8425,
+      "step": 19823
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.344246111940507e-10,
+      "loss": 0.9188,
+      "step": 19824
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2339446930974562e-10,
+      "loss": 0.9007,
+      "step": 19825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1263011006578127e-10,
+      "loss": 0.8688,
+      "step": 19826
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.021315337497054e-10,
+      "loss": 0.9027,
+      "step": 19827
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9189874064018398e-10,
+      "loss": 0.7438,
+      "step": 19828
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8193173100922168e-10,
+      "loss": 0.8379,
+      "step": 19829
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7223050512216176e-10,
+      "loss": 0.7888,
+      "step": 19830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.627950632354658e-10,
+      "loss": 0.8437,
+      "step": 19831
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5362540560115436e-10,
+      "loss": 0.8287,
+      "step": 19832
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.447215324623663e-10,
+      "loss": 0.7908,
+      "step": 19833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3608344405557915e-10,
+      "loss": 0.8515,
+      "step": 19834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2771114061060907e-10,
+      "loss": 0.8396,
+      "step": 19835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1960462235061087e-10,
+      "loss": 0.7786,
+      "step": 19836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1176388949096783e-10,
+      "loss": 0.8851,
+      "step": 19837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0418894223929166e-10,
+      "loss": 0.8278,
+      "step": 19838
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.687978079764293e-11,
+      "loss": 0.8784,
+      "step": 19839
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.983640535920046e-11,
+      "loss": 0.9314,
+      "step": 19840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.305881611270217e-11,
+      "loss": 0.8773,
+      "step": 19841
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.654701323689395e-11,
+      "loss": 0.9205,
+      "step": 19842
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.030099690608083e-11,
+      "loss": 0.8334,
+      "step": 19843
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.432076728457581e-11,
+      "loss": 0.792,
+      "step": 19844
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.860632453336124e-11,
+      "loss": 0.956,
+      "step": 19845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.315766880342743e-11,
+      "loss": 0.8634,
+      "step": 19846
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.797480023910339e-11,
+      "loss": 0.8664,
+      "step": 19847
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3057718979167e-11,
+      "loss": 0.8647,
+      "step": 19848
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8406425152404116e-11,
+      "loss": 0.8797,
+      "step": 19849
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.402091888538017e-11,
+      "loss": 0.8702,
+      "step": 19850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.990120029133792e-11,
+      "loss": 0.8589,
+      "step": 19851
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6047269482409875e-11,
+      "loss": 0.8363,
+      "step": 19852
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2459126560736566e-11,
+      "loss": 0.8328,
+      "step": 19853
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.913677161957672e-11,
+      "loss": 0.7028,
+      "step": 19854
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.608020474996863e-11,
+      "loss": 0.8884,
+      "step": 19855
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3289426030738128e-11,
+      "loss": 0.8577,
+      "step": 19856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0764435538490602e-11,
+      "loss": 0.8907,
+      "step": 19857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.505233338729213e-12,
+      "loss": 0.8668,
+      "step": 19858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.511819491406002e-12,
+      "loss": 0.8217,
+      "step": 19859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7841940498116746e-12,
+      "loss": 0.8466,
+      "step": 19860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.322357059465375e-12,
+      "loss": 0.8268,
+      "step": 19861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.126308560335133e-12,
+      "loss": 0.8914,
+      "step": 19862
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1960485835071922e-12,
+      "loss": 0.8056,
+      "step": 19863
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.315771545166825e-13,
+      "loss": 0.8393,
+      "step": 19864
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.328942889067264e-13,
+      "loss": 0.7708,
+      "step": 19865
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.5343,
+      "step": 19866
+    },
+    {
+      "epoch": 1.0,
+      "step": 19866,
+      "total_flos": 5341988136697856.0,
+      "train_loss": 0.9188437570555146,
+      "train_runtime": 43095.5965,
+      "train_samples_per_second": 29.502,
+      "train_steps_per_second": 0.461
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 19866,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 50000,
+  "total_flos": 5341988136697856.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}