diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,88210 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999532239340373,
+  "eval_steps": 500,
+  "global_step": 14697,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.267573696145125e-08,
+      "loss": 3.0737,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.53514739229025e-08,
+      "loss": 4.1281,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.802721088435375e-08,
+      "loss": 3.818,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.0702947845805e-08,
+      "loss": 3.0824,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1337868480725624e-07,
+      "loss": 3.1784,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.360544217687075e-07,
+      "loss": 2.6682,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5873015873015874e-07,
+      "loss": 4.1884,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8140589569161e-07,
+      "loss": 3.8919,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0408163265306121e-07,
+      "loss": 2.5557,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.267573696145125e-07,
+      "loss": 3.3745,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4943310657596374e-07,
+      "loss": 3.1709,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.72108843537415e-07,
+      "loss": 2.8285,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9478458049886624e-07,
+      "loss": 3.2384,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.174603174603175e-07,
+      "loss": 3.1554,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.401360544217688e-07,
+      "loss": 3.5385,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6281179138322e-07,
+      "loss": 4.0767,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8548752834467123e-07,
+      "loss": 2.6847,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0816326530612243e-07,
+      "loss": 3.801,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.3083900226757373e-07,
+      "loss": 2.619,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.53514739229025e-07,
+      "loss": 4.0516,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.7619047619047623e-07,
+      "loss": 3.5127,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.988662131519275e-07,
+      "loss": 3.5965,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.215419501133788e-07,
+      "loss": 3.3556,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.4421768707483e-07,
+      "loss": 3.3882,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.668934240362813e-07,
+      "loss": 2.8489,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.895691609977325e-07,
+      "loss": 3.0267,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.122448979591837e-07,
+      "loss": 3.3852,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.34920634920635e-07,
+      "loss": 3.0874,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.575963718820862e-07,
+      "loss": 2.9957,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.802721088435376e-07,
+      "loss": 3.5983,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.029478458049888e-07,
+      "loss": 2.8416,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2562358276644e-07,
+      "loss": 3.2739,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.482993197278913e-07,
+      "loss": 3.3285,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.709750566893425e-07,
+      "loss": 2.9394,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.936507936507937e-07,
+      "loss": 2.5542,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.163265306122449e-07,
+      "loss": 3.2555,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.390022675736963e-07,
+      "loss": 2.7448,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.616780045351475e-07,
+      "loss": 2.7339,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.843537414965988e-07,
+      "loss": 3.0543,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.0702947845805e-07,
+      "loss": 3.3138,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.297052154195012e-07,
+      "loss": 3.0896,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.523809523809525e-07,
+      "loss": 2.4997,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.750566893424039e-07,
+      "loss": 2.734,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.97732426303855e-07,
+      "loss": 2.2318,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0204081632653063e-06,
+      "loss": 2.4609,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0430839002267576e-06,
+      "loss": 2.0296,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0657596371882086e-06,
+      "loss": 2.4072,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.08843537414966e-06,
+      "loss": 1.8861,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.111111111111111e-06,
+      "loss": 1.6955,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1337868480725626e-06,
+      "loss": 1.6461,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1564625850340136e-06,
+      "loss": 2.1036,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.179138321995465e-06,
+      "loss": 1.6062,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2018140589569162e-06,
+      "loss": 1.8117,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2244897959183673e-06,
+      "loss": 1.5118,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2471655328798186e-06,
+      "loss": 1.6371,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.26984126984127e-06,
+      "loss": 1.6709,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2925170068027212e-06,
+      "loss": 1.8728,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3151927437641723e-06,
+      "loss": 1.6665,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3378684807256236e-06,
+      "loss": 1.7888,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3605442176870751e-06,
+      "loss": 1.8609,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3832199546485262e-06,
+      "loss": 1.5921,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4058956916099775e-06,
+      "loss": 1.5959,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4285714285714286e-06,
+      "loss": 1.4104,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.45124716553288e-06,
+      "loss": 1.4761,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.473922902494331e-06,
+      "loss": 1.46,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4965986394557825e-06,
+      "loss": 1.5858,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5192743764172338e-06,
+      "loss": 1.4599,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.541950113378685e-06,
+      "loss": 1.4588,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5646258503401362e-06,
+      "loss": 1.701,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5873015873015873e-06,
+      "loss": 1.4829,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6099773242630386e-06,
+      "loss": 1.2778,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6326530612244897e-06,
+      "loss": 1.3621,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6553287981859412e-06,
+      "loss": 1.2578,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6780045351473925e-06,
+      "loss": 1.2194,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7006802721088436e-06,
+      "loss": 0.9722,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.723356009070295e-06,
+      "loss": 1.0816,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.746031746031746e-06,
+      "loss": 1.2608,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7687074829931975e-06,
+      "loss": 1.2289,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7913832199546488e-06,
+      "loss": 1.0393,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8140589569161e-06,
+      "loss": 1.2266,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8367346938775512e-06,
+      "loss": 1.302,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8594104308390023e-06,
+      "loss": 1.0839,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8820861678004538e-06,
+      "loss": 1.0299,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.904761904761905e-06,
+      "loss": 1.1148,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.927437641723356e-06,
+      "loss": 0.9842,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9501133786848077e-06,
+      "loss": 1.2896,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9727891156462586e-06,
+      "loss": 1.2472,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99546485260771e-06,
+      "loss": 1.4597,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.018140589569161e-06,
+      "loss": 1.5361,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0408163265306125e-06,
+      "loss": 1.1304,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0634920634920634e-06,
+      "loss": 1.1436,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.086167800453515e-06,
+      "loss": 1.0142,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1088435374149664e-06,
+      "loss": 0.9359,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1315192743764173e-06,
+      "loss": 0.8073,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1541950113378686e-06,
+      "loss": 1.2182,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.17687074829932e-06,
+      "loss": 1.0334,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.199546485260771e-06,
+      "loss": 1.1066,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.9271,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.244897959183674e-06,
+      "loss": 1.4945,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.267573696145125e-06,
+      "loss": 1.2441,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.290249433106576e-06,
+      "loss": 1.037,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3129251700680273e-06,
+      "loss": 0.9802,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3356009070294786e-06,
+      "loss": 1.1354,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.35827664399093e-06,
+      "loss": 1.3285,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.380952380952381e-06,
+      "loss": 1.1345,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4036281179138325e-06,
+      "loss": 0.8447,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.426303854875284e-06,
+      "loss": 1.3145,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4489795918367347e-06,
+      "loss": 0.8902,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4716553287981864e-06,
+      "loss": 1.0027,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4943310657596373e-06,
+      "loss": 0.9319,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5170068027210886e-06,
+      "loss": 1.4418,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.53968253968254e-06,
+      "loss": 1.1004,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5623582766439916e-06,
+      "loss": 0.8312,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5850340136054425e-06,
+      "loss": 1.1035,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6077097505668938e-06,
+      "loss": 1.6079,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6303854875283447e-06,
+      "loss": 0.9468,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6530612244897964e-06,
+      "loss": 1.2445,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6757369614512473e-06,
+      "loss": 1.2862,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6984126984126986e-06,
+      "loss": 1.0655,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7210884353741503e-06,
+      "loss": 1.1146,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.743764172335601e-06,
+      "loss": 1.1339,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7664399092970525e-06,
+      "loss": 0.9769,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7891156462585034e-06,
+      "loss": 0.929,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.811791383219955e-06,
+      "loss": 0.8992,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.834467120181406e-06,
+      "loss": 1.4347,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 1.1451,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.879818594104309e-06,
+      "loss": 1.0267,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.90249433106576e-06,
+      "loss": 1.0714,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.925170068027211e-06,
+      "loss": 0.9355,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.947845804988662e-06,
+      "loss": 0.8705,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9705215419501138e-06,
+      "loss": 0.878,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.993197278911565e-06,
+      "loss": 1.0497,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.015873015873016e-06,
+      "loss": 0.8883,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0385487528344677e-06,
+      "loss": 1.1595,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0612244897959185e-06,
+      "loss": 0.8867,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.08390022675737e-06,
+      "loss": 1.2283,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1065759637188207e-06,
+      "loss": 1.4963,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1292517006802725e-06,
+      "loss": 0.8936,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1519274376417238e-06,
+      "loss": 1.0299,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1746031746031746e-06,
+      "loss": 0.8415,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1972789115646264e-06,
+      "loss": 1.0461,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2199546485260772e-06,
+      "loss": 1.0349,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2426303854875285e-06,
+      "loss": 0.8666,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2653061224489794e-06,
+      "loss": 1.1409,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.287981859410431e-06,
+      "loss": 1.3864,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3106575963718824e-06,
+      "loss": 1.0527,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 1.0451,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.356009070294785e-06,
+      "loss": 1.2838,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3786848072562364e-06,
+      "loss": 0.8661,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4013605442176872e-06,
+      "loss": 1.1002,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4240362811791385e-06,
+      "loss": 1.0703,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.44671201814059e-06,
+      "loss": 1.0562,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.469387755102041e-06,
+      "loss": 1.0003,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.492063492063492e-06,
+      "loss": 1.3293,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5147392290249437e-06,
+      "loss": 0.9623,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.537414965986395e-06,
+      "loss": 1.036,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.560090702947846e-06,
+      "loss": 1.2849,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5827664399092976e-06,
+      "loss": 1.0581,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6054421768707485e-06,
+      "loss": 1.0738,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6281179138322e-06,
+      "loss": 0.6251,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6507936507936507e-06,
+      "loss": 0.8815,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6734693877551024e-06,
+      "loss": 1.1345,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6961451247165537e-06,
+      "loss": 1.0384,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7188208616780046e-06,
+      "loss": 0.7732,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7414965986394563e-06,
+      "loss": 1.1157,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7641723356009076e-06,
+      "loss": 0.8675,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7868480725623585e-06,
+      "loss": 1.0125,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.80952380952381e-06,
+      "loss": 0.8833,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.832199546485261e-06,
+      "loss": 0.9328,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.854875283446712e-06,
+      "loss": 1.0435,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.877551020408164e-06,
+      "loss": 0.8955,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9002267573696154e-06,
+      "loss": 0.8991,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.922902494331066e-06,
+      "loss": 1.269,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.945578231292517e-06,
+      "loss": 1.0404,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.968253968253968e-06,
+      "loss": 1.0794,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.99092970521542e-06,
+      "loss": 1.1312,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.013605442176871e-06,
+      "loss": 0.8122,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.036281179138322e-06,
+      "loss": 1.014,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.058956916099774e-06,
+      "loss": 0.9509,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.081632653061225e-06,
+      "loss": 1.1741,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.104308390022676e-06,
+      "loss": 0.9657,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.126984126984127e-06,
+      "loss": 0.7954,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1496598639455785e-06,
+      "loss": 0.9968,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.17233560090703e-06,
+      "loss": 0.9238,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.195011337868481e-06,
+      "loss": 1.1057,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.217687074829933e-06,
+      "loss": 0.7711,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.240362811791384e-06,
+      "loss": 1.1153,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.263038548752835e-06,
+      "loss": 0.9374,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2857142857142855e-06,
+      "loss": 0.8809,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.308390022675737e-06,
+      "loss": 1.2112,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.331065759637189e-06,
+      "loss": 1.0382,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.35374149659864e-06,
+      "loss": 0.9379,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3764172335600915e-06,
+      "loss": 0.9179,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.399092970521542e-06,
+      "loss": 1.0088,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.421768707482993e-06,
+      "loss": 1.0734,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.9722,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.467120181405896e-06,
+      "loss": 1.0435,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.489795918367348e-06,
+      "loss": 1.1202,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5124716553287985e-06,
+      "loss": 1.1269,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.53514739229025e-06,
+      "loss": 1.2964,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.557823129251701e-06,
+      "loss": 0.7935,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.580498866213152e-06,
+      "loss": 0.7529,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.603174603174604e-06,
+      "loss": 1.1087,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6258503401360546e-06,
+      "loss": 0.7707,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.648526077097506e-06,
+      "loss": 1.0553,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.671201814058957e-06,
+      "loss": 0.6881,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.693877551020409e-06,
+      "loss": 1.3186,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.71655328798186e-06,
+      "loss": 1.0069,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.739229024943311e-06,
+      "loss": 1.0935,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.761904761904762e-06,
+      "loss": 1.0582,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.784580498866213e-06,
+      "loss": 1.2832,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.807256235827665e-06,
+      "loss": 0.8895,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.829931972789116e-06,
+      "loss": 0.81,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.852607709750568e-06,
+      "loss": 0.9181,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8752834467120185e-06,
+      "loss": 1.0627,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.897959183673469e-06,
+      "loss": 0.9392,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.920634920634921e-06,
+      "loss": 0.9142,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.943310657596373e-06,
+      "loss": 0.6597,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.965986394557824e-06,
+      "loss": 0.8956,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9886621315192745e-06,
+      "loss": 0.9985,
+      "step": 220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0113378684807254e-06,
+      "loss": 1.137,
+      "step": 221
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.034013605442177e-06,
+      "loss": 1.1757,
+      "step": 222
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.056689342403629e-06,
+      "loss": 1.1944,
+      "step": 223
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.07936507936508e-06,
+      "loss": 0.6863,
+      "step": 224
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.1020408163265315e-06,
+      "loss": 1.2971,
+      "step": 225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.124716553287983e-06,
+      "loss": 0.977,
+      "step": 226
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.147392290249433e-06,
+      "loss": 1.1083,
+      "step": 227
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.170068027210885e-06,
+      "loss": 1.2573,
+      "step": 228
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.192743764172336e-06,
+      "loss": 0.903,
+      "step": 229
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2154195011337876e-06,
+      "loss": 0.9919,
+      "step": 230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2380952380952384e-06,
+      "loss": 0.9819,
+      "step": 231
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.260770975056689e-06,
+      "loss": 1.1921,
+      "step": 232
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.283446712018141e-06,
+      "loss": 1.0607,
+      "step": 233
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.306122448979593e-06,
+      "loss": 0.8522,
+      "step": 234
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.328798185941043e-06,
+      "loss": 0.8657,
+      "step": 235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3514739229024945e-06,
+      "loss": 0.7995,
+      "step": 236
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.374149659863946e-06,
+      "loss": 1.0714,
+      "step": 237
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.396825396825397e-06,
+      "loss": 1.1683,
+      "step": 238
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.419501133786849e-06,
+      "loss": 0.8349,
+      "step": 239
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.442176870748301e-06,
+      "loss": 0.8768,
+      "step": 240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.464852607709751e-06,
+      "loss": 1.1139,
+      "step": 241
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.487528344671202e-06,
+      "loss": 1.0874,
+      "step": 242
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.510204081632653e-06,
+      "loss": 1.1501,
+      "step": 243
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.532879818594105e-06,
+      "loss": 0.9968,
+      "step": 244
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.555555555555557e-06,
+      "loss": 0.9319,
+      "step": 245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.578231292517007e-06,
+      "loss": 1.0045,
+      "step": 246
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.6009070294784584e-06,
+      "loss": 1.1612,
+      "step": 247
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.62358276643991e-06,
+      "loss": 1.0186,
+      "step": 248
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.646258503401361e-06,
+      "loss": 0.7991,
+      "step": 249
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.668934240362812e-06,
+      "loss": 0.8565,
+      "step": 250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.691609977324264e-06,
+      "loss": 1.02,
+      "step": 251
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 0.8,
+      "step": 252
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.736961451247166e-06,
+      "loss": 0.9538,
+      "step": 253
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.759637188208618e-06,
+      "loss": 0.9287,
+      "step": 254
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.782312925170068e-06,
+      "loss": 0.6778,
+      "step": 255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.80498866213152e-06,
+      "loss": 0.8517,
+      "step": 256
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8276643990929714e-06,
+      "loss": 1.1485,
+      "step": 257
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.850340136054422e-06,
+      "loss": 0.7058,
+      "step": 258
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.873015873015874e-06,
+      "loss": 0.9235,
+      "step": 259
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.895691609977324e-06,
+      "loss": 0.7059,
+      "step": 260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.918367346938776e-06,
+      "loss": 1.1026,
+      "step": 261
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9410430839002275e-06,
+      "loss": 1.0484,
+      "step": 262
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.963718820861678e-06,
+      "loss": 1.0375,
+      "step": 263
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.98639455782313e-06,
+      "loss": 1.0361,
+      "step": 264
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.009070294784581e-06,
+      "loss": 0.9471,
+      "step": 265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.031746031746032e-06,
+      "loss": 0.9347,
+      "step": 266
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.054421768707484e-06,
+      "loss": 0.9149,
+      "step": 267
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.077097505668935e-06,
+      "loss": 1.1553,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.099773242630385e-06,
+      "loss": 1.1913,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.122448979591837e-06,
+      "loss": 0.8657,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.145124716553289e-06,
+      "loss": 1.112,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.16780045351474e-06,
+      "loss": 0.8022,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.1904761904761914e-06,
+      "loss": 1.0001,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.2131519274376415e-06,
+      "loss": 0.9189,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.235827664399093e-06,
+      "loss": 1.1097,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.258503401360545e-06,
+      "loss": 1.085,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.281179138321996e-06,
+      "loss": 1.0355,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.3038548752834475e-06,
+      "loss": 0.9859,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.326530612244899e-06,
+      "loss": 0.8633,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.349206349206349e-06,
+      "loss": 0.9385,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.371882086167801e-06,
+      "loss": 0.6106,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.394557823129253e-06,
+      "loss": 0.9754,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.417233560090704e-06,
+      "loss": 0.831,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4399092970521545e-06,
+      "loss": 0.8306,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.462585034013606e-06,
+      "loss": 1.1183,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.485260770975057e-06,
+      "loss": 0.7311,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.507936507936509e-06,
+      "loss": 0.6582,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.530612244897959e-06,
+      "loss": 0.9443,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.5532879818594106e-06,
+      "loss": 1.0159,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.575963718820862e-06,
+      "loss": 0.9257,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.598639455782313e-06,
+      "loss": 1.0245,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.621315192743765e-06,
+      "loss": 0.7989,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.643990929705217e-06,
+      "loss": 0.8054,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.842,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.689342403628118e-06,
+      "loss": 1.2006,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.71201814058957e-06,
+      "loss": 0.8469,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.734693877551021e-06,
+      "loss": 0.9169,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.757369614512473e-06,
+      "loss": 0.5965,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.780045351473924e-06,
+      "loss": 1.2393,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.8027210884353745e-06,
+      "loss": 0.9747,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.825396825396826e-06,
+      "loss": 1.2515,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.848072562358277e-06,
+      "loss": 0.6831,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.870748299319728e-06,
+      "loss": 0.858,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.89342403628118e-06,
+      "loss": 0.9282,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.9160997732426305e-06,
+      "loss": 1.058,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.938775510204082e-06,
+      "loss": 1.0098,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.961451247165534e-06,
+      "loss": 0.9523,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.984126984126984e-06,
+      "loss": 0.9447,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.006802721088436e-06,
+      "loss": 1.1989,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0294784580498875e-06,
+      "loss": 0.9452,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.052154195011338e-06,
+      "loss": 1.0367,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.07482993197279e-06,
+      "loss": 1.1805,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.097505668934242e-06,
+      "loss": 0.8579,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.120181405895692e-06,
+      "loss": 0.9877,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.1428571428571436e-06,
+      "loss": 1.0693,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.165532879818595e-06,
+      "loss": 1.0348,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.188208616780046e-06,
+      "loss": 1.0461,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.210884353741497e-06,
+      "loss": 1.0413,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.233560090702948e-06,
+      "loss": 1.1227,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2562358276644e-06,
+      "loss": 0.8693,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.278911564625851e-06,
+      "loss": 0.7695,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.301587301587301e-06,
+      "loss": 0.8716,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.324263038548753e-06,
+      "loss": 0.8587,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.346938775510205e-06,
+      "loss": 0.8692,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.369614512471656e-06,
+      "loss": 1.1463,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.3922902494331075e-06,
+      "loss": 1.0531,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.414965986394559e-06,
+      "loss": 1.1806,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.437641723356009e-06,
+      "loss": 0.9669,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.460317460317461e-06,
+      "loss": 1.0766,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.482993197278913e-06,
+      "loss": 0.9923,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.5056689342403635e-06,
+      "loss": 1.0806,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.528344671201815e-06,
+      "loss": 1.0513,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.551020408163265e-06,
+      "loss": 0.953,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.573696145124717e-06,
+      "loss": 0.8556,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.596371882086169e-06,
+      "loss": 1.0723,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.61904761904762e-06,
+      "loss": 0.6511,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.64172335600907e-06,
+      "loss": 0.8854,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.664399092970522e-06,
+      "loss": 0.7344,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.687074829931972e-06,
+      "loss": 1.1343,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.709750566893424e-06,
+      "loss": 1.1291,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.732426303854876e-06,
+      "loss": 0.6505,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.755102040816327e-06,
+      "loss": 0.9934,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.77777777777778e-06,
+      "loss": 1.1583,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.800453514739231e-06,
+      "loss": 0.9731,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.823129251700681e-06,
+      "loss": 0.6715,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.845804988662133e-06,
+      "loss": 0.8165,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.868480725623583e-06,
+      "loss": 0.8421,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.891156462585034e-06,
+      "loss": 1.0149,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.913832199546486e-06,
+      "loss": 0.7794,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.936507936507936e-06,
+      "loss": 1.0508,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.959183673469388e-06,
+      "loss": 1.1174,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.98185941043084e-06,
+      "loss": 0.868,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.004535147392291e-06,
+      "loss": 1.0808,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.027210884353741e-06,
+      "loss": 0.7908,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.049886621315193e-06,
+      "loss": 0.9823,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.072562358276645e-06,
+      "loss": 0.7633,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.095238095238097e-06,
+      "loss": 0.9756,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.117913832199548e-06,
+      "loss": 0.8296,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.140589569160998e-06,
+      "loss": 0.8389,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.16326530612245e-06,
+      "loss": 1.0999,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.1859410430839e-06,
+      "loss": 1.003,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.208616780045352e-06,
+      "loss": 0.9822,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.231292517006804e-06,
+      "loss": 0.8909,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.253968253968254e-06,
+      "loss": 0.9852,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.276643990929705e-06,
+      "loss": 0.6985,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.299319727891157e-06,
+      "loss": 0.9759,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.321995464852609e-06,
+      "loss": 1.1392,
+      "step": 367
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.34467120181406e-06,
+      "loss": 1.0631,
+      "step": 368
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.36734693877551e-06,
+      "loss": 1.106,
+      "step": 369
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.390022675736962e-06,
+      "loss": 1.0656,
+      "step": 370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.412698412698414e-06,
+      "loss": 0.8535,
+      "step": 371
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.435374149659866e-06,
+      "loss": 1.0529,
+      "step": 372
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.458049886621316e-06,
+      "loss": 0.8903,
+      "step": 373
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.480725623582767e-06,
+      "loss": 0.9366,
+      "step": 374
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.503401360544217e-06,
+      "loss": 0.3792,
+      "step": 375
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.52607709750567e-06,
+      "loss": 0.6759,
+      "step": 376
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.548752834467121e-06,
+      "loss": 0.7296,
+      "step": 377
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.792,
+      "step": 378
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.594104308390023e-06,
+      "loss": 0.8419,
+      "step": 379
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.616780045351474e-06,
+      "loss": 0.8687,
+      "step": 380
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.639455782312926e-06,
+      "loss": 0.8054,
+      "step": 381
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.662131519274378e-06,
+      "loss": 1.1622,
+      "step": 382
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.68480725623583e-06,
+      "loss": 0.761,
+      "step": 383
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.70748299319728e-06,
+      "loss": 0.9406,
+      "step": 384
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.730158730158731e-06,
+      "loss": 0.9431,
+      "step": 385
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.752834467120183e-06,
+      "loss": 0.8541,
+      "step": 386
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.775510204081633e-06,
+      "loss": 1.0308,
+      "step": 387
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.798185941043085e-06,
+      "loss": 1.0083,
+      "step": 388
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.820861678004537e-06,
+      "loss": 1.0319,
+      "step": 389
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.843537414965987e-06,
+      "loss": 0.9403,
+      "step": 390
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.866213151927438e-06,
+      "loss": 1.1072,
+      "step": 391
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.5402,
+      "step": 392
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.91156462585034e-06,
+      "loss": 0.9419,
+      "step": 393
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.934240362811792e-06,
+      "loss": 0.9959,
+      "step": 394
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.956916099773243e-06,
+      "loss": 1.0007,
+      "step": 395
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.979591836734695e-06,
+      "loss": 0.974,
+      "step": 396
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.002267573696147e-06,
+      "loss": 1.0185,
+      "step": 397
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.024943310657597e-06,
+      "loss": 0.8747,
+      "step": 398
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.047619047619049e-06,
+      "loss": 0.7549,
+      "step": 399
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.0702947845805e-06,
+      "loss": 1.0628,
+      "step": 400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.09297052154195e-06,
+      "loss": 0.7613,
+      "step": 401
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.115646258503402e-06,
+      "loss": 0.7078,
+      "step": 402
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.138321995464854e-06,
+      "loss": 1.1163,
+      "step": 403
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.160997732426304e-06,
+      "loss": 1.0022,
+      "step": 404
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.183673469387756e-06,
+      "loss": 0.9554,
+      "step": 405
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.206349206349207e-06,
+      "loss": 0.9446,
+      "step": 406
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.229024943310657e-06,
+      "loss": 0.7689,
+      "step": 407
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.251700680272109e-06,
+      "loss": 0.8617,
+      "step": 408
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.274376417233561e-06,
+      "loss": 1.1826,
+      "step": 409
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.297052154195013e-06,
+      "loss": 1.129,
+      "step": 410
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.319727891156464e-06,
+      "loss": 0.768,
+      "step": 411
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.342403628117914e-06,
+      "loss": 0.9167,
+      "step": 412
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.365079365079366e-06,
+      "loss": 0.8756,
+      "step": 413
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.387755102040818e-06,
+      "loss": 1.0135,
+      "step": 414
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.410430839002268e-06,
+      "loss": 0.9855,
+      "step": 415
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.43310657596372e-06,
+      "loss": 0.9621,
+      "step": 416
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.455782312925171e-06,
+      "loss": 1.0338,
+      "step": 417
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.478458049886621e-06,
+      "loss": 0.8916,
+      "step": 418
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.501133786848073e-06,
+      "loss": 1.0151,
+      "step": 419
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.523809523809525e-06,
+      "loss": 0.9086,
+      "step": 420
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.546485260770976e-06,
+      "loss": 0.9562,
+      "step": 421
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.569160997732427e-06,
+      "loss": 0.9247,
+      "step": 422
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.591836734693878e-06,
+      "loss": 1.125,
+      "step": 423
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.61451247165533e-06,
+      "loss": 0.7238,
+      "step": 424
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.637188208616782e-06,
+      "loss": 0.8903,
+      "step": 425
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.659863945578232e-06,
+      "loss": 0.7305,
+      "step": 426
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.682539682539683e-06,
+      "loss": 1.0001,
+      "step": 427
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.705215419501135e-06,
+      "loss": 0.9099,
+      "step": 428
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.727891156462585e-06,
+      "loss": 0.9049,
+      "step": 429
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.750566893424037e-06,
+      "loss": 0.7605,
+      "step": 430
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.773242630385489e-06,
+      "loss": 0.7064,
+      "step": 431
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.795918367346939e-06,
+      "loss": 0.8302,
+      "step": 432
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.81859410430839e-06,
+      "loss": 0.9455,
+      "step": 433
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.841269841269842e-06,
+      "loss": 0.6436,
+      "step": 434
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.863945578231294e-06,
+      "loss": 0.7888,
+      "step": 435
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.886621315192746e-06,
+      "loss": 1.1496,
+      "step": 436
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.909297052154196e-06,
+      "loss": 1.193,
+      "step": 437
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.931972789115647e-06,
+      "loss": 1.0166,
+      "step": 438
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.954648526077099e-06,
+      "loss": 1.167,
+      "step": 439
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977324263038549e-06,
+      "loss": 1.0068,
+      "step": 440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 1.0286,
+      "step": 441
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999878592819e-06,
+      "loss": 0.7708,
+      "step": 442
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999999514371279e-06,
+      "loss": 1.0215,
+      "step": 443
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998907335398e-06,
+      "loss": 1.0562,
+      "step": 444
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999998057485206e-06,
+      "loss": 1.0448,
+      "step": 445
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999996964820744e-06,
+      "loss": 0.9986,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999995629342067e-06,
+      "loss": 0.9375,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999994051049235e-06,
+      "loss": 1.0149,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99999222994233e-06,
+      "loss": 0.9252,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999990166021438e-06,
+      "loss": 0.6102,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999987859286661e-06,
+      "loss": 0.9785,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999985309738107e-06,
+      "loss": 1.0115,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999982517375903e-06,
+      "loss": 0.9496,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999979482200185e-06,
+      "loss": 0.5953,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999976204211099e-06,
+      "loss": 0.8409,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999972683408804e-06,
+      "loss": 1.1119,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99996891979347e-06,
+      "loss": 0.9484,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999964913365284e-06,
+      "loss": 1.0565,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999960664124435e-06,
+      "loss": 0.814,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999956172071135e-06,
+      "loss": 0.8956,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999951437205597e-06,
+      "loss": 0.8738,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999946459528055e-06,
+      "loss": 0.951,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999941239038748e-06,
+      "loss": 0.9999,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999935775737931e-06,
+      "loss": 1.0647,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999930069625869e-06,
+      "loss": 1.0806,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999924120702838e-06,
+      "loss": 0.7878,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99991792896913e-06,
+      "loss": 0.986,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999911494425041e-06,
+      "loss": 0.9739,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999904817070886e-06,
+      "loss": 0.736,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999897896906992e-06,
+      "loss": 0.8963,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99989073393369e-06,
+      "loss": 1.0341,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99988332815133e-06,
+      "loss": 0.9557,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999875679560272e-06,
+      "loss": 0.8746,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999867788160888e-06,
+      "loss": 1.027,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99985965395356e-06,
+      "loss": 0.808,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999851276938684e-06,
+      "loss": 0.9882,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999842657116667e-06,
+      "loss": 0.8865,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999833794487925e-06,
+      "loss": 0.876,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999824689052891e-06,
+      "loss": 1.1876,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999815340812007e-06,
+      "loss": 0.9536,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999805749765727e-06,
+      "loss": 0.7778,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999795915914517e-06,
+      "loss": 1.1396,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999785839258851e-06,
+      "loss": 1.1222,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999775519799222e-06,
+      "loss": 0.9308,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99976495753613e-06,
+      "loss": 0.9865,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99975415247009e-06,
+      "loss": 0.794,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999743104601624e-06,
+      "loss": 0.9191,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99973181393127e-06,
+      "loss": 0.8211,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999720280459576e-06,
+      "loss": 1.0052,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9997085041871e-06,
+      "loss": 0.8997,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999696485114418e-06,
+      "loss": 0.6046,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999684223242111e-06,
+      "loss": 0.829,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999671718570774e-06,
+      "loss": 0.7757,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999658971101016e-06,
+      "loss": 0.9034,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999645980833454e-06,
+      "loss": 0.8246,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999632747768722e-06,
+      "loss": 0.7995,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99961927190746e-06,
+      "loss": 1.2295,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999605553250322e-06,
+      "loss": 0.8449,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999591591797977e-06,
+      "loss": 1.016,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999577387551101e-06,
+      "loss": 1.1252,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999562940510383e-06,
+      "loss": 0.7827,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999548250676527e-06,
+      "loss": 0.8807,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999533318050245e-06,
+      "loss": 1.0494,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999518142632263e-06,
+      "loss": 0.8017,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999502724423316e-06,
+      "loss": 1.3116,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999487063424154e-06,
+      "loss": 0.9011,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999471159635538e-06,
+      "loss": 0.8571,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99945501305824e-06,
+      "loss": 0.6268,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999438623693044e-06,
+      "loss": 0.7921,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999421991540746e-06,
+      "loss": 0.9595,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999405116602154e-06,
+      "loss": 0.9109,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999387998878088e-06,
+      "loss": 0.9417,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999370638369377e-06,
+      "loss": 0.9619,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999353035076864e-06,
+      "loss": 0.9294,
+      "step": 514
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999335189001408e-06,
+      "loss": 1.0481,
+      "step": 515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999317100143873e-06,
+      "loss": 0.8685,
+      "step": 516
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999298768505136e-06,
+      "loss": 0.6911,
+      "step": 517
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99928019408609e-06,
+      "loss": 0.6866,
+      "step": 518
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999261376887634e-06,
+      "loss": 0.8125,
+      "step": 519
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999242316910685e-06,
+      "loss": 1.084,
+      "step": 520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999223014156167e-06,
+      "loss": 0.6803,
+      "step": 521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999203468625017e-06,
+      "loss": 0.5298,
+      "step": 522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999183680318185e-06,
+      "loss": 0.7797,
+      "step": 523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999163649236632e-06,
+      "loss": 0.9839,
+      "step": 524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999143375381331e-06,
+      "loss": 1.0552,
+      "step": 525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999122858753265e-06,
+      "loss": 0.8215,
+      "step": 526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999102099353432e-06,
+      "loss": 0.7418,
+      "step": 527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99908109718284e-06,
+      "loss": 1.0466,
+      "step": 528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999059852242508e-06,
+      "loss": 1.0271,
+      "step": 529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999038364533469e-06,
+      "loss": 1.0975,
+      "step": 530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.999016634056764e-06,
+      "loss": 1.1038,
+      "step": 531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99899466081345e-06,
+      "loss": 1.1738,
+      "step": 532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998972444804596e-06,
+      "loss": 0.9639,
+      "step": 533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998949986031277e-06,
+      "loss": 0.9396,
+      "step": 534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998927284494586e-06,
+      "loss": 0.9038,
+      "step": 535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998904340195626e-06,
+      "loss": 0.8264,
+      "step": 536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99888115313551e-06,
+      "loss": 0.8689,
+      "step": 537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998857723315363e-06,
+      "loss": 1.1806,
+      "step": 538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998834050736326e-06,
+      "loss": 0.8824,
+      "step": 539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998810135399545e-06,
+      "loss": 0.931,
+      "step": 540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998785977306185e-06,
+      "loss": 0.8786,
+      "step": 541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998761576457417e-06,
+      "loss": 0.9607,
+      "step": 542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998736932854425e-06,
+      "loss": 0.8477,
+      "step": 543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998712046498408e-06,
+      "loss": 0.9265,
+      "step": 544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998686917390575e-06,
+      "loss": 1.1029,
+      "step": 545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998661545532142e-06,
+      "loss": 1.0182,
+      "step": 546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998635930924346e-06,
+      "loss": 1.1153,
+      "step": 547
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998610073568428e-06,
+      "loss": 1.1893,
+      "step": 548
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998583973465647e-06,
+      "loss": 0.8113,
+      "step": 549
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998557630617266e-06,
+      "loss": 0.8888,
+      "step": 550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998531045024567e-06,
+      "loss": 0.8428,
+      "step": 551
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99850421668884e-06,
+      "loss": 0.6468,
+      "step": 552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998477145611389e-06,
+      "loss": 0.9844,
+      "step": 553
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998449831793528e-06,
+      "loss": 0.8977,
+      "step": 554
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998422275236584e-06,
+      "loss": 0.9396,
+      "step": 555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998394475941895e-06,
+      "loss": 1.0258,
+      "step": 556
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99836643391081e-06,
+      "loss": 0.6159,
+      "step": 557
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998338149144693e-06,
+      "loss": 1.2241,
+      "step": 558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998309621644915e-06,
+      "loss": 0.8942,
+      "step": 559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998280851412862e-06,
+      "loss": 1.0843,
+      "step": 560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998251838449932e-06,
+      "loss": 0.7246,
+      "step": 561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998222582757533e-06,
+      "loss": 0.814,
+      "step": 562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99819308433709e-06,
+      "loss": 0.6943,
+      "step": 563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998163343190028e-06,
+      "loss": 0.9997,
+      "step": 564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998133359317797e-06,
+      "loss": 0.6625,
+      "step": 565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99810313272185e-06,
+      "loss": 1.1172,
+      "step": 566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998072663403657e-06,
+      "loss": 1.2199,
+      "step": 567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998041951364696e-06,
+      "loss": 1.0306,
+      "step": 568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99801099660646e-06,
+      "loss": 1.0597,
+      "step": 569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997979799130453e-06,
+      "loss": 0.7157,
+      "step": 570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997948358938188e-06,
+      "loss": 1.003,
+      "step": 571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997916676031192e-06,
+      "loss": 1.1984,
+      "step": 572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997884750411004e-06,
+      "loss": 0.7911,
+      "step": 573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997852582079174e-06,
+      "loss": 0.8022,
+      "step": 574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997820171037266e-06,
+      "loss": 1.1303,
+      "step": 575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997787517286852e-06,
+      "loss": 0.7424,
+      "step": 576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99775462082952e-06,
+      "loss": 0.805,
+      "step": 577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997721481666863e-06,
+      "loss": 0.8207,
+      "step": 578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997688099800496e-06,
+      "loss": 1.015,
+      "step": 579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997654475232035e-06,
+      "loss": 0.9507,
+      "step": 580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997620607963116e-06,
+      "loss": 0.7859,
+      "step": 581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997586497995383e-06,
+      "loss": 0.8426,
+      "step": 582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997552145330494e-06,
+      "loss": 0.9555,
+      "step": 583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997517549970116e-06,
+      "loss": 1.0098,
+      "step": 584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997482711915926e-06,
+      "loss": 0.7383,
+      "step": 585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997447631169622e-06,
+      "loss": 1.0562,
+      "step": 586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9974123077329e-06,
+      "loss": 0.9193,
+      "step": 587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997376741607482e-06,
+      "loss": 1.0952,
+      "step": 588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997340932795094e-06,
+      "loss": 0.9795,
+      "step": 589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997304881297474e-06,
+      "loss": 1.0383,
+      "step": 590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997268587116372e-06,
+      "loss": 0.8487,
+      "step": 591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997232050253549e-06,
+      "loss": 1.1559,
+      "step": 592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997195270710784e-06,
+      "loss": 0.8201,
+      "step": 593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99715824848986e-06,
+      "loss": 0.8703,
+      "step": 594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997120983592576e-06,
+      "loss": 1.0464,
+      "step": 595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99708347602074e-06,
+      "loss": 0.7406,
+      "step": 596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997045725776175e-06,
+      "loss": 0.9317,
+      "step": 597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.997007732860716e-06,
+      "loss": 0.9453,
+      "step": 598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996969497276204e-06,
+      "loss": 0.6378,
+      "step": 599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996931019024498e-06,
+      "loss": 0.9311,
+      "step": 600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996892298107466e-06,
+      "loss": 0.7054,
+      "step": 601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99685333452699e-06,
+      "loss": 0.7005,
+      "step": 602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99681412828496e-06,
+      "loss": 0.8484,
+      "step": 603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996774679383281e-06,
+      "loss": 0.7731,
+      "step": 604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996734987823869e-06,
+      "loss": 0.8296,
+      "step": 605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996695053608651e-06,
+      "loss": 1.1222,
+      "step": 606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996654876739566e-06,
+      "loss": 0.9667,
+      "step": 607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996614457218568e-06,
+      "loss": 1.1136,
+      "step": 608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996573795047616e-06,
+      "loss": 0.7122,
+      "step": 609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996532890228687e-06,
+      "loss": 0.8282,
+      "step": 610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996491742763766e-06,
+      "loss": 1.0208,
+      "step": 611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996450352654852e-06,
+      "loss": 1.018,
+      "step": 612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996408719903956e-06,
+      "loss": 0.9424,
+      "step": 613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996366844513097e-06,
+      "loss": 0.8931,
+      "step": 614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996324726484313e-06,
+      "loss": 0.7995,
+      "step": 615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996282365819644e-06,
+      "loss": 0.6003,
+      "step": 616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996239762521152e-06,
+      "loss": 0.9366,
+      "step": 617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996196916590902e-06,
+      "loss": 0.7289,
+      "step": 618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996153828030978e-06,
+      "loss": 0.5565,
+      "step": 619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99611049684347e-06,
+      "loss": 0.7998,
+      "step": 620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996066923030484e-06,
+      "loss": 0.8477,
+      "step": 621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.996023106594135e-06,
+      "loss": 0.7558,
+      "step": 622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995979047536553e-06,
+      "loss": 1.3387,
+      "step": 623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995934745859873e-06,
+      "loss": 1.1974,
+      "step": 624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995890201566251e-06,
+      "loss": 0.9241,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995845414657849e-06,
+      "loss": 0.4868,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99580038513684e-06,
+      "loss": 1.0591,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995755113005414e-06,
+      "loss": 0.8194,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995709598265765e-06,
+      "loss": 0.666,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995663840920109e-06,
+      "loss": 0.8259,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995617840970664e-06,
+      "loss": 1.0881,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995571598419667e-06,
+      "loss": 0.6999,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99552511326936e-06,
+      "loss": 0.7324,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995478385522002e-06,
+      "loss": 0.8462,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995431415179864e-06,
+      "loss": 1.1179,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995384202245224e-06,
+      "loss": 0.8413,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995336746720376e-06,
+      "loss": 1.1809,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995289048607627e-06,
+      "loss": 0.9695,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99524110790929e-06,
+      "loss": 0.6834,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995192924627694e-06,
+      "loss": 1.0083,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995144498765179e-06,
+      "loss": 0.8407,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995095830324099e-06,
+      "loss": 0.7773,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.995046919306814e-06,
+      "loss": 0.9561,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.9949977657157e-06,
+      "loss": 0.9485,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994948369553146e-06,
+      "loss": 0.7437,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99489873082155e-06,
+      "loss": 1.0295,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994848849523322e-06,
+      "loss": 0.6941,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994798725660884e-06,
+      "loss": 1.0064,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99474835923667e-06,
+      "loss": 0.9345,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994697750253127e-06,
+      "loss": 0.6706,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994646898712713e-06,
+      "loss": 0.9667,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994595804617897e-06,
+      "loss": 0.8079,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994544467971159e-06,
+      "loss": 1.0069,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994492888774996e-06,
+      "loss": 0.9708,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994441067031906e-06,
+      "loss": 0.8045,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994389002744414e-06,
+      "loss": 0.913,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994336695915041e-06,
+      "loss": 0.6229,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994284146546332e-06,
+      "loss": 0.8022,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994231354640835e-06,
+      "loss": 1.0338,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994178320201117e-06,
+      "loss": 1.0191,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.994125043229753e-06,
+      "loss": 0.9216,
+      "step": 661
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99407152372933e-06,
+      "loss": 0.8117,
+      "step": 662
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.994017761702446e-06,
+      "loss": 0.891,
+      "step": 663
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993963757151713e-06,
+      "loss": 0.7839,
+      "step": 664
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993909510079752e-06,
+      "loss": 0.9883,
+      "step": 665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993855020489198e-06,
+      "loss": 0.9923,
+      "step": 666
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9938002883827e-06,
+      "loss": 0.687,
+      "step": 667
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993745313762913e-06,
+      "loss": 0.9995,
+      "step": 668
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993690096632507e-06,
+      "loss": 0.8745,
+      "step": 669
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993634636994165e-06,
+      "loss": 0.8541,
+      "step": 670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993578934850579e-06,
+      "loss": 0.9272,
+      "step": 671
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993522990204453e-06,
+      "loss": 0.7993,
+      "step": 672
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993466803058505e-06,
+      "loss": 0.8555,
+      "step": 673
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993410373415465e-06,
+      "loss": 0.918,
+      "step": 674
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993353701278072e-06,
+      "loss": 1.0979,
+      "step": 675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993296786649078e-06,
+      "loss": 0.8213,
+      "step": 676
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993239629531247e-06,
+      "loss": 1.0264,
+      "step": 677
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993182229927356e-06,
+      "loss": 0.9114,
+      "step": 678
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993124587840191e-06,
+      "loss": 0.799,
+      "step": 679
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99306670327255e-06,
+      "loss": 1.0365,
+      "step": 680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.993008576227248e-06,
+      "loss": 1.0552,
+      "step": 681
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992950206707104e-06,
+      "loss": 1.0555,
+      "step": 682
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992891594714954e-06,
+      "loss": 0.9426,
+      "step": 683
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992832740253646e-06,
+      "loss": 0.6923,
+      "step": 684
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992773643326034e-06,
+      "loss": 1.0317,
+      "step": 685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992714303934993e-06,
+      "loss": 1.0452,
+      "step": 686
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992654722083401e-06,
+      "loss": 0.9683,
+      "step": 687
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992594897774155e-06,
+      "loss": 0.9734,
+      "step": 688
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992534831010155e-06,
+      "loss": 0.8736,
+      "step": 689
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992474521794322e-06,
+      "loss": 0.89,
+      "step": 690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992413970129584e-06,
+      "loss": 1.0117,
+      "step": 691
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99235317601888e-06,
+      "loss": 0.7857,
+      "step": 692
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992292139465166e-06,
+      "loss": 0.8763,
+      "step": 693
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992230860471402e-06,
+      "loss": 0.863,
+      "step": 694
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992169339040567e-06,
+      "loss": 0.8634,
+      "step": 695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.992107575175647e-06,
+      "loss": 0.8397,
+      "step": 696
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99204556887964e-06,
+      "loss": 0.9133,
+      "step": 697
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99198332015556e-06,
+      "loss": 0.8809,
+      "step": 698
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991920829006429e-06,
+      "loss": 1.0297,
+      "step": 699
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991858095435284e-06,
+      "loss": 0.7653,
+      "step": 700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991795119445165e-06,
+      "loss": 0.6905,
+      "step": 701
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991731901039137e-06,
+      "loss": 1.0048,
+      "step": 702
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991668440220269e-06,
+      "loss": 1.0553,
+      "step": 703
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991604736991638e-06,
+      "loss": 0.9565,
+      "step": 704
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991540791356342e-06,
+      "loss": 0.9506,
+      "step": 705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991476603317486e-06,
+      "loss": 1.1772,
+      "step": 706
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991412172878186e-06,
+      "loss": 1.1097,
+      "step": 707
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991347500041571e-06,
+      "loss": 0.7317,
+      "step": 708
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991282584810784e-06,
+      "loss": 1.1012,
+      "step": 709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991217427188975e-06,
+      "loss": 1.0166,
+      "step": 710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991152027179307e-06,
+      "loss": 0.679,
+      "step": 711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99108638478496e-06,
+      "loss": 1.0204,
+      "step": 712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.991020500009118e-06,
+      "loss": 0.9321,
+      "step": 713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990954372854985e-06,
+      "loss": 0.96,
+      "step": 714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990888003325766e-06,
+      "loss": 0.8274,
+      "step": 715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990821391424689e-06,
+      "loss": 0.9132,
+      "step": 716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99075453715499e-06,
+      "loss": 0.7669,
+      "step": 717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99068744051991e-06,
+      "loss": 0.8302,
+      "step": 718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990620101522713e-06,
+      "loss": 0.8255,
+      "step": 719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990552520166664e-06,
+      "loss": 0.917,
+      "step": 720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99048469645505e-06,
+      "loss": 0.8114,
+      "step": 721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990416630391161e-06,
+      "loss": 1.1028,
+      "step": 722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990348321978304e-06,
+      "loss": 0.6827,
+      "step": 723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990279771219796e-06,
+      "loss": 0.8701,
+      "step": 724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990210978118967e-06,
+      "loss": 1.0165,
+      "step": 725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990141942679157e-06,
+      "loss": 0.9197,
+      "step": 726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990072664903717e-06,
+      "loss": 0.9114,
+      "step": 727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.990003144796014e-06,
+      "loss": 0.9476,
+      "step": 728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989933382359423e-06,
+      "loss": 0.8018,
+      "step": 729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989863377597331e-06,
+      "loss": 0.9666,
+      "step": 730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989793130513138e-06,
+      "loss": 0.9666,
+      "step": 731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989722641110258e-06,
+      "loss": 0.4707,
+      "step": 732
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98965190939211e-06,
+      "loss": 1.012,
+      "step": 733
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989580935362133e-06,
+      "loss": 0.8088,
+      "step": 734
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98950971902377e-06,
+      "loss": 0.9398,
+      "step": 735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989438260380481e-06,
+      "loss": 0.9767,
+      "step": 736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989366559435736e-06,
+      "loss": 0.9475,
+      "step": 737
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989294616193018e-06,
+      "loss": 1.2218,
+      "step": 738
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98922243065582e-06,
+      "loss": 0.9277,
+      "step": 739
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989150002827647e-06,
+      "loss": 0.9842,
+      "step": 740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989077332712019e-06,
+      "loss": 0.8288,
+      "step": 741
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.989004420312461e-06,
+      "loss": 0.9884,
+      "step": 742
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988931265632517e-06,
+      "loss": 1.1002,
+      "step": 743
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98885786867574e-06,
+      "loss": 1.1019,
+      "step": 744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988784229445689e-06,
+      "loss": 0.9135,
+      "step": 745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988710347945946e-06,
+      "loss": 1.0187,
+      "step": 746
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988636224180097e-06,
+      "loss": 1.0357,
+      "step": 747
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98856185815174e-06,
+      "loss": 0.9085,
+      "step": 748
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98848724986449e-06,
+      "loss": 0.7701,
+      "step": 749
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988412399321965e-06,
+      "loss": 0.7242,
+      "step": 750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988337306527807e-06,
+      "loss": 1.1952,
+      "step": 751
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988261971485657e-06,
+      "loss": 0.9736,
+      "step": 752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988186394199175e-06,
+      "loss": 0.9072,
+      "step": 753
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988110574672031e-06,
+      "loss": 1.1,
+      "step": 754
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.988034512907909e-06,
+      "loss": 0.9729,
+      "step": 755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9879582089105e-06,
+      "loss": 0.556,
+      "step": 756
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987881662683514e-06,
+      "loss": 0.7109,
+      "step": 757
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987804874230662e-06,
+      "loss": 0.689,
+      "step": 758
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987727843555677e-06,
+      "loss": 0.7439,
+      "step": 759
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9876505706623e-06,
+      "loss": 0.9961,
+      "step": 760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98757305555428e-06,
+      "loss": 0.9773,
+      "step": 761
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987495298235386e-06,
+      "loss": 0.7225,
+      "step": 762
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987417298709393e-06,
+      "loss": 0.8906,
+      "step": 763
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987339056980086e-06,
+      "loss": 1.0509,
+      "step": 764
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987260573051268e-06,
+      "loss": 0.8076,
+      "step": 765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987181846926751e-06,
+      "loss": 0.7733,
+      "step": 766
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987102878610353e-06,
+      "loss": 1.0183,
+      "step": 767
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.987023668105913e-06,
+      "loss": 0.7147,
+      "step": 768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986944215417276e-06,
+      "loss": 0.8549,
+      "step": 769
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986864520548303e-06,
+      "loss": 0.5676,
+      "step": 770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986784583502863e-06,
+      "loss": 0.9589,
+      "step": 771
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986704404284834e-06,
+      "loss": 1.1171,
+      "step": 772
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986623982898118e-06,
+      "loss": 0.8307,
+      "step": 773
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986543319346613e-06,
+      "loss": 0.9294,
+      "step": 774
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98646241363424e-06,
+      "loss": 0.9143,
+      "step": 775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986381265764927e-06,
+      "loss": 1.042,
+      "step": 776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986299875742612e-06,
+      "loss": 0.8004,
+      "step": 777
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986218243571253e-06,
+      "loss": 0.7599,
+      "step": 778
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98613636925481e-06,
+      "loss": 1.0906,
+      "step": 779
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.986054252797264e-06,
+      "loss": 0.9139,
+      "step": 780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985971894202596e-06,
+      "loss": 0.9434,
+      "step": 781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98588929347481e-06,
+      "loss": 1.0159,
+      "step": 782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985806450617916e-06,
+      "loss": 0.8714,
+      "step": 783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985723365635939e-06,
+      "loss": 0.9299,
+      "step": 784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98564003853291e-06,
+      "loss": 0.8276,
+      "step": 785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98555646931288e-06,
+      "loss": 1.0971,
+      "step": 786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985472657979904e-06,
+      "loss": 0.7184,
+      "step": 787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985388604538055e-06,
+      "loss": 0.6482,
+      "step": 788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985304308991412e-06,
+      "loss": 0.9537,
+      "step": 789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98521977134407e-06,
+      "loss": 0.8789,
+      "step": 790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985134991600135e-06,
+      "loss": 0.9967,
+      "step": 791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.985049969763721e-06,
+      "loss": 0.7109,
+      "step": 792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98496470583896e-06,
+      "loss": 0.7495,
+      "step": 793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984879199829994e-06,
+      "loss": 0.8907,
+      "step": 794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984793451740972e-06,
+      "loss": 1.0264,
+      "step": 795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98470746157606e-06,
+      "loss": 0.9792,
+      "step": 796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984621229339434e-06,
+      "loss": 1.2087,
+      "step": 797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98453475503528e-06,
+      "loss": 0.961,
+      "step": 798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9844480386678e-06,
+      "loss": 0.7172,
+      "step": 799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984361080241204e-06,
+      "loss": 0.8242,
+      "step": 800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984273879759713e-06,
+      "loss": 0.7281,
+      "step": 801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984186437227564e-06,
+      "loss": 0.7914,
+      "step": 802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984098752649004e-06,
+      "loss": 0.736,
+      "step": 803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.984010826028289e-06,
+      "loss": 0.633,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98392265736969e-06,
+      "loss": 0.843,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983834246677489e-06,
+      "loss": 0.6566,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.98374559395598e-06,
+      "loss": 0.8687,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.983656699209467e-06,
+      "loss": 0.9731,
+      "step": 808
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98356756244227e-06,
+      "loss": 0.9285,
+      "step": 809
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983478183658712e-06,
+      "loss": 0.7114,
+      "step": 810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983388562863138e-06,
+      "loss": 0.874,
+      "step": 811
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983298700059898e-06,
+      "loss": 0.7764,
+      "step": 812
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983208595253359e-06,
+      "loss": 0.8247,
+      "step": 813
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.983118248447894e-06,
+      "loss": 0.9693,
+      "step": 814
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98302765964789e-06,
+      "loss": 0.791,
+      "step": 815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982936828857749e-06,
+      "loss": 0.9437,
+      "step": 816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98284575608188e-06,
+      "loss": 1.0417,
+      "step": 817
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982754441324707e-06,
+      "loss": 0.8688,
+      "step": 818
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982662884590662e-06,
+      "loss": 0.8816,
+      "step": 819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982571085884195e-06,
+      "loss": 0.9904,
+      "step": 820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98247904520976e-06,
+      "loss": 0.6892,
+      "step": 821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98238676257183e-06,
+      "loss": 1.1648,
+      "step": 822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982294237974885e-06,
+      "loss": 0.8184,
+      "step": 823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98220147142342e-06,
+      "loss": 1.142,
+      "step": 824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982108462921938e-06,
+      "loss": 0.937,
+      "step": 825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.982015212474955e-06,
+      "loss": 0.9604,
+      "step": 826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981921720087002e-06,
+      "loss": 0.8462,
+      "step": 827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981827985762618e-06,
+      "loss": 1.1554,
+      "step": 828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981734009506353e-06,
+      "loss": 0.7734,
+      "step": 829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981639791322777e-06,
+      "loss": 1.1122,
+      "step": 830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981545331216458e-06,
+      "loss": 0.5724,
+      "step": 831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981450629191988e-06,
+      "loss": 0.7601,
+      "step": 832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981355685253965e-06,
+      "loss": 0.9869,
+      "step": 833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981260499407e-06,
+      "loss": 0.802,
+      "step": 834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.981165071655712e-06,
+      "loss": 1.0743,
+      "step": 835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98106940200474e-06,
+      "loss": 1.1156,
+      "step": 836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980973490458728e-06,
+      "loss": 0.654,
+      "step": 837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980877337022334e-06,
+      "loss": 0.5341,
+      "step": 838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980780941700227e-06,
+      "loss": 0.5768,
+      "step": 839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98068430449709e-06,
+      "loss": 0.932,
+      "step": 840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980587425417612e-06,
+      "loss": 0.7007,
+      "step": 841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980490304466503e-06,
+      "loss": 0.912,
+      "step": 842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980392941648476e-06,
+      "loss": 0.8129,
+      "step": 843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.98029533696826e-06,
+      "loss": 0.852,
+      "step": 844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980197490430592e-06,
+      "loss": 1.0018,
+      "step": 845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980099402040231e-06,
+      "loss": 0.6849,
+      "step": 846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.980001071801933e-06,
+      "loss": 0.9769,
+      "step": 847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979902499720477e-06,
+      "loss": 0.888,
+      "step": 848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979803685800651e-06,
+      "loss": 0.9107,
+      "step": 849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979704630047251e-06,
+      "loss": 0.7805,
+      "step": 850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979605332465088e-06,
+      "loss": 0.8358,
+      "step": 851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979505793058985e-06,
+      "loss": 0.7062,
+      "step": 852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979406011833775e-06,
+      "loss": 0.6887,
+      "step": 853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979305988794304e-06,
+      "loss": 0.7493,
+      "step": 854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97920572394543e-06,
+      "loss": 0.5542,
+      "step": 855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.979105217292023e-06,
+      "loss": 0.8126,
+      "step": 856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97900446883896e-06,
+      "loss": 0.8511,
+      "step": 857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978903478591137e-06,
+      "loss": 0.8154,
+      "step": 858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978802246553458e-06,
+      "loss": 0.9594,
+      "step": 859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978700772730838e-06,
+      "loss": 0.7871,
+      "step": 860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978599057128207e-06,
+      "loss": 0.5975,
+      "step": 861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9784970997505e-06,
+      "loss": 0.6185,
+      "step": 862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978394900602676e-06,
+      "loss": 0.9218,
+      "step": 863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978292459689692e-06,
+      "loss": 1.1037,
+      "step": 864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978189777016523e-06,
+      "loss": 0.8095,
+      "step": 865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.978086852588157e-06,
+      "loss": 0.6856,
+      "step": 866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977983686409595e-06,
+      "loss": 0.8892,
+      "step": 867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977880278485842e-06,
+      "loss": 0.9921,
+      "step": 868
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977776628821923e-06,
+      "loss": 1.1702,
+      "step": 869
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97767273742287e-06,
+      "loss": 0.8783,
+      "step": 870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97756860429373e-06,
+      "loss": 1.1596,
+      "step": 871
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977464229439557e-06,
+      "loss": 0.8342,
+      "step": 872
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977359612865424e-06,
+      "loss": 1.0359,
+      "step": 873
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977254754576409e-06,
+      "loss": 0.9948,
+      "step": 874
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977149654577602e-06,
+      "loss": 0.8403,
+      "step": 875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.977044312874111e-06,
+      "loss": 0.8605,
+      "step": 876
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976938729471049e-06,
+      "loss": 0.9936,
+      "step": 877
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976832904373546e-06,
+      "loss": 0.9939,
+      "step": 878
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97672683758674e-06,
+      "loss": 0.8771,
+      "step": 879
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97662052911578e-06,
+      "loss": 0.7188,
+      "step": 880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976513978965829e-06,
+      "loss": 0.9249,
+      "step": 881
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976407187142066e-06,
+      "loss": 0.7348,
+      "step": 882
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97630015364967e-06,
+      "loss": 0.9338,
+      "step": 883
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976192878493844e-06,
+      "loss": 1.0085,
+      "step": 884
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.976085361679797e-06,
+      "loss": 1.0025,
+      "step": 885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97597760321275e-06,
+      "loss": 0.639,
+      "step": 886
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975869603097934e-06,
+      "loss": 0.924,
+      "step": 887
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975761361340596e-06,
+      "loss": 0.846,
+      "step": 888
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975652877945991e-06,
+      "loss": 0.7529,
+      "step": 889
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97554415291939e-06,
+      "loss": 0.8563,
+      "step": 890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975435186266069e-06,
+      "loss": 0.7845,
+      "step": 891
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975325977991323e-06,
+      "loss": 0.9205,
+      "step": 892
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975216528100453e-06,
+      "loss": 0.8585,
+      "step": 893
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.975106836598777e-06,
+      "loss": 0.7845,
+      "step": 894
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97499690349162e-06,
+      "loss": 0.7377,
+      "step": 895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97488672878432e-06,
+      "loss": 0.9445,
+      "step": 896
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97477631248223e-06,
+      "loss": 0.823,
+      "step": 897
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97466565459071e-06,
+      "loss": 0.8876,
+      "step": 898
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974554755115134e-06,
+      "loss": 0.864,
+      "step": 899
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974443614060889e-06,
+      "loss": 0.8641,
+      "step": 900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974332231433372e-06,
+      "loss": 0.8034,
+      "step": 901
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97422060723799e-06,
+      "loss": 0.8655,
+      "step": 902
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.974108741480167e-06,
+      "loss": 1.0799,
+      "step": 903
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973996634165332e-06,
+      "loss": 0.9288,
+      "step": 904
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973884285298932e-06,
+      "loss": 0.654,
+      "step": 905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973771694886421e-06,
+      "loss": 0.8955,
+      "step": 906
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97365886293327e-06,
+      "loss": 1.0048,
+      "step": 907
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973545789444954e-06,
+      "loss": 0.921,
+      "step": 908
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973432474426968e-06,
+      "loss": 0.7134,
+      "step": 909
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973318917884811e-06,
+      "loss": 0.8356,
+      "step": 910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973205119824001e-06,
+      "loss": 1.0431,
+      "step": 911
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.973091080250063e-06,
+      "loss": 0.703,
+      "step": 912
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972976799168537e-06,
+      "loss": 0.923,
+      "step": 913
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97286227658497e-06,
+      "loss": 0.9369,
+      "step": 914
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972747512504925e-06,
+      "loss": 0.892,
+      "step": 915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972632506933975e-06,
+      "loss": 0.9426,
+      "step": 916
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972517259877706e-06,
+      "loss": 1.0259,
+      "step": 917
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972401771341711e-06,
+      "loss": 0.9894,
+      "step": 918
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972286041331602e-06,
+      "loss": 0.9155,
+      "step": 919
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972170069853e-06,
+      "loss": 0.7154,
+      "step": 920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972053856911534e-06,
+      "loss": 0.7764,
+      "step": 921
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97193740251285e-06,
+      "loss": 1.1794,
+      "step": 922
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971820706662601e-06,
+      "loss": 0.8321,
+      "step": 923
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971703769366456e-06,
+      "loss": 0.8042,
+      "step": 924
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971586590630094e-06,
+      "loss": 1.0626,
+      "step": 925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971469170459202e-06,
+      "loss": 0.5372,
+      "step": 926
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971351508859488e-06,
+      "loss": 0.8777,
+      "step": 927
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97123360583666e-06,
+      "loss": 0.912,
+      "step": 928
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97111546139645e-06,
+      "loss": 0.7869,
+      "step": 929
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97099707554459e-06,
+      "loss": 0.6234,
+      "step": 930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970878448286832e-06,
+      "loss": 0.8101,
+      "step": 931
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970759579628934e-06,
+      "loss": 0.7453,
+      "step": 932
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970640469576674e-06,
+      "loss": 0.8642,
+      "step": 933
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97052111813583e-06,
+      "loss": 0.5998,
+      "step": 934
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970401525312203e-06,
+      "loss": 0.8027,
+      "step": 935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970281691111598e-06,
+      "loss": 0.9096,
+      "step": 936
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970161615539837e-06,
+      "loss": 0.9558,
+      "step": 937
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970041298602748e-06,
+      "loss": 0.9493,
+      "step": 938
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969920740306176e-06,
+      "loss": 1.0839,
+      "step": 939
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969799940655975e-06,
+      "loss": 0.9094,
+      "step": 940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969678899658012e-06,
+      "loss": 0.8007,
+      "step": 941
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969557617318163e-06,
+      "loss": 0.6995,
+      "step": 942
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969436093642321e-06,
+      "loss": 0.821,
+      "step": 943
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969314328636386e-06,
+      "loss": 0.8389,
+      "step": 944
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969192322306271e-06,
+      "loss": 1.0772,
+      "step": 945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969070074657901e-06,
+      "loss": 1.0473,
+      "step": 946
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968947585697214e-06,
+      "loss": 0.9075,
+      "step": 947
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968824855430157e-06,
+      "loss": 0.9522,
+      "step": 948
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96870188386269e-06,
+      "loss": 0.7876,
+      "step": 949
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968578671000784e-06,
+      "loss": 0.6698,
+      "step": 950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968455216850426e-06,
+      "loss": 1.0314,
+      "step": 951
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96833152141761e-06,
+      "loss": 0.8518,
+      "step": 952
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96820758470834e-06,
+      "loss": 0.788,
+      "step": 953
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968083406728637e-06,
+      "loss": 0.9662,
+      "step": 954
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.967958987484532e-06,
+      "loss": 1.0834,
+      "step": 955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967834326982066e-06,
+      "loss": 0.6006,
+      "step": 956
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967709425227295e-06,
+      "loss": 0.9836,
+      "step": 957
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967584282226281e-06,
+      "loss": 0.6001,
+      "step": 958
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967458897985105e-06,
+      "loss": 0.9284,
+      "step": 959
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967333272509854e-06,
+      "loss": 0.8488,
+      "step": 960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967207405806628e-06,
+      "loss": 0.7975,
+      "step": 961
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.967081297881541e-06,
+      "loss": 0.8351,
+      "step": 962
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966954948740717e-06,
+      "loss": 0.7796,
+      "step": 963
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96682835839029e-06,
+      "loss": 0.794,
+      "step": 964
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96670152683641e-06,
+      "loss": 0.6236,
+      "step": 965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966574454085236e-06,
+      "loss": 0.6932,
+      "step": 966
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966447140142938e-06,
+      "loss": 1.1135,
+      "step": 967
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9663195850157e-06,
+      "loss": 0.7784,
+      "step": 968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.966191788709716e-06,
+      "loss": 0.8298,
+      "step": 969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96606375123119e-06,
+      "loss": 0.9059,
+      "step": 970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965935472586344e-06,
+      "loss": 0.7616,
+      "step": 971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965806952781402e-06,
+      "loss": 0.8778,
+      "step": 972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965678191822612e-06,
+      "loss": 1.0371,
+      "step": 973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96554918971622e-06,
+      "loss": 1.0207,
+      "step": 974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965419946468496e-06,
+      "loss": 1.0085,
+      "step": 975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965290462085714e-06,
+      "loss": 0.7069,
+      "step": 976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965160736574164e-06,
+      "loss": 0.7451,
+      "step": 977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.965030769940144e-06,
+      "loss": 0.7624,
+      "step": 978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964900562189967e-06,
+      "loss": 0.9137,
+      "step": 979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964770113329953e-06,
+      "loss": 0.8597,
+      "step": 980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964639423366442e-06,
+      "loss": 0.6523,
+      "step": 981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964508492305777e-06,
+      "loss": 0.6447,
+      "step": 982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964377320154318e-06,
+      "loss": 0.8247,
+      "step": 983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964245906918434e-06,
+      "loss": 0.95,
+      "step": 984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.964114252604508e-06,
+      "loss": 0.8897,
+      "step": 985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963982357218933e-06,
+      "loss": 0.8203,
+      "step": 986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963850220768112e-06,
+      "loss": 0.7757,
+      "step": 987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963717843258467e-06,
+      "loss": 0.7768,
+      "step": 988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96358522469642e-06,
+      "loss": 0.9281,
+      "step": 989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96345236508842e-06,
+      "loss": 0.9078,
+      "step": 990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96331926444091e-06,
+      "loss": 0.7585,
+      "step": 991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96318592276036e-06,
+      "loss": 0.9591,
+      "step": 992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.963052340053242e-06,
+      "loss": 0.73,
+      "step": 993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962918516326045e-06,
+      "loss": 0.7862,
+      "step": 994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962784451585267e-06,
+      "loss": 0.8513,
+      "step": 995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962650145837419e-06,
+      "loss": 0.8993,
+      "step": 996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962515599089022e-06,
+      "loss": 0.7956,
+      "step": 997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962380811346612e-06,
+      "loss": 0.9562,
+      "step": 998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962245782616734e-06,
+      "loss": 1.0018,
+      "step": 999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.962110512905947e-06,
+      "loss": 0.7094,
+      "step": 1000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961975002220816e-06,
+      "loss": 0.9404,
+      "step": 1001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961839250567925e-06,
+      "loss": 0.992,
+      "step": 1002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961703257953863e-06,
+      "loss": 0.6723,
+      "step": 1003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96156702438524e-06,
+      "loss": 0.794,
+      "step": 1004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961430549868668e-06,
+      "loss": 0.635,
+      "step": 1005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961293834410774e-06,
+      "loss": 0.5754,
+      "step": 1006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961156878018199e-06,
+      "loss": 0.9945,
+      "step": 1007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961019680697593e-06,
+      "loss": 1.0653,
+      "step": 1008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960882242455621e-06,
+      "loss": 0.8014,
+      "step": 1009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960744563298955e-06,
+      "loss": 0.9883,
+      "step": 1010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960606643234278e-06,
+      "loss": 1.0897,
+      "step": 1011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960468482268295e-06,
+      "loss": 0.6663,
+      "step": 1012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960330080407712e-06,
+      "loss": 0.7559,
+      "step": 1013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960191437659251e-06,
+      "loss": 0.7644,
+      "step": 1014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960052554029643e-06,
+      "loss": 0.6677,
+      "step": 1015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959913429525632e-06,
+      "loss": 0.7873,
+      "step": 1016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959774064153977e-06,
+      "loss": 0.6822,
+      "step": 1017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959634457921446e-06,
+      "loss": 0.9185,
+      "step": 1018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959494610834817e-06,
+      "loss": 0.689,
+      "step": 1019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959354522900881e-06,
+      "loss": 1.0119,
+      "step": 1020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959214194126442e-06,
+      "loss": 1.1086,
+      "step": 1021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959073624518316e-06,
+      "loss": 0.8281,
+      "step": 1022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958932814083328e-06,
+      "loss": 0.8328,
+      "step": 1023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958791762828317e-06,
+      "loss": 0.671,
+      "step": 1024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958650470760131e-06,
+      "loss": 0.8074,
+      "step": 1025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958508937885633e-06,
+      "loss": 0.9857,
+      "step": 1026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958367164211698e-06,
+      "loss": 0.8624,
+      "step": 1027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958225149745207e-06,
+      "loss": 0.8205,
+      "step": 1028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95808289449306e-06,
+      "loss": 0.7773,
+      "step": 1029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957940398462162e-06,
+      "loss": 0.8059,
+      "step": 1030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957797661659436e-06,
+      "loss": 1.0314,
+      "step": 1031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957654684091813e-06,
+      "loss": 0.7638,
+      "step": 1032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957511465766236e-06,
+      "loss": 1.0358,
+      "step": 1033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957368006689661e-06,
+      "loss": 0.8404,
+      "step": 1034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957224306869053e-06,
+      "loss": 0.7965,
+      "step": 1035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95708036631139e-06,
+      "loss": 0.8219,
+      "step": 1036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956936185023668e-06,
+      "loss": 0.9272,
+      "step": 1037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95679176301288e-06,
+      "loss": 1.0783,
+      "step": 1038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956647100286045e-06,
+      "loss": 0.7829,
+      "step": 1039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956502196850188e-06,
+      "loss": 1.0005,
+      "step": 1040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956357052712347e-06,
+      "loss": 0.7066,
+      "step": 1041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956211667879566e-06,
+      "loss": 0.8861,
+      "step": 1042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95606604235891e-06,
+      "loss": 1.0976,
+      "step": 1043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955920176157448e-06,
+      "loss": 0.9447,
+      "step": 1044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955774069282265e-06,
+      "loss": 0.7886,
+      "step": 1045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955627721740454e-06,
+      "loss": 1.0236,
+      "step": 1046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95548113353913e-06,
+      "loss": 0.7422,
+      "step": 1047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9553343046854e-06,
+      "loss": 0.8181,
+      "step": 1048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955187235186403e-06,
+      "loss": 0.6173,
+      "step": 1049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95503992504928e-06,
+      "loss": 0.8782,
+      "step": 1050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95489237428118e-06,
+      "loss": 0.9491,
+      "step": 1051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954744582889274e-06,
+      "loss": 0.8633,
+      "step": 1052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954596550880735e-06,
+      "loss": 0.8697,
+      "step": 1053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954448278262756e-06,
+      "loss": 0.6922,
+      "step": 1054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954299765042534e-06,
+      "loss": 0.8053,
+      "step": 1055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954151011227284e-06,
+      "loss": 0.9417,
+      "step": 1056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954002016824226e-06,
+      "loss": 0.9301,
+      "step": 1057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9538527818406e-06,
+      "loss": 0.8463,
+      "step": 1058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953703306283654e-06,
+      "loss": 0.9243,
+      "step": 1059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95355359016064e-06,
+      "loss": 0.5158,
+      "step": 1060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953403633478835e-06,
+      "loss": 1.1643,
+      "step": 1061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953253436245519e-06,
+      "loss": 0.7633,
+      "step": 1062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953102998467985e-06,
+      "loss": 0.8574,
+      "step": 1063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952952320153541e-06,
+      "loss": 0.8423,
+      "step": 1064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952801401309504e-06,
+      "loss": 0.7767,
+      "step": 1065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952650241943203e-06,
+      "loss": 0.744,
+      "step": 1066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952498842061978e-06,
+      "loss": 0.7337,
+      "step": 1067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952347201673181e-06,
+      "loss": 0.7559,
+      "step": 1068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952195320784175e-06,
+      "loss": 0.8845,
+      "step": 1069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952043199402338e-06,
+      "loss": 0.8228,
+      "step": 1070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951890837535058e-06,
+      "loss": 0.724,
+      "step": 1071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951738235189734e-06,
+      "loss": 1.0015,
+      "step": 1072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951585392373775e-06,
+      "loss": 0.6596,
+      "step": 1073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951432309094605e-06,
+      "loss": 1.0465,
+      "step": 1074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951278985359656e-06,
+      "loss": 0.9023,
+      "step": 1075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.951125421176376e-06,
+      "loss": 0.7877,
+      "step": 1076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950971616552222e-06,
+      "loss": 0.7498,
+      "step": 1077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950817571494664e-06,
+      "loss": 0.7632,
+      "step": 1078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95066328601118e-06,
+      "loss": 0.9069,
+      "step": 1079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950508760109265e-06,
+      "loss": 0.5187,
+      "step": 1080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950353993796424e-06,
+      "loss": 0.6443,
+      "step": 1081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950198987080172e-06,
+      "loss": 0.8618,
+      "step": 1082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.950043739968032e-06,
+      "loss": 0.6364,
+      "step": 1083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949888252467552e-06,
+      "loss": 0.7836,
+      "step": 1084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949732524586276e-06,
+      "loss": 0.723,
+      "step": 1085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.94957655633177e-06,
+      "loss": 0.7709,
+      "step": 1086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949420347711605e-06,
+      "loss": 0.8074,
+      "step": 1087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949263898733372e-06,
+      "loss": 0.9737,
+      "step": 1088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.949107209404664e-06,
+      "loss": 0.9392,
+      "step": 1089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948950279733094e-06,
+      "loss": 0.8589,
+      "step": 1090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948793109726277e-06,
+      "loss": 0.8206,
+      "step": 1091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948635699391853e-06,
+      "loss": 0.9632,
+      "step": 1092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948478048737464e-06,
+      "loss": 0.734,
+      "step": 1093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948320157770763e-06,
+      "loss": 1.1518,
+      "step": 1094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948162026499419e-06,
+      "loss": 0.9827,
+      "step": 1095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.948003654931111e-06,
+      "loss": 0.6939,
+      "step": 1096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947845043073533e-06,
+      "loss": 1.0374,
+      "step": 1097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947686190934385e-06,
+      "loss": 1.1073,
+      "step": 1098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947527098521382e-06,
+      "loss": 0.6708,
+      "step": 1099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947367765842248e-06,
+      "loss": 0.8562,
+      "step": 1100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947208192904722e-06,
+      "loss": 0.766,
+      "step": 1101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.947048379716557e-06,
+      "loss": 0.7519,
+      "step": 1102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94688832628551e-06,
+      "loss": 1.144,
+      "step": 1103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946728032619351e-06,
+      "loss": 0.7051,
+      "step": 1104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946567498725871e-06,
+      "loss": 0.8475,
+      "step": 1105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946406724612861e-06,
+      "loss": 0.9129,
+      "step": 1106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946245710288132e-06,
+      "loss": 0.6664,
+      "step": 1107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9460844557595e-06,
+      "loss": 0.6729,
+      "step": 1108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9459229610348e-06,
+      "loss": 0.9135,
+      "step": 1109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945761226121868e-06,
+      "loss": 0.7301,
+      "step": 1110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945599251028565e-06,
+      "loss": 0.6459,
+      "step": 1111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945437035762755e-06,
+      "loss": 0.8683,
+      "step": 1112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945274580332316e-06,
+      "loss": 0.7698,
+      "step": 1113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945111884745135e-06,
+      "loss": 0.7874,
+      "step": 1114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944948949009114e-06,
+      "loss": 0.7209,
+      "step": 1115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944785773132168e-06,
+      "loss": 0.7286,
+      "step": 1116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944622357122218e-06,
+      "loss": 0.7983,
+      "step": 1117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944458700987202e-06,
+      "loss": 0.8733,
+      "step": 1118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944294804735067e-06,
+      "loss": 0.7891,
+      "step": 1119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94413066837377e-06,
+      "loss": 0.7978,
+      "step": 1120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943966291911286e-06,
+      "loss": 0.9014,
+      "step": 1121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943801675355597e-06,
+      "loss": 0.9267,
+      "step": 1122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943636818714695e-06,
+      "loss": 0.6922,
+      "step": 1123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943471721996588e-06,
+      "loss": 0.7722,
+      "step": 1124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94330638520929e-06,
+      "loss": 0.8368,
+      "step": 1125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943140808360834e-06,
+      "loss": 0.7319,
+      "step": 1126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.94297499145926e-06,
+      "loss": 0.7688,
+      "step": 1127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942808934512617e-06,
+      "loss": 0.5421,
+      "step": 1128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942642637528977e-06,
+      "loss": 0.4896,
+      "step": 1129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942476100516408e-06,
+      "loss": 1.0237,
+      "step": 1130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942309323483003e-06,
+      "loss": 0.7317,
+      "step": 1131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942142306436856e-06,
+      "loss": 0.7485,
+      "step": 1132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941975049386084e-06,
+      "loss": 0.9418,
+      "step": 1133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941807552338805e-06,
+      "loss": 0.794,
+      "step": 1134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941639815303154e-06,
+      "loss": 0.8286,
+      "step": 1135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941471838287276e-06,
+      "loss": 0.8997,
+      "step": 1136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941303621299332e-06,
+      "loss": 0.6749,
+      "step": 1137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.941135164347486e-06,
+      "loss": 1.0116,
+      "step": 1138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940966467439923e-06,
+      "loss": 1.0687,
+      "step": 1139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940797530584835e-06,
+      "loss": 0.794,
+      "step": 1140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940628353790423e-06,
+      "loss": 0.9918,
+      "step": 1141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940458937064905e-06,
+      "loss": 1.0887,
+      "step": 1142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940289280416509e-06,
+      "loss": 1.1518,
+      "step": 1143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.940119383853473e-06,
+      "loss": 0.9407,
+      "step": 1144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939949247384046e-06,
+      "loss": 1.0415,
+      "step": 1145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939778871016493e-06,
+      "loss": 0.6786,
+      "step": 1146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939608254759085e-06,
+      "loss": 1.0658,
+      "step": 1147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939437398620113e-06,
+      "loss": 0.7134,
+      "step": 1148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939266302607869e-06,
+      "loss": 1.0137,
+      "step": 1149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.939094966730664e-06,
+      "loss": 1.056,
+      "step": 1150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938923390996817e-06,
+      "loss": 0.9235,
+      "step": 1151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938751575414663e-06,
+      "loss": 0.8259,
+      "step": 1152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938579519992545e-06,
+      "loss": 0.8496,
+      "step": 1153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938407224738816e-06,
+      "loss": 0.6761,
+      "step": 1154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938234689661846e-06,
+      "loss": 0.9534,
+      "step": 1155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.938061914770013e-06,
+      "loss": 0.9373,
+      "step": 1156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937888900071707e-06,
+      "loss": 0.7732,
+      "step": 1157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93771564557533e-06,
+      "loss": 0.7213,
+      "step": 1158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937542151289296e-06,
+      "loss": 0.7919,
+      "step": 1159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937368417222033e-06,
+      "loss": 0.951,
+      "step": 1160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937194443381972e-06,
+      "loss": 0.6158,
+      "step": 1161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.937020229777568e-06,
+      "loss": 0.6294,
+      "step": 1162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936845776417277e-06,
+      "loss": 0.7708,
+      "step": 1163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936671083309573e-06,
+      "loss": 0.6076,
+      "step": 1164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936496150462938e-06,
+      "loss": 0.9868,
+      "step": 1165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936320977885868e-06,
+      "loss": 0.9148,
+      "step": 1166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.936145565586871e-06,
+      "loss": 0.8865,
+      "step": 1167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935969913574464e-06,
+      "loss": 1.1216,
+      "step": 1168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93579402185718e-06,
+      "loss": 1.2224,
+      "step": 1169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935617890443557e-06,
+      "loss": 0.9285,
+      "step": 1170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93544151934215e-06,
+      "loss": 0.9465,
+      "step": 1171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935264908561523e-06,
+      "loss": 0.8467,
+      "step": 1172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.935088058110256e-06,
+      "loss": 0.8267,
+      "step": 1173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934910967996936e-06,
+      "loss": 0.5918,
+      "step": 1174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934733638230161e-06,
+      "loss": 0.6267,
+      "step": 1175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934556068818543e-06,
+      "loss": 0.5588,
+      "step": 1176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934378259770708e-06,
+      "loss": 1.0092,
+      "step": 1177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934200211095288e-06,
+      "loss": 0.8186,
+      "step": 1178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.934021922800931e-06,
+      "loss": 0.8727,
+      "step": 1179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933843394896295e-06,
+      "loss": 1.0618,
+      "step": 1180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93366462739005e-06,
+      "loss": 0.7807,
+      "step": 1181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933485620290875e-06,
+      "loss": 0.7461,
+      "step": 1182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.933306373607468e-06,
+      "loss": 0.7335,
+      "step": 1183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93312688734853e-06,
+      "loss": 0.6926,
+      "step": 1184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932947161522779e-06,
+      "loss": 0.7632,
+      "step": 1185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932767196138941e-06,
+      "loss": 0.6778,
+      "step": 1186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932586991205758e-06,
+      "loss": 0.7215,
+      "step": 1187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932406546731981e-06,
+      "loss": 0.8416,
+      "step": 1188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93222586272637e-06,
+      "loss": 1.2677,
+      "step": 1189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.932044939197702e-06,
+      "loss": 0.8013,
+      "step": 1190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931863776154766e-06,
+      "loss": 0.9362,
+      "step": 1191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931682373606354e-06,
+      "loss": 0.79,
+      "step": 1192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931500731561279e-06,
+      "loss": 0.8983,
+      "step": 1193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.93131885002836e-06,
+      "loss": 0.8119,
+      "step": 1194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.931136729016432e-06,
+      "loss": 0.845,
+      "step": 1195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930954368534338e-06,
+      "loss": 0.7024,
+      "step": 1196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930771768590934e-06,
+      "loss": 0.8395,
+      "step": 1197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930588929195087e-06,
+      "loss": 0.7988,
+      "step": 1198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930405850355677e-06,
+      "loss": 0.9255,
+      "step": 1199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930222532081597e-06,
+      "loss": 0.7134,
+      "step": 1200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.930038974381744e-06,
+      "loss": 0.8739,
+      "step": 1201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929855177265037e-06,
+      "loss": 1.0463,
+      "step": 1202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9296711407404e-06,
+      "loss": 0.8927,
+      "step": 1203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92948686481677e-06,
+      "loss": 1.0424,
+      "step": 1204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929302349503098e-06,
+      "loss": 0.6025,
+      "step": 1205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.929117594808341e-06,
+      "loss": 0.8313,
+      "step": 1206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928932600741474e-06,
+      "loss": 0.8489,
+      "step": 1207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92874736731148e-06,
+      "loss": 0.5644,
+      "step": 1208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928561894527354e-06,
+      "loss": 0.7937,
+      "step": 1209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928376182398104e-06,
+      "loss": 0.878,
+      "step": 1210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928190230932745e-06,
+      "loss": 0.9616,
+      "step": 1211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.928004040140315e-06,
+      "loss": 0.6883,
+      "step": 1212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92781761002985e-06,
+      "loss": 0.8137,
+      "step": 1213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927630940610405e-06,
+      "loss": 0.6985,
+      "step": 1214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927444031891044e-06,
+      "loss": 1.0319,
+      "step": 1215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927256883880848e-06,
+      "loss": 0.9193,
+      "step": 1216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.927069496588902e-06,
+      "loss": 0.6618,
+      "step": 1217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926881870024306e-06,
+      "loss": 1.0353,
+      "step": 1218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926694004196172e-06,
+      "loss": 0.9377,
+      "step": 1219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926505899113623e-06,
+      "loss": 0.599,
+      "step": 1220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926317554785796e-06,
+      "loss": 0.8359,
+      "step": 1221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.926128971221836e-06,
+      "loss": 0.8736,
+      "step": 1222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.9259401484309e-06,
+      "loss": 0.6876,
+      "step": 1223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92575108642216e-06,
+      "loss": 0.8773,
+      "step": 1224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925561785204797e-06,
+      "loss": 0.8233,
+      "step": 1225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925372244788004e-06,
+      "loss": 0.9906,
+      "step": 1226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.925182465180983e-06,
+      "loss": 0.6834,
+      "step": 1227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924992446392954e-06,
+      "loss": 0.6929,
+      "step": 1228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92480218843314e-06,
+      "loss": 0.7417,
+      "step": 1229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924611691310788e-06,
+      "loss": 0.8692,
+      "step": 1230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924420955035143e-06,
+      "loss": 1.0428,
+      "step": 1231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924229979615469e-06,
+      "loss": 0.941,
+      "step": 1232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924038765061042e-06,
+      "loss": 0.8537,
+      "step": 1233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923847311381144e-06,
+      "loss": 1.026,
+      "step": 1234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923655618585078e-06,
+      "loss": 1.0497,
+      "step": 1235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92346368668215e-06,
+      "loss": 0.861,
+      "step": 1236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92327151568168e-06,
+      "loss": 0.8958,
+      "step": 1237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.923079105593003e-06,
+      "loss": 0.5782,
+      "step": 1238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92288645642546e-06,
+      "loss": 0.9092,
+      "step": 1239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922693568188408e-06,
+      "loss": 0.9412,
+      "step": 1240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922500440891217e-06,
+      "loss": 0.7434,
+      "step": 1241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92230707454326e-06,
+      "loss": 0.854,
+      "step": 1242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922113469153933e-06,
+      "loss": 0.9188,
+      "step": 1243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921919624732634e-06,
+      "loss": 0.737,
+      "step": 1244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921725541288779e-06,
+      "loss": 0.8823,
+      "step": 1245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921531218831793e-06,
+      "loss": 1.0138,
+      "step": 1246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.92133665737111e-06,
+      "loss": 0.663,
+      "step": 1247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921141856916183e-06,
+      "loss": 0.6462,
+      "step": 1248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.920946817476469e-06,
+      "loss": 0.847,
+      "step": 1249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92075153906144e-06,
+      "loss": 0.7486,
+      "step": 1250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92055602168058e-06,
+      "loss": 0.9885,
+      "step": 1251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.920360265343386e-06,
+      "loss": 0.9004,
+      "step": 1252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.92016427005936e-06,
+      "loss": 0.9709,
+      "step": 1253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919968035838024e-06,
+      "loss": 0.8071,
+      "step": 1254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919771562688905e-06,
+      "loss": 1.0899,
+      "step": 1255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919574850621544e-06,
+      "loss": 0.8821,
+      "step": 1256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919377899645497e-06,
+      "loss": 0.929,
+      "step": 1257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.919180709770327e-06,
+      "loss": 0.9237,
+      "step": 1258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918983281005607e-06,
+      "loss": 0.6323,
+      "step": 1259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918785613360931e-06,
+      "loss": 0.655,
+      "step": 1260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918587706845892e-06,
+      "loss": 0.8533,
+      "step": 1261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918389561470105e-06,
+      "loss": 0.9484,
+      "step": 1262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.918191177243192e-06,
+      "loss": 0.7075,
+      "step": 1263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917992554174786e-06,
+      "loss": 0.5748,
+      "step": 1264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917793692274533e-06,
+      "loss": 0.8031,
+      "step": 1265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91759459155209e-06,
+      "loss": 1.1287,
+      "step": 1266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917395252017126e-06,
+      "loss": 0.8946,
+      "step": 1267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.917195673679323e-06,
+      "loss": 0.8173,
+      "step": 1268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916995856548371e-06,
+      "loss": 1.0591,
+      "step": 1269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916795800633974e-06,
+      "loss": 0.9199,
+      "step": 1270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916595505945848e-06,
+      "loss": 0.536,
+      "step": 1271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916394972493722e-06,
+      "loss": 0.6357,
+      "step": 1272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.916194200287329e-06,
+      "loss": 0.6588,
+      "step": 1273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915993189336423e-06,
+      "loss": 0.7631,
+      "step": 1274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915791939650767e-06,
+      "loss": 0.8786,
+      "step": 1275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915590451240131e-06,
+      "loss": 0.8977,
+      "step": 1276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915388724114301e-06,
+      "loss": 0.6734,
+      "step": 1277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.915186758283072e-06,
+      "loss": 0.546,
+      "step": 1278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914984553756257e-06,
+      "loss": 0.7614,
+      "step": 1279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91478211054367e-06,
+      "loss": 0.6044,
+      "step": 1280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914579428655144e-06,
+      "loss": 0.9366,
+      "step": 1281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914376508100524e-06,
+      "loss": 0.734,
+      "step": 1282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914173348889663e-06,
+      "loss": 1.0406,
+      "step": 1283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913969951032426e-06,
+      "loss": 0.5803,
+      "step": 1284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913766314538689e-06,
+      "loss": 0.7999,
+      "step": 1285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913562439418345e-06,
+      "loss": 0.8616,
+      "step": 1286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913358325681292e-06,
+      "loss": 0.8566,
+      "step": 1287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.913153973337445e-06,
+      "loss": 0.7733,
+      "step": 1288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912949382396728e-06,
+      "loss": 1.0935,
+      "step": 1289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912744552869072e-06,
+      "loss": 1.0416,
+      "step": 1290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912539484764429e-06,
+      "loss": 0.9843,
+      "step": 1291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912334178092754e-06,
+      "loss": 0.7878,
+      "step": 1292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91212863286402e-06,
+      "loss": 1.1566,
+      "step": 1293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91192284908821e-06,
+      "loss": 0.9487,
+      "step": 1294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911716826775313e-06,
+      "loss": 0.7518,
+      "step": 1295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911510565935335e-06,
+      "loss": 0.8058,
+      "step": 1296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911304066578298e-06,
+      "loss": 0.7855,
+      "step": 1297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.911097328714223e-06,
+      "loss": 0.9528,
+      "step": 1298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910890352353154e-06,
+      "loss": 1.1755,
+      "step": 1299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910683137505142e-06,
+      "loss": 1.044,
+      "step": 1300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91047568418025e-06,
+      "loss": 0.7008,
+      "step": 1301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91026799238855e-06,
+      "loss": 0.7528,
+      "step": 1302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.91006006214013e-06,
+      "loss": 0.8762,
+      "step": 1303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909851893445088e-06,
+      "loss": 0.9458,
+      "step": 1304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909643486313533e-06,
+      "loss": 0.6604,
+      "step": 1305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909434840755588e-06,
+      "loss": 1.1497,
+      "step": 1306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909225956781382e-06,
+      "loss": 0.7207,
+      "step": 1307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.909016834401058e-06,
+      "loss": 0.675,
+      "step": 1308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908807473624776e-06,
+      "loss": 1.042,
+      "step": 1309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9085978744627e-06,
+      "loss": 0.8551,
+      "step": 1310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908388036925011e-06,
+      "loss": 0.9712,
+      "step": 1311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908177961021896e-06,
+      "loss": 0.7517,
+      "step": 1312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90796764676356e-06,
+      "loss": 0.9759,
+      "step": 1313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907757094160217e-06,
+      "loss": 0.5987,
+      "step": 1314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907546303222088e-06,
+      "loss": 0.7593,
+      "step": 1315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907335273959413e-06,
+      "loss": 0.8807,
+      "step": 1316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907124006382438e-06,
+      "loss": 1.1629,
+      "step": 1317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906912500501427e-06,
+      "loss": 0.9671,
+      "step": 1318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906700756326647e-06,
+      "loss": 0.7937,
+      "step": 1319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90648877386838e-06,
+      "loss": 1.0419,
+      "step": 1320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906276553136924e-06,
+      "loss": 0.9864,
+      "step": 1321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.906064094142584e-06,
+      "loss": 1.0472,
+      "step": 1322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905851396895679e-06,
+      "loss": 0.8342,
+      "step": 1323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905638461406534e-06,
+      "loss": 1.0198,
+      "step": 1324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905425287685493e-06,
+      "loss": 0.8497,
+      "step": 1325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905211875742908e-06,
+      "loss": 0.5586,
+      "step": 1326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904998225589142e-06,
+      "loss": 0.7167,
+      "step": 1327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904784337234572e-06,
+      "loss": 0.7572,
+      "step": 1328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904570210689584e-06,
+      "loss": 0.7116,
+      "step": 1329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904355845964574e-06,
+      "loss": 1.0476,
+      "step": 1330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.904141243069958e-06,
+      "loss": 0.5988,
+      "step": 1331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903926402016153e-06,
+      "loss": 0.94,
+      "step": 1332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903711322813594e-06,
+      "loss": 0.7721,
+      "step": 1333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903496005472726e-06,
+      "loss": 0.8121,
+      "step": 1334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.903280450004004e-06,
+      "loss": 1.1958,
+      "step": 1335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.9030646564179e-06,
+      "loss": 0.6916,
+      "step": 1336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902848624724887e-06,
+      "loss": 0.6941,
+      "step": 1337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902632354935462e-06,
+      "loss": 0.8177,
+      "step": 1338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902415847060125e-06,
+      "loss": 0.4313,
+      "step": 1339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.902199101109392e-06,
+      "loss": 0.9845,
+      "step": 1340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901982117093786e-06,
+      "loss": 0.8674,
+      "step": 1341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901764895023846e-06,
+      "loss": 0.9091,
+      "step": 1342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90154743491012e-06,
+      "loss": 0.6487,
+      "step": 1343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901329736763171e-06,
+      "loss": 0.8106,
+      "step": 1344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.901111800593569e-06,
+      "loss": 1.2027,
+      "step": 1345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900893626411898e-06,
+      "loss": 0.8861,
+      "step": 1346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900675214228752e-06,
+      "loss": 0.7492,
+      "step": 1347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.90045656405474e-06,
+      "loss": 0.7995,
+      "step": 1348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900237675900479e-06,
+      "loss": 0.7019,
+      "step": 1349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.900018549776598e-06,
+      "loss": 0.6972,
+      "step": 1350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89979918569374e-06,
+      "loss": 0.7851,
+      "step": 1351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899579583662557e-06,
+      "loss": 0.7723,
+      "step": 1352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899359743693715e-06,
+      "loss": 0.827,
+      "step": 1353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.899139665797888e-06,
+      "loss": 0.6377,
+      "step": 1354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898919349985765e-06,
+      "loss": 0.7577,
+      "step": 1355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898698796268045e-06,
+      "loss": 0.7592,
+      "step": 1356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898478004655437e-06,
+      "loss": 0.9238,
+      "step": 1357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898256975158664e-06,
+      "loss": 0.7968,
+      "step": 1358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.898035707788462e-06,
+      "loss": 0.955,
+      "step": 1359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897814202555577e-06,
+      "loss": 0.9149,
+      "step": 1360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89759245947076e-06,
+      "loss": 0.9361,
+      "step": 1361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897370478544785e-06,
+      "loss": 0.6359,
+      "step": 1362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.897148259788429e-06,
+      "loss": 0.9583,
+      "step": 1363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896925803212487e-06,
+      "loss": 0.8129,
+      "step": 1364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896703108827758e-06,
+      "loss": 0.7813,
+      "step": 1365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89648017664506e-06,
+      "loss": 0.56,
+      "step": 1366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896257006675217e-06,
+      "loss": 1.0453,
+      "step": 1367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.896033598929069e-06,
+      "loss": 0.8787,
+      "step": 1368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895809953417464e-06,
+      "loss": 0.8617,
+      "step": 1369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895586070151263e-06,
+      "loss": 0.83,
+      "step": 1370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.895361949141336e-06,
+      "loss": 1.0208,
+      "step": 1371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89513759039857e-06,
+      "loss": 0.9647,
+      "step": 1372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894912993933861e-06,
+      "loss": 0.7944,
+      "step": 1373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894688159758114e-06,
+      "loss": 0.8138,
+      "step": 1374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894463087882249e-06,
+      "loss": 1.0175,
+      "step": 1375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894237778317195e-06,
+      "loss": 0.8368,
+      "step": 1376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.894012231073895e-06,
+      "loss": 1.0481,
+      "step": 1377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8937864461633e-06,
+      "loss": 0.724,
+      "step": 1378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893560423596378e-06,
+      "loss": 0.9015,
+      "step": 1379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893334163384102e-06,
+      "loss": 0.8872,
+      "step": 1380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.893107665537462e-06,
+      "loss": 0.9744,
+      "step": 1381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892880930067457e-06,
+      "loss": 0.9706,
+      "step": 1382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892653956985097e-06,
+      "loss": 0.8053,
+      "step": 1383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892426746301407e-06,
+      "loss": 0.8181,
+      "step": 1384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.892199298027416e-06,
+      "loss": 0.8666,
+      "step": 1385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891971612174176e-06,
+      "loss": 0.8158,
+      "step": 1386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891743688752738e-06,
+      "loss": 0.9765,
+      "step": 1387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891515527774176e-06,
+      "loss": 0.8229,
+      "step": 1388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.891287129249565e-06,
+      "loss": 0.6566,
+      "step": 1389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.89105849319e-06,
+      "loss": 1.079,
+      "step": 1390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890829619606584e-06,
+      "loss": 0.8399,
+      "step": 1391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890600508510431e-06,
+      "loss": 0.6804,
+      "step": 1392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890371159912668e-06,
+      "loss": 0.7599,
+      "step": 1393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.890141573824433e-06,
+      "loss": 0.8032,
+      "step": 1394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.889911750256873e-06,
+      "loss": 0.6801,
+      "step": 1395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.88968168922115e-06,
+      "loss": 0.9124,
+      "step": 1396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88945139072844e-06,
+      "loss": 0.7654,
+      "step": 1397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88922085478992e-06,
+      "loss": 0.844,
+      "step": 1398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888990081416792e-06,
+      "loss": 0.5321,
+      "step": 1399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88875907062026e-06,
+      "loss": 0.6922,
+      "step": 1400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888527822411543e-06,
+      "loss": 0.829,
+      "step": 1401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888296336801871e-06,
+      "loss": 0.8021,
+      "step": 1402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.888064613802485e-06,
+      "loss": 0.7166,
+      "step": 1403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88783265342464e-06,
+      "loss": 0.5991,
+      "step": 1404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.8876004556796e-06,
+      "loss": 0.7992,
+      "step": 1405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88736802057864e-06,
+      "loss": 0.8085,
+      "step": 1406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.887135348133048e-06,
+      "loss": 0.7295,
+      "step": 1407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886902438354125e-06,
+      "loss": 0.9169,
+      "step": 1408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886669291253178e-06,
+      "loss": 0.9429,
+      "step": 1409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886435906841534e-06,
+      "loss": 0.8517,
+      "step": 1410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.886202285130524e-06,
+      "loss": 0.9033,
+      "step": 1411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885968426131492e-06,
+      "loss": 0.9336,
+      "step": 1412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885734329855798e-06,
+      "loss": 1.1096,
+      "step": 1413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885499996314809e-06,
+      "loss": 1.2068,
+      "step": 1414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885265425519906e-06,
+      "loss": 0.7516,
+      "step": 1415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.885030617482478e-06,
+      "loss": 0.7307,
+      "step": 1416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.88479557221393e-06,
+      "loss": 0.7371,
+      "step": 1417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884560289725677e-06,
+      "loss": 0.8061,
+      "step": 1418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884324770029143e-06,
+      "loss": 0.6568,
+      "step": 1419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.884089013135766e-06,
+      "loss": 0.7454,
+      "step": 1420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883853019056995e-06,
+      "loss": 0.8092,
+      "step": 1421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883616787804292e-06,
+      "loss": 0.7172,
+      "step": 1422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883380319389129e-06,
+      "loss": 0.54,
+      "step": 1423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.883143613822986e-06,
+      "loss": 0.8886,
+      "step": 1424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882906671117363e-06,
+      "loss": 0.6219,
+      "step": 1425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882669491283763e-06,
+      "loss": 0.6323,
+      "step": 1426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882432074333705e-06,
+      "loss": 0.9396,
+      "step": 1427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.882194420278721e-06,
+      "loss": 0.9965,
+      "step": 1428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881956529130348e-06,
+      "loss": 0.868,
+      "step": 1429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881718400900143e-06,
+      "loss": 0.8097,
+      "step": 1430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881480035599667e-06,
+      "loss": 0.7971,
+      "step": 1431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881241433240498e-06,
+      "loss": 0.6642,
+      "step": 1432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.881002593834221e-06,
+      "loss": 0.8718,
+      "step": 1433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880763517392436e-06,
+      "loss": 0.9562,
+      "step": 1434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880524203926753e-06,
+      "loss": 0.962,
+      "step": 1435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880284653448796e-06,
+      "loss": 0.9912,
+      "step": 1436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.880044865970192e-06,
+      "loss": 0.5986,
+      "step": 1437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879804841502594e-06,
+      "loss": 0.9547,
+      "step": 1438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879564580057652e-06,
+      "loss": 0.7017,
+      "step": 1439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879324081647035e-06,
+      "loss": 0.9695,
+      "step": 1440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.879083346282424e-06,
+      "loss": 0.8208,
+      "step": 1441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87884237397551e-06,
+      "loss": 0.8397,
+      "step": 1442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878601164737993e-06,
+      "loss": 0.8371,
+      "step": 1443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878359718581588e-06,
+      "loss": 0.6956,
+      "step": 1444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.878118035518021e-06,
+      "loss": 0.9694,
+      "step": 1445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877876115559028e-06,
+      "loss": 0.7275,
+      "step": 1446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877633958716358e-06,
+      "loss": 0.6112,
+      "step": 1447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.87739156500177e-06,
+      "loss": 0.9038,
+      "step": 1448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.877148934427037e-06,
+      "loss": 0.457,
+      "step": 1449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876906067003938e-06,
+      "loss": 0.6759,
+      "step": 1450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876662962744273e-06,
+      "loss": 0.8209,
+      "step": 1451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876419621659843e-06,
+      "loss": 0.8761,
+      "step": 1452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.876176043762467e-06,
+      "loss": 1.0798,
+      "step": 1453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875932229063974e-06,
+      "loss": 0.8007,
+      "step": 1454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875688177576203e-06,
+      "loss": 0.9161,
+      "step": 1455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875443889311009e-06,
+      "loss": 1.0259,
+      "step": 1456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.875199364280255e-06,
+      "loss": 0.9205,
+      "step": 1457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874954602495811e-06,
+      "loss": 0.7104,
+      "step": 1458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874709603969567e-06,
+      "loss": 0.8995,
+      "step": 1459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874464368713423e-06,
+      "loss": 0.7893,
+      "step": 1460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.874218896739285e-06,
+      "loss": 0.736,
+      "step": 1461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873973188059073e-06,
+      "loss": 0.6554,
+      "step": 1462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873727242684723e-06,
+      "loss": 0.9516,
+      "step": 1463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873481060628175e-06,
+      "loss": 0.8498,
+      "step": 1464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.873234641901387e-06,
+      "loss": 0.7389,
+      "step": 1465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872987986516325e-06,
+      "loss": 1.1149,
+      "step": 1466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872741094484965e-06,
+      "loss": 0.7903,
+      "step": 1467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872493965819303e-06,
+      "loss": 0.8348,
+      "step": 1468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.872246600531333e-06,
+      "loss": 0.7916,
+      "step": 1469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871998998633072e-06,
+      "loss": 0.8442,
+      "step": 1470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871751160136544e-06,
+      "loss": 1.0155,
+      "step": 1471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871503085053782e-06,
+      "loss": 0.7861,
+      "step": 1472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871254773396837e-06,
+      "loss": 0.865,
+      "step": 1473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.871006225177767e-06,
+      "loss": 0.8566,
+      "step": 1474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870757440408638e-06,
+      "loss": 0.9117,
+      "step": 1475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870508419101536e-06,
+      "loss": 1.0782,
+      "step": 1476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870259161268553e-06,
+      "loss": 0.5062,
+      "step": 1477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.870009666921796e-06,
+      "loss": 0.8945,
+      "step": 1478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869759936073377e-06,
+      "loss": 0.5835,
+      "step": 1479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869509968735426e-06,
+      "loss": 0.8232,
+      "step": 1480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869259764920081e-06,
+      "loss": 0.6941,
+      "step": 1481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.869009324639495e-06,
+      "loss": 1.0919,
+      "step": 1482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868758647905828e-06,
+      "loss": 0.8191,
+      "step": 1483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868507734731252e-06,
+      "loss": 0.9317,
+      "step": 1484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868256585127956e-06,
+      "loss": 0.803,
+      "step": 1485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.868005199108133e-06,
+      "loss": 0.5518,
+      "step": 1486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867753576683993e-06,
+      "loss": 0.842,
+      "step": 1487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867501717867757e-06,
+      "loss": 0.8325,
+      "step": 1488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.867249622671653e-06,
+      "loss": 0.7312,
+      "step": 1489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866997291107925e-06,
+      "loss": 1.0498,
+      "step": 1490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866744723188826e-06,
+      "loss": 0.7292,
+      "step": 1491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866491918926622e-06,
+      "loss": 0.9523,
+      "step": 1492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.866238878333589e-06,
+      "loss": 0.7334,
+      "step": 1493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865985601422018e-06,
+      "loss": 0.8362,
+      "step": 1494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865732088204206e-06,
+      "loss": 0.7347,
+      "step": 1495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.865478338692467e-06,
+      "loss": 1.0892,
+      "step": 1496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86522435289912e-06,
+      "loss": 0.9865,
+      "step": 1497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864970130836501e-06,
+      "loss": 0.8071,
+      "step": 1498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864715672516958e-06,
+      "loss": 0.9619,
+      "step": 1499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864460977952847e-06,
+      "loss": 0.706,
+      "step": 1500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.864206047156536e-06,
+      "loss": 0.9824,
+      "step": 1501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863950880140405e-06,
+      "loss": 0.646,
+      "step": 1502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863695476916846e-06,
+      "loss": 0.8706,
+      "step": 1503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863439837498261e-06,
+      "loss": 1.1034,
+      "step": 1504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.863183961897067e-06,
+      "loss": 0.9724,
+      "step": 1505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862927850125689e-06,
+      "loss": 0.597,
+      "step": 1506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862671502196563e-06,
+      "loss": 0.86,
+      "step": 1507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.862414918122141e-06,
+      "loss": 1.0535,
+      "step": 1508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.86215809791488e-06,
+      "loss": 0.9823,
+      "step": 1509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861901041587253e-06,
+      "loss": 0.7604,
+      "step": 1510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861643749151748e-06,
+      "loss": 0.6316,
+      "step": 1511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861386220620853e-06,
+      "loss": 1.0414,
+      "step": 1512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.861128456007076e-06,
+      "loss": 0.5838,
+      "step": 1513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860870455322939e-06,
+      "loss": 0.9099,
+      "step": 1514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860612218580966e-06,
+      "loss": 0.7335,
+      "step": 1515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860353745793702e-06,
+      "loss": 0.9423,
+      "step": 1516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.860095036973697e-06,
+      "loss": 0.906,
+      "step": 1517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859836092133512e-06,
+      "loss": 0.709,
+      "step": 1518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859576911285728e-06,
+      "loss": 0.8972,
+      "step": 1519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859317494442927e-06,
+      "loss": 0.7122,
+      "step": 1520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.859057841617709e-06,
+      "loss": 0.9809,
+      "step": 1521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858797952822683e-06,
+      "loss": 1.0717,
+      "step": 1522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85853782807047e-06,
+      "loss": 0.7228,
+      "step": 1523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858277467373702e-06,
+      "loss": 0.7335,
+      "step": 1524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.858016870745025e-06,
+      "loss": 0.8653,
+      "step": 1525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85775603819709e-06,
+      "loss": 0.7268,
+      "step": 1526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857494969742568e-06,
+      "loss": 0.996,
+      "step": 1527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.857233665394134e-06,
+      "loss": 0.9037,
+      "step": 1528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85697212516448e-06,
+      "loss": 0.9698,
+      "step": 1529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856710349066307e-06,
+      "loss": 1.0955,
+      "step": 1530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856448337112327e-06,
+      "loss": 0.8512,
+      "step": 1531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.856186089315264e-06,
+      "loss": 0.7141,
+      "step": 1532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855923605687854e-06,
+      "loss": 0.6733,
+      "step": 1533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855660886242843e-06,
+      "loss": 0.7378,
+      "step": 1534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85539793099299e-06,
+      "loss": 0.7111,
+      "step": 1535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.855134739951064e-06,
+      "loss": 0.7747,
+      "step": 1536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854871313129847e-06,
+      "loss": 0.8696,
+      "step": 1537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854607650542133e-06,
+      "loss": 0.8211,
+      "step": 1538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854343752200725e-06,
+      "loss": 1.0595,
+      "step": 1539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.854079618118438e-06,
+      "loss": 0.8444,
+      "step": 1540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853815248308101e-06,
+      "loss": 0.8036,
+      "step": 1541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.85355064278255e-06,
+      "loss": 0.67,
+      "step": 1542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.853285801554637e-06,
+      "loss": 1.0718,
+      "step": 1543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.853020724637223e-06,
+      "loss": 1.0944,
+      "step": 1544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85275541204318e-06,
+      "loss": 0.9024,
+      "step": 1545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852489863785395e-06,
+      "loss": 0.8101,
+      "step": 1546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.852224079876761e-06,
+      "loss": 0.7336,
+      "step": 1547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851958060330186e-06,
+      "loss": 0.9019,
+      "step": 1548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851691805158589e-06,
+      "loss": 0.5997,
+      "step": 1549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851425314374898e-06,
+      "loss": 1.0487,
+      "step": 1550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.851158587992059e-06,
+      "loss": 0.6457,
+      "step": 1551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850891626023021e-06,
+      "loss": 0.6386,
+      "step": 1552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.85062442848075e-06,
+      "loss": 0.9152,
+      "step": 1553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850356995378223e-06,
+      "loss": 0.5706,
+      "step": 1554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.850089326728425e-06,
+      "loss": 0.7749,
+      "step": 1555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849821422544357e-06,
+      "loss": 1.0384,
+      "step": 1556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849553282839025e-06,
+      "loss": 1.0293,
+      "step": 1557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849284907625458e-06,
+      "loss": 0.721,
+      "step": 1558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.849016296916682e-06,
+      "loss": 0.7882,
+      "step": 1559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848747450725747e-06,
+      "loss": 0.8099,
+      "step": 1560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848478369065703e-06,
+      "loss": 0.8029,
+      "step": 1561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.848209051949621e-06,
+      "loss": 0.711,
+      "step": 1562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847939499390581e-06,
+      "loss": 0.7245,
+      "step": 1563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84766971140167e-06,
+      "loss": 0.935,
+      "step": 1564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847399687995993e-06,
+      "loss": 0.9809,
+      "step": 1565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.847129429186662e-06,
+      "loss": 0.825,
+      "step": 1566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846858934986798e-06,
+      "loss": 0.8217,
+      "step": 1567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846588205409542e-06,
+      "loss": 1.042,
+      "step": 1568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84631724046804e-06,
+      "loss": 0.9674,
+      "step": 1569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.846046040175448e-06,
+      "loss": 0.9332,
+      "step": 1570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84577460454494e-06,
+      "loss": 0.7706,
+      "step": 1571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.845502933589697e-06,
+      "loss": 0.8333,
+      "step": 1572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84523102732291e-06,
+      "loss": 0.8569,
+      "step": 1573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844958885757786e-06,
+      "loss": 0.8235,
+      "step": 1574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844686508907538e-06,
+      "loss": 1.0181,
+      "step": 1575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844413896785395e-06,
+      "loss": 0.8643,
+      "step": 1576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.844141049404598e-06,
+      "loss": 1.0036,
+      "step": 1577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843867966778394e-06,
+      "loss": 0.8,
+      "step": 1578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843594648920047e-06,
+      "loss": 0.5674,
+      "step": 1579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843321095842829e-06,
+      "loss": 0.9534,
+      "step": 1580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.843047307560024e-06,
+      "loss": 0.8983,
+      "step": 1581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842773284084928e-06,
+      "loss": 1.0454,
+      "step": 1582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84249902543085e-06,
+      "loss": 1.028,
+      "step": 1583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.842224531611106e-06,
+      "loss": 0.8575,
+      "step": 1584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841949802639031e-06,
+      "loss": 0.6909,
+      "step": 1585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841674838527962e-06,
+      "loss": 0.4943,
+      "step": 1586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.841399639291252e-06,
+      "loss": 0.9226,
+      "step": 1587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.84112420494227e-06,
+      "loss": 0.8297,
+      "step": 1588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840848535494385e-06,
+      "loss": 0.9232,
+      "step": 1589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840572630960992e-06,
+      "loss": 0.7672,
+      "step": 1590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840296491355485e-06,
+      "loss": 0.839,
+      "step": 1591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.840020116691274e-06,
+      "loss": 0.8266,
+      "step": 1592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839743506981783e-06,
+      "loss": 0.8122,
+      "step": 1593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839466662240443e-06,
+      "loss": 0.7827,
+      "step": 1594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.839189582480697e-06,
+      "loss": 0.987,
+      "step": 1595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838912267716006e-06,
+      "loss": 0.6064,
+      "step": 1596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838634717959831e-06,
+      "loss": 0.7428,
+      "step": 1597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838356933225656e-06,
+      "loss": 1.0867,
+      "step": 1598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.838078913526968e-06,
+      "loss": 0.858,
+      "step": 1599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83780065887727e-06,
+      "loss": 0.5934,
+      "step": 1600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.837522169290072e-06,
+      "loss": 0.5301,
+      "step": 1601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.8372434447789e-06,
+      "loss": 0.6662,
+      "step": 1602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836964485357291e-06,
+      "loss": 0.9465,
+      "step": 1603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83668529103879e-06,
+      "loss": 0.8109,
+      "step": 1604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836405861836957e-06,
+      "loss": 0.8168,
+      "step": 1605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.836126197765361e-06,
+      "loss": 0.6545,
+      "step": 1606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835846298837583e-06,
+      "loss": 0.5649,
+      "step": 1607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835566165067219e-06,
+      "loss": 0.8155,
+      "step": 1608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835285796467867e-06,
+      "loss": 0.8559,
+      "step": 1609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.835005193053146e-06,
+      "loss": 0.7702,
+      "step": 1610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834724354836684e-06,
+      "loss": 0.9077,
+      "step": 1611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834443281832118e-06,
+      "loss": 0.8364,
+      "step": 1612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.834161974053098e-06,
+      "loss": 0.97,
+      "step": 1613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833880431513284e-06,
+      "loss": 0.7112,
+      "step": 1614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83359865422635e-06,
+      "loss": 0.6344,
+      "step": 1615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83331664220598e-06,
+      "loss": 0.8362,
+      "step": 1616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.833034395465866e-06,
+      "loss": 1.136,
+      "step": 1617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83275191401972e-06,
+      "loss": 0.7374,
+      "step": 1618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832469197881255e-06,
+      "loss": 0.9086,
+      "step": 1619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.832186247064205e-06,
+      "loss": 0.5564,
+      "step": 1620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831903061582309e-06,
+      "loss": 0.8334,
+      "step": 1621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831619641449318e-06,
+      "loss": 0.5092,
+      "step": 1622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.831335986678998e-06,
+      "loss": 0.806,
+      "step": 1623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.83105209728512e-06,
+      "loss": 0.8792,
+      "step": 1624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830767973281477e-06,
+      "loss": 1.0307,
+      "step": 1625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830483614681863e-06,
+      "loss": 0.8615,
+      "step": 1626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.830199021500085e-06,
+      "loss": 0.7433,
+      "step": 1627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829914193749968e-06,
+      "loss": 0.7294,
+      "step": 1628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829629131445342e-06,
+      "loss": 1.0593,
+      "step": 1629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.829343834600052e-06,
+      "loss": 0.9085,
+      "step": 1630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82905830322795e-06,
+      "loss": 0.9676,
+      "step": 1631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828772537342904e-06,
+      "loss": 0.8876,
+      "step": 1632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82848653695879e-06,
+      "loss": 0.6924,
+      "step": 1633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.828200302089502e-06,
+      "loss": 0.7758,
+      "step": 1634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827913832748933e-06,
+      "loss": 0.9006,
+      "step": 1635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827627128951001e-06,
+      "loss": 0.689,
+      "step": 1636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827340190709625e-06,
+      "loss": 0.9191,
+      "step": 1637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.827053018038743e-06,
+      "loss": 0.8458,
+      "step": 1638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826765610952298e-06,
+      "loss": 0.73,
+      "step": 1639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82647796946425e-06,
+      "loss": 0.4518,
+      "step": 1640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.826190093588564e-06,
+      "loss": 0.6951,
+      "step": 1641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825901983339223e-06,
+      "loss": 0.7001,
+      "step": 1642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825613638730219e-06,
+      "loss": 0.9062,
+      "step": 1643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825325059775553e-06,
+      "loss": 0.9064,
+      "step": 1644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.825036246489239e-06,
+      "loss": 0.6323,
+      "step": 1645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824747198885306e-06,
+      "loss": 0.9559,
+      "step": 1646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.824457916977785e-06,
+      "loss": 0.9275,
+      "step": 1647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82416840078073e-06,
+      "loss": 0.7431,
+      "step": 1648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823878650308196e-06,
+      "loss": 0.9544,
+      "step": 1649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823588665574259e-06,
+      "loss": 0.7601,
+      "step": 1650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823298446592998e-06,
+      "loss": 0.7786,
+      "step": 1651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.823007993378509e-06,
+      "loss": 0.6866,
+      "step": 1652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822717305944895e-06,
+      "loss": 0.6809,
+      "step": 1653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822426384306274e-06,
+      "loss": 0.6683,
+      "step": 1654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.822135228476774e-06,
+      "loss": 0.8672,
+      "step": 1655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821843838470536e-06,
+      "loss": 0.6753,
+      "step": 1656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.821552214301705e-06,
+      "loss": 0.9829,
+      "step": 1657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.82126035598445e-06,
+      "loss": 0.9219,
+      "step": 1658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820968263532942e-06,
+      "loss": 1.0237,
+      "step": 1659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820675936961364e-06,
+      "loss": 0.9055,
+      "step": 1660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820383376283914e-06,
+      "loss": 0.7923,
+      "step": 1661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.820090581514799e-06,
+      "loss": 0.8492,
+      "step": 1662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819797552668238e-06,
+      "loss": 0.9602,
+      "step": 1663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81950428975846e-06,
+      "loss": 0.7721,
+      "step": 1664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.819210792799711e-06,
+      "loss": 0.7885,
+      "step": 1665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81891706180624e-06,
+      "loss": 0.6724,
+      "step": 1666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818623096792312e-06,
+      "loss": 0.7474,
+      "step": 1667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818328897772205e-06,
+      "loss": 0.727,
+      "step": 1668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.818034464760201e-06,
+      "loss": 0.7057,
+      "step": 1669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817739797770606e-06,
+      "loss": 0.8905,
+      "step": 1670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817444896817724e-06,
+      "loss": 0.6207,
+      "step": 1671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.817149761915879e-06,
+      "loss": 0.6645,
+      "step": 1672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816854393079402e-06,
+      "loss": 0.5058,
+      "step": 1673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816558790322638e-06,
+      "loss": 0.8868,
+      "step": 1674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.816262953659944e-06,
+      "loss": 0.7567,
+      "step": 1675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815966883105681e-06,
+      "loss": 0.8299,
+      "step": 1676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815670578674232e-06,
+      "loss": 0.5062,
+      "step": 1677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815374040379987e-06,
+      "loss": 0.9223,
+      "step": 1678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.815077268237342e-06,
+      "loss": 0.9367,
+      "step": 1679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814780262260714e-06,
+      "loss": 0.7847,
+      "step": 1680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814483022464524e-06,
+      "loss": 0.7767,
+      "step": 1681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.814185548863208e-06,
+      "loss": 0.9759,
+      "step": 1682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81388784147121e-06,
+      "loss": 0.9917,
+      "step": 1683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813589900302992e-06,
+      "loss": 0.6222,
+      "step": 1684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813291725373017e-06,
+      "loss": 1.0244,
+      "step": 1685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81299331669577e-06,
+      "loss": 0.7436,
+      "step": 1686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812694674285738e-06,
+      "loss": 1.0008,
+      "step": 1687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812395798157427e-06,
+      "loss": 0.6756,
+      "step": 1688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.812096688325354e-06,
+      "loss": 0.9117,
+      "step": 1689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.81179734480404e-06,
+      "loss": 1.2417,
+      "step": 1690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811497767608022e-06,
+      "loss": 0.8763,
+      "step": 1691
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.811197956751851e-06,
+      "loss": 0.7213,
+      "step": 1692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810897912250086e-06,
+      "loss": 0.8387,
+      "step": 1693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810597634117297e-06,
+      "loss": 1.0898,
+      "step": 1694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.810297122368067e-06,
+      "loss": 0.8159,
+      "step": 1695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809996377016989e-06,
+      "loss": 1.0434,
+      "step": 1696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80969539807867e-06,
+      "loss": 0.9937,
+      "step": 1697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.809394185567724e-06,
+      "loss": 0.5457,
+      "step": 1698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80909273949878e-06,
+      "loss": 0.7491,
+      "step": 1699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808791059886478e-06,
+      "loss": 0.8798,
+      "step": 1700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808489146745466e-06,
+      "loss": 0.7931,
+      "step": 1701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.808187000090407e-06,
+      "loss": 1.0017,
+      "step": 1702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807884619935975e-06,
+      "loss": 0.8403,
+      "step": 1703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.807582006296855e-06,
+      "loss": 0.7808,
+      "step": 1704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80727915918774e-06,
+      "loss": 1.0552,
+      "step": 1705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806976078623337e-06,
+      "loss": 0.8449,
+      "step": 1706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806672764618368e-06,
+      "loss": 0.9202,
+      "step": 1707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80636921718756e-06,
+      "loss": 0.658,
+      "step": 1708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.806065436345655e-06,
+      "loss": 0.8241,
+      "step": 1709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805761422107407e-06,
+      "loss": 0.8708,
+      "step": 1710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.805457174487576e-06,
+      "loss": 0.694,
+      "step": 1711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80515269350094e-06,
+      "loss": 0.8352,
+      "step": 1712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804847979162286e-06,
+      "loss": 0.7002,
+      "step": 1713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.80454303148641e-06,
+      "loss": 0.7281,
+      "step": 1714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.804237850488123e-06,
+      "loss": 0.9901,
+      "step": 1715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803932436182242e-06,
+      "loss": 0.7919,
+      "step": 1716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803626788583603e-06,
+      "loss": 0.8624,
+      "step": 1717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803320907707047e-06,
+      "loss": 0.8315,
+      "step": 1718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.803014793567429e-06,
+      "loss": 0.7015,
+      "step": 1719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802708446179611e-06,
+      "loss": 0.9455,
+      "step": 1720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802401865558477e-06,
+      "loss": 0.8675,
+      "step": 1721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.802095051718912e-06,
+      "loss": 0.8935,
+      "step": 1722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801788004675814e-06,
+      "loss": 0.8808,
+      "step": 1723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801480724444096e-06,
+      "loss": 0.7885,
+      "step": 1724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.801173211038682e-06,
+      "loss": 0.7681,
+      "step": 1725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800865464474503e-06,
+      "loss": 0.9834,
+      "step": 1726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800557484766505e-06,
+      "loss": 0.6873,
+      "step": 1727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.800249271929645e-06,
+      "loss": 0.8759,
+      "step": 1728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799940825978888e-06,
+      "loss": 0.6269,
+      "step": 1729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.799632146929216e-06,
+      "loss": 0.9915,
+      "step": 1730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79932323479562e-06,
+      "loss": 0.8078,
+      "step": 1731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7990140895931e-06,
+      "loss": 0.8713,
+      "step": 1732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.798704711336669e-06,
+      "loss": 0.813,
+      "step": 1733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79839510004135e-06,
+      "loss": 0.8316,
+      "step": 1734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79808525572218e-06,
+      "loss": 1.1043,
+      "step": 1735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797775178394207e-06,
+      "loss": 0.9942,
+      "step": 1736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797464868072489e-06,
+      "loss": 0.7319,
+      "step": 1737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.797154324772092e-06,
+      "loss": 1.0098,
+      "step": 1738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796843548508101e-06,
+      "loss": 0.723,
+      "step": 1739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796532539295607e-06,
+      "loss": 0.733,
+      "step": 1740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.796221297149716e-06,
+      "loss": 0.6864,
+      "step": 1741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795909822085537e-06,
+      "loss": 0.562,
+      "step": 1742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.795598114118199e-06,
+      "loss": 0.7045,
+      "step": 1743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79528617326284e-06,
+      "loss": 0.8708,
+      "step": 1744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794973999534609e-06,
+      "loss": 0.7412,
+      "step": 1745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794661592948665e-06,
+      "loss": 0.8887,
+      "step": 1746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794348953520181e-06,
+      "loss": 0.797,
+      "step": 1747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.794036081264337e-06,
+      "loss": 1.0368,
+      "step": 1748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79372297619633e-06,
+      "loss": 0.7039,
+      "step": 1749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793409638331364e-06,
+      "loss": 1.2106,
+      "step": 1750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.793096067684653e-06,
+      "loss": 1.1307,
+      "step": 1751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79278226427143e-06,
+      "loss": 0.7109,
+      "step": 1752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79246822810693e-06,
+      "loss": 0.5511,
+      "step": 1753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.792153959206405e-06,
+      "loss": 0.8536,
+      "step": 1754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791839457585118e-06,
+      "loss": 0.8245,
+      "step": 1755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791524723258338e-06,
+      "loss": 0.9464,
+      "step": 1756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.791209756241356e-06,
+      "loss": 1.018,
+      "step": 1757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.79089455654946e-06,
+      "loss": 0.7922,
+      "step": 1758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790579124197963e-06,
+      "loss": 0.6636,
+      "step": 1759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.790263459202182e-06,
+      "loss": 1.0902,
+      "step": 1760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789947561577445e-06,
+      "loss": 0.6871,
+      "step": 1761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.789631431339094e-06,
+      "loss": 0.7884,
+      "step": 1762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78931506850248e-06,
+      "loss": 0.8098,
+      "step": 1763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788998473082968e-06,
+      "loss": 0.8514,
+      "step": 1764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788681645095935e-06,
+      "loss": 0.6603,
+      "step": 1765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78836458455676e-06,
+      "loss": 0.5845,
+      "step": 1766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.788047291480846e-06,
+      "loss": 0.8759,
+      "step": 1767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787729765883602e-06,
+      "loss": 0.8628,
+      "step": 1768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787412007780445e-06,
+      "loss": 0.9737,
+      "step": 1769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.787094017186807e-06,
+      "loss": 0.7698,
+      "step": 1770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786775794118132e-06,
+      "loss": 0.7676,
+      "step": 1771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786457338589871e-06,
+      "loss": 1.1777,
+      "step": 1772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.786138650617494e-06,
+      "loss": 1.0153,
+      "step": 1773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785819730216474e-06,
+      "loss": 1.0251,
+      "step": 1774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785500577402299e-06,
+      "loss": 0.6996,
+      "step": 1775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.785181192190467e-06,
+      "loss": 0.8876,
+      "step": 1776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784861574596489e-06,
+      "loss": 0.7156,
+      "step": 1777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784541724635888e-06,
+      "loss": 0.8937,
+      "step": 1778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784221642324194e-06,
+      "loss": 0.9093,
+      "step": 1779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783901327676955e-06,
+      "loss": 0.5598,
+      "step": 1780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783580780709723e-06,
+      "loss": 1.0186,
+      "step": 1781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.783260001438066e-06,
+      "loss": 0.8255,
+      "step": 1782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782938989877562e-06,
+      "loss": 0.8119,
+      "step": 1783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.782617746043801e-06,
+      "loss": 0.7477,
+      "step": 1784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78229626995238e-06,
+      "loss": 0.8947,
+      "step": 1785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781974561618917e-06,
+      "loss": 0.7349,
+      "step": 1786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.78165262105903e-06,
+      "loss": 0.9143,
+      "step": 1787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781330448288355e-06,
+      "loss": 0.7262,
+      "step": 1788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.781008043322537e-06,
+      "loss": 0.692,
+      "step": 1789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780685406177233e-06,
+      "loss": 0.612,
+      "step": 1790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780362536868113e-06,
+      "loss": 0.8336,
+      "step": 1791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.780039435410855e-06,
+      "loss": 0.7866,
+      "step": 1792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779716101821149e-06,
+      "loss": 0.6387,
+      "step": 1793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779392536114698e-06,
+      "loss": 0.9994,
+      "step": 1794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.779068738307215e-06,
+      "loss": 0.9629,
+      "step": 1795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778744708414425e-06,
+      "loss": 1.1775,
+      "step": 1796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778420446452063e-06,
+      "loss": 0.7971,
+      "step": 1797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.778095952435879e-06,
+      "loss": 0.7388,
+      "step": 1798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777771226381628e-06,
+      "loss": 0.9275,
+      "step": 1799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777446268305079e-06,
+      "loss": 0.9199,
+      "step": 1800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.777121078222015e-06,
+      "loss": 0.9252,
+      "step": 1801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776795656148229e-06,
+      "loss": 0.8357,
+      "step": 1802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77647000209952e-06,
+      "loss": 0.8729,
+      "step": 1803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.776144116091708e-06,
+      "loss": 0.6,
+      "step": 1804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775817998140615e-06,
+      "loss": 0.9685,
+      "step": 1805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775491648262084e-06,
+      "loss": 0.9594,
+      "step": 1806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.775165066471954e-06,
+      "loss": 0.8771,
+      "step": 1807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774838252786094e-06,
+      "loss": 0.567,
+      "step": 1808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774511207220369e-06,
+      "loss": 0.924,
+      "step": 1809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.774183929790665e-06,
+      "loss": 0.4881,
+      "step": 1810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773856420512876e-06,
+      "loss": 0.8908,
+      "step": 1811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.773528679402901e-06,
+      "loss": 1.0318,
+      "step": 1812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77320070647666e-06,
+      "loss": 0.7875,
+      "step": 1813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772872501750083e-06,
+      "loss": 1.0187,
+      "step": 1814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772544065239105e-06,
+      "loss": 0.8418,
+      "step": 1815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.772215396959674e-06,
+      "loss": 0.725,
+      "step": 1816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771886496927756e-06,
+      "loss": 1.0381,
+      "step": 1817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.77155736515932e-06,
+      "loss": 0.8079,
+      "step": 1818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.771228001670351e-06,
+      "loss": 0.8186,
+      "step": 1819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770898406476843e-06,
+      "loss": 0.7949,
+      "step": 1820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770568579594803e-06,
+      "loss": 0.8208,
+      "step": 1821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.770238521040248e-06,
+      "loss": 1.0209,
+      "step": 1822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769908230829205e-06,
+      "loss": 0.9451,
+      "step": 1823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.769577708977716e-06,
+      "loss": 0.7303,
+      "step": 1824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76924695550183e-06,
+      "loss": 0.9143,
+      "step": 1825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768915970417614e-06,
+      "loss": 0.8622,
+      "step": 1826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768584753741134e-06,
+      "loss": 0.7379,
+      "step": 1827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.768253305488483e-06,
+      "loss": 0.6088,
+      "step": 1828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767921625675751e-06,
+      "loss": 0.7486,
+      "step": 1829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767589714319049e-06,
+      "loss": 0.7312,
+      "step": 1830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.767257571434492e-06,
+      "loss": 0.885,
+      "step": 1831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766925197038213e-06,
+      "loss": 0.9705,
+      "step": 1832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.766592591146353e-06,
+      "loss": 1.0891,
+      "step": 1833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.76625975377506e-06,
+      "loss": 0.6874,
+      "step": 1834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765926684940503e-06,
+      "loss": 0.9923,
+      "step": 1835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.765593384658855e-06,
+      "loss": 0.8904,
+      "step": 1836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.7652598529463e-06,
+      "loss": 0.7589,
+      "step": 1837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76492608981904e-06,
+      "loss": 0.8327,
+      "step": 1838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764592095293278e-06,
+      "loss": 0.7996,
+      "step": 1839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.764257869385235e-06,
+      "loss": 0.7659,
+      "step": 1840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763923412111146e-06,
+      "loss": 0.6928,
+      "step": 1841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763588723487245e-06,
+      "loss": 0.9513,
+      "step": 1842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.763253803529796e-06,
+      "loss": 0.7355,
+      "step": 1843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762918652255055e-06,
+      "loss": 0.9714,
+      "step": 1844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762583269679304e-06,
+      "loss": 1.0141,
+      "step": 1845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.762247655818827e-06,
+      "loss": 0.9561,
+      "step": 1846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.76191181068992e-06,
+      "loss": 0.9069,
+      "step": 1847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761575734308896e-06,
+      "loss": 0.8112,
+      "step": 1848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.761239426692077e-06,
+      "loss": 0.7017,
+      "step": 1849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760902887855794e-06,
+      "loss": 1.1308,
+      "step": 1850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760566117816387e-06,
+      "loss": 0.8878,
+      "step": 1851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.760229116590214e-06,
+      "loss": 0.7188,
+      "step": 1852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75989188419364e-06,
+      "loss": 0.7374,
+      "step": 1853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759554420643043e-06,
+      "loss": 0.9403,
+      "step": 1854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.759216725954807e-06,
+      "loss": 0.9337,
+      "step": 1855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758878800145337e-06,
+      "loss": 1.0517,
+      "step": 1856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.758540643231041e-06,
+      "loss": 0.6168,
+      "step": 1857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75820225522834e-06,
+      "loss": 0.9116,
+      "step": 1858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757863636153669e-06,
+      "loss": 0.7643,
+      "step": 1859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75752478602347e-06,
+      "loss": 1.0918,
+      "step": 1860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.757185704854199e-06,
+      "loss": 0.9043,
+      "step": 1861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756846392662326e-06,
+      "loss": 0.7633,
+      "step": 1862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756506849464327e-06,
+      "loss": 0.8788,
+      "step": 1863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.756167075276688e-06,
+      "loss": 0.9702,
+      "step": 1864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755827070115915e-06,
+      "loss": 0.8111,
+      "step": 1865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755486833998515e-06,
+      "loss": 0.7928,
+      "step": 1866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.755146366941014e-06,
+      "loss": 0.7645,
+      "step": 1867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754805668959944e-06,
+      "loss": 0.6834,
+      "step": 1868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754464740071852e-06,
+      "loss": 0.9704,
+      "step": 1869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75412358029329e-06,
+      "loss": 0.9788,
+      "step": 1870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753782189640833e-06,
+      "loss": 0.6818,
+      "step": 1871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753440568131056e-06,
+      "loss": 0.6878,
+      "step": 1872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.753098715780546e-06,
+      "loss": 0.754,
+      "step": 1873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.75275663260591e-06,
+      "loss": 0.6097,
+      "step": 1874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752414318623759e-06,
+      "loss": 0.8762,
+      "step": 1875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.752071773850715e-06,
+      "loss": 0.9519,
+      "step": 1876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751728998303413e-06,
+      "loss": 0.9747,
+      "step": 1877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751385991998501e-06,
+      "loss": 0.7667,
+      "step": 1878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.751042754952636e-06,
+      "loss": 0.9566,
+      "step": 1879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750699287182484e-06,
+      "loss": 0.6781,
+      "step": 1880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750355588704728e-06,
+      "loss": 0.6838,
+      "step": 1881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.750011659536057e-06,
+      "loss": 0.9906,
+      "step": 1882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749667499693174e-06,
+      "loss": 0.8128,
+      "step": 1883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.749323109192792e-06,
+      "loss": 0.9353,
+      "step": 1884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748978488051637e-06,
+      "loss": 0.6764,
+      "step": 1885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748633636286443e-06,
+      "loss": 1.0961,
+      "step": 1886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.748288553913958e-06,
+      "loss": 0.8103,
+      "step": 1887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747943240950939e-06,
+      "loss": 0.9164,
+      "step": 1888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.747597697414157e-06,
+      "loss": 0.7505,
+      "step": 1889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74725192332039e-06,
+      "loss": 0.9344,
+      "step": 1890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.746905918686436e-06,
+      "loss": 0.7571,
+      "step": 1891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74655968352909e-06,
+      "loss": 0.9854,
+      "step": 1892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74621321786517e-06,
+      "loss": 1.1534,
+      "step": 1893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745866521711502e-06,
+      "loss": 0.8137,
+      "step": 1894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745519595084922e-06,
+      "loss": 0.9215,
+      "step": 1895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.745172438002276e-06,
+      "loss": 0.8025,
+      "step": 1896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744825050480425e-06,
+      "loss": 0.5914,
+      "step": 1897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.74447743253624e-06,
+      "loss": 0.8293,
+      "step": 1898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.744129584186599e-06,
+      "loss": 0.6871,
+      "step": 1899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743781505448398e-06,
+      "loss": 0.7114,
+      "step": 1900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743433196338537e-06,
+      "loss": 0.6703,
+      "step": 1901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.743084656873933e-06,
+      "loss": 0.8597,
+      "step": 1902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742735887071515e-06,
+      "loss": 0.6946,
+      "step": 1903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742386886948215e-06,
+      "loss": 0.9575,
+      "step": 1904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.742037656520984e-06,
+      "loss": 0.6534,
+      "step": 1905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741688195806781e-06,
+      "loss": 0.9136,
+      "step": 1906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.741338504822577e-06,
+      "loss": 0.8743,
+      "step": 1907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740988583585356e-06,
+      "loss": 0.9557,
+      "step": 1908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740638432112108e-06,
+      "loss": 1.0307,
+      "step": 1909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.740288050419839e-06,
+      "loss": 0.8591,
+      "step": 1910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739937438525565e-06,
+      "loss": 0.793,
+      "step": 1911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739586596446311e-06,
+      "loss": 0.8566,
+      "step": 1912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.739235524199117e-06,
+      "loss": 1.0195,
+      "step": 1913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738884221801031e-06,
+      "loss": 0.6215,
+      "step": 1914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738532689269111e-06,
+      "loss": 0.5626,
+      "step": 1915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.738180926620434e-06,
+      "loss": 0.8778,
+      "step": 1916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737828933872076e-06,
+      "loss": 0.7109,
+      "step": 1917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737476711041136e-06,
+      "loss": 0.8663,
+      "step": 1918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.737124258144716e-06,
+      "loss": 1.0726,
+      "step": 1919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736771575199932e-06,
+      "loss": 0.875,
+      "step": 1920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.736418662223915e-06,
+      "loss": 0.7775,
+      "step": 1921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7360655192338e-06,
+      "loss": 0.48,
+      "step": 1922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735712146246737e-06,
+      "loss": 0.891,
+      "step": 1923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735358543279889e-06,
+      "loss": 0.9432,
+      "step": 1924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.735004710350424e-06,
+      "loss": 0.7028,
+      "step": 1925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73465064747553e-06,
+      "loss": 0.7943,
+      "step": 1926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.734296354672397e-06,
+      "loss": 0.6101,
+      "step": 1927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733941831958231e-06,
+      "loss": 1.1245,
+      "step": 1928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733587079350254e-06,
+      "loss": 0.8599,
+      "step": 1929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.733232096865686e-06,
+      "loss": 0.8556,
+      "step": 1930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.73287688452177e-06,
+      "loss": 0.9311,
+      "step": 1931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732521442335758e-06,
+      "loss": 0.849,
+      "step": 1932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.732165770324906e-06,
+      "loss": 0.9809,
+      "step": 1933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731809868506491e-06,
+      "loss": 0.445,
+      "step": 1934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731453736897796e-06,
+      "loss": 0.6728,
+      "step": 1935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.731097375516113e-06,
+      "loss": 0.6997,
+      "step": 1936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730740784378755e-06,
+      "loss": 0.7214,
+      "step": 1937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730383963503028e-06,
+      "loss": 0.7965,
+      "step": 1938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.730026912906269e-06,
+      "loss": 1.0103,
+      "step": 1939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729669632605814e-06,
+      "loss": 0.797,
+      "step": 1940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.729312122619013e-06,
+      "loss": 0.8471,
+      "step": 1941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72895438296323e-06,
+      "loss": 0.7439,
+      "step": 1942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.728596413655837e-06,
+      "loss": 0.8343,
+      "step": 1943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72823821471422e-06,
+      "loss": 1.2427,
+      "step": 1944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727879786155767e-06,
+      "loss": 0.5425,
+      "step": 1945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727521127997892e-06,
+      "loss": 0.8505,
+      "step": 1946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.727162240258009e-06,
+      "loss": 0.5575,
+      "step": 1947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726803122953547e-06,
+      "loss": 0.7084,
+      "step": 1948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.726443776101947e-06,
+      "loss": 0.7837,
+      "step": 1949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72608419972066e-06,
+      "loss": 0.7308,
+      "step": 1950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725724393827145e-06,
+      "loss": 0.9642,
+      "step": 1951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725364358438878e-06,
+      "loss": 1.0301,
+      "step": 1952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.725004093573343e-06,
+      "loss": 0.8047,
+      "step": 1953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.724643599248035e-06,
+      "loss": 0.8059,
+      "step": 1954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.72428287548046e-06,
+      "loss": 0.6284,
+      "step": 1955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723921922288138e-06,
+      "loss": 0.6338,
+      "step": 1956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723560739688596e-06,
+      "loss": 0.8534,
+      "step": 1957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.723199327699376e-06,
+      "loss": 0.7438,
+      "step": 1958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722837686338025e-06,
+      "loss": 0.9898,
+      "step": 1959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.722475815622108e-06,
+      "loss": 0.8113,
+      "step": 1960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7221137155692e-06,
+      "loss": 0.9699,
+      "step": 1961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721751386196885e-06,
+      "loss": 0.9383,
+      "step": 1962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721388827522757e-06,
+      "loss": 0.5943,
+      "step": 1963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.721026039564423e-06,
+      "loss": 0.811,
+      "step": 1964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720663022339503e-06,
+      "loss": 0.8126,
+      "step": 1965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.720299775865625e-06,
+      "loss": 0.642,
+      "step": 1966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719936300160428e-06,
+      "loss": 1.0667,
+      "step": 1967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.719572595241566e-06,
+      "loss": 0.964,
+      "step": 1968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.7192086611267e-06,
+      "loss": 0.9078,
+      "step": 1969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718844497833504e-06,
+      "loss": 0.969,
+      "step": 1970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718480105379663e-06,
+      "loss": 0.9648,
+      "step": 1971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.718115483782873e-06,
+      "loss": 0.7068,
+      "step": 1972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717750633060841e-06,
+      "loss": 1.0629,
+      "step": 1973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717385553231284e-06,
+      "loss": 0.8189,
+      "step": 1974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.717020244311932e-06,
+      "loss": 0.8162,
+      "step": 1975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716654706320528e-06,
+      "loss": 0.6346,
+      "step": 1976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.716288939274818e-06,
+      "loss": 0.896,
+      "step": 1977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71592294319257e-06,
+      "loss": 0.776,
+      "step": 1978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715556718091556e-06,
+      "loss": 1.0011,
+      "step": 1979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.715190263989562e-06,
+      "loss": 0.8946,
+      "step": 1980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.71482358090438e-06,
+      "loss": 0.9698,
+      "step": 1981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714456668853822e-06,
+      "loss": 0.8374,
+      "step": 1982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.714089527855705e-06,
+      "loss": 0.6964,
+      "step": 1983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.713722157927856e-06,
+      "loss": 0.7351,
+      "step": 1984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.713354559088119e-06,
+      "loss": 0.7167,
+      "step": 1985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712986731354344e-06,
+      "loss": 0.7314,
+      "step": 1986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.712618674744393e-06,
+      "loss": 0.7412,
+      "step": 1987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71225038927614e-06,
+      "loss": 0.5705,
+      "step": 1988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711881874967471e-06,
+      "loss": 0.8611,
+      "step": 1989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.711513131836283e-06,
+      "loss": 0.7881,
+      "step": 1990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.71114415990048e-06,
+      "loss": 0.7255,
+      "step": 1991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710774959177983e-06,
+      "loss": 0.666,
+      "step": 1992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710405529686722e-06,
+      "loss": 0.6331,
+      "step": 1993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.710035871444636e-06,
+      "loss": 0.8153,
+      "step": 1994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709665984469676e-06,
+      "loss": 0.9131,
+      "step": 1995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.709295868779807e-06,
+      "loss": 0.6414,
+      "step": 1996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708925524393e-06,
+      "loss": 0.7746,
+      "step": 1997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.708554951327243e-06,
+      "loss": 0.5684,
+      "step": 1998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70818414960053e-06,
+      "loss": 0.9202,
+      "step": 1999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70781311923087e-06,
+      "loss": 0.8435,
+      "step": 2000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707441860236278e-06,
+      "loss": 0.9542,
+      "step": 2001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.707070372634788e-06,
+      "loss": 0.9778,
+      "step": 2002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706698656444438e-06,
+      "loss": 1.0596,
+      "step": 2003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.706326711683277e-06,
+      "loss": 0.8701,
+      "step": 2004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705954538369375e-06,
+      "loss": 0.762,
+      "step": 2005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705582136520798e-06,
+      "loss": 0.7772,
+      "step": 2006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.705209506155635e-06,
+      "loss": 0.9933,
+      "step": 2007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704836647291981e-06,
+      "loss": 0.9516,
+      "step": 2008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.704463559947944e-06,
+      "loss": 0.7959,
+      "step": 2009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70409024414164e-06,
+      "loss": 0.823,
+      "step": 2010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703716699891202e-06,
+      "loss": 1.0642,
+      "step": 2011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.703342927214766e-06,
+      "loss": 0.8092,
+      "step": 2012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702968926130486e-06,
+      "loss": 0.8025,
+      "step": 2013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702594696656525e-06,
+      "loss": 0.547,
+      "step": 2014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.702220238811056e-06,
+      "loss": 0.6773,
+      "step": 2015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.701845552612261e-06,
+      "loss": 0.8546,
+      "step": 2016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.70147063807834e-06,
+      "loss": 0.7155,
+      "step": 2017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.7010954952275e-06,
+      "loss": 0.7062,
+      "step": 2018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700720124077957e-06,
+      "loss": 0.744,
+      "step": 2019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.700344524647938e-06,
+      "loss": 0.9228,
+      "step": 2020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699968696955688e-06,
+      "loss": 0.7709,
+      "step": 2021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699592641019455e-06,
+      "loss": 0.822,
+      "step": 2022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.699216356857502e-06,
+      "loss": 0.9963,
+      "step": 2023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698839844488104e-06,
+      "loss": 0.8253,
+      "step": 2024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698463103929542e-06,
+      "loss": 0.9408,
+      "step": 2025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.698086135200116e-06,
+      "loss": 0.6094,
+      "step": 2026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697708938318129e-06,
+      "loss": 1.0087,
+      "step": 2027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.697331513301903e-06,
+      "loss": 0.7391,
+      "step": 2028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696953860169761e-06,
+      "loss": 0.4714,
+      "step": 2029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696575978940048e-06,
+      "loss": 0.9209,
+      "step": 2030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.696197869631111e-06,
+      "loss": 0.9704,
+      "step": 2031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695819532261316e-06,
+      "loss": 0.8569,
+      "step": 2032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695440966849034e-06,
+      "loss": 0.9763,
+      "step": 2033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.695062173412648e-06,
+      "loss": 0.7237,
+      "step": 2034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694683151970556e-06,
+      "loss": 0.6214,
+      "step": 2035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.694303902541164e-06,
+      "loss": 0.7256,
+      "step": 2036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693924425142886e-06,
+      "loss": 0.7895,
+      "step": 2037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.693544719794157e-06,
+      "loss": 0.9256,
+      "step": 2038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69316478651341e-06,
+      "loss": 1.0789,
+      "step": 2039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692784625319098e-06,
+      "loss": 0.8429,
+      "step": 2040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.692404236229684e-06,
+      "loss": 0.7768,
+      "step": 2041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69202361926364e-06,
+      "loss": 0.676,
+      "step": 2042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69164277443945e-06,
+      "loss": 0.6451,
+      "step": 2043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.691261701775607e-06,
+      "loss": 0.9267,
+      "step": 2044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690880401290621e-06,
+      "loss": 1.0644,
+      "step": 2045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.690498873003006e-06,
+      "loss": 0.9998,
+      "step": 2046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.69011711693129e-06,
+      "loss": 0.7237,
+      "step": 2047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689735133094013e-06,
+      "loss": 0.9984,
+      "step": 2048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689352921509725e-06,
+      "loss": 1.0097,
+      "step": 2049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68897048219699e-06,
+      "loss": 0.8001,
+      "step": 2050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688587815174375e-06,
+      "loss": 0.6947,
+      "step": 2051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.688204920460467e-06,
+      "loss": 1.0277,
+      "step": 2052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687821798073858e-06,
+      "loss": 0.7933,
+      "step": 2053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.687438448033157e-06,
+      "loss": 0.7835,
+      "step": 2054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68705487035698e-06,
+      "loss": 0.8893,
+      "step": 2055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68667106506395e-06,
+      "loss": 1.1264,
+      "step": 2056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.686287032172712e-06,
+      "loss": 0.9481,
+      "step": 2057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685902771701913e-06,
+      "loss": 0.6725,
+      "step": 2058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685518283670212e-06,
+      "loss": 0.9603,
+      "step": 2059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.685133568096284e-06,
+      "loss": 0.7316,
+      "step": 2060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68474862499881e-06,
+      "loss": 1.0833,
+      "step": 2061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.684363454396483e-06,
+      "loss": 0.7822,
+      "step": 2062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683978056308012e-06,
+      "loss": 0.9808,
+      "step": 2063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683592430752107e-06,
+      "loss": 0.4879,
+      "step": 2064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.683206577747501e-06,
+      "loss": 0.6138,
+      "step": 2065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682820497312928e-06,
+      "loss": 0.7479,
+      "step": 2066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.68243418946714e-06,
+      "loss": 1.0039,
+      "step": 2067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.682047654228895e-06,
+      "loss": 0.7583,
+      "step": 2068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681660891616967e-06,
+      "loss": 1.0853,
+      "step": 2069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.681273901650134e-06,
+      "loss": 0.8156,
+      "step": 2070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680886684347194e-06,
+      "loss": 0.7272,
+      "step": 2071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680499239726948e-06,
+      "loss": 0.771,
+      "step": 2072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.680111567808212e-06,
+      "loss": 0.6543,
+      "step": 2073
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679723668609815e-06,
+      "loss": 0.9037,
+      "step": 2074
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.679335542150592e-06,
+      "loss": 0.9519,
+      "step": 2075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678947188449393e-06,
+      "loss": 0.739,
+      "step": 2076
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678558607525075e-06,
+      "loss": 0.6412,
+      "step": 2077
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.678169799396512e-06,
+      "loss": 0.9228,
+      "step": 2078
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677780764082583e-06,
+      "loss": 0.687,
+      "step": 2079
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677391501602183e-06,
+      "loss": 0.8436,
+      "step": 2080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.677002011974213e-06,
+      "loss": 0.7731,
+      "step": 2081
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676612295217588e-06,
+      "loss": 0.7882,
+      "step": 2082
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.676222351351236e-06,
+      "loss": 1.0695,
+      "step": 2083
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675832180394093e-06,
+      "loss": 0.7636,
+      "step": 2084
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675441782365107e-06,
+      "loss": 0.6598,
+      "step": 2085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.675051157283236e-06,
+      "loss": 0.8659,
+      "step": 2086
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.674660305167451e-06,
+      "loss": 0.643,
+      "step": 2087
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.67426922603673e-06,
+      "loss": 0.8117,
+      "step": 2088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673877919910069e-06,
+      "loss": 0.6709,
+      "step": 2089
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673486386806469e-06,
+      "loss": 0.8179,
+      "step": 2090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.673094626744944e-06,
+      "loss": 0.8708,
+      "step": 2091
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672702639744517e-06,
+      "loss": 0.9196,
+      "step": 2092
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.672310425824228e-06,
+      "loss": 0.5774,
+      "step": 2093
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671917985003121e-06,
+      "loss": 0.7918,
+      "step": 2094
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.671525317300254e-06,
+      "loss": 0.8035,
+      "step": 2095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6711324227347e-06,
+      "loss": 0.7536,
+      "step": 2096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670739301325534e-06,
+      "loss": 0.8033,
+      "step": 2097
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.670345953091851e-06,
+      "loss": 0.8841,
+      "step": 2098
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669952378052749e-06,
+      "loss": 0.539,
+      "step": 2099
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669558576227347e-06,
+      "loss": 1.0218,
+      "step": 2100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.669164547634763e-06,
+      "loss": 0.9548,
+      "step": 2101
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668770292294135e-06,
+      "loss": 0.7447,
+      "step": 2102
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.668375810224611e-06,
+      "loss": 0.74,
+      "step": 2103
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667981101445346e-06,
+      "loss": 0.8092,
+      "step": 2104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667586165975507e-06,
+      "loss": 0.6195,
+      "step": 2105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.667191003834276e-06,
+      "loss": 0.9806,
+      "step": 2106
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66679561504084e-06,
+      "loss": 0.7263,
+      "step": 2107
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666399999614403e-06,
+      "loss": 0.957,
+      "step": 2108
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.666004157574177e-06,
+      "loss": 0.8119,
+      "step": 2109
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665608088939384e-06,
+      "loss": 0.8051,
+      "step": 2110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.665211793729259e-06,
+      "loss": 0.9,
+      "step": 2111
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664815271963045e-06,
+      "loss": 0.7762,
+      "step": 2112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664418523660004e-06,
+      "loss": 0.8568,
+      "step": 2113
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.664021548839398e-06,
+      "loss": 0.6508,
+      "step": 2114
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663624347520506e-06,
+      "loss": 0.7283,
+      "step": 2115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.663226919722617e-06,
+      "loss": 0.7422,
+      "step": 2116
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662829265465033e-06,
+      "loss": 0.8176,
+      "step": 2117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662431384767067e-06,
+      "loss": 0.6015,
+      "step": 2118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.662033277648036e-06,
+      "loss": 0.8345,
+      "step": 2119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.661634944127274e-06,
+      "loss": 0.5801,
+      "step": 2120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.66123638422413e-06,
+      "loss": 0.7407,
+      "step": 2121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660837597957954e-06,
+      "loss": 0.9132,
+      "step": 2122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660438585348117e-06,
+      "loss": 0.9284,
+      "step": 2123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.660039346413994e-06,
+      "loss": 0.7878,
+      "step": 2124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.659639881174972e-06,
+      "loss": 0.777,
+      "step": 2125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65924018965045e-06,
+      "loss": 0.7901,
+      "step": 2126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.65884027185984e-06,
+      "loss": 0.7819,
+      "step": 2127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658440127822562e-06,
+      "loss": 0.8876,
+      "step": 2128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.658039757558048e-06,
+      "loss": 0.9392,
+      "step": 2129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.657639161085743e-06,
+      "loss": 0.9375,
+      "step": 2130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.6572383384251e-06,
+      "loss": 0.8077,
+      "step": 2131
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656837289595582e-06,
+      "loss": 0.7238,
+      "step": 2132
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65643601461667e-06,
+      "loss": 0.6544,
+      "step": 2133
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.656034513507846e-06,
+      "loss": 0.7119,
+      "step": 2134
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65563278628861e-06,
+      "loss": 0.836,
+      "step": 2135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65523083297847e-06,
+      "loss": 0.6813,
+      "step": 2136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65482865359695e-06,
+      "loss": 1.0351,
+      "step": 2137
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654426248163577e-06,
+      "loss": 0.9405,
+      "step": 2138
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.654023616697895e-06,
+      "loss": 0.63,
+      "step": 2139
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.653620759219452e-06,
+      "loss": 1.0623,
+      "step": 2140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65321767574782e-06,
+      "loss": 0.7917,
+      "step": 2141
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.65281436630257e-06,
+      "loss": 0.8541,
+      "step": 2142
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652410830903285e-06,
+      "loss": 0.8199,
+      "step": 2143
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.652007069569566e-06,
+      "loss": 0.9738,
+      "step": 2144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651603082321019e-06,
+      "loss": 0.9813,
+      "step": 2145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.651198869177263e-06,
+      "loss": 0.682,
+      "step": 2146
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650794430157928e-06,
+      "loss": 0.8703,
+      "step": 2147
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.650389765282655e-06,
+      "loss": 0.6819,
+      "step": 2148
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649984874571095e-06,
+      "loss": 0.7639,
+      "step": 2149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649579758042911e-06,
+      "loss": 0.89,
+      "step": 2150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.649174415717776e-06,
+      "loss": 0.9371,
+      "step": 2151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648768847615375e-06,
+      "loss": 0.7333,
+      "step": 2152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.648363053755406e-06,
+      "loss": 0.8529,
+      "step": 2153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64795703415757e-06,
+      "loss": 0.5367,
+      "step": 2154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64755078884159e-06,
+      "loss": 0.8287,
+      "step": 2155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64714431782719e-06,
+      "loss": 1.0513,
+      "step": 2156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646737621134112e-06,
+      "loss": 1.0887,
+      "step": 2157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.646330698782107e-06,
+      "loss": 0.7148,
+      "step": 2158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645923550790936e-06,
+      "loss": 0.7344,
+      "step": 2159
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64551617718037e-06,
+      "loss": 1.2593,
+      "step": 2160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.645108577970191e-06,
+      "loss": 0.6206,
+      "step": 2161
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644700753180197e-06,
+      "loss": 0.675,
+      "step": 2162
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.644292702830189e-06,
+      "loss": 0.5256,
+      "step": 2163
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643884426939986e-06,
+      "loss": 0.663,
+      "step": 2164
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643475925529415e-06,
+      "loss": 0.9951,
+      "step": 2165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.643067198618312e-06,
+      "loss": 1.0749,
+      "step": 2166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.642658246226527e-06,
+      "loss": 1.0296,
+      "step": 2167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.64224906837392e-06,
+      "loss": 0.7895,
+      "step": 2168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641839665080363e-06,
+      "loss": 1.0021,
+      "step": 2169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641430036365738e-06,
+      "loss": 0.7575,
+      "step": 2170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.641020182249934e-06,
+      "loss": 0.9362,
+      "step": 2171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640610102752858e-06,
+      "loss": 0.8566,
+      "step": 2172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.640199797894423e-06,
+      "loss": 0.9361,
+      "step": 2173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639789267694557e-06,
+      "loss": 0.6733,
+      "step": 2174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.639378512173194e-06,
+      "loss": 0.9475,
+      "step": 2175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638967531350282e-06,
+      "loss": 0.9251,
+      "step": 2176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63855632524578e-06,
+      "loss": 0.7341,
+      "step": 2177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.638144893879657e-06,
+      "loss": 0.7644,
+      "step": 2178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637733237271894e-06,
+      "loss": 0.8834,
+      "step": 2179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.637321355442481e-06,
+      "loss": 0.6666,
+      "step": 2180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636909248411421e-06,
+      "loss": 0.8633,
+      "step": 2181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636496916198728e-06,
+      "loss": 1.0083,
+      "step": 2182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.636084358824422e-06,
+      "loss": 0.774,
+      "step": 2183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635671576308543e-06,
+      "loss": 0.8026,
+      "step": 2184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.635258568671135e-06,
+      "loss": 0.5858,
+      "step": 2185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634845335932255e-06,
+      "loss": 0.9049,
+      "step": 2186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634431878111969e-06,
+      "loss": 0.7214,
+      "step": 2187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.634018195230358e-06,
+      "loss": 0.8273,
+      "step": 2188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633604287307509e-06,
+      "loss": 1.0516,
+      "step": 2189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.633190154363527e-06,
+      "loss": 0.8446,
+      "step": 2190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63277579641852e-06,
+      "loss": 0.6359,
+      "step": 2191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.63236121349261e-06,
+      "loss": 0.982,
+      "step": 2192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631946405605932e-06,
+      "loss": 0.8386,
+      "step": 2193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631531372778632e-06,
+      "loss": 1.0731,
+      "step": 2194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.631116115030861e-06,
+      "loss": 0.4001,
+      "step": 2195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630700632382787e-06,
+      "loss": 0.8189,
+      "step": 2196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.630284924854587e-06,
+      "loss": 0.6289,
+      "step": 2197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62986899246645e-06,
+      "loss": 0.8328,
+      "step": 2198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629452835238573e-06,
+      "loss": 0.8286,
+      "step": 2199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.629036453191167e-06,
+      "loss": 0.9808,
+      "step": 2200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628619846344453e-06,
+      "loss": 0.8874,
+      "step": 2201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.628203014718661e-06,
+      "loss": 0.8481,
+      "step": 2202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627785958334036e-06,
+      "loss": 0.8248,
+      "step": 2203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.627368677210828e-06,
+      "loss": 0.8168,
+      "step": 2204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626951171369306e-06,
+      "loss": 0.6104,
+      "step": 2205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626533440829741e-06,
+      "loss": 0.9311,
+      "step": 2206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.626115485612422e-06,
+      "loss": 1.0962,
+      "step": 2207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625697305737645e-06,
+      "loss": 0.7339,
+      "step": 2208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.625278901225716e-06,
+      "loss": 0.7545,
+      "step": 2209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624860272096958e-06,
+      "loss": 0.8225,
+      "step": 2210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.6244414183717e-06,
+      "loss": 0.8458,
+      "step": 2211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.624022340070279e-06,
+      "loss": 0.8877,
+      "step": 2212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62360303721305e-06,
+      "loss": 1.014,
+      "step": 2213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.623183509820376e-06,
+      "loss": 0.6988,
+      "step": 2214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.622763757912627e-06,
+      "loss": 0.8778,
+      "step": 2215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62234378151019e-06,
+      "loss": 0.5333,
+      "step": 2216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621923580633462e-06,
+      "loss": 0.9959,
+      "step": 2217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621503155302845e-06,
+      "loss": 1.0336,
+      "step": 2218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.621082505538757e-06,
+      "loss": 0.7335,
+      "step": 2219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.62066163136163e-06,
+      "loss": 1.0016,
+      "step": 2220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.620240532791899e-06,
+      "loss": 0.6315,
+      "step": 2221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619819209850011e-06,
+      "loss": 0.8483,
+      "step": 2222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.619397662556434e-06,
+      "loss": 0.9382,
+      "step": 2223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618975890931635e-06,
+      "loss": 0.8465,
+      "step": 2224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618553894996098e-06,
+      "loss": 0.6843,
+      "step": 2225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.618131674770315e-06,
+      "loss": 1.0002,
+      "step": 2226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61770923027479e-06,
+      "loss": 0.9101,
+      "step": 2227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.617286561530039e-06,
+      "loss": 0.6957,
+      "step": 2228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616863668556588e-06,
+      "loss": 0.7444,
+      "step": 2229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616440551374977e-06,
+      "loss": 0.6932,
+      "step": 2230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.616017210005747e-06,
+      "loss": 0.6596,
+      "step": 2231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.61559364446946e-06,
+      "loss": 0.5015,
+      "step": 2232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.615169854786688e-06,
+      "loss": 1.0436,
+      "step": 2233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614745840978008e-06,
+      "loss": 0.9283,
+      "step": 2234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.614321603064014e-06,
+      "loss": 1.0053,
+      "step": 2235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613897141065305e-06,
+      "loss": 0.7792,
+      "step": 2236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613472455002498e-06,
+      "loss": 0.6907,
+      "step": 2237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.613047544896215e-06,
+      "loss": 0.9016,
+      "step": 2238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612622410767088e-06,
+      "loss": 0.6567,
+      "step": 2239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.612197052635768e-06,
+      "loss": 0.9091,
+      "step": 2240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611771470522908e-06,
+      "loss": 0.7488,
+      "step": 2241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.611345664449177e-06,
+      "loss": 0.7384,
+      "step": 2242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610919634435252e-06,
+      "loss": 0.8001,
+      "step": 2243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610493380501825e-06,
+      "loss": 0.8366,
+      "step": 2244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.610066902669593e-06,
+      "loss": 0.6159,
+      "step": 2245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60964020095927e-06,
+      "loss": 0.9783,
+      "step": 2246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.609213275391573e-06,
+      "loss": 0.8793,
+      "step": 2247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608786125987239e-06,
+      "loss": 0.7304,
+      "step": 2248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.608358752767013e-06,
+      "loss": 0.8228,
+      "step": 2249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607931155751646e-06,
+      "loss": 0.5092,
+      "step": 2250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607503334961904e-06,
+      "loss": 0.5726,
+      "step": 2251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.607075290418561e-06,
+      "loss": 0.8545,
+      "step": 2252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606647022142411e-06,
+      "loss": 0.9464,
+      "step": 2253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.606218530154243e-06,
+      "loss": 1.0933,
+      "step": 2254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605789814474873e-06,
+      "loss": 0.7383,
+      "step": 2255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.605360875125119e-06,
+      "loss": 0.6072,
+      "step": 2256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60493171212581e-06,
+      "loss": 0.7293,
+      "step": 2257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604502325497785e-06,
+      "loss": 0.6727,
+      "step": 2258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.604072715261902e-06,
+      "loss": 0.6814,
+      "step": 2259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60364288143902e-06,
+      "loss": 0.9334,
+      "step": 2260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.603212824050015e-06,
+      "loss": 0.9657,
+      "step": 2261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602782543115772e-06,
+      "loss": 0.7394,
+      "step": 2262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.602352038657183e-06,
+      "loss": 0.8737,
+      "step": 2263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.60192131069516e-06,
+      "loss": 0.6538,
+      "step": 2264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601490359250616e-06,
+      "loss": 1.1493,
+      "step": 2265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.601059184344483e-06,
+      "loss": 0.6663,
+      "step": 2266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600627785997696e-06,
+      "loss": 0.7046,
+      "step": 2267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.600196164231209e-06,
+      "loss": 0.9921,
+      "step": 2268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599764319065979e-06,
+      "loss": 0.8555,
+      "step": 2269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.599332250522978e-06,
+      "loss": 0.7332,
+      "step": 2270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598899958623192e-06,
+      "loss": 0.8903,
+      "step": 2271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598467443387612e-06,
+      "loss": 0.6327,
+      "step": 2272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.598034704837242e-06,
+      "loss": 0.8601,
+      "step": 2273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597601742993097e-06,
+      "loss": 0.93,
+      "step": 2274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.597168557876203e-06,
+      "loss": 0.8033,
+      "step": 2275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596735149507598e-06,
+      "loss": 0.9036,
+      "step": 2276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.596301517908329e-06,
+      "loss": 0.7842,
+      "step": 2277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.595867663099454e-06,
+      "loss": 0.7321,
+      "step": 2278
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59543358510204e-06,
+      "loss": 0.8677,
+      "step": 2279
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594999283937171e-06,
+      "loss": 0.7803,
+      "step": 2280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594564759625936e-06,
+      "loss": 0.8413,
+      "step": 2281
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.594130012189438e-06,
+      "loss": 0.8271,
+      "step": 2282
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593695041648787e-06,
+      "loss": 0.6479,
+      "step": 2283
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.593259848025107e-06,
+      "loss": 0.6197,
+      "step": 2284
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592824431339536e-06,
+      "loss": 0.6372,
+      "step": 2285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.592388791613215e-06,
+      "loss": 0.9146,
+      "step": 2286
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5919529288673e-06,
+      "loss": 0.8812,
+      "step": 2287
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591516843122958e-06,
+      "loss": 0.5007,
+      "step": 2288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.591080534401371e-06,
+      "loss": 0.824,
+      "step": 2289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.59064400272372e-06,
+      "loss": 0.8782,
+      "step": 2290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.590207248111211e-06,
+      "loss": 0.7401,
+      "step": 2291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589770270585053e-06,
+      "loss": 0.785,
+      "step": 2292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.589333070166463e-06,
+      "loss": 0.8362,
+      "step": 2293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588895646876672e-06,
+      "loss": 0.7095,
+      "step": 2294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588458000736929e-06,
+      "loss": 0.8945,
+      "step": 2295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.588020131768482e-06,
+      "loss": 0.5376,
+      "step": 2296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.587582039992598e-06,
+      "loss": 0.6539,
+      "step": 2297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58714372543055e-06,
+      "loss": 0.6431,
+      "step": 2298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.586705188103625e-06,
+      "loss": 0.8559,
+      "step": 2299
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58626642803312e-06,
+      "loss": 0.546,
+      "step": 2300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58582744524034e-06,
+      "loss": 0.8747,
+      "step": 2301
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.585388239746606e-06,
+      "loss": 0.8975,
+      "step": 2302
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584948811573246e-06,
+      "loss": 0.8522,
+      "step": 2303
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584509160741599e-06,
+      "loss": 0.4743,
+      "step": 2304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.584069287273019e-06,
+      "loss": 0.8368,
+      "step": 2305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583629191188864e-06,
+      "loss": 0.9226,
+      "step": 2306
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.583188872510508e-06,
+      "loss": 0.6596,
+      "step": 2307
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582748331259333e-06,
+      "loss": 0.7426,
+      "step": 2308
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.582307567456732e-06,
+      "loss": 0.6021,
+      "step": 2309
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.581866581124114e-06,
+      "loss": 0.7856,
+      "step": 2310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58142537228289e-06,
+      "loss": 0.8655,
+      "step": 2311
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.58098394095449e-06,
+      "loss": 0.5982,
+      "step": 2312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580542287160348e-06,
+      "loss": 0.6816,
+      "step": 2313
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.580100410921915e-06,
+      "loss": 0.8468,
+      "step": 2314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579658312260648e-06,
+      "loss": 0.9371,
+      "step": 2315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.579215991198017e-06,
+      "loss": 0.7966,
+      "step": 2316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5787734477555e-06,
+      "loss": 1.1634,
+      "step": 2317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.578330681954593e-06,
+      "loss": 0.5521,
+      "step": 2318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577887693816794e-06,
+      "loss": 0.8526,
+      "step": 2319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577444483363618e-06,
+      "loss": 0.9552,
+      "step": 2320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.577001050616588e-06,
+      "loss": 0.6664,
+      "step": 2321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.576557395597237e-06,
+      "loss": 0.7603,
+      "step": 2322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57611351832711e-06,
+      "loss": 0.964,
+      "step": 2323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.575669418827766e-06,
+      "loss": 0.9588,
+      "step": 2324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.57522509712077e-06,
+      "loss": 0.8784,
+      "step": 2325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5747805532277e-06,
+      "loss": 0.8402,
+      "step": 2326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.574335787170144e-06,
+      "loss": 0.8797,
+      "step": 2327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5738907989697e-06,
+      "loss": 0.7637,
+      "step": 2328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.573445588647978e-06,
+      "loss": 0.8396,
+      "step": 2329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5730001562266e-06,
+      "loss": 0.9433,
+      "step": 2330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572554501727198e-06,
+      "loss": 0.7922,
+      "step": 2331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.572108625171413e-06,
+      "loss": 0.976,
+      "step": 2332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571662526580898e-06,
+      "loss": 0.6321,
+      "step": 2333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.571216205977316e-06,
+      "loss": 0.6745,
+      "step": 2334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570769663382344e-06,
+      "loss": 1.0895,
+      "step": 2335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.570322898817668e-06,
+      "loss": 0.5317,
+      "step": 2336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56987591230498e-06,
+      "loss": 0.761,
+      "step": 2337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56942870386599e-06,
+      "loss": 1.0483,
+      "step": 2338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568981273522415e-06,
+      "loss": 1.0582,
+      "step": 2339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568533621295982e-06,
+      "loss": 0.7035,
+      "step": 2340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.568085747208434e-06,
+      "loss": 0.8647,
+      "step": 2341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567637651281516e-06,
+      "loss": 0.6508,
+      "step": 2342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.567189333536994e-06,
+      "loss": 0.7264,
+      "step": 2343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566740793996637e-06,
+      "loss": 0.6043,
+      "step": 2344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.566292032682228e-06,
+      "loss": 0.8341,
+      "step": 2345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56584304961556e-06,
+      "loss": 0.788,
+      "step": 2346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.565393844818435e-06,
+      "loss": 0.707,
+      "step": 2347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56494441831267e-06,
+      "loss": 0.6969,
+      "step": 2348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.564494770120089e-06,
+      "loss": 0.7756,
+      "step": 2349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56404490026253e-06,
+      "loss": 0.748,
+      "step": 2350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.56359480876184e-06,
+      "loss": 0.8113,
+      "step": 2351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.563144495639874e-06,
+      "loss": 0.8867,
+      "step": 2352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562693960918504e-06,
+      "loss": 0.5524,
+      "step": 2353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.562243204619606e-06,
+      "loss": 0.956,
+      "step": 2354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561792226765072e-06,
+      "loss": 0.6632,
+      "step": 2355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.561341027376801e-06,
+      "loss": 0.6599,
+      "step": 2356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560889606476708e-06,
+      "loss": 0.8867,
+      "step": 2357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.560437964086713e-06,
+      "loss": 0.851,
+      "step": 2358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559986100228746e-06,
+      "loss": 0.8536,
+      "step": 2359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559534014924756e-06,
+      "loss": 0.8492,
+      "step": 2360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.559081708196696e-06,
+      "loss": 0.7411,
+      "step": 2361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558629180066531e-06,
+      "loss": 0.9065,
+      "step": 2362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.558176430556236e-06,
+      "loss": 0.8648,
+      "step": 2363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5577234596878e-06,
+      "loss": 0.8455,
+      "step": 2364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.557270267483219e-06,
+      "loss": 0.976,
+      "step": 2365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556816853964501e-06,
+      "loss": 0.7399,
+      "step": 2366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.556363219153664e-06,
+      "loss": 0.9195,
+      "step": 2367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555909363072742e-06,
+      "loss": 0.7409,
+      "step": 2368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555455285743772e-06,
+      "loss": 0.6713,
+      "step": 2369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.555000987188806e-06,
+      "loss": 0.8382,
+      "step": 2370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554546467429906e-06,
+      "loss": 1.0834,
+      "step": 2371
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.554091726489147e-06,
+      "loss": 0.7917,
+      "step": 2372
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55363676438861e-06,
+      "loss": 0.6911,
+      "step": 2373
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.553181581150391e-06,
+      "loss": 0.7031,
+      "step": 2374
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.552726176796591e-06,
+      "loss": 0.7382,
+      "step": 2375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.55227055134933e-06,
+      "loss": 0.9283,
+      "step": 2376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551814704830734e-06,
+      "loss": 0.7419,
+      "step": 2377
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.551358637262939e-06,
+      "loss": 0.6213,
+      "step": 2378
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550902348668093e-06,
+      "loss": 0.5296,
+      "step": 2379
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.550445839068356e-06,
+      "loss": 0.9224,
+      "step": 2380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549989108485896e-06,
+      "loss": 0.8277,
+      "step": 2381
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549532156942894e-06,
+      "loss": 0.7746,
+      "step": 2382
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.549074984461538e-06,
+      "loss": 0.8867,
+      "step": 2383
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548617591064034e-06,
+      "loss": 0.9181,
+      "step": 2384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.548159976772593e-06,
+      "loss": 0.7418,
+      "step": 2385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.547702141609437e-06,
+      "loss": 0.8159,
+      "step": 2386
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.5472440855968e-06,
+      "loss": 0.7179,
+      "step": 2387
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546785808756926e-06,
+      "loss": 0.95,
+      "step": 2388
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.546327311112072e-06,
+      "loss": 1.0373,
+      "step": 2389
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545868592684503e-06,
+      "loss": 0.8898,
+      "step": 2390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.545409653496496e-06,
+      "loss": 0.8563,
+      "step": 2391
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544950493570338e-06,
+      "loss": 0.7657,
+      "step": 2392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544491112928327e-06,
+      "loss": 0.573,
+      "step": 2393
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.544031511592772e-06,
+      "loss": 0.8059,
+      "step": 2394
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.543571689585992e-06,
+      "loss": 0.8501,
+      "step": 2395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.54311164693032e-06,
+      "loss": 0.7127,
+      "step": 2396
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542651383648092e-06,
+      "loss": 0.5784,
+      "step": 2397
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.542190899761664e-06,
+      "loss": 0.5711,
+      "step": 2398
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541730195293398e-06,
+      "loss": 0.8784,
+      "step": 2399
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.541269270265665e-06,
+      "loss": 0.8089,
+      "step": 2400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540808124700848e-06,
+      "loss": 0.8645,
+      "step": 2401
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.540346758621347e-06,
+      "loss": 0.7983,
+      "step": 2402
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539885172049563e-06,
+      "loss": 0.8245,
+      "step": 2403
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.539423365007912e-06,
+      "loss": 0.7089,
+      "step": 2404
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538961337518822e-06,
+      "loss": 1.0679,
+      "step": 2405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53849908960473e-06,
+      "loss": 0.774,
+      "step": 2406
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.538036621288083e-06,
+      "loss": 0.6402,
+      "step": 2407
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537573932591341e-06,
+      "loss": 1.015,
+      "step": 2408
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.537111023536973e-06,
+      "loss": 0.7428,
+      "step": 2409
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53664789414746e-06,
+      "loss": 0.8781,
+      "step": 2410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.536184544445292e-06,
+      "loss": 0.7183,
+      "step": 2411
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.535720974452973e-06,
+      "loss": 0.7516,
+      "step": 2412
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53525718419301e-06,
+      "loss": 1.0253,
+      "step": 2413
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53479317368793e-06,
+      "loss": 0.8616,
+      "step": 2414
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.534328942960267e-06,
+      "loss": 0.726,
+      "step": 2415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533864492032562e-06,
+      "loss": 0.8727,
+      "step": 2416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.533399820927375e-06,
+      "loss": 0.7669,
+      "step": 2417
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532934929667269e-06,
+      "loss": 0.8207,
+      "step": 2418
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.53246981827482e-06,
+      "loss": 1.2203,
+      "step": 2419
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.532004486772612e-06,
+      "loss": 0.9208,
+      "step": 2420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531538935183252e-06,
+      "loss": 1.0153,
+      "step": 2421
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.531073163529338e-06,
+      "loss": 0.8848,
+      "step": 2422
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530607171833499e-06,
+      "loss": 0.781,
+      "step": 2423
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.530140960118358e-06,
+      "loss": 0.5688,
+      "step": 2424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.529674528406556e-06,
+      "loss": 0.7238,
+      "step": 2425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.529207876720747e-06,
+      "loss": 0.675,
+      "step": 2426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528741005083593e-06,
+      "loss": 0.9103,
+      "step": 2427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.528273913517765e-06,
+      "loss": 0.6439,
+      "step": 2428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527806602045947e-06,
+      "loss": 0.8617,
+      "step": 2429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.527339070690833e-06,
+      "loss": 0.7708,
+      "step": 2430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526871319475128e-06,
+      "loss": 0.7768,
+      "step": 2431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.526403348421545e-06,
+      "loss": 0.7891,
+      "step": 2432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525935157552813e-06,
+      "loss": 0.8739,
+      "step": 2433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.525466746891666e-06,
+      "loss": 1.0838,
+      "step": 2434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524998116460856e-06,
+      "loss": 1.2553,
+      "step": 2435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524529266283136e-06,
+      "loss": 0.8358,
+      "step": 2436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.524060196381277e-06,
+      "loss": 0.5758,
+      "step": 2437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.523590906778058e-06,
+      "loss": 0.6736,
+      "step": 2438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.52312139749627e-06,
+      "loss": 0.7204,
+      "step": 2439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522651668558713e-06,
+      "loss": 0.6038,
+      "step": 2440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.522181719988196e-06,
+      "loss": 0.9485,
+      "step": 2441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521711551807544e-06,
+      "loss": 0.8341,
+      "step": 2442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.521241164039589e-06,
+      "loss": 0.9683,
+      "step": 2443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520770556707177e-06,
+      "loss": 0.8708,
+      "step": 2444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.520299729833157e-06,
+      "loss": 0.8191,
+      "step": 2445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.519828683440397e-06,
+      "loss": 0.5763,
+      "step": 2446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51935741755177e-06,
+      "loss": 0.8571,
+      "step": 2447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518885932190166e-06,
+      "loss": 0.9275,
+      "step": 2448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.518414227378477e-06,
+      "loss": 0.6166,
+      "step": 2449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517942303139613e-06,
+      "loss": 0.9487,
+      "step": 2450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.517470159496492e-06,
+      "loss": 0.8894,
+      "step": 2451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51699779647204e-06,
+      "loss": 0.8882,
+      "step": 2452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516525214089202e-06,
+      "loss": 0.6095,
+      "step": 2453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.516052412370922e-06,
+      "loss": 0.8058,
+      "step": 2454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515579391340164e-06,
+      "loss": 0.8478,
+      "step": 2455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.515106151019899e-06,
+      "loss": 0.8558,
+      "step": 2456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514632691433108e-06,
+      "loss": 0.7545,
+      "step": 2457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.514159012602782e-06,
+      "loss": 0.8396,
+      "step": 2458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513685114551927e-06,
+      "loss": 0.8093,
+      "step": 2459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.513210997303556e-06,
+      "loss": 0.9765,
+      "step": 2460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512736660880694e-06,
+      "loss": 0.9535,
+      "step": 2461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.512262105306375e-06,
+      "loss": 0.7488,
+      "step": 2462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511787330603645e-06,
+      "loss": 0.9891,
+      "step": 2463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.511312336795561e-06,
+      "loss": 0.726,
+      "step": 2464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.51083712390519e-06,
+      "loss": 1.0289,
+      "step": 2465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.510361691955607e-06,
+      "loss": 0.6172,
+      "step": 2466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509886040969906e-06,
+      "loss": 0.7694,
+      "step": 2467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.509410170971181e-06,
+      "loss": 0.5229,
+      "step": 2468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508934081982545e-06,
+      "loss": 0.7545,
+      "step": 2469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.508457774027116e-06,
+      "loss": 0.862,
+      "step": 2470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507981247128024e-06,
+      "loss": 0.8689,
+      "step": 2471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507504501308412e-06,
+      "loss": 0.921,
+      "step": 2472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.507027536591436e-06,
+      "loss": 0.7683,
+      "step": 2473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50655035300025e-06,
+      "loss": 0.8711,
+      "step": 2474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.506072950558036e-06,
+      "loss": 0.81,
+      "step": 2475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505595329287974e-06,
+      "loss": 0.6473,
+      "step": 2476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.505117489213257e-06,
+      "loss": 0.5947,
+      "step": 2477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504639430357092e-06,
+      "loss": 0.7956,
+      "step": 2478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.504161152742696e-06,
+      "loss": 0.7342,
+      "step": 2479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503682656393298e-06,
+      "loss": 0.8283,
+      "step": 2480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.503203941332128e-06,
+      "loss": 0.6005,
+      "step": 2481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.502725007582439e-06,
+      "loss": 0.8501,
+      "step": 2482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50224585516749e-06,
+      "loss": 0.7279,
+      "step": 2483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.501766484110546e-06,
+      "loss": 0.8462,
+      "step": 2484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.50128689443489e-06,
+      "loss": 0.9358,
+      "step": 2485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500807086163808e-06,
+      "loss": 0.8313,
+      "step": 2486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.500327059320606e-06,
+      "loss": 0.518,
+      "step": 2487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499846813928595e-06,
+      "loss": 1.003,
+      "step": 2488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.499366350011093e-06,
+      "loss": 0.8532,
+      "step": 2489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498885667591435e-06,
+      "loss": 0.6634,
+      "step": 2490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.498404766692966e-06,
+      "loss": 0.925,
+      "step": 2491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497923647339039e-06,
+      "loss": 0.5769,
+      "step": 2492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.497442309553017e-06,
+      "loss": 0.7874,
+      "step": 2493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496960753358276e-06,
+      "loss": 0.8712,
+      "step": 2494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.496478978778202e-06,
+      "loss": 0.8834,
+      "step": 2495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495996985836192e-06,
+      "loss": 0.5367,
+      "step": 2496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495514774555649e-06,
+      "loss": 0.8405,
+      "step": 2497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.495032344959998e-06,
+      "loss": 0.827,
+      "step": 2498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.494549697072662e-06,
+      "loss": 0.7163,
+      "step": 2499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49406683091708e-06,
+      "loss": 0.8267,
+      "step": 2500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.4935837465167e-06,
+      "loss": 0.7316,
+      "step": 2501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.493100443894986e-06,
+      "loss": 0.867,
+      "step": 2502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492616923075406e-06,
+      "loss": 1.0404,
+      "step": 2503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.492133184081442e-06,
+      "loss": 0.6566,
+      "step": 2504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.491649226936586e-06,
+      "loss": 0.8961,
+      "step": 2505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.49116505166434e-06,
+      "loss": 0.745,
+      "step": 2506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490680658288215e-06,
+      "loss": 0.9568,
+      "step": 2507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.490196046831739e-06,
+      "loss": 0.4623,
+      "step": 2508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48971121731844e-06,
+      "loss": 0.7939,
+      "step": 2509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48922616977187e-06,
+      "loss": 0.7963,
+      "step": 2510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488740904215578e-06,
+      "loss": 0.8743,
+      "step": 2511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.488255420673134e-06,
+      "loss": 0.6966,
+      "step": 2512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487769719168112e-06,
+      "loss": 1.0448,
+      "step": 2513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.487283799724099e-06,
+      "loss": 0.6168,
+      "step": 2514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486797662364695e-06,
+      "loss": 1.1269,
+      "step": 2515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.486311307113506e-06,
+      "loss": 0.7744,
+      "step": 2516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.485824733994151e-06,
+      "loss": 0.7958,
+      "step": 2517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48533794303026e-06,
+      "loss": 0.6859,
+      "step": 2518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484850934245475e-06,
+      "loss": 0.7583,
+      "step": 2519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.484363707663443e-06,
+      "loss": 0.5743,
+      "step": 2520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483876263307825e-06,
+      "loss": 0.8673,
+      "step": 2521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.483388601202297e-06,
+      "loss": 0.6952,
+      "step": 2522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.482900721370538e-06,
+      "loss": 0.8304,
+      "step": 2523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48241262383624e-06,
+      "loss": 0.7431,
+      "step": 2524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481924308623109e-06,
+      "loss": 0.9559,
+      "step": 2525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.481435775754857e-06,
+      "loss": 0.7291,
+      "step": 2526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.48094702525521e-06,
+      "loss": 0.8859,
+      "step": 2527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.480458057147904e-06,
+      "loss": 0.963,
+      "step": 2528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47996887145668e-06,
+      "loss": 0.6549,
+      "step": 2529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.479479468205299e-06,
+      "loss": 0.9697,
+      "step": 2530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478989847417526e-06,
+      "loss": 0.8505,
+      "step": 2531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47850000911714e-06,
+      "loss": 0.897,
+      "step": 2532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.478009953327928e-06,
+      "loss": 0.8862,
+      "step": 2533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477519680073686e-06,
+      "loss": 0.7437,
+      "step": 2534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.477029189378226e-06,
+      "loss": 0.8351,
+      "step": 2535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476538481265366e-06,
+      "loss": 0.7378,
+      "step": 2536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.476047555758938e-06,
+      "loss": 0.8533,
+      "step": 2537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.475556412882782e-06,
+      "loss": 1.2686,
+      "step": 2538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47506505266075e-06,
+      "loss": 0.9708,
+      "step": 2539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.474573475116701e-06,
+      "loss": 0.6693,
+      "step": 2540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47408168027451e-06,
+      "loss": 0.7698,
+      "step": 2541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473589668158061e-06,
+      "loss": 0.9527,
+      "step": 2542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.473097438791243e-06,
+      "loss": 0.8488,
+      "step": 2543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.472604992197966e-06,
+      "loss": 0.7062,
+      "step": 2544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.47211232840214e-06,
+      "loss": 0.7764,
+      "step": 2545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471619447427692e-06,
+      "loss": 0.7498,
+      "step": 2546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.471126349298557e-06,
+      "loss": 0.9579,
+      "step": 2547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470633034038684e-06,
+      "loss": 0.899,
+      "step": 2548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.470139501672025e-06,
+      "loss": 0.9694,
+      "step": 2549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46964575222255e-06,
+      "loss": 0.8154,
+      "step": 2550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.469151785714237e-06,
+      "loss": 0.7263,
+      "step": 2551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468657602171076e-06,
+      "loss": 0.7917,
+      "step": 2552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.468163201617063e-06,
+      "loss": 0.9135,
+      "step": 2553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467668584076209e-06,
+      "loss": 0.5649,
+      "step": 2554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.467173749572531e-06,
+      "loss": 0.9238,
+      "step": 2555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.466678698130064e-06,
+      "loss": 0.9009,
+      "step": 2556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46618342977285e-06,
+      "loss": 0.5382,
+      "step": 2557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465687944524936e-06,
+      "loss": 0.9724,
+      "step": 2558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.465192242410388e-06,
+      "loss": 0.7862,
+      "step": 2559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464696323453275e-06,
+      "loss": 0.7156,
+      "step": 2560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.464200187677683e-06,
+      "loss": 0.8531,
+      "step": 2561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463703835107705e-06,
+      "loss": 0.8603,
+      "step": 2562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.463207265767445e-06,
+      "loss": 0.7608,
+      "step": 2563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.462710479681018e-06,
+      "loss": 1.0955,
+      "step": 2564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.46221347687255e-06,
+      "loss": 0.9745,
+      "step": 2565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461716257366175e-06,
+      "loss": 0.8272,
+      "step": 2566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.461218821186045e-06,
+      "loss": 0.5443,
+      "step": 2567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460721168356309e-06,
+      "loss": 0.5762,
+      "step": 2568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.460223298901138e-06,
+      "loss": 0.9945,
+      "step": 2569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459725212844713e-06,
+      "loss": 0.6908,
+      "step": 2570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.459226910211219e-06,
+      "loss": 0.6355,
+      "step": 2571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.458728391024854e-06,
+      "loss": 1.0865,
+      "step": 2572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45822965530983e-06,
+      "loss": 0.7698,
+      "step": 2573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457730703090367e-06,
+      "loss": 0.6703,
+      "step": 2574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.457231534390694e-06,
+      "loss": 0.5709,
+      "step": 2575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456732149235053e-06,
+      "loss": 0.7312,
+      "step": 2576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.456232547647695e-06,
+      "loss": 0.6422,
+      "step": 2577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.455732729652884e-06,
+      "loss": 0.7368,
+      "step": 2578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45523269527489e-06,
+      "loss": 1.0639,
+      "step": 2579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454732444537997e-06,
+      "loss": 0.9127,
+      "step": 2580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.454231977466499e-06,
+      "loss": 0.7867,
+      "step": 2581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.4537312940847e-06,
+      "loss": 0.8216,
+      "step": 2582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.453230394416914e-06,
+      "loss": 0.7111,
+      "step": 2583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45272927848747e-06,
+      "loss": 0.8257,
+      "step": 2584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452227946320697e-06,
+      "loss": 0.7793,
+      "step": 2585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451726397940946e-06,
+      "loss": 0.711,
+      "step": 2586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.451224633372572e-06,
+      "loss": 0.8253,
+      "step": 2587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.45072265263994e-06,
+      "loss": 0.8526,
+      "step": 2588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.450220455767432e-06,
+      "loss": 0.9394,
+      "step": 2589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449718042779434e-06,
+      "loss": 0.945,
+      "step": 2590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.449215413700345e-06,
+      "loss": 0.8168,
+      "step": 2591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448712568554572e-06,
+      "loss": 0.9575,
+      "step": 2592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.448209507366537e-06,
+      "loss": 0.7682,
+      "step": 2593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44770623016067e-06,
+      "loss": 0.7256,
+      "step": 2594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.447202736961411e-06,
+      "loss": 0.7354,
+      "step": 2595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446699027793212e-06,
+      "loss": 0.9557,
+      "step": 2596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.446195102680531e-06,
+      "loss": 0.9033,
+      "step": 2597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445690961647844e-06,
+      "loss": 1.0154,
+      "step": 2598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.445186604719632e-06,
+      "loss": 1.0009,
+      "step": 2599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.44468203192039e-06,
+      "loss": 0.8052,
+      "step": 2600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.444177243274619e-06,
+      "loss": 0.7981,
+      "step": 2601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443672238806832e-06,
+      "loss": 0.565,
+      "step": 2602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.443167018541557e-06,
+      "loss": 0.6624,
+      "step": 2603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442661582503326e-06,
+      "loss": 0.935,
+      "step": 2604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.442155930716686e-06,
+      "loss": 0.6772,
+      "step": 2605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441650063206193e-06,
+      "loss": 0.6961,
+      "step": 2606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.441143979996413e-06,
+      "loss": 0.9192,
+      "step": 2607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440637681111922e-06,
+      "loss": 0.8395,
+      "step": 2608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.440131166577309e-06,
+      "loss": 0.8566,
+      "step": 2609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43962443641717e-06,
+      "loss": 0.657,
+      "step": 2610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.439117490656114e-06,
+      "loss": 0.6504,
+      "step": 2611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43861032931876e-06,
+      "loss": 0.7775,
+      "step": 2612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438102952429737e-06,
+      "loss": 0.546,
+      "step": 2613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437595360013686e-06,
+      "loss": 0.8031,
+      "step": 2614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.437087552095254e-06,
+      "loss": 0.6398,
+      "step": 2615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436579528699105e-06,
+      "loss": 0.7248,
+      "step": 2616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.436071289849909e-06,
+      "loss": 0.6136,
+      "step": 2617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.435562835572347e-06,
+      "loss": 0.9388,
+      "step": 2618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43505416589111e-06,
+      "loss": 0.5158,
+      "step": 2619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434545280830901e-06,
+      "loss": 0.8332,
+      "step": 2620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.434036180416435e-06,
+      "loss": 0.9185,
+      "step": 2621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.433526864672434e-06,
+      "loss": 0.8518,
+      "step": 2622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43301733362363e-06,
+      "loss": 0.8461,
+      "step": 2623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43250758729477e-06,
+      "loss": 0.7006,
+      "step": 2624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.43199762571061e-06,
+      "loss": 0.8689,
+      "step": 2625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.431487448895909e-06,
+      "loss": 0.4546,
+      "step": 2626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430977056875449e-06,
+      "loss": 0.8765,
+      "step": 2627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.430466449674014e-06,
+      "loss": 0.9301,
+      "step": 2628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429955627316399e-06,
+      "loss": 0.9162,
+      "step": 2629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.429444589827412e-06,
+      "loss": 0.7272,
+      "step": 2630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428933337231872e-06,
+      "loss": 0.8207,
+      "step": 2631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.428421869554607e-06,
+      "loss": 0.5887,
+      "step": 2632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42791018682045e-06,
+      "loss": 0.8268,
+      "step": 2633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.427398289054257e-06,
+      "loss": 0.9723,
+      "step": 2634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426886176280883e-06,
+      "loss": 0.9129,
+      "step": 2635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.426373848525199e-06,
+      "loss": 0.9578,
+      "step": 2636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425861305812083e-06,
+      "loss": 0.9186,
+      "step": 2637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.425348548166429e-06,
+      "loss": 0.7142,
+      "step": 2638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424835575613136e-06,
+      "loss": 0.9262,
+      "step": 2639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.424322388177116e-06,
+      "loss": 0.6061,
+      "step": 2640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423808985883289e-06,
+      "loss": 0.7818,
+      "step": 2641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423295368756591e-06,
+      "loss": 0.9482,
+      "step": 2642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.42278153682196e-06,
+      "loss": 0.7205,
+      "step": 2643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.422267490104354e-06,
+      "loss": 0.6854,
+      "step": 2644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421753228628734e-06,
+      "loss": 0.6331,
+      "step": 2645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.421238752420075e-06,
+      "loss": 1.0151,
+      "step": 2646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420724061503359e-06,
+      "loss": 0.8886,
+      "step": 2647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.420209155903583e-06,
+      "loss": 0.6542,
+      "step": 2648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419694035645753e-06,
+      "loss": 0.7863,
+      "step": 2649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.419178700754881e-06,
+      "loss": 1.0319,
+      "step": 2650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.418663151255997e-06,
+      "loss": 0.913,
+      "step": 2651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41814738717414e-06,
+      "loss": 0.8091,
+      "step": 2652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417631408534349e-06,
+      "loss": 0.7145,
+      "step": 2653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.417115215361687e-06,
+      "loss": 0.647,
+      "step": 2654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416598807681221e-06,
+      "loss": 0.6161,
+      "step": 2655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.416082185518029e-06,
+      "loss": 0.9618,
+      "step": 2656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.415565348897199e-06,
+      "loss": 0.6583,
+      "step": 2657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.41504829784383e-06,
+      "loss": 0.8827,
+      "step": 2658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414531032383033e-06,
+      "loss": 0.8581,
+      "step": 2659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.414013552539926e-06,
+      "loss": 0.846,
+      "step": 2660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.413495858339641e-06,
+      "loss": 0.8423,
+      "step": 2661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412977949807319e-06,
+      "loss": 0.938,
+      "step": 2662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.412459826968109e-06,
+      "loss": 0.7347,
+      "step": 2663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411941489847172e-06,
+      "loss": 0.8622,
+      "step": 2664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.411422938469683e-06,
+      "loss": 0.6162,
+      "step": 2665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410904172860821e-06,
+      "loss": 1.153,
+      "step": 2666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.410385193045783e-06,
+      "loss": 0.7138,
+      "step": 2667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40986599904977e-06,
+      "loss": 0.7475,
+      "step": 2668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.409346590897994e-06,
+      "loss": 0.5277,
+      "step": 2669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40882696861568e-06,
+      "loss": 0.666,
+      "step": 2670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.408307132228066e-06,
+      "loss": 0.9104,
+      "step": 2671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.407787081760388e-06,
+      "loss": 0.8111,
+      "step": 2672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40726681723791e-06,
+      "loss": 0.6916,
+      "step": 2673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406746338685895e-06,
+      "loss": 1.0661,
+      "step": 2674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.406225646129617e-06,
+      "loss": 0.791,
+      "step": 2675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405704739594361e-06,
+      "loss": 0.7627,
+      "step": 2676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.405183619105429e-06,
+      "loss": 0.9086,
+      "step": 2677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404662284688124e-06,
+      "loss": 0.7631,
+      "step": 2678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.404140736367764e-06,
+      "loss": 0.8137,
+      "step": 2679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403618974169679e-06,
+      "loss": 0.9074,
+      "step": 2680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.403096998119206e-06,
+      "loss": 0.7982,
+      "step": 2681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402574808241692e-06,
+      "loss": 0.7565,
+      "step": 2682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.402052404562498e-06,
+      "loss": 0.7814,
+      "step": 2683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401529787106993e-06,
+      "loss": 0.8661,
+      "step": 2684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.401006955900555e-06,
+      "loss": 0.8135,
+      "step": 2685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.400483910968579e-06,
+      "loss": 0.7413,
+      "step": 2686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399960652336462e-06,
+      "loss": 0.9824,
+      "step": 2687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.399437180029614e-06,
+      "loss": 0.5628,
+      "step": 2688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398913494073459e-06,
+      "loss": 0.5834,
+      "step": 2689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.398389594493426e-06,
+      "loss": 0.7108,
+      "step": 2690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.397865481314959e-06,
+      "loss": 0.5491,
+      "step": 2691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39734115456351e-06,
+      "loss": 0.7121,
+      "step": 2692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.396816614264543e-06,
+      "loss": 0.7452,
+      "step": 2693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39629186044353e-06,
+      "loss": 0.5167,
+      "step": 2694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395766893125952e-06,
+      "loss": 0.6872,
+      "step": 2695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395241712337307e-06,
+      "loss": 0.959,
+      "step": 2696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394716318103098e-06,
+      "loss": 1.0495,
+      "step": 2697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.394190710448839e-06,
+      "loss": 0.8482,
+      "step": 2698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393664889400056e-06,
+      "loss": 0.6609,
+      "step": 2699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.393138854982283e-06,
+      "loss": 0.6313,
+      "step": 2700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392612607221067e-06,
+      "loss": 0.9621,
+      "step": 2701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.392086146141963e-06,
+      "loss": 0.8717,
+      "step": 2702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.391559471770537e-06,
+      "loss": 0.7593,
+      "step": 2703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.39103258413237e-06,
+      "loss": 0.8027,
+      "step": 2704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.390505483253045e-06,
+      "loss": 0.7727,
+      "step": 2705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38997816915816e-06,
+      "loss": 0.775,
+      "step": 2706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.389450641873323e-06,
+      "loss": 0.6746,
+      "step": 2707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388922901424154e-06,
+      "loss": 0.7136,
+      "step": 2708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.388394947836278e-06,
+      "loss": 0.6435,
+      "step": 2709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38786678113534e-06,
+      "loss": 0.9752,
+      "step": 2710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.387338401346982e-06,
+      "loss": 0.6377,
+      "step": 2711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38680980849687e-06,
+      "loss": 0.6925,
+      "step": 2712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.386281002610669e-06,
+      "loss": 1.2425,
+      "step": 2713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385751983714061e-06,
+      "loss": 0.8987,
+      "step": 2714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.385222751832739e-06,
+      "loss": 0.8244,
+      "step": 2715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384693306992403e-06,
+      "loss": 0.5636,
+      "step": 2716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.384163649218761e-06,
+      "loss": 0.629,
+      "step": 2717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38363377853754e-06,
+      "loss": 0.8247,
+      "step": 2718
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.383103694974467e-06,
+      "loss": 0.6832,
+      "step": 2719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382573398555289e-06,
+      "loss": 0.7542,
+      "step": 2720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.382042889305754e-06,
+      "loss": 1.0019,
+      "step": 2721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.38151216725163e-06,
+      "loss": 0.7214,
+      "step": 2722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380981232418685e-06,
+      "loss": 0.6922,
+      "step": 2723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.380450084832707e-06,
+      "loss": 0.5716,
+      "step": 2724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379918724519488e-06,
+      "loss": 0.7167,
+      "step": 2725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.379387151504836e-06,
+      "loss": 0.633,
+      "step": 2726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37885536581456e-06,
+      "loss": 0.5747,
+      "step": 2727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.378323367474488e-06,
+      "loss": 0.8394,
+      "step": 2728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377791156510456e-06,
+      "loss": 0.8407,
+      "step": 2729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.377258732948307e-06,
+      "loss": 0.6724,
+      "step": 2730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.376726096813901e-06,
+      "loss": 0.668,
+      "step": 2731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3761932481331e-06,
+      "loss": 0.6921,
+      "step": 2732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375660186931785e-06,
+      "loss": 0.7545,
+      "step": 2733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.375126913235838e-06,
+      "loss": 0.9178,
+      "step": 2734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374593427071163e-06,
+      "loss": 0.7883,
+      "step": 2735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.374059728463663e-06,
+      "loss": 0.6787,
+      "step": 2736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.373525817439255e-06,
+      "loss": 0.679,
+      "step": 2737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372991694023869e-06,
+      "loss": 0.7943,
+      "step": 2738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.372457358243445e-06,
+      "loss": 0.9248,
+      "step": 2739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.37192281012393e-06,
+      "loss": 0.962,
+      "step": 2740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.371388049691283e-06,
+      "loss": 0.7901,
+      "step": 2741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370853076971473e-06,
+      "loss": 0.6565,
+      "step": 2742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.370317891990482e-06,
+      "loss": 0.8431,
+      "step": 2743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3697824947743e-06,
+      "loss": 0.647,
+      "step": 2744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.369246885348926e-06,
+      "loss": 0.8911,
+      "step": 2745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.36871106374037e-06,
+      "loss": 0.9159,
+      "step": 2746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.368175029974656e-06,
+      "loss": 0.7927,
+      "step": 2747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367638784077814e-06,
+      "loss": 0.5545,
+      "step": 2748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.367102326075881e-06,
+      "loss": 0.9704,
+      "step": 2749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366565655994917e-06,
+      "loss": 0.7449,
+      "step": 2750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.366028773860981e-06,
+      "loss": 0.6506,
+      "step": 2751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.365491679700143e-06,
+      "loss": 0.5736,
+      "step": 2752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364954373538486e-06,
+      "loss": 0.9148,
+      "step": 2753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.364416855402107e-06,
+      "loss": 0.5983,
+      "step": 2754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363879125317108e-06,
+      "loss": 0.9832,
+      "step": 2755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.363341183309601e-06,
+      "loss": 1.0159,
+      "step": 2756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362803029405712e-06,
+      "loss": 0.7796,
+      "step": 2757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.362264663631574e-06,
+      "loss": 0.5515,
+      "step": 2758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361726086013331e-06,
+      "loss": 0.7754,
+      "step": 2759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.361187296577142e-06,
+      "loss": 0.8215,
+      "step": 2760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360648295349165e-06,
+      "loss": 0.91,
+      "step": 2761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.360109082355582e-06,
+      "loss": 0.9807,
+      "step": 2762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.359569657622573e-06,
+      "loss": 0.9504,
+      "step": 2763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35903002117634e-06,
+      "loss": 0.6805,
+      "step": 2764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.358490173043086e-06,
+      "loss": 0.714,
+      "step": 2765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357950113249027e-06,
+      "loss": 0.6624,
+      "step": 2766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.357409841820394e-06,
+      "loss": 0.9309,
+      "step": 2767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.356869358783418e-06,
+      "loss": 0.7908,
+      "step": 2768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35632866416435e-06,
+      "loss": 0.8915,
+      "step": 2769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355787757989448e-06,
+      "loss": 0.756,
+      "step": 2770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.355246640284979e-06,
+      "loss": 0.9525,
+      "step": 2771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35470531107722e-06,
+      "loss": 0.861,
+      "step": 2772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35416377039246e-06,
+      "loss": 0.9166,
+      "step": 2773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.353622018257002e-06,
+      "loss": 0.9007,
+      "step": 2774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.35308005469715e-06,
+      "loss": 0.7672,
+      "step": 2775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.352537879739224e-06,
+      "loss": 0.7933,
+      "step": 2776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351995493409556e-06,
+      "loss": 0.7701,
+      "step": 2777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.351452895734483e-06,
+      "loss": 0.9533,
+      "step": 2778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350910086740356e-06,
+      "loss": 0.8087,
+      "step": 2779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.350367066453539e-06,
+      "loss": 0.8821,
+      "step": 2780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349823834900396e-06,
+      "loss": 0.6401,
+      "step": 2781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.349280392107314e-06,
+      "loss": 0.7796,
+      "step": 2782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34873673810068e-06,
+      "loss": 0.9905,
+      "step": 2783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.348192872906896e-06,
+      "loss": 0.6587,
+      "step": 2784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347648796552375e-06,
+      "loss": 0.9128,
+      "step": 2785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.347104509063538e-06,
+      "loss": 0.6756,
+      "step": 2786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34656001046682e-06,
+      "loss": 0.8383,
+      "step": 2787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34601530078866e-06,
+      "loss": 0.7785,
+      "step": 2788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34547038005551e-06,
+      "loss": 0.9243,
+      "step": 2789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.344925248293837e-06,
+      "loss": 0.5099,
+      "step": 2790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34437990553011e-06,
+      "loss": 0.9471,
+      "step": 2791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343834351790815e-06,
+      "loss": 0.7967,
+      "step": 2792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.343288587102444e-06,
+      "loss": 0.7217,
+      "step": 2793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342742611491502e-06,
+      "loss": 0.6991,
+      "step": 2794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.342196424984503e-06,
+      "loss": 0.6963,
+      "step": 2795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34165002760797e-06,
+      "loss": 0.694,
+      "step": 2796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.341103419388443e-06,
+      "loss": 0.575,
+      "step": 2797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.34055660035246e-06,
+      "loss": 0.4585,
+      "step": 2798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.340009570526578e-06,
+      "loss": 0.772,
+      "step": 2799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.339462329937365e-06,
+      "loss": 0.6205,
+      "step": 2800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.338914878611394e-06,
+      "loss": 0.8993,
+      "step": 2801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33836721657525e-06,
+      "loss": 0.7913,
+      "step": 2802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337819343855531e-06,
+      "loss": 0.6295,
+      "step": 2803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.337271260478847e-06,
+      "loss": 0.6744,
+      "step": 2804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336722966471806e-06,
+      "loss": 0.5089,
+      "step": 2805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.336174461861041e-06,
+      "loss": 0.8865,
+      "step": 2806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335625746673186e-06,
+      "loss": 0.7425,
+      "step": 2807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.335076820934889e-06,
+      "loss": 0.6367,
+      "step": 2808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.334527684672809e-06,
+      "loss": 0.8004,
+      "step": 2809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333978337913614e-06,
+      "loss": 0.7713,
+      "step": 2810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.333428780683977e-06,
+      "loss": 0.8874,
+      "step": 2811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332879013010591e-06,
+      "loss": 0.8268,
+      "step": 2812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.332329034920151e-06,
+      "loss": 0.7654,
+      "step": 2813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.331778846439368e-06,
+      "loss": 0.8309,
+      "step": 2814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.33122844759496e-06,
+      "loss": 0.5801,
+      "step": 2815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330677838413656e-06,
+      "loss": 0.877,
+      "step": 2816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.330127018922195e-06,
+      "loss": 0.6759,
+      "step": 2817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329575989147325e-06,
+      "loss": 0.5327,
+      "step": 2818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.329024749115807e-06,
+      "loss": 0.7743,
+      "step": 2819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.328473298854413e-06,
+      "loss": 0.9544,
+      "step": 2820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327921638389918e-06,
+      "loss": 0.8484,
+      "step": 2821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.327369767749115e-06,
+      "loss": 1.0587,
+      "step": 2822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.326817686958806e-06,
+      "loss": 0.7931,
+      "step": 2823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3262653960458e-06,
+      "loss": 0.5594,
+      "step": 2824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325712895036916e-06,
+      "loss": 0.6246,
+      "step": 2825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.325160183958987e-06,
+      "loss": 0.5795,
+      "step": 2826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324607262838856e-06,
+      "loss": 0.5826,
+      "step": 2827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.324054131703371e-06,
+      "loss": 0.6886,
+      "step": 2828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.323500790579395e-06,
+      "loss": 0.949,
+      "step": 2829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.3229472394938e-06,
+      "loss": 0.8945,
+      "step": 2830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.32239347847347e-06,
+      "loss": 1.1193,
+      "step": 2831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321839507545294e-06,
+      "loss": 0.9327,
+      "step": 2832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.321285326736175e-06,
+      "loss": 0.7023,
+      "step": 2833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320730936073028e-06,
+      "loss": 1.0286,
+      "step": 2834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.320176335582772e-06,
+      "loss": 0.876,
+      "step": 2835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319621525292343e-06,
+      "loss": 0.8843,
+      "step": 2836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.319066505228682e-06,
+      "loss": 0.9592,
+      "step": 2837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.318511275418745e-06,
+      "loss": 0.5789,
+      "step": 2838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317955835889493e-06,
+      "loss": 0.9745,
+      "step": 2839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.317400186667903e-06,
+      "loss": 0.7723,
+      "step": 2840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316844327780955e-06,
+      "loss": 0.7598,
+      "step": 2841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.316288259255646e-06,
+      "loss": 0.6526,
+      "step": 2842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315731981118979e-06,
+      "loss": 0.6746,
+      "step": 2843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.315175493397968e-06,
+      "loss": 0.8143,
+      "step": 2844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314618796119638e-06,
+      "loss": 0.6816,
+      "step": 2845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.314061889311026e-06,
+      "loss": 0.7657,
+      "step": 2846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.313504772999173e-06,
+      "loss": 0.6027,
+      "step": 2847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312947447211138e-06,
+      "loss": 0.5952,
+      "step": 2848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.312389911973985e-06,
+      "loss": 0.7872,
+      "step": 2849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311832167314788e-06,
+      "loss": 0.7182,
+      "step": 2850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.311274213260633e-06,
+      "loss": 0.8544,
+      "step": 2851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310716049838618e-06,
+      "loss": 0.8175,
+      "step": 2852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.310157677075847e-06,
+      "loss": 0.6161,
+      "step": 2853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309599094999438e-06,
+      "loss": 0.6006,
+      "step": 2854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.309040303636517e-06,
+      "loss": 0.7322,
+      "step": 2855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.308481303014216e-06,
+      "loss": 0.8233,
+      "step": 2856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307922093159688e-06,
+      "loss": 0.6476,
+      "step": 2857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.307362674100088e-06,
+      "loss": 0.8992,
+      "step": 2858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30680304586258e-06,
+      "loss": 0.7695,
+      "step": 2859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.306243208474346e-06,
+      "loss": 0.9717,
+      "step": 2860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.305683161962569e-06,
+      "loss": 0.8367,
+      "step": 2861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.30512290635445e-06,
+      "loss": 0.8305,
+      "step": 2862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304562441677192e-06,
+      "loss": 0.8406,
+      "step": 2863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.304001767958017e-06,
+      "loss": 0.6891,
+      "step": 2864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.303440885224153e-06,
+      "loss": 0.9301,
+      "step": 2865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.302879793502835e-06,
+      "loss": 0.8945,
+      "step": 2866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.302318492821312e-06,
+      "loss": 1.1265,
+      "step": 2867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.301756983206845e-06,
+      "loss": 0.5185,
+      "step": 2868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.3011952646867e-06,
+      "loss": 0.6319,
+      "step": 2869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300633337288156e-06,
+      "loss": 0.9918,
+      "step": 2870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.300071201038503e-06,
+      "loss": 0.7236,
+      "step": 2871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29950885596504e-06,
+      "loss": 0.7208,
+      "step": 2872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298946302095074e-06,
+      "loss": 0.9502,
+      "step": 2873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.298383539455925e-06,
+      "loss": 0.6949,
+      "step": 2874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297820568074924e-06,
+      "loss": 0.7905,
+      "step": 2875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.297257387979409e-06,
+      "loss": 0.8309,
+      "step": 2876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296693999196729e-06,
+      "loss": 0.8846,
+      "step": 2877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.296130401754247e-06,
+      "loss": 0.8504,
+      "step": 2878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295566595679327e-06,
+      "loss": 1.0185,
+      "step": 2879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.295002580999355e-06,
+      "loss": 0.4863,
+      "step": 2880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29443835774172e-06,
+      "loss": 0.875,
+      "step": 2881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29387392593382e-06,
+      "loss": 0.9698,
+      "step": 2882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.293309285603066e-06,
+      "loss": 0.9301,
+      "step": 2883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.29274443677688e-06,
+      "loss": 0.6444,
+      "step": 2884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.292179379482691e-06,
+      "loss": 0.7891,
+      "step": 2885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291614113747942e-06,
+      "loss": 1.088,
+      "step": 2886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.291048639600082e-06,
+      "loss": 0.69,
+      "step": 2887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.290482957066574e-06,
+      "loss": 0.7909,
+      "step": 2888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289917066174887e-06,
+      "loss": 0.9417,
+      "step": 2889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.289350966952502e-06,
+      "loss": 0.9196,
+      "step": 2890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28878465942691e-06,
+      "loss": 0.8597,
+      "step": 2891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.288218143625618e-06,
+      "loss": 0.8294,
+      "step": 2892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28765141957613e-06,
+      "loss": 0.8397,
+      "step": 2893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.287084487305975e-06,
+      "loss": 0.8189,
+      "step": 2894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28651734684268e-06,
+      "loss": 0.4846,
+      "step": 2895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.285949998213788e-06,
+      "loss": 0.7199,
+      "step": 2896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28538244144685e-06,
+      "loss": 0.7062,
+      "step": 2897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284814676569431e-06,
+      "loss": 0.7943,
+      "step": 2898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.284246703609103e-06,
+      "loss": 0.9732,
+      "step": 2899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283678522593445e-06,
+      "loss": 0.731,
+      "step": 2900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.283110133550053e-06,
+      "loss": 0.719,
+      "step": 2901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.282541536506527e-06,
+      "loss": 0.7239,
+      "step": 2902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.281972731490483e-06,
+      "loss": 0.7833,
+      "step": 2903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.28140371852954e-06,
+      "loss": 0.8464,
+      "step": 2904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280834497651334e-06,
+      "loss": 0.9455,
+      "step": 2905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.280265068883505e-06,
+      "loss": 0.621,
+      "step": 2906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27969543225371e-06,
+      "loss": 0.8767,
+      "step": 2907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.279125587789609e-06,
+      "loss": 0.5832,
+      "step": 2908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.278555535518876e-06,
+      "loss": 0.8249,
+      "step": 2909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277985275469194e-06,
+      "loss": 0.7817,
+      "step": 2910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.277414807668259e-06,
+      "loss": 0.8777,
+      "step": 2911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276844132143771e-06,
+      "loss": 0.6519,
+      "step": 2912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.276273248923446e-06,
+      "loss": 0.9878,
+      "step": 2913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.275702158035007e-06,
+      "loss": 0.8452,
+      "step": 2914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27513085950619e-06,
+      "loss": 0.7005,
+      "step": 2915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.274559353364734e-06,
+      "loss": 0.9362,
+      "step": 2916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.273987639638396e-06,
+      "loss": 0.9825,
+      "step": 2917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27341571835494e-06,
+      "loss": 0.7505,
+      "step": 2918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272843589542143e-06,
+      "loss": 1.0364,
+      "step": 2919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.272271253227782e-06,
+      "loss": 0.5028,
+      "step": 2920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271698709439658e-06,
+      "loss": 0.767,
+      "step": 2921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.271125958205571e-06,
+      "loss": 0.681,
+      "step": 2922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.27055299955334e-06,
+      "loss": 0.8354,
+      "step": 2923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269979833510783e-06,
+      "loss": 0.6547,
+      "step": 2924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.269406460105742e-06,
+      "loss": 0.8095,
+      "step": 2925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268832879366056e-06,
+      "loss": 0.6687,
+      "step": 2926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.268259091319583e-06,
+      "loss": 0.6668,
+      "step": 2927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267685095994186e-06,
+      "loss": 0.823,
+      "step": 2928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.267110893417739e-06,
+      "loss": 0.8543,
+      "step": 2929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.266536483618132e-06,
+      "loss": 0.8477,
+      "step": 2930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265961866623253e-06,
+      "loss": 0.8215,
+      "step": 2931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.265387042461012e-06,
+      "loss": 0.8021,
+      "step": 2932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264812011159321e-06,
+      "loss": 0.9718,
+      "step": 2933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.264236772746108e-06,
+      "loss": 0.5551,
+      "step": 2934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263661327249306e-06,
+      "loss": 0.8087,
+      "step": 2935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.263085674696864e-06,
+      "loss": 1.0378,
+      "step": 2936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.262509815116732e-06,
+      "loss": 0.7049,
+      "step": 2937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261933748536879e-06,
+      "loss": 0.9057,
+      "step": 2938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.261357474985279e-06,
+      "loss": 0.6818,
+      "step": 2939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.260780994489918e-06,
+      "loss": 0.7975,
+      "step": 2940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.26020430707879e-06,
+      "loss": 0.7898,
+      "step": 2941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259627412779905e-06,
+      "loss": 0.7033,
+      "step": 2942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.259050311621274e-06,
+      "loss": 1.025,
+      "step": 2943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.258473003630923e-06,
+      "loss": 1.0539,
+      "step": 2944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257895488836893e-06,
+      "loss": 0.7086,
+      "step": 2945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.257317767267223e-06,
+      "loss": 0.9902,
+      "step": 2946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25673983894997e-06,
+      "loss": 1.0664,
+      "step": 2947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.256161703913205e-06,
+      "loss": 0.6535,
+      "step": 2948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.255583362184998e-06,
+      "loss": 0.7425,
+      "step": 2949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25500481379344e-06,
+      "loss": 0.5052,
+      "step": 2950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.254426058766624e-06,
+      "loss": 0.7401,
+      "step": 2951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253847097132656e-06,
+      "loss": 0.7847,
+      "step": 2952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.253267928919652e-06,
+      "loss": 0.9735,
+      "step": 2953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.252688554155738e-06,
+      "loss": 0.8609,
+      "step": 2954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25210897286905e-06,
+      "loss": 0.666,
+      "step": 2955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.251529185087735e-06,
+      "loss": 0.6973,
+      "step": 2956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.25094919083995e-06,
+      "loss": 0.6947,
+      "step": 2957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.250368990153861e-06,
+      "loss": 0.8089,
+      "step": 2958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.249788583057641e-06,
+      "loss": 0.7145,
+      "step": 2959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24920796957948e-06,
+      "loss": 0.671,
+      "step": 2960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248627149747573e-06,
+      "loss": 0.7263,
+      "step": 2961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.248046123590126e-06,
+      "loss": 0.8996,
+      "step": 2962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.247464891135353e-06,
+      "loss": 0.6072,
+      "step": 2963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246883452411486e-06,
+      "loss": 0.7171,
+      "step": 2964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.246301807446757e-06,
+      "loss": 0.7498,
+      "step": 2965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.245719956269412e-06,
+      "loss": 0.7675,
+      "step": 2966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.24513789890771e-06,
+      "loss": 0.8795,
+      "step": 2967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.244555635389915e-06,
+      "loss": 0.466,
+      "step": 2968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243973165744306e-06,
+      "loss": 0.8753,
+      "step": 2969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.243390489999166e-06,
+      "loss": 0.7933,
+      "step": 2970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242807608182795e-06,
+      "loss": 0.7927,
+      "step": 2971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.242224520323498e-06,
+      "loss": 0.6596,
+      "step": 2972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.241641226449592e-06,
+      "loss": 0.7675,
+      "step": 2973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2410577265894e-06,
+      "loss": 1.0128,
+      "step": 2974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.240474020771262e-06,
+      "loss": 0.9161,
+      "step": 2975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239890109023524e-06,
+      "loss": 0.9403,
+      "step": 2976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.239305991374541e-06,
+      "loss": 0.8297,
+      "step": 2977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238721667852681e-06,
+      "loss": 0.7886,
+      "step": 2978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.238137138486318e-06,
+      "loss": 0.7221,
+      "step": 2979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.237552403303841e-06,
+      "loss": 0.7344,
+      "step": 2980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.236967462333649e-06,
+      "loss": 0.9281,
+      "step": 2981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23638231560414e-06,
+      "loss": 0.61,
+      "step": 2982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235796963143737e-06,
+      "loss": 0.576,
+      "step": 2983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.235211404980866e-06,
+      "loss": 1.1444,
+      "step": 2984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.234625641143962e-06,
+      "loss": 0.5696,
+      "step": 2985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23403967166147e-06,
+      "loss": 0.835,
+      "step": 2986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.233453496561851e-06,
+      "loss": 0.781,
+      "step": 2987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232867115873566e-06,
+      "loss": 0.7868,
+      "step": 2988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.232280529625092e-06,
+      "loss": 0.9641,
+      "step": 2989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.23169373784492e-06,
+      "loss": 0.8509,
+      "step": 2990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.231106740561541e-06,
+      "loss": 0.7816,
+      "step": 2991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.230519537803466e-06,
+      "loss": 0.911,
+      "step": 2992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229932129599206e-06,
+      "loss": 0.8272,
+      "step": 2993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.229344515977293e-06,
+      "loss": 1.0203,
+      "step": 2994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228756696966256e-06,
+      "loss": 0.8432,
+      "step": 2995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.228168672594648e-06,
+      "loss": 0.9673,
+      "step": 2996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.227580442891022e-06,
+      "loss": 0.7926,
+      "step": 2997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226992007883945e-06,
+      "loss": 0.7935,
+      "step": 2998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.226403367601993e-06,
+      "loss": 0.6613,
+      "step": 2999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.22581452207375e-06,
+      "loss": 0.8878,
+      "step": 3000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.225225471327815e-06,
+      "loss": 0.8011,
+      "step": 3001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.224636215392794e-06,
+      "loss": 0.7108,
+      "step": 3002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.2240467542973e-06,
+      "loss": 0.5499,
+      "step": 3003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.223457088069962e-06,
+      "loss": 0.9411,
+      "step": 3004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222867216739413e-06,
+      "loss": 0.6976,
+      "step": 3005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.222277140334301e-06,
+      "loss": 0.663,
+      "step": 3006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221686858883282e-06,
+      "loss": 0.7523,
+      "step": 3007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.221096372415021e-06,
+      "loss": 0.9241,
+      "step": 3008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.220505680958194e-06,
+      "loss": 0.7291,
+      "step": 3009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219914784541485e-06,
+      "loss": 0.8081,
+      "step": 3010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.219323683193592e-06,
+      "loss": 0.8279,
+      "step": 3011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218732376943219e-06,
+      "loss": 0.6404,
+      "step": 3012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.218140865819083e-06,
+      "loss": 0.7531,
+      "step": 3013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.217549149849907e-06,
+      "loss": 0.9032,
+      "step": 3014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.21695722906443e-06,
+      "loss": 1.0221,
+      "step": 3015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.216365103491393e-06,
+      "loss": 0.6047,
+      "step": 3016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215772773159556e-06,
+      "loss": 0.8767,
+      "step": 3017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.215180238097682e-06,
+      "loss": 0.4491,
+      "step": 3018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.214587498334544e-06,
+      "loss": 0.7913,
+      "step": 3019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213994553898931e-06,
+      "loss": 0.9119,
+      "step": 3020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.213401404819636e-06,
+      "loss": 0.8675,
+      "step": 3021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212808051125465e-06,
+      "loss": 0.8812,
+      "step": 3022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.212214492845232e-06,
+      "loss": 1.0523,
+      "step": 3023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.211620730007763e-06,
+      "loss": 0.8604,
+      "step": 3024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.21102676264189e-06,
+      "loss": 0.9199,
+      "step": 3025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.210432590776461e-06,
+      "loss": 0.7322,
+      "step": 3026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20983821444033e-06,
+      "loss": 0.6603,
+      "step": 3027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.209243633662362e-06,
+      "loss": 0.6289,
+      "step": 3028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20864884847143e-06,
+      "loss": 0.7991,
+      "step": 3029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.208053858896418e-06,
+      "loss": 0.8466,
+      "step": 3030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.207458664966224e-06,
+      "loss": 0.7865,
+      "step": 3031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206863266709749e-06,
+      "loss": 0.7596,
+      "step": 3032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.206267664155906e-06,
+      "loss": 0.8838,
+      "step": 3033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205671857333625e-06,
+      "loss": 1.0003,
+      "step": 3034
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.205075846271833e-06,
+      "loss": 0.8034,
+      "step": 3035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.20447963099948e-06,
+      "loss": 0.7335,
+      "step": 3036
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203883211545517e-06,
+      "loss": 0.846,
+      "step": 3037
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.203286587938908e-06,
+      "loss": 0.9611,
+      "step": 3038
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202689760208627e-06,
+      "loss": 0.6433,
+      "step": 3039
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.202092728383656e-06,
+      "loss": 0.7913,
+      "step": 3040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.201495492492994e-06,
+      "loss": 0.6931,
+      "step": 3041
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200898052565638e-06,
+      "loss": 0.6174,
+      "step": 3042
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.200300408630605e-06,
+      "loss": 1.06,
+      "step": 3043
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199702560716917e-06,
+      "loss": 0.9015,
+      "step": 3044
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.199104508853607e-06,
+      "loss": 1.1991,
+      "step": 3045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19850625306972e-06,
+      "loss": 0.7959,
+      "step": 3046
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197907793394308e-06,
+      "loss": 0.784,
+      "step": 3047
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.197309129856433e-06,
+      "loss": 0.7565,
+      "step": 3048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196710262485168e-06,
+      "loss": 0.8135,
+      "step": 3049
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.196111191309598e-06,
+      "loss": 1.042,
+      "step": 3050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.195511916358813e-06,
+      "loss": 0.8639,
+      "step": 3051
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.194912437661916e-06,
+      "loss": 1.1413,
+      "step": 3052
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19431275524802e-06,
+      "loss": 0.5397,
+      "step": 3053
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.193712869146248e-06,
+      "loss": 0.7104,
+      "step": 3054
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19311277938573e-06,
+      "loss": 0.8902,
+      "step": 3055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19251248599561e-06,
+      "loss": 0.8482,
+      "step": 3056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191911989005038e-06,
+      "loss": 0.7712,
+      "step": 3057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.191311288443179e-06,
+      "loss": 0.6233,
+      "step": 3058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.190710384339203e-06,
+      "loss": 0.6658,
+      "step": 3059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.19010927672229e-06,
+      "loss": 0.7774,
+      "step": 3060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.189507965621634e-06,
+      "loss": 0.6782,
+      "step": 3061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188906451066438e-06,
+      "loss": 0.8704,
+      "step": 3062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.188304733085907e-06,
+      "loss": 0.746,
+      "step": 3063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187702811709267e-06,
+      "loss": 1.0151,
+      "step": 3064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.187100686965749e-06,
+      "loss": 0.7217,
+      "step": 3065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18649835888459e-06,
+      "loss": 0.6457,
+      "step": 3066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.185895827495047e-06,
+      "loss": 0.846,
+      "step": 3067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.185293092826375e-06,
+      "loss": 0.7229,
+      "step": 3068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.18469015490785e-06,
+      "loss": 0.7373,
+      "step": 3069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.184087013768746e-06,
+      "loss": 0.8381,
+      "step": 3070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.183483669438358e-06,
+      "loss": 0.9174,
+      "step": 3071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182880121945983e-06,
+      "loss": 0.7724,
+      "step": 3072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.182276371320936e-06,
+      "loss": 0.5781,
+      "step": 3073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181672417592529e-06,
+      "loss": 0.8374,
+      "step": 3074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.181068260790101e-06,
+      "loss": 1.0157,
+      "step": 3075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.180463900942983e-06,
+      "loss": 0.8461,
+      "step": 3076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17985933808053e-06,
+      "loss": 0.5998,
+      "step": 3077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1792545722321e-06,
+      "loss": 0.4892,
+      "step": 3078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17864960342706e-06,
+      "loss": 0.9846,
+      "step": 3079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.178044431694793e-06,
+      "loss": 0.9216,
+      "step": 3080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.177439057064684e-06,
+      "loss": 0.5662,
+      "step": 3081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.176833479566131e-06,
+      "loss": 0.824,
+      "step": 3082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17622769922855e-06,
+      "loss": 1.0461,
+      "step": 3083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17562171608135e-06,
+      "loss": 0.9178,
+      "step": 3084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.175015530153968e-06,
+      "loss": 0.809,
+      "step": 3085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.174409141475833e-06,
+      "loss": 0.7591,
+      "step": 3086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173802550076402e-06,
+      "loss": 0.5417,
+      "step": 3087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.173195755985127e-06,
+      "loss": 0.6878,
+      "step": 3088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.172588759231477e-06,
+      "loss": 0.6491,
+      "step": 3089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.17198155984493e-06,
+      "loss": 0.9769,
+      "step": 3090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.171374157854973e-06,
+      "loss": 0.8181,
+      "step": 3091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170766553291102e-06,
+      "loss": 0.4071,
+      "step": 3092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.170158746182827e-06,
+      "loss": 0.8214,
+      "step": 3093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.169550736559662e-06,
+      "loss": 0.5843,
+      "step": 3094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168942524451137e-06,
+      "loss": 0.8345,
+      "step": 3095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.168334109886785e-06,
+      "loss": 0.955,
+      "step": 3096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.167725492896153e-06,
+      "loss": 0.7141,
+      "step": 3097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.1671166735088e-06,
+      "loss": 0.8725,
+      "step": 3098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.166507651754289e-06,
+      "loss": 0.7895,
+      "step": 3099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165898427662196e-06,
+      "loss": 0.9183,
+      "step": 3100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.165289001262108e-06,
+      "loss": 0.7281,
+      "step": 3101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16467937258362e-06,
+      "loss": 0.9828,
+      "step": 3102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.164069541656337e-06,
+      "loss": 0.8724,
+      "step": 3103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.163459508509876e-06,
+      "loss": 0.9459,
+      "step": 3104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.162849273173857e-06,
+      "loss": 0.6426,
+      "step": 3105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.16223883567792e-06,
+      "loss": 0.7081,
+      "step": 3106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161628196051709e-06,
+      "loss": 0.9526,
+      "step": 3107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.161017354324877e-06,
+      "loss": 0.6554,
+      "step": 3108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.160406310527087e-06,
+      "loss": 0.603,
+      "step": 3109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159795064688016e-06,
+      "loss": 0.7492,
+      "step": 3110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.159183616837345e-06,
+      "loss": 0.6672,
+      "step": 3111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15857196700477e-06,
+      "loss": 0.9307,
+      "step": 3112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157960115219993e-06,
+      "loss": 0.9172,
+      "step": 3113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.157348061512728e-06,
+      "loss": 0.8879,
+      "step": 3114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156735805912697e-06,
+      "loss": 0.5562,
+      "step": 3115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.156123348449634e-06,
+      "loss": 0.8377,
+      "step": 3116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.155510689153282e-06,
+      "loss": 0.8022,
+      "step": 3117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154897828053393e-06,
+      "loss": 0.765,
+      "step": 3118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.154284765179728e-06,
+      "loss": 0.8612,
+      "step": 3119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.15367150056206e-06,
+      "loss": 0.7176,
+      "step": 3120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.153058034230174e-06,
+      "loss": 0.66,
+      "step": 3121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.152444366213856e-06,
+      "loss": 0.5844,
+      "step": 3122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151830496542912e-06,
+      "loss": 0.3518,
+      "step": 3123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.151216425247152e-06,
+      "loss": 0.9361,
+      "step": 3124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.150602152356394e-06,
+      "loss": 0.8591,
+      "step": 3125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.149987677900475e-06,
+      "loss": 0.6919,
+      "step": 3126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14937300190923e-06,
+      "loss": 0.7586,
+      "step": 3127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14875812441251e-06,
+      "loss": 0.7195,
+      "step": 3128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.148143045440181e-06,
+      "loss": 0.9819,
+      "step": 3129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.147527765022107e-06,
+      "loss": 0.8908,
+      "step": 3130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14691228318817e-06,
+      "loss": 0.9861,
+      "step": 3131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.146296599968258e-06,
+      "loss": 0.9245,
+      "step": 3132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.145680715392274e-06,
+      "loss": 0.7087,
+      "step": 3133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14506462949012e-06,
+      "loss": 0.6638,
+      "step": 3134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.144448342291724e-06,
+      "loss": 0.5994,
+      "step": 3135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14383185382701e-06,
+      "loss": 0.8538,
+      "step": 3136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.143215164125915e-06,
+      "loss": 0.9106,
+      "step": 3137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.142598273218389e-06,
+      "loss": 0.8653,
+      "step": 3138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.14198118113439e-06,
+      "loss": 0.6588,
+      "step": 3139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.141363887903887e-06,
+      "loss": 0.7287,
+      "step": 3140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140746393556853e-06,
+      "loss": 1.0434,
+      "step": 3141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.140128698123284e-06,
+      "loss": 0.8905,
+      "step": 3142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.139510801633167e-06,
+      "loss": 0.774,
+      "step": 3143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138892704116515e-06,
+      "loss": 0.8517,
+      "step": 3144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.138274405603342e-06,
+      "loss": 0.6624,
+      "step": 3145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137655906123678e-06,
+      "loss": 0.6881,
+      "step": 3146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.137037205707552e-06,
+      "loss": 0.8246,
+      "step": 3147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.136418304385019e-06,
+      "loss": 0.8622,
+      "step": 3148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135799202186128e-06,
+      "loss": 0.9133,
+      "step": 3149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.135179899140947e-06,
+      "loss": 0.8466,
+      "step": 3150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.134560395279547e-06,
+      "loss": 0.7292,
+      "step": 3151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.13394069063202e-06,
+      "loss": 0.7142,
+      "step": 3152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.133320785228457e-06,
+      "loss": 0.8636,
+      "step": 3153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132700679098962e-06,
+      "loss": 0.6319,
+      "step": 3154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.132080372273649e-06,
+      "loss": 0.7503,
+      "step": 3155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.131459864782641e-06,
+      "loss": 0.8358,
+      "step": 3156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130839156656075e-06,
+      "loss": 0.6714,
+      "step": 3157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.130218247924093e-06,
+      "loss": 0.7327,
+      "step": 3158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.129597138616845e-06,
+      "loss": 0.8484,
+      "step": 3159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.128975828764498e-06,
+      "loss": 0.9849,
+      "step": 3160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.128354318397223e-06,
+      "loss": 0.7101,
+      "step": 3161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.127732607545201e-06,
+      "loss": 0.7246,
+      "step": 3162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.127110696238626e-06,
+      "loss": 1.0372,
+      "step": 3163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1264885845077e-06,
+      "loss": 0.9554,
+      "step": 3164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125866272382633e-06,
+      "loss": 0.7588,
+      "step": 3165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.125243759893646e-06,
+      "loss": 0.7237,
+      "step": 3166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.124621047070974e-06,
+      "loss": 0.7707,
+      "step": 3167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123998133944854e-06,
+      "loss": 0.7371,
+      "step": 3168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.123375020545534e-06,
+      "loss": 0.8456,
+      "step": 3169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122751706903279e-06,
+      "loss": 1.0,
+      "step": 3170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.122128193048358e-06,
+      "loss": 0.8585,
+      "step": 3171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.12150447901105e-06,
+      "loss": 0.8489,
+      "step": 3172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120880564821642e-06,
+      "loss": 0.7467,
+      "step": 3173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.120256450510436e-06,
+      "loss": 0.8037,
+      "step": 3174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119632136107741e-06,
+      "loss": 0.7984,
+      "step": 3175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.119007621643874e-06,
+      "loss": 0.7374,
+      "step": 3176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.118382907149164e-06,
+      "loss": 0.7217,
+      "step": 3177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117757992653949e-06,
+      "loss": 0.8611,
+      "step": 3178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.117132878188576e-06,
+      "loss": 0.5834,
+      "step": 3179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.116507563783402e-06,
+      "loss": 0.7103,
+      "step": 3180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115882049468795e-06,
+      "loss": 0.9787,
+      "step": 3181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.115256335275133e-06,
+      "loss": 0.7283,
+      "step": 3182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.1146304212328e-06,
+      "loss": 0.9013,
+      "step": 3183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.114004307372192e-06,
+      "loss": 0.929,
+      "step": 3184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.113377993723716e-06,
+      "loss": 0.4261,
+      "step": 3185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.11275148031779e-06,
+      "loss": 0.8127,
+      "step": 3186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.112124767184835e-06,
+      "loss": 0.7201,
+      "step": 3187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.111497854355287e-06,
+      "loss": 0.6649,
+      "step": 3188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110870741859594e-06,
+      "loss": 0.763,
+      "step": 3189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.110243429728207e-06,
+      "loss": 0.7431,
+      "step": 3190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.109615917991592e-06,
+      "loss": 0.7237,
+      "step": 3191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108988206680221e-06,
+      "loss": 0.5715,
+      "step": 3192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.108360295824576e-06,
+      "loss": 0.7514,
+      "step": 3193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107732185455155e-06,
+      "loss": 0.9217,
+      "step": 3194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.107103875602458e-06,
+      "loss": 0.767,
+      "step": 3195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.106475366297e-06,
+      "loss": 0.9384,
+      "step": 3196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105846657569298e-06,
+      "loss": 0.6414,
+      "step": 3197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.105217749449887e-06,
+      "loss": 0.6168,
+      "step": 3198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.10458864196931e-06,
+      "loss": 0.8721,
+      "step": 3199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103959335158116e-06,
+      "loss": 0.6869,
+      "step": 3200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.103329829046866e-06,
+      "loss": 0.9431,
+      "step": 3201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102700123666133e-06,
+      "loss": 0.8155,
+      "step": 3202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.102070219046495e-06,
+      "loss": 0.5555,
+      "step": 3203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.101440115218543e-06,
+      "loss": 0.6756,
+      "step": 3204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100809812212875e-06,
+      "loss": 0.6619,
+      "step": 3205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.100179310060104e-06,
+      "loss": 0.8969,
+      "step": 3206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.099548608790844e-06,
+      "loss": 0.447,
+      "step": 3207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.098917708435728e-06,
+      "loss": 0.9845,
+      "step": 3208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.098286609025392e-06,
+      "loss": 0.6559,
+      "step": 3209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097655310590484e-06,
+      "loss": 0.7518,
+      "step": 3210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.097023813161664e-06,
+      "loss": 0.7548,
+      "step": 3211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.096392116769597e-06,
+      "loss": 0.7453,
+      "step": 3212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09576022144496e-06,
+      "loss": 0.572,
+      "step": 3213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.09512812721844e-06,
+      "loss": 1.0035,
+      "step": 3214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.094495834120735e-06,
+      "loss": 0.8936,
+      "step": 3215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093863342182549e-06,
+      "loss": 1.0364,
+      "step": 3216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.093230651434596e-06,
+      "loss": 0.7761,
+      "step": 3217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.092597761907607e-06,
+      "loss": 0.6794,
+      "step": 3218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091964673632311e-06,
+      "loss": 0.6554,
+      "step": 3219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.091331386639456e-06,
+      "loss": 0.8672,
+      "step": 3220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090697900959796e-06,
+      "loss": 0.8933,
+      "step": 3221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.090064216624093e-06,
+      "loss": 0.6407,
+      "step": 3222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08943033366312e-06,
+      "loss": 0.8089,
+      "step": 3223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088796252107665e-06,
+      "loss": 0.7565,
+      "step": 3224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.088161971988517e-06,
+      "loss": 0.8028,
+      "step": 3225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.087527493336476e-06,
+      "loss": 0.8821,
+      "step": 3226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08689281618236e-06,
+      "loss": 0.7671,
+      "step": 3227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.086257940556988e-06,
+      "loss": 0.8147,
+      "step": 3228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08562286649119e-06,
+      "loss": 0.8633,
+      "step": 3229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08498759401581e-06,
+      "loss": 0.4978,
+      "step": 3230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.084352123161695e-06,
+      "loss": 1.0506,
+      "step": 3231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08371645395971e-06,
+      "loss": 0.8637,
+      "step": 3232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08308058644072e-06,
+      "loss": 0.8195,
+      "step": 3233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.082444520635608e-06,
+      "loss": 0.9969,
+      "step": 3234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.08180825657526e-06,
+      "loss": 0.8257,
+      "step": 3235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.081171794290577e-06,
+      "loss": 0.9503,
+      "step": 3236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.080535133812469e-06,
+      "loss": 0.7599,
+      "step": 3237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079898275171852e-06,
+      "loss": 0.7933,
+      "step": 3238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.079261218399652e-06,
+      "loss": 0.9208,
+      "step": 3239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.078623963526811e-06,
+      "loss": 0.7971,
+      "step": 3240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077986510584273e-06,
+      "loss": 1.0546,
+      "step": 3241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.077348859602992e-06,
+      "loss": 0.6949,
+      "step": 3242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076711010613939e-06,
+      "loss": 0.9059,
+      "step": 3243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.076072963648087e-06,
+      "loss": 0.9061,
+      "step": 3244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.075434718736423e-06,
+      "loss": 0.8478,
+      "step": 3245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074796275909941e-06,
+      "loss": 0.8017,
+      "step": 3246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.074157635199646e-06,
+      "loss": 0.7428,
+      "step": 3247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.073518796636548e-06,
+      "loss": 0.8774,
+      "step": 3248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.07287976025168e-06,
+      "loss": 0.8663,
+      "step": 3249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.072240526076067e-06,
+      "loss": 0.7855,
+      "step": 3250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.071601094140757e-06,
+      "loss": 0.8848,
+      "step": 3251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0709614644768e-06,
+      "loss": 0.8676,
+      "step": 3252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.07032163711526e-06,
+      "loss": 0.6965,
+      "step": 3253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069681612087206e-06,
+      "loss": 0.9211,
+      "step": 3254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.069041389423722e-06,
+      "loss": 1.052,
+      "step": 3255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0684009691559e-06,
+      "loss": 1.0053,
+      "step": 3256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067760351314838e-06,
+      "loss": 0.5751,
+      "step": 3257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.067119535931648e-06,
+      "loss": 0.8875,
+      "step": 3258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06647852303745e-06,
+      "loss": 0.8861,
+      "step": 3259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065837312663371e-06,
+      "loss": 0.8475,
+      "step": 3260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.065195904840553e-06,
+      "loss": 0.6673,
+      "step": 3261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.064554299600143e-06,
+      "loss": 0.9999,
+      "step": 3262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0639124969733e-06,
+      "loss": 0.9491,
+      "step": 3263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.06327049699119e-06,
+      "loss": 0.7793,
+      "step": 3264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.062628299684992e-06,
+      "loss": 0.829,
+      "step": 3265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061985905085891e-06,
+      "loss": 0.8775,
+      "step": 3266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.061343313225088e-06,
+      "loss": 1.0557,
+      "step": 3267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.060700524133786e-06,
+      "loss": 0.7681,
+      "step": 3268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0600575378432e-06,
+      "loss": 0.6597,
+      "step": 3269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.059414354384555e-06,
+      "loss": 0.9112,
+      "step": 3270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.058770973789087e-06,
+      "loss": 0.8737,
+      "step": 3271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.058127396088042e-06,
+      "loss": 0.8302,
+      "step": 3272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.057483621312671e-06,
+      "loss": 0.7533,
+      "step": 3273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05683964949424e-06,
+      "loss": 0.7793,
+      "step": 3274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05619548066402e-06,
+      "loss": 0.8859,
+      "step": 3275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.055551114853296e-06,
+      "loss": 0.7888,
+      "step": 3276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05490655209336e-06,
+      "loss": 0.8334,
+      "step": 3277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.054261792415511e-06,
+      "loss": 0.6842,
+      "step": 3278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.053616835851062e-06,
+      "loss": 0.9823,
+      "step": 3279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.052971682431335e-06,
+      "loss": 0.6686,
+      "step": 3280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.05232633218766e-06,
+      "loss": 1.0696,
+      "step": 3281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051680785151375e-06,
+      "loss": 0.7546,
+      "step": 3282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.051035041353833e-06,
+      "loss": 0.9357,
+      "step": 3283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.050389100826391e-06,
+      "loss": 0.9385,
+      "step": 3284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04974296360042e-06,
+      "loss": 0.6746,
+      "step": 3285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.049096629707292e-06,
+      "loss": 0.7239,
+      "step": 3286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.048450099178403e-06,
+      "loss": 0.4764,
+      "step": 3287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047803372045146e-06,
+      "loss": 0.805,
+      "step": 3288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.047156448338927e-06,
+      "loss": 0.9225,
+      "step": 3289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.046509328091167e-06,
+      "loss": 0.7956,
+      "step": 3290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045862011333286e-06,
+      "loss": 1.0268,
+      "step": 3291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.045214498096724e-06,
+      "loss": 0.5477,
+      "step": 3292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.044566788412926e-06,
+      "loss": 0.6321,
+      "step": 3293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043918882313344e-06,
+      "loss": 0.6913,
+      "step": 3294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.043270779829445e-06,
+      "loss": 0.8526,
+      "step": 3295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.0426224809927e-06,
+      "loss": 0.7941,
+      "step": 3296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.041973985834595e-06,
+      "loss": 0.7575,
+      "step": 3297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04132529438662e-06,
+      "loss": 0.9638,
+      "step": 3298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.04067640668028e-06,
+      "loss": 0.7529,
+      "step": 3299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.040027322747085e-06,
+      "loss": 0.7418,
+      "step": 3300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.039378042618556e-06,
+      "loss": 0.5632,
+      "step": 3301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038728566326225e-06,
+      "loss": 0.9373,
+      "step": 3302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.038078893901634e-06,
+      "loss": 0.8711,
+      "step": 3303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.03742902537633e-06,
+      "loss": 0.6104,
+      "step": 3304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.036778960781874e-06,
+      "loss": 0.974,
+      "step": 3305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.036128700149836e-06,
+      "loss": 0.7658,
+      "step": 3306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.035478243511791e-06,
+      "loss": 0.8356,
+      "step": 3307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03482759089933e-06,
+      "loss": 0.8863,
+      "step": 3308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.034176742344049e-06,
+      "loss": 0.8405,
+      "step": 3309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.033525697877558e-06,
+      "loss": 0.6875,
+      "step": 3310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03287445753147e-06,
+      "loss": 0.5915,
+      "step": 3311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.032223021337415e-06,
+      "loss": 0.921,
+      "step": 3312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.031571389327023e-06,
+      "loss": 0.8315,
+      "step": 3313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.030919561531943e-06,
+      "loss": 1.2107,
+      "step": 3314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.03026753798383e-06,
+      "loss": 0.6734,
+      "step": 3315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.029615318714348e-06,
+      "loss": 0.6276,
+      "step": 3316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.028962903755168e-06,
+      "loss": 0.8463,
+      "step": 3317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.028310293137976e-06,
+      "loss": 0.8414,
+      "step": 3318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027657486894463e-06,
+      "loss": 0.6682,
+      "step": 3319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.027004485056332e-06,
+      "loss": 0.9215,
+      "step": 3320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.026351287655294e-06,
+      "loss": 0.8121,
+      "step": 3321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025697894723071e-06,
+      "loss": 0.8446,
+      "step": 3322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.025044306291392e-06,
+      "loss": 0.8096,
+      "step": 3323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.024390522391999e-06,
+      "loss": 0.8548,
+      "step": 3324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023736543056641e-06,
+      "loss": 0.948,
+      "step": 3325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.023082368317077e-06,
+      "loss": 0.6648,
+      "step": 3326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.022427998205076e-06,
+      "loss": 0.747,
+      "step": 3327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021773432752416e-06,
+      "loss": 1.048,
+      "step": 3328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.021118671990883e-06,
+      "loss": 0.7655,
+      "step": 3329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.020463715952276e-06,
+      "loss": 0.7496,
+      "step": 3330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.0198085646684e-06,
+      "loss": 0.7979,
+      "step": 3331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.019153218171072e-06,
+      "loss": 0.9031,
+      "step": 3332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01849767649212e-06,
+      "loss": 0.9687,
+      "step": 3333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017841939663375e-06,
+      "loss": 0.7507,
+      "step": 3334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.017186007716681e-06,
+      "loss": 0.9667,
+      "step": 3335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.016529880683896e-06,
+      "loss": 0.7569,
+      "step": 3336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01587355859688e-06,
+      "loss": 0.9093,
+      "step": 3337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.015217041487507e-06,
+      "loss": 0.821,
+      "step": 3338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.014560329387661e-06,
+      "loss": 1.0476,
+      "step": 3339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.013903422329232e-06,
+      "loss": 0.7869,
+      "step": 3340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01324632034412e-06,
+      "loss": 0.8826,
+      "step": 3341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.012589023464238e-06,
+      "loss": 0.8571,
+      "step": 3342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011931531721508e-06,
+      "loss": 0.744,
+      "step": 3343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.011273845147854e-06,
+      "loss": 0.7544,
+      "step": 3344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.01061596377522e-06,
+      "loss": 1.0801,
+      "step": 3345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009957887635552e-06,
+      "loss": 0.8659,
+      "step": 3346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.009299616760809e-06,
+      "loss": 0.6559,
+      "step": 3347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00864115118296e-06,
+      "loss": 0.6628,
+      "step": 3348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00798249093398e-06,
+      "loss": 0.7447,
+      "step": 3349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.007323636045857e-06,
+      "loss": 0.8364,
+      "step": 3350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.006664586550585e-06,
+      "loss": 0.5874,
+      "step": 3351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00600534248017e-06,
+      "loss": 0.8663,
+      "step": 3352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.005345903866627e-06,
+      "loss": 0.6847,
+      "step": 3353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00468627074198e-06,
+      "loss": 0.8412,
+      "step": 3354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.004026443138265e-06,
+      "loss": 0.9472,
+      "step": 3355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00336642108752e-06,
+      "loss": 0.7578,
+      "step": 3356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002706204621802e-06,
+      "loss": 0.7907,
+      "step": 3357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.002045793773173e-06,
+      "loss": 1.0478,
+      "step": 3358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.001385188573702e-06,
+      "loss": 0.8244,
+      "step": 3359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00072438905547e-06,
+      "loss": 0.8416,
+      "step": 3360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.00006339525057e-06,
+      "loss": 0.7387,
+      "step": 3361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.999402207191099e-06,
+      "loss": 0.7734,
+      "step": 3362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998740824909169e-06,
+      "loss": 0.5308,
+      "step": 3363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.998079248436895e-06,
+      "loss": 0.6372,
+      "step": 3364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.997417477806408e-06,
+      "loss": 0.3033,
+      "step": 3365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996755513049844e-06,
+      "loss": 0.8638,
+      "step": 3366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.996093354199349e-06,
+      "loss": 0.8406,
+      "step": 3367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.995431001287082e-06,
+      "loss": 0.9167,
+      "step": 3368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994768454345207e-06,
+      "loss": 0.58,
+      "step": 3369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.994105713405901e-06,
+      "loss": 0.8956,
+      "step": 3370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.993442778501344e-06,
+      "loss": 0.8994,
+      "step": 3371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.992779649663736e-06,
+      "loss": 0.8621,
+      "step": 3372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.992116326925277e-06,
+      "loss": 0.7152,
+      "step": 3373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.991452810318183e-06,
+      "loss": 0.5978,
+      "step": 3374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.99078909987467e-06,
+      "loss": 0.546,
+      "step": 3375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.990125195626974e-06,
+      "loss": 0.5605,
+      "step": 3376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.989461097607338e-06,
+      "loss": 0.7098,
+      "step": 3377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988796805848008e-06,
+      "loss": 0.8251,
+      "step": 3378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.988132320381246e-06,
+      "loss": 0.7893,
+      "step": 3379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.987467641239323e-06,
+      "loss": 0.6739,
+      "step": 3380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.986802768454514e-06,
+      "loss": 0.7816,
+      "step": 3381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98613770205911e-06,
+      "loss": 0.9512,
+      "step": 3382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.985472442085408e-06,
+      "loss": 0.5766,
+      "step": 3383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984806988565716e-06,
+      "loss": 0.8549,
+      "step": 3384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.984141341532346e-06,
+      "loss": 0.6501,
+      "step": 3385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98347550101763e-06,
+      "loss": 0.5892,
+      "step": 3386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982809467053897e-06,
+      "loss": 0.9995,
+      "step": 3387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.982143239673496e-06,
+      "loss": 0.921,
+      "step": 3388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.981476818908778e-06,
+      "loss": 0.8221,
+      "step": 3389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.98081020479211e-06,
+      "loss": 0.9105,
+      "step": 3390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.980143397355861e-06,
+      "loss": 0.9486,
+      "step": 3391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.979476396632414e-06,
+      "loss": 0.5889,
+      "step": 3392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978809202654161e-06,
+      "loss": 0.8425,
+      "step": 3393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.978141815453503e-06,
+      "loss": 0.8795,
+      "step": 3394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.977474235062851e-06,
+      "loss": 0.7285,
+      "step": 3395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976806461514622e-06,
+      "loss": 0.6873,
+      "step": 3396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.976138494841246e-06,
+      "loss": 0.6463,
+      "step": 3397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.975470335075164e-06,
+      "loss": 0.8925,
+      "step": 3398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97480198224882e-06,
+      "loss": 0.687,
+      "step": 3399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.974133436394674e-06,
+      "loss": 0.8247,
+      "step": 3400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.973464697545191e-06,
+      "loss": 0.6317,
+      "step": 3401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.972795765732847e-06,
+      "loss": 0.9776,
+      "step": 3402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.972126640990129e-06,
+      "loss": 0.7185,
+      "step": 3403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.971457323349528e-06,
+      "loss": 0.6252,
+      "step": 3404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.970787812843552e-06,
+      "loss": 1.0264,
+      "step": 3405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.97011810950471e-06,
+      "loss": 0.7385,
+      "step": 3406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.969448213365528e-06,
+      "loss": 0.8736,
+      "step": 3407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96877812445854e-06,
+      "loss": 0.5398,
+      "step": 3408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96810784281628e-06,
+      "loss": 0.616,
+      "step": 3409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.967437368471307e-06,
+      "loss": 0.8205,
+      "step": 3410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.966766701456177e-06,
+      "loss": 0.8376,
+      "step": 3411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.966095841803459e-06,
+      "loss": 0.5937,
+      "step": 3412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.965424789545735e-06,
+      "loss": 0.8782,
+      "step": 3413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96475354471559e-06,
+      "loss": 0.7871,
+      "step": 3414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.964082107345622e-06,
+      "loss": 0.7061,
+      "step": 3415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96341047746844e-06,
+      "loss": 0.7667,
+      "step": 3416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.96273865511666e-06,
+      "loss": 0.9042,
+      "step": 3417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.962066640322905e-06,
+      "loss": 0.8267,
+      "step": 3418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.961394433119812e-06,
+      "loss": 0.6974,
+      "step": 3419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960722033540025e-06,
+      "loss": 0.8381,
+      "step": 3420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.960049441616199e-06,
+      "loss": 1.1,
+      "step": 3421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.959376657380993e-06,
+      "loss": 0.7616,
+      "step": 3422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.958703680867083e-06,
+      "loss": 0.7185,
+      "step": 3423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.95803051210715e-06,
+      "loss": 0.8652,
+      "step": 3424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.957357151133886e-06,
+      "loss": 0.7105,
+      "step": 3425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956683597979987e-06,
+      "loss": 0.8968,
+      "step": 3426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.956009852678168e-06,
+      "loss": 0.568,
+      "step": 3427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.955335915261146e-06,
+      "loss": 0.6856,
+      "step": 3428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.954661785761648e-06,
+      "loss": 0.7244,
+      "step": 3429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953987464212411e-06,
+      "loss": 1.0419,
+      "step": 3430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.953312950646185e-06,
+      "loss": 0.9157,
+      "step": 3431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.952638245095726e-06,
+      "loss": 0.9837,
+      "step": 3432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951963347593797e-06,
+      "loss": 0.9067,
+      "step": 3433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.951288258173176e-06,
+      "loss": 1.0841,
+      "step": 3434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.950612976866643e-06,
+      "loss": 0.8252,
+      "step": 3435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.949937503706999e-06,
+      "loss": 0.8185,
+      "step": 3436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94926183872704e-06,
+      "loss": 0.8552,
+      "step": 3437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94858598195958e-06,
+      "loss": 0.9002,
+      "step": 3438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947909933437442e-06,
+      "loss": 0.5804,
+      "step": 3439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.947233693193456e-06,
+      "loss": 0.9939,
+      "step": 3440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94655726126046e-06,
+      "loss": 0.7831,
+      "step": 3441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.94588063767131e-06,
+      "loss": 0.6817,
+      "step": 3442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.945203822458856e-06,
+      "loss": 0.6625,
+      "step": 3443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.944526815655974e-06,
+      "loss": 0.8493,
+      "step": 3444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943849617295536e-06,
+      "loss": 1.0122,
+      "step": 3445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.943172227410431e-06,
+      "loss": 0.874,
+      "step": 3446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.942494646033555e-06,
+      "loss": 0.5947,
+      "step": 3447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941816873197812e-06,
+      "loss": 0.831,
+      "step": 3448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.941138908936118e-06,
+      "loss": 0.68,
+      "step": 3449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.940460753281397e-06,
+      "loss": 0.7856,
+      "step": 3450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939782406266581e-06,
+      "loss": 0.8292,
+      "step": 3451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.939103867924614e-06,
+      "loss": 0.5291,
+      "step": 3452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.938425138288445e-06,
+      "loss": 0.8091,
+      "step": 3453
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.93774621739104e-06,
+      "loss": 0.7559,
+      "step": 3454
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.937067105265363e-06,
+      "loss": 0.679,
+      "step": 3455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.936387801944397e-06,
+      "loss": 0.7434,
+      "step": 3456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935708307461134e-06,
+      "loss": 0.747,
+      "step": 3457
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.935028621848566e-06,
+      "loss": 0.63,
+      "step": 3458
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.934348745139703e-06,
+      "loss": 1.0661,
+      "step": 3459
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.933668677367563e-06,
+      "loss": 0.8328,
+      "step": 3460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.932988418565174e-06,
+      "loss": 0.7653,
+      "step": 3461
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.932307968765566e-06,
+      "loss": 0.939,
+      "step": 3462
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.931627328001787e-06,
+      "loss": 0.8501,
+      "step": 3463
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930946496306889e-06,
+      "loss": 0.941,
+      "step": 3464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.930265473713939e-06,
+      "loss": 0.787,
+      "step": 3465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.929584260256004e-06,
+      "loss": 0.666,
+      "step": 3466
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928902855966169e-06,
+      "loss": 0.7742,
+      "step": 3467
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.928221260877526e-06,
+      "loss": 0.6592,
+      "step": 3468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92753947502317e-06,
+      "loss": 0.7052,
+      "step": 3469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926857498436216e-06,
+      "loss": 0.9025,
+      "step": 3470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.926175331149783e-06,
+      "loss": 0.5727,
+      "step": 3471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.925492973196993e-06,
+      "loss": 0.4807,
+      "step": 3472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924810424610988e-06,
+      "loss": 0.9234,
+      "step": 3473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.924127685424914e-06,
+      "loss": 0.7458,
+      "step": 3474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.923444755671926e-06,
+      "loss": 0.745,
+      "step": 3475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92276163538519e-06,
+      "loss": 0.6697,
+      "step": 3476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.92207832459788e-06,
+      "loss": 0.8614,
+      "step": 3477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.921394823343178e-06,
+      "loss": 1.0822,
+      "step": 3478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920711131654278e-06,
+      "loss": 0.8805,
+      "step": 3479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.920027249564382e-06,
+      "loss": 0.6023,
+      "step": 3480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.9193431771067e-06,
+      "loss": 0.7928,
+      "step": 3481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.918658914314458e-06,
+      "loss": 0.9441,
+      "step": 3482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917974461220877e-06,
+      "loss": 0.7193,
+      "step": 3483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.917289817859203e-06,
+      "loss": 0.7078,
+      "step": 3484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91660498426268e-06,
+      "loss": 0.6815,
+      "step": 3485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91591996046457e-06,
+      "loss": 0.7402,
+      "step": 3486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.915234746498134e-06,
+      "loss": 0.9516,
+      "step": 3487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.914549342396652e-06,
+      "loss": 0.6275,
+      "step": 3488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.91386374819341e-06,
+      "loss": 0.774,
+      "step": 3489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.913177963921698e-06,
+      "loss": 0.7614,
+      "step": 3490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.912491989614823e-06,
+      "loss": 0.8392,
+      "step": 3491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911805825306097e-06,
+      "loss": 0.587,
+      "step": 3492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.911119471028842e-06,
+      "loss": 0.7842,
+      "step": 3493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.910432926816391e-06,
+      "loss": 0.8194,
+      "step": 3494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909746192702082e-06,
+      "loss": 0.7597,
+      "step": 3495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.909059268719267e-06,
+      "loss": 0.8685,
+      "step": 3496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.908372154901302e-06,
+      "loss": 0.6077,
+      "step": 3497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90768485128156e-06,
+      "loss": 0.8347,
+      "step": 3498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.906997357893412e-06,
+      "loss": 1.0905,
+      "step": 3499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90630967477025e-06,
+      "loss": 0.6488,
+      "step": 3500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.905621801945467e-06,
+      "loss": 0.8397,
+      "step": 3501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90493373945247e-06,
+      "loss": 0.7683,
+      "step": 3502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.904245487324673e-06,
+      "loss": 0.9762,
+      "step": 3503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.903557045595499e-06,
+      "loss": 0.9147,
+      "step": 3504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.90286841429838e-06,
+      "loss": 0.7967,
+      "step": 3505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.902179593466758e-06,
+      "loss": 0.8561,
+      "step": 3506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.901490583134087e-06,
+      "loss": 0.8693,
+      "step": 3507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900801383333822e-06,
+      "loss": 0.9927,
+      "step": 3508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.900111994099437e-06,
+      "loss": 0.7564,
+      "step": 3509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.899422415464409e-06,
+      "loss": 0.8694,
+      "step": 3510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898732647462226e-06,
+      "loss": 0.8373,
+      "step": 3511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.898042690126387e-06,
+      "loss": 0.6385,
+      "step": 3512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.897352543490396e-06,
+      "loss": 0.61,
+      "step": 3513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.896662207587767e-06,
+      "loss": 0.5654,
+      "step": 3514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89597168245203e-06,
+      "loss": 0.8764,
+      "step": 3515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.895280968116713e-06,
+      "loss": 0.7047,
+      "step": 3516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.894590064615363e-06,
+      "loss": 0.5908,
+      "step": 3517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89389897198153e-06,
+      "loss": 0.7922,
+      "step": 3518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.893207690248776e-06,
+      "loss": 0.6506,
+      "step": 3519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.892516219450674e-06,
+      "loss": 0.7667,
+      "step": 3520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891824559620801e-06,
+      "loss": 0.723,
+      "step": 3521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.891132710792747e-06,
+      "loss": 1.0714,
+      "step": 3522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.89044067300011e-06,
+      "loss": 0.7069,
+      "step": 3523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.889748446276497e-06,
+      "loss": 1.043,
+      "step": 3524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.889056030655526e-06,
+      "loss": 0.9177,
+      "step": 3525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88836342617082e-06,
+      "loss": 0.7063,
+      "step": 3526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.887670632856016e-06,
+      "loss": 0.7205,
+      "step": 3527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88697765074476e-06,
+      "loss": 0.5665,
+      "step": 3528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8862844798707e-06,
+      "loss": 0.7842,
+      "step": 3529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8855911202675e-06,
+      "loss": 0.5873,
+      "step": 3530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884897571968835e-06,
+      "loss": 0.6037,
+      "step": 3531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.884203835008383e-06,
+      "loss": 0.8823,
+      "step": 3532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.883509909419831e-06,
+      "loss": 0.779,
+      "step": 3533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882815795236886e-06,
+      "loss": 0.72,
+      "step": 3534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.882121492493248e-06,
+      "loss": 1.0199,
+      "step": 3535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88142700122264e-06,
+      "loss": 0.5812,
+      "step": 3536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.880732321458785e-06,
+      "loss": 0.5918,
+      "step": 3537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.88003745323542e-06,
+      "loss": 0.8134,
+      "step": 3538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87934239658629e-06,
+      "loss": 0.7708,
+      "step": 3539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.878647151545148e-06,
+      "loss": 1.2344,
+      "step": 3540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.877951718145757e-06,
+      "loss": 1.0034,
+      "step": 3541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87725609642189e-06,
+      "loss": 0.9421,
+      "step": 3542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87656028640733e-06,
+      "loss": 1.0379,
+      "step": 3543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875864288135864e-06,
+      "loss": 1.0368,
+      "step": 3544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.875168101641294e-06,
+      "loss": 0.7116,
+      "step": 3545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.87447172695743e-06,
+      "loss": 0.8117,
+      "step": 3546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873775164118086e-06,
+      "loss": 0.7391,
+      "step": 3547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.873078413157094e-06,
+      "loss": 0.6867,
+      "step": 3548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.872381474108284e-06,
+      "loss": 0.8448,
+      "step": 3549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.871684347005508e-06,
+      "loss": 0.8203,
+      "step": 3550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870987031882616e-06,
+      "loss": 0.6987,
+      "step": 3551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.870289528773474e-06,
+      "loss": 0.6772,
+      "step": 3552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.869591837711952e-06,
+      "loss": 0.807,
+      "step": 3553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868893958731937e-06,
+      "loss": 0.6977,
+      "step": 3554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.868195891867315e-06,
+      "loss": 0.6965,
+      "step": 3555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.867497637151987e-06,
+      "loss": 0.9165,
+      "step": 3556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866799194619864e-06,
+      "loss": 0.8168,
+      "step": 3557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.866100564304862e-06,
+      "loss": 0.8998,
+      "step": 3558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.865401746240913e-06,
+      "loss": 0.8156,
+      "step": 3559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864702740461947e-06,
+      "loss": 0.9343,
+      "step": 3560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.864003547001916e-06,
+      "loss": 0.8518,
+      "step": 3561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.863304165894774e-06,
+      "loss": 0.6846,
+      "step": 3562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.862604597174477e-06,
+      "loss": 0.6905,
+      "step": 3563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86190484087501e-06,
+      "loss": 0.953,
+      "step": 3564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.861204897030346e-06,
+      "loss": 0.807,
+      "step": 3565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.86050476567448e-06,
+      "loss": 0.7143,
+      "step": 3566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859804446841413e-06,
+      "loss": 0.7354,
+      "step": 3567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.859103940565154e-06,
+      "loss": 0.6392,
+      "step": 3568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85840324687972e-06,
+      "loss": 0.7781,
+      "step": 3569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.857702365819141e-06,
+      "loss": 0.9718,
+      "step": 3570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.857001297417452e-06,
+      "loss": 0.7385,
+      "step": 3571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.8563000417087e-06,
+      "loss": 0.8832,
+      "step": 3572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85559859872694e-06,
+      "loss": 0.5637,
+      "step": 3573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.854896968506233e-06,
+      "loss": 0.6334,
+      "step": 3574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.854195151080656e-06,
+      "loss": 0.984,
+      "step": 3575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.85349314648429e-06,
+      "loss": 0.6126,
+      "step": 3576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852790954751229e-06,
+      "loss": 0.8921,
+      "step": 3577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.852088575915568e-06,
+      "loss": 0.7543,
+      "step": 3578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.851386010011421e-06,
+      "loss": 0.536,
+      "step": 3579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.850683257072904e-06,
+      "loss": 0.6033,
+      "step": 3580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.849980317134145e-06,
+      "loss": 0.7212,
+      "step": 3581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.849277190229284e-06,
+      "loss": 0.87,
+      "step": 3582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84857387639246e-06,
+      "loss": 0.7097,
+      "step": 3583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.847870375657837e-06,
+      "loss": 0.7508,
+      "step": 3584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.847166688059572e-06,
+      "loss": 0.7472,
+      "step": 3585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.846462813631841e-06,
+      "loss": 0.7666,
+      "step": 3586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845758752408826e-06,
+      "loss": 0.7138,
+      "step": 3587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.845054504424717e-06,
+      "loss": 0.6643,
+      "step": 3588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.844350069713716e-06,
+      "loss": 0.681,
+      "step": 3589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84364544831003e-06,
+      "loss": 0.6286,
+      "step": 3590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84294064024788e-06,
+      "loss": 0.7362,
+      "step": 3591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.842235645561492e-06,
+      "loss": 0.7976,
+      "step": 3592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.841530464285105e-06,
+      "loss": 0.5242,
+      "step": 3593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84082509645296e-06,
+      "loss": 0.7862,
+      "step": 3594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.840119542099315e-06,
+      "loss": 0.8005,
+      "step": 3595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.839413801258434e-06,
+      "loss": 0.9428,
+      "step": 3596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838707873964587e-06,
+      "loss": 0.4965,
+      "step": 3597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.838001760252059e-06,
+      "loss": 0.9065,
+      "step": 3598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.83729546015514e-06,
+      "loss": 0.768,
+      "step": 3599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.836588973708129e-06,
+      "loss": 0.8205,
+      "step": 3600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835882300945336e-06,
+      "loss": 0.6039,
+      "step": 3601
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.835175441901078e-06,
+      "loss": 0.9413,
+      "step": 3602
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.834468396609683e-06,
+      "loss": 0.8135,
+      "step": 3603
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833761165105488e-06,
+      "loss": 0.9643,
+      "step": 3604
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.833053747422837e-06,
+      "loss": 0.7238,
+      "step": 3605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.832346143596082e-06,
+      "loss": 0.8512,
+      "step": 3606
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83163835365959e-06,
+      "loss": 0.6574,
+      "step": 3607
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.830930377647733e-06,
+      "loss": 0.9369,
+      "step": 3608
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.83022221559489e-06,
+      "loss": 0.7192,
+      "step": 3609
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.829513867535454e-06,
+      "loss": 0.6346,
+      "step": 3610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828805333503822e-06,
+      "loss": 0.8723,
+      "step": 3611
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.828096613534405e-06,
+      "loss": 0.6541,
+      "step": 3612
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.827387707661617e-06,
+      "loss": 0.7863,
+      "step": 3613
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.826678615919889e-06,
+      "loss": 0.7002,
+      "step": 3614
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825969338343652e-06,
+      "loss": 0.7666,
+      "step": 3615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.825259874967353e-06,
+      "loss": 0.9006,
+      "step": 3616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.824550225825445e-06,
+      "loss": 0.6919,
+      "step": 3617
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82384039095239e-06,
+      "loss": 0.6791,
+      "step": 3618
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.823130370382662e-06,
+      "loss": 0.6641,
+      "step": 3619
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82242016415074e-06,
+      "loss": 0.8727,
+      "step": 3620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.821709772291113e-06,
+      "loss": 0.5844,
+      "step": 3621
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.82099919483828e-06,
+      "loss": 0.6591,
+      "step": 3622
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.820288431826751e-06,
+      "loss": 0.5858,
+      "step": 3623
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.819577483291038e-06,
+      "loss": 0.9407,
+      "step": 3624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81886634926567e-06,
+      "loss": 0.7105,
+      "step": 3625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.818155029785183e-06,
+      "loss": 0.8361,
+      "step": 3626
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.817443524884119e-06,
+      "loss": 0.8648,
+      "step": 3627
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81673183459703e-06,
+      "loss": 0.8624,
+      "step": 3628
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.816019958958477e-06,
+      "loss": 0.9501,
+      "step": 3629
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.815307898003032e-06,
+      "loss": 0.8722,
+      "step": 3630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.814595651765276e-06,
+      "loss": 0.6757,
+      "step": 3631
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.813883220279797e-06,
+      "loss": 0.8151,
+      "step": 3632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81317060358119e-06,
+      "loss": 0.7776,
+      "step": 3633
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.812457801704067e-06,
+      "loss": 0.8323,
+      "step": 3634
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811744814683039e-06,
+      "loss": 0.9094,
+      "step": 3635
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.811031642552732e-06,
+      "loss": 0.7697,
+      "step": 3636
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.81031828534778e-06,
+      "loss": 0.7022,
+      "step": 3637
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.809604743102828e-06,
+      "loss": 0.7888,
+      "step": 3638
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808891015852523e-06,
+      "loss": 0.8378,
+      "step": 3639
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.808177103631529e-06,
+      "loss": 0.7302,
+      "step": 3640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.807463006474514e-06,
+      "loss": 0.8384,
+      "step": 3641
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806748724416156e-06,
+      "loss": 0.758,
+      "step": 3642
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.806034257491148e-06,
+      "loss": 0.8887,
+      "step": 3643
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.805319605734178e-06,
+      "loss": 0.7686,
+      "step": 3644
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.804604769179958e-06,
+      "loss": 0.4305,
+      "step": 3645
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803889747863201e-06,
+      "loss": 0.9857,
+      "step": 3646
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.803174541818629e-06,
+      "loss": 0.5489,
+      "step": 3647
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.802459151080976e-06,
+      "loss": 0.7684,
+      "step": 3648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.80174357568498e-06,
+      "loss": 0.9301,
+      "step": 3649
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.801027815665398e-06,
+      "loss": 1.1709,
+      "step": 3650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.800311871056983e-06,
+      "loss": 0.8655,
+      "step": 3651
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.799595741894508e-06,
+      "loss": 0.7452,
+      "step": 3652
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798879428212748e-06,
+      "loss": 0.8015,
+      "step": 3653
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.798162930046488e-06,
+      "loss": 0.647,
+      "step": 3654
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.797446247430525e-06,
+      "loss": 0.7784,
+      "step": 3655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796729380399665e-06,
+      "loss": 0.8288,
+      "step": 3656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.796012328988716e-06,
+      "loss": 0.7352,
+      "step": 3657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.795295093232505e-06,
+      "loss": 1.1749,
+      "step": 3658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.794577673165861e-06,
+      "loss": 0.9927,
+      "step": 3659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793860068823624e-06,
+      "loss": 0.7668,
+      "step": 3660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.793142280240643e-06,
+      "loss": 0.8439,
+      "step": 3661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792424307451775e-06,
+      "loss": 0.8278,
+      "step": 3662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.791706150491887e-06,
+      "loss": 0.625,
+      "step": 3663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790987809395857e-06,
+      "loss": 0.6971,
+      "step": 3664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790269284198568e-06,
+      "loss": 0.7124,
+      "step": 3665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.789550574934914e-06,
+      "loss": 0.6923,
+      "step": 3666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788831681639798e-06,
+      "loss": 0.9911,
+      "step": 3667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.788112604348129e-06,
+      "loss": 0.7725,
+      "step": 3668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78739334309483e-06,
+      "loss": 0.743,
+      "step": 3669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.786673897914831e-06,
+      "loss": 0.6978,
+      "step": 3670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785954268843066e-06,
+      "loss": 0.6862,
+      "step": 3671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78523445591449e-06,
+      "loss": 0.8729,
+      "step": 3672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78451445916405e-06,
+      "loss": 0.8451,
+      "step": 3673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783794278626718e-06,
+      "loss": 0.6909,
+      "step": 3674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783073914337465e-06,
+      "loss": 0.7156,
+      "step": 3675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.782353366331276e-06,
+      "loss": 1.0054,
+      "step": 3676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.78163263464314e-06,
+      "loss": 0.7226,
+      "step": 3677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780911719308059e-06,
+      "loss": 1.091,
+      "step": 3678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.780190620361044e-06,
+      "loss": 0.7533,
+      "step": 3679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779469337837112e-06,
+      "loss": 0.6699,
+      "step": 3680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778747871771293e-06,
+      "loss": 0.8616,
+      "step": 3681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.778026222198618e-06,
+      "loss": 0.8501,
+      "step": 3682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77730438915414e-06,
+      "loss": 0.8045,
+      "step": 3683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.776582372672908e-06,
+      "loss": 0.7506,
+      "step": 3684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.775860172789985e-06,
+      "loss": 0.6769,
+      "step": 3685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.775137789540447e-06,
+      "loss": 0.7837,
+      "step": 3686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.774415222959371e-06,
+      "loss": 0.7369,
+      "step": 3687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77369247308185e-06,
+      "loss": 0.8954,
+      "step": 3688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772969539942981e-06,
+      "loss": 0.5817,
+      "step": 3689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.772246423577871e-06,
+      "loss": 0.9358,
+      "step": 3690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77152312402164e-06,
+      "loss": 0.6203,
+      "step": 3691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770799641309408e-06,
+      "loss": 0.8863,
+      "step": 3692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770075975476316e-06,
+      "loss": 0.8115,
+      "step": 3693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7693521265575e-06,
+      "loss": 1.1272,
+      "step": 3694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76862809458812e-06,
+      "loss": 0.8836,
+      "step": 3695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76790387960333e-06,
+      "loss": 0.742,
+      "step": 3696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.767179481638303e-06,
+      "loss": 0.8903,
+      "step": 3697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766454900728221e-06,
+      "loss": 0.7385,
+      "step": 3698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.765730136908266e-06,
+      "loss": 0.9829,
+      "step": 3699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.76500519021364e-06,
+      "loss": 0.9556,
+      "step": 3700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764280060679543e-06,
+      "loss": 0.8108,
+      "step": 3701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.763554748341194e-06,
+      "loss": 0.9563,
+      "step": 3702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762829253233813e-06,
+      "loss": 0.8434,
+      "step": 3703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762103575392634e-06,
+      "loss": 0.47,
+      "step": 3704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7613777148529e-06,
+      "loss": 0.6963,
+      "step": 3705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760651671649854e-06,
+      "loss": 0.8227,
+      "step": 3706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759925445818762e-06,
+      "loss": 0.7831,
+      "step": 3707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.759199037394888e-06,
+      "loss": 0.9558,
+      "step": 3708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758472446413508e-06,
+      "loss": 0.745,
+      "step": 3709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757745672909911e-06,
+      "loss": 0.9376,
+      "step": 3710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.757018716919388e-06,
+      "loss": 0.7164,
+      "step": 3711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.756291578477243e-06,
+      "loss": 0.6363,
+      "step": 3712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755564257618787e-06,
+      "loss": 0.6835,
+      "step": 3713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754836754379342e-06,
+      "loss": 0.6555,
+      "step": 3714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.754109068794237e-06,
+      "loss": 0.7462,
+      "step": 3715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.753381200898812e-06,
+      "loss": 0.995,
+      "step": 3716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.752653150728412e-06,
+      "loss": 0.8769,
+      "step": 3717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751924918318394e-06,
+      "loss": 0.5681,
+      "step": 3718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751196503704123e-06,
+      "loss": 0.8188,
+      "step": 3719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.750467906920974e-06,
+      "loss": 0.8172,
+      "step": 3720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749739128004329e-06,
+      "loss": 0.8221,
+      "step": 3721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74901016698958e-06,
+      "loss": 0.6874,
+      "step": 3722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.748281023912127e-06,
+      "loss": 0.6388,
+      "step": 3723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747551698807379e-06,
+      "loss": 0.6153,
+      "step": 3724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746822191710755e-06,
+      "loss": 0.6729,
+      "step": 3725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.746092502657681e-06,
+      "loss": 0.8655,
+      "step": 3726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745362631683594e-06,
+      "loss": 0.8347,
+      "step": 3727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.744632578823936e-06,
+      "loss": 0.8483,
+      "step": 3728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.743902344114163e-06,
+      "loss": 0.7891,
+      "step": 3729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.743171927589738e-06,
+      "loss": 0.6123,
+      "step": 3730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.74244132928613e-06,
+      "loss": 0.773,
+      "step": 3731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.741710549238818e-06,
+      "loss": 0.8434,
+      "step": 3732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740979587483296e-06,
+      "loss": 0.9018,
+      "step": 3733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.740248444055055e-06,
+      "loss": 0.8251,
+      "step": 3734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.739517118989606e-06,
+      "loss": 0.9297,
+      "step": 3735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.738785612322463e-06,
+      "loss": 0.8221,
+      "step": 3736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.738053924089149e-06,
+      "loss": 0.706,
+      "step": 3737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.737322054325197e-06,
+      "loss": 0.6082,
+      "step": 3738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73659000306615e-06,
+      "loss": 1.0135,
+      "step": 3739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735857770347558e-06,
+      "loss": 0.5797,
+      "step": 3740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.735125356204982e-06,
+      "loss": 0.8519,
+      "step": 3741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.734392760673985e-06,
+      "loss": 0.8077,
+      "step": 3742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.73365998379015e-06,
+      "loss": 0.6406,
+      "step": 3743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732927025589058e-06,
+      "loss": 0.9645,
+      "step": 3744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.732193886106308e-06,
+      "loss": 0.6558,
+      "step": 3745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.731460565377499e-06,
+      "loss": 1.204,
+      "step": 3746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.730727063438246e-06,
+      "loss": 0.9336,
+      "step": 3747
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.729993380324168e-06,
+      "loss": 0.7056,
+      "step": 3748
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.729259516070899e-06,
+      "loss": 0.7348,
+      "step": 3749
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.72852547071407e-06,
+      "loss": 0.6993,
+      "step": 3750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727791244289335e-06,
+      "loss": 0.8755,
+      "step": 3751
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.727056836832347e-06,
+      "loss": 1.0313,
+      "step": 3752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.726322248378775e-06,
+      "loss": 0.9901,
+      "step": 3753
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725587478964286e-06,
+      "loss": 0.742,
+      "step": 3754
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724852528624567e-06,
+      "loss": 0.6191,
+      "step": 3755
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.724117397395309e-06,
+      "loss": 0.909,
+      "step": 3756
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723382085312212e-06,
+      "loss": 0.8252,
+      "step": 3757
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.722646592410984e-06,
+      "loss": 0.7638,
+      "step": 3758
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721910918727342e-06,
+      "loss": 0.8276,
+      "step": 3759
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721175064297014e-06,
+      "loss": 0.7421,
+      "step": 3760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.720439029155733e-06,
+      "loss": 0.9497,
+      "step": 3761
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719702813339248e-06,
+      "loss": 0.7387,
+      "step": 3762
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718966416883307e-06,
+      "loss": 0.7317,
+      "step": 3763
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.718229839823672e-06,
+      "loss": 0.8058,
+      "step": 3764
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.717493082196114e-06,
+      "loss": 0.6606,
+      "step": 3765
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716756144036413e-06,
+      "loss": 0.8252,
+      "step": 3766
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.716019025380356e-06,
+      "loss": 0.8673,
+      "step": 3767
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715281726263741e-06,
+      "loss": 0.865,
+      "step": 3768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.714544246722369e-06,
+      "loss": 0.5946,
+      "step": 3769
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71380658679206e-06,
+      "loss": 0.8007,
+      "step": 3770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.713068746508633e-06,
+      "loss": 1.0691,
+      "step": 3771
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71233072590792e-06,
+      "loss": 0.617,
+      "step": 3772
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.711592525025762e-06,
+      "loss": 0.5793,
+      "step": 3773
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710854143898008e-06,
+      "loss": 0.6926,
+      "step": 3774
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.710115582560519e-06,
+      "loss": 0.9971,
+      "step": 3775
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.709376841049157e-06,
+      "loss": 0.7195,
+      "step": 3776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708637919399798e-06,
+      "loss": 1.0714,
+      "step": 3777
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.707898817648329e-06,
+      "loss": 0.7932,
+      "step": 3778
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70715953583064e-06,
+      "loss": 0.5052,
+      "step": 3779
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706420073982636e-06,
+      "loss": 0.8735,
+      "step": 3780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.705680432140225e-06,
+      "loss": 0.7077,
+      "step": 3781
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704940610339326e-06,
+      "loss": 0.5577,
+      "step": 3782
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704200608615865e-06,
+      "loss": 0.5493,
+      "step": 3783
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.703460427005783e-06,
+      "loss": 0.5124,
+      "step": 3784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702720065545024e-06,
+      "loss": 0.7964,
+      "step": 3785
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.70197952426954e-06,
+      "loss": 0.8935,
+      "step": 3786
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.701238803215296e-06,
+      "loss": 0.9042,
+      "step": 3787
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700497902418264e-06,
+      "loss": 0.734,
+      "step": 3788
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69975682191442e-06,
+      "loss": 0.8872,
+      "step": 3789
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.699015561739757e-06,
+      "loss": 0.8657,
+      "step": 3790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.698274121930272e-06,
+      "loss": 1.0523,
+      "step": 3791
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.697532502521973e-06,
+      "loss": 0.8712,
+      "step": 3792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69679070355087e-06,
+      "loss": 0.7408,
+      "step": 3793
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696048725052992e-06,
+      "loss": 0.8942,
+      "step": 3794
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.69530656706437e-06,
+      "loss": 0.8514,
+      "step": 3795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.694564229621045e-06,
+      "loss": 0.7029,
+      "step": 3796
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693821712759067e-06,
+      "loss": 0.4683,
+      "step": 3797
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693079016514497e-06,
+      "loss": 0.7149,
+      "step": 3798
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.692336140923397e-06,
+      "loss": 0.8966,
+      "step": 3799
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.691593086021848e-06,
+      "loss": 0.7696,
+      "step": 3800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690849851845933e-06,
+      "loss": 0.7259,
+      "step": 3801
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.690106438431746e-06,
+      "loss": 0.5387,
+      "step": 3802
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.689362845815393e-06,
+      "loss": 0.9414,
+      "step": 3803
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.688619074032977e-06,
+      "loss": 0.7012,
+      "step": 3804
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687875123120623e-06,
+      "loss": 0.7596,
+      "step": 3805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.687130993114459e-06,
+      "loss": 0.8244,
+      "step": 3806
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68638668405062e-06,
+      "loss": 0.7412,
+      "step": 3807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.685642195965255e-06,
+      "loss": 1.0375,
+      "step": 3808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684897528894517e-06,
+      "loss": 0.7644,
+      "step": 3809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.684152682874568e-06,
+      "loss": 0.7465,
+      "step": 3810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.683407657941581e-06,
+      "loss": 0.955,
+      "step": 3811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.682662454131736e-06,
+      "loss": 0.7372,
+      "step": 3812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.681917071481224e-06,
+      "loss": 0.6263,
+      "step": 3813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.68117151002624e-06,
+      "loss": 0.5622,
+      "step": 3814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.680425769802993e-06,
+      "loss": 1.0615,
+      "step": 3815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.679679850847698e-06,
+      "loss": 0.7242,
+      "step": 3816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678933753196577e-06,
+      "loss": 0.8379,
+      "step": 3817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.678187476885864e-06,
+      "loss": 0.6307,
+      "step": 3818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.677441021951802e-06,
+      "loss": 0.8688,
+      "step": 3819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.676694388430637e-06,
+      "loss": 0.5662,
+      "step": 3820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675947576358632e-06,
+      "loss": 0.8504,
+      "step": 3821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.675200585772053e-06,
+      "loss": 0.9127,
+      "step": 3822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.674453416707174e-06,
+      "loss": 0.742,
+      "step": 3823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.67370606920028e-06,
+      "loss": 0.7572,
+      "step": 3824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672958543287666e-06,
+      "loss": 0.6768,
+      "step": 3825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.672210839005635e-06,
+      "loss": 0.8846,
+      "step": 3826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.671462956390496e-06,
+      "loss": 0.9188,
+      "step": 3827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.670714895478566e-06,
+      "loss": 0.5918,
+      "step": 3828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669966656306176e-06,
+      "loss": 0.9344,
+      "step": 3829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.669218238909665e-06,
+      "loss": 0.841,
+      "step": 3830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.668469643325373e-06,
+      "loss": 0.6876,
+      "step": 3831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.667720869589655e-06,
+      "loss": 0.875,
+      "step": 3832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666971917738876e-06,
+      "loss": 0.9052,
+      "step": 3833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.666222787809407e-06,
+      "loss": 0.7694,
+      "step": 3834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.665473479837627e-06,
+      "loss": 0.6346,
+      "step": 3835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.664723993859923e-06,
+      "loss": 0.6302,
+      "step": 3836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663974329912695e-06,
+      "loss": 0.6413,
+      "step": 3837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.663224488032348e-06,
+      "loss": 0.6502,
+      "step": 3838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.662474468255295e-06,
+      "loss": 1.0035,
+      "step": 3839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.661724270617961e-06,
+      "loss": 0.8829,
+      "step": 3840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660973895156777e-06,
+      "loss": 0.6566,
+      "step": 3841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.660223341908183e-06,
+      "loss": 0.7722,
+      "step": 3842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.659472610908628e-06,
+      "loss": 0.7452,
+      "step": 3843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65872170219457e-06,
+      "loss": 0.8091,
+      "step": 3844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.657970615802475e-06,
+      "loss": 0.7692,
+      "step": 3845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65721935176882e-06,
+      "loss": 0.8881,
+      "step": 3846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.656467910130086e-06,
+      "loss": 0.7621,
+      "step": 3847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.655716290922764e-06,
+      "loss": 0.5665,
+      "step": 3848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.65496449418336e-06,
+      "loss": 0.786,
+      "step": 3849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.654212519948376e-06,
+      "loss": 0.5528,
+      "step": 3850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.653460368254338e-06,
+      "loss": 0.9025,
+      "step": 3851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.652708039137767e-06,
+      "loss": 1.076,
+      "step": 3852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651955532635201e-06,
+      "loss": 0.9659,
+      "step": 3853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.651202848783182e-06,
+      "loss": 0.8671,
+      "step": 3854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.650449987618263e-06,
+      "loss": 0.7534,
+      "step": 3855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.649696949177006e-06,
+      "loss": 0.6097,
+      "step": 3856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64894373349598e-06,
+      "loss": 0.7502,
+      "step": 3857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.648190340611766e-06,
+      "loss": 0.8818,
+      "step": 3858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.647436770560946e-06,
+      "loss": 0.6934,
+      "step": 3859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.646683023380118e-06,
+      "loss": 0.7758,
+      "step": 3860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645929099105886e-06,
+      "loss": 0.6139,
+      "step": 3861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.645174997774863e-06,
+      "loss": 0.9523,
+      "step": 3862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.644420719423672e-06,
+      "loss": 0.8202,
+      "step": 3863
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.643666264088939e-06,
+      "loss": 0.5893,
+      "step": 3864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642911631807306e-06,
+      "loss": 0.8869,
+      "step": 3865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.642156822615418e-06,
+      "loss": 0.4292,
+      "step": 3866
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.641401836549931e-06,
+      "loss": 0.6258,
+      "step": 3867
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.640646673647511e-06,
+      "loss": 0.8648,
+      "step": 3868
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639891333944829e-06,
+      "loss": 0.8715,
+      "step": 3869
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.639135817478566e-06,
+      "loss": 0.7127,
+      "step": 3870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.638380124285414e-06,
+      "loss": 0.5901,
+      "step": 3871
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.637624254402072e-06,
+      "loss": 0.689,
+      "step": 3872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636868207865244e-06,
+      "loss": 1.1721,
+      "step": 3873
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.636111984711648e-06,
+      "loss": 0.8958,
+      "step": 3874
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.63535558497801e-06,
+      "loss": 0.7309,
+      "step": 3875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.634599008701059e-06,
+      "loss": 0.4729,
+      "step": 3876
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633842255917539e-06,
+      "loss": 0.5091,
+      "step": 3877
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.633085326664202e-06,
+      "loss": 0.5263,
+      "step": 3878
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.632328220977801e-06,
+      "loss": 0.7261,
+      "step": 3879
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.631570938895109e-06,
+      "loss": 0.6488,
+      "step": 3880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630813480452898e-06,
+      "loss": 0.726,
+      "step": 3881
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.630055845687953e-06,
+      "loss": 0.8367,
+      "step": 3882
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.62929803463707e-06,
+      "loss": 0.9236,
+      "step": 3883
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.628540047337046e-06,
+      "loss": 0.6184,
+      "step": 3884
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627781883824694e-06,
+      "loss": 0.625,
+      "step": 3885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.627023544136832e-06,
+      "loss": 0.6085,
+      "step": 3886
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.626265028310287e-06,
+      "loss": 0.9096,
+      "step": 3887
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.625506336381895e-06,
+      "loss": 0.7175,
+      "step": 3888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.624747468388497e-06,
+      "loss": 0.7456,
+      "step": 3889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.623988424366952e-06,
+      "loss": 0.6355,
+      "step": 3890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.623229204354117e-06,
+      "loss": 0.575,
+      "step": 3891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.622469808386863e-06,
+      "loss": 0.9465,
+      "step": 3892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.621710236502068e-06,
+      "loss": 0.6129,
+      "step": 3893
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.620950488736619e-06,
+      "loss": 0.8308,
+      "step": 3894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.620190565127413e-06,
+      "loss": 0.6079,
+      "step": 3895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.619430465711353e-06,
+      "loss": 0.78,
+      "step": 3896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61867019052535e-06,
+      "loss": 0.7543,
+      "step": 3897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617909739606329e-06,
+      "loss": 0.717,
+      "step": 3898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.617149112991217e-06,
+      "loss": 0.9501,
+      "step": 3899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.616388310716952e-06,
+      "loss": 0.8145,
+      "step": 3900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.615627332820482e-06,
+      "loss": 0.8853,
+      "step": 3901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.61486617933876e-06,
+      "loss": 0.9179,
+      "step": 3902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.614104850308753e-06,
+      "loss": 0.8368,
+      "step": 3903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.613343345767433e-06,
+      "loss": 0.6711,
+      "step": 3904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.612581665751778e-06,
+      "loss": 0.7976,
+      "step": 3905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611819810298778e-06,
+      "loss": 0.9436,
+      "step": 3906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.611057779445434e-06,
+      "loss": 0.9381,
+      "step": 3907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.610295573228748e-06,
+      "loss": 0.6696,
+      "step": 3908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60953319168574e-06,
+      "loss": 0.9321,
+      "step": 3909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60877063485343e-06,
+      "loss": 0.6153,
+      "step": 3910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.608007902768849e-06,
+      "loss": 0.628,
+      "step": 3911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.607244995469039e-06,
+      "loss": 0.5606,
+      "step": 3912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.606481912991052e-06,
+      "loss": 0.6004,
+      "step": 3913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.60571865537194e-06,
+      "loss": 0.7831,
+      "step": 3914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604955222648772e-06,
+      "loss": 0.7235,
+      "step": 3915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.604191614858623e-06,
+      "loss": 0.8244,
+      "step": 3916
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.603427832038574e-06,
+      "loss": 0.8718,
+      "step": 3917
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.602663874225717e-06,
+      "loss": 0.5703,
+      "step": 3918
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.601899741457153e-06,
+      "loss": 0.7386,
+      "step": 3919
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.601135433769991e-06,
+      "loss": 0.7714,
+      "step": 3920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.600370951201345e-06,
+      "loss": 0.5768,
+      "step": 3921
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.599606293788344e-06,
+      "loss": 0.4855,
+      "step": 3922
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.59884146156812e-06,
+      "loss": 0.7209,
+      "step": 3923
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.598076454577815e-06,
+      "loss": 0.6796,
+      "step": 3924
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.597311272854581e-06,
+      "loss": 0.7795,
+      "step": 3925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.596545916435579e-06,
+      "loss": 1.0339,
+      "step": 3926
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595780385357972e-06,
+      "loss": 0.7472,
+      "step": 3927
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.595014679658942e-06,
+      "loss": 1.1041,
+      "step": 3928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.594248799375671e-06,
+      "loss": 0.5968,
+      "step": 3929
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.593482744545351e-06,
+      "loss": 0.7667,
+      "step": 3930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.592716515205187e-06,
+      "loss": 0.6796,
+      "step": 3931
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591950111392386e-06,
+      "loss": 0.7775,
+      "step": 3932
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.591183533144172e-06,
+      "loss": 0.9092,
+      "step": 3933
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.590416780497766e-06,
+      "loss": 0.8651,
+      "step": 3934
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.589649853490407e-06,
+      "loss": 1.0905,
+      "step": 3935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588882752159339e-06,
+      "loss": 0.7649,
+      "step": 3936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.588115476541815e-06,
+      "loss": 0.8529,
+      "step": 3937
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.587348026675095e-06,
+      "loss": 0.6432,
+      "step": 3938
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.586580402596448e-06,
+      "loss": 0.8456,
+      "step": 3939
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.585812604343154e-06,
+      "loss": 0.8837,
+      "step": 3940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.585044631952498e-06,
+      "loss": 0.4728,
+      "step": 3941
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.584276485461775e-06,
+      "loss": 0.6228,
+      "step": 3942
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.58350816490829e-06,
+      "loss": 0.749,
+      "step": 3943
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.582739670329354e-06,
+      "loss": 0.8786,
+      "step": 3944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.581971001762287e-06,
+      "loss": 0.6881,
+      "step": 3945
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.581202159244417e-06,
+      "loss": 0.7636,
+      "step": 3946
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.580433142813082e-06,
+      "loss": 0.981,
+      "step": 3947
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.579663952505628e-06,
+      "loss": 0.8833,
+      "step": 3948
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.57889458835941e-06,
+      "loss": 0.7302,
+      "step": 3949
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.578125050411788e-06,
+      "loss": 0.9813,
+      "step": 3950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.577355338700133e-06,
+      "loss": 0.6665,
+      "step": 3951
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.576585453261827e-06,
+      "loss": 0.7799,
+      "step": 3952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575815394134257e-06,
+      "loss": 0.743,
+      "step": 3953
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.575045161354818e-06,
+      "loss": 0.9334,
+      "step": 3954
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.574274754960916e-06,
+      "loss": 0.6023,
+      "step": 3955
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.573504174989964e-06,
+      "loss": 0.7158,
+      "step": 3956
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.572733421479382e-06,
+      "loss": 0.6913,
+      "step": 3957
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571962494466602e-06,
+      "loss": 0.6255,
+      "step": 3958
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.571191393989062e-06,
+      "loss": 0.7085,
+      "step": 3959
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.570420120084208e-06,
+      "loss": 0.8081,
+      "step": 3960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.569648672789496e-06,
+      "loss": 0.788,
+      "step": 3961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.568877052142388e-06,
+      "loss": 0.6524,
+      "step": 3962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56810525818036e-06,
+      "loss": 0.906,
+      "step": 3963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56733329094089e-06,
+      "loss": 0.6424,
+      "step": 3964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.566561150461466e-06,
+      "loss": 0.904,
+      "step": 3965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565788836779587e-06,
+      "loss": 0.9014,
+      "step": 3966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.565016349932757e-06,
+      "loss": 0.6678,
+      "step": 3967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.564243689958493e-06,
+      "loss": 0.7119,
+      "step": 3968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.563470856894316e-06,
+      "loss": 0.9799,
+      "step": 3969
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.562697850777755e-06,
+      "loss": 1.121,
+      "step": 3970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.561924671646353e-06,
+      "loss": 0.7765,
+      "step": 3971
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.561151319537656e-06,
+      "loss": 0.6936,
+      "step": 3972
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.56037779448922e-06,
+      "loss": 0.6181,
+      "step": 3973
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.559604096538611e-06,
+      "loss": 0.7789,
+      "step": 3974
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5588302257234e-06,
+      "loss": 0.6628,
+      "step": 3975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.558056182081169e-06,
+      "loss": 0.7566,
+      "step": 3976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.557281965649508e-06,
+      "loss": 0.8581,
+      "step": 3977
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.556507576466017e-06,
+      "loss": 0.7841,
+      "step": 3978
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.5557330145683e-06,
+      "loss": 1.066,
+      "step": 3979
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554958279993973e-06,
+      "loss": 0.5347,
+      "step": 3980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.554183372780659e-06,
+      "loss": 0.9205,
+      "step": 3981
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55340829296599e-06,
+      "loss": 0.4864,
+      "step": 3982
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.552633040587605e-06,
+      "loss": 0.7387,
+      "step": 3983
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551857615683154e-06,
+      "loss": 0.6056,
+      "step": 3984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.551082018290295e-06,
+      "loss": 0.8977,
+      "step": 3985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.55030624844669e-06,
+      "loss": 0.6826,
+      "step": 3986
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.549530306190015e-06,
+      "loss": 0.7673,
+      "step": 3987
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.54875419155795e-06,
+      "loss": 0.5773,
+      "step": 3988
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547977904588188e-06,
+      "loss": 0.5643,
+      "step": 3989
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.547201445318425e-06,
+      "loss": 0.7921,
+      "step": 3990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.54642481378637e-06,
+      "loss": 0.7364,
+      "step": 3991
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.545648010029737e-06,
+      "loss": 0.5085,
+      "step": 3992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.54487103408625e-06,
+      "loss": 0.735,
+      "step": 3993
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.544093885993643e-06,
+      "loss": 0.6999,
+      "step": 3994
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.543316565789657e-06,
+      "loss": 0.8172,
+      "step": 3995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.542539073512038e-06,
+      "loss": 1.0095,
+      "step": 3996
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.541761409198544e-06,
+      "loss": 0.681,
+      "step": 3997
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540983572886941e-06,
+      "loss": 0.6003,
+      "step": 3998
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.540205564615002e-06,
+      "loss": 0.7743,
+      "step": 3999
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.539427384420513e-06,
+      "loss": 0.7121,
+      "step": 4000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53864903234126e-06,
+      "loss": 1.1382,
+      "step": 4001
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537870508415046e-06,
+      "loss": 0.8681,
+      "step": 4002
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.537091812679675e-06,
+      "loss": 0.8201,
+      "step": 4003
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.536312945172964e-06,
+      "loss": 0.6325,
+      "step": 4004
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.535533905932739e-06,
+      "loss": 0.9102,
+      "step": 4005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53475469499683e-06,
+      "loss": 0.7239,
+      "step": 4006
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.533975312403076e-06,
+      "loss": 0.7807,
+      "step": 4007
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.53319575818933e-06,
+      "loss": 0.8955,
+      "step": 4008
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.532416032393447e-06,
+      "loss": 0.6272,
+      "step": 4009
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.531636135053296e-06,
+      "loss": 0.8003,
+      "step": 4010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530856066206746e-06,
+      "loss": 0.796,
+      "step": 4011
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.530075825891684e-06,
+      "loss": 1.0019,
+      "step": 4012
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.529295414145997e-06,
+      "loss": 0.5288,
+      "step": 4013
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.528514831007587e-06,
+      "loss": 0.5219,
+      "step": 4014
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.527734076514357e-06,
+      "loss": 0.5948,
+      "step": 4015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526953150704229e-06,
+      "loss": 0.893,
+      "step": 4016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.526172053615122e-06,
+      "loss": 0.789,
+      "step": 4017
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.525390785284971e-06,
+      "loss": 0.6356,
+      "step": 4018
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.524609345751715e-06,
+      "loss": 0.7954,
+      "step": 4019
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523827735053303e-06,
+      "loss": 0.7843,
+      "step": 4020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.523045953227693e-06,
+      "loss": 1.0736,
+      "step": 4021
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.522264000312853e-06,
+      "loss": 0.7978,
+      "step": 4022
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.521481876346751e-06,
+      "loss": 1.0838,
+      "step": 4023
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.520699581367374e-06,
+      "loss": 0.6703,
+      "step": 4024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51991711541271e-06,
+      "loss": 0.7242,
+      "step": 4025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.519134478520761e-06,
+      "loss": 0.9891,
+      "step": 4026
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51835167072953e-06,
+      "loss": 0.6493,
+      "step": 4027
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.517568692077033e-06,
+      "loss": 0.9289,
+      "step": 4028
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.516785542601298e-06,
+      "loss": 0.3263,
+      "step": 4029
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.516002222340353e-06,
+      "loss": 0.931,
+      "step": 4030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.51521873133224e-06,
+      "loss": 0.8169,
+      "step": 4031
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.514435069615005e-06,
+      "loss": 0.7308,
+      "step": 4032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.513651237226707e-06,
+      "loss": 0.8441,
+      "step": 4033
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.512867234205414e-06,
+      "loss": 0.9335,
+      "step": 4034
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.512083060589193e-06,
+      "loss": 0.8007,
+      "step": 4035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.511298716416127e-06,
+      "loss": 0.6839,
+      "step": 4036
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.510514201724313e-06,
+      "loss": 0.7402,
+      "step": 4037
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.509729516551842e-06,
+      "loss": 0.7976,
+      "step": 4038
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.508944660936822e-06,
+      "loss": 0.4925,
+      "step": 4039
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.50815963491737e-06,
+      "loss": 0.8887,
+      "step": 4040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.507374438531606e-06,
+      "loss": 0.5093,
+      "step": 4041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.506589071817666e-06,
+      "loss": 0.6228,
+      "step": 4042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505803534813686e-06,
+      "loss": 0.7906,
+      "step": 4043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.505017827557814e-06,
+      "loss": 0.7585,
+      "step": 4044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.504231950088206e-06,
+      "loss": 1.0533,
+      "step": 4045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.50344590244303e-06,
+      "loss": 0.8854,
+      "step": 4046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.502659684660455e-06,
+      "loss": 0.7552,
+      "step": 4047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501873296778663e-06,
+      "loss": 0.7437,
+      "step": 4048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.501086738835843e-06,
+      "loss": 0.862,
+      "step": 4049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.500300010870195e-06,
+      "loss": 0.804,
+      "step": 4050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49951311291992e-06,
+      "loss": 0.9385,
+      "step": 4051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.498726045023235e-06,
+      "loss": 0.8491,
+      "step": 4052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497938807218365e-06,
+      "loss": 0.6285,
+      "step": 4053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.497151399543535e-06,
+      "loss": 0.8657,
+      "step": 4054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.496363822036987e-06,
+      "loss": 0.7906,
+      "step": 4055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49557607473697e-06,
+      "loss": 0.555,
+      "step": 4056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.494788157681733e-06,
+      "loss": 0.6101,
+      "step": 4057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.494000070909543e-06,
+      "loss": 0.9056,
+      "step": 4058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.493211814458674e-06,
+      "loss": 0.8412,
+      "step": 4059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.492423388367403e-06,
+      "loss": 0.5056,
+      "step": 4060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49163479267402e-06,
+      "loss": 0.9143,
+      "step": 4061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.49084602741682e-06,
+      "loss": 0.7074,
+      "step": 4062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.490057092634108e-06,
+      "loss": 0.9643,
+      "step": 4063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.489267988364198e-06,
+      "loss": 0.9504,
+      "step": 4064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48847871464541e-06,
+      "loss": 0.9296,
+      "step": 4065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.487689271516072e-06,
+      "loss": 0.8124,
+      "step": 4066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.486899659014527e-06,
+      "loss": 0.6753,
+      "step": 4067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.486109877179115e-06,
+      "loss": 0.8122,
+      "step": 4068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.485319926048194e-06,
+      "loss": 0.4455,
+      "step": 4069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.484529805660125e-06,
+      "loss": 0.6346,
+      "step": 4070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.483739516053276e-06,
+      "loss": 0.7724,
+      "step": 4071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.48294905726603e-06,
+      "loss": 0.8031,
+      "step": 4072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.482158429336769e-06,
+      "loss": 0.9718,
+      "step": 4073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.481367632303894e-06,
+      "loss": 0.7912,
+      "step": 4074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.480576666205805e-06,
+      "loss": 0.9985,
+      "step": 4075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.479785531080913e-06,
+      "loss": 0.7737,
+      "step": 4076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.478994226967638e-06,
+      "loss": 0.6043,
+      "step": 4077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.47820275390441e-06,
+      "loss": 0.781,
+      "step": 4078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.477411111929666e-06,
+      "loss": 1.0083,
+      "step": 4079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.476619301081845e-06,
+      "loss": 0.6956,
+      "step": 4080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475827321399406e-06,
+      "loss": 0.7293,
+      "step": 4081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.475035172920804e-06,
+      "loss": 0.7823,
+      "step": 4082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.474242855684514e-06,
+      "loss": 0.8421,
+      "step": 4083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.473450369729007e-06,
+      "loss": 0.7937,
+      "step": 4084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.472657715092773e-06,
+      "loss": 0.7945,
+      "step": 4085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.471864891814304e-06,
+      "loss": 0.7589,
+      "step": 4086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.471071899932101e-06,
+      "loss": 1.0116,
+      "step": 4087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.470278739484677e-06,
+      "loss": 0.7922,
+      "step": 4088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.469485410510545e-06,
+      "loss": 0.5942,
+      "step": 4089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.468691913048237e-06,
+      "loss": 0.6488,
+      "step": 4090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467898247136284e-06,
+      "loss": 1.1418,
+      "step": 4091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.467104412813228e-06,
+      "loss": 0.6805,
+      "step": 4092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.466310410117622e-06,
+      "loss": 0.9756,
+      "step": 4093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.465516239088025e-06,
+      "loss": 0.9195,
+      "step": 4094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.464721899763003e-06,
+      "loss": 0.8178,
+      "step": 4095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.463927392181133e-06,
+      "loss": 0.9376,
+      "step": 4096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.463132716380996e-06,
+      "loss": 0.8071,
+      "step": 4097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.462337872401186e-06,
+      "loss": 0.7538,
+      "step": 4098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.461542860280302e-06,
+      "loss": 1.0002,
+      "step": 4099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.460747680056953e-06,
+      "loss": 0.6369,
+      "step": 4100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459952331769755e-06,
+      "loss": 0.5504,
+      "step": 4101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.459156815457333e-06,
+      "loss": 0.6916,
+      "step": 4102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.458361131158315e-06,
+      "loss": 0.7908,
+      "step": 4103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.457565278911349e-06,
+      "loss": 0.6839,
+      "step": 4104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.456769258755078e-06,
+      "loss": 0.7526,
+      "step": 4105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.45597307072816e-06,
+      "loss": 0.502,
+      "step": 4106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.455176714869262e-06,
+      "loss": 0.9095,
+      "step": 4107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.454380191217058e-06,
+      "loss": 0.7306,
+      "step": 4108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.453583499810227e-06,
+      "loss": 0.7805,
+      "step": 4109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.452786640687461e-06,
+      "loss": 0.7564,
+      "step": 4110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.451989613887455e-06,
+      "loss": 1.0544,
+      "step": 4111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.451192419448918e-06,
+      "loss": 0.8568,
+      "step": 4112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.450395057410561e-06,
+      "loss": 0.6045,
+      "step": 4113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.449597527811108e-06,
+      "loss": 0.6356,
+      "step": 4114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448799830689289e-06,
+      "loss": 1.1364,
+      "step": 4115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.448001966083844e-06,
+      "loss": 0.8344,
+      "step": 4116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.447203934033515e-06,
+      "loss": 0.5987,
+      "step": 4117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.446405734577062e-06,
+      "loss": 0.8012,
+      "step": 4118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.445607367753245e-06,
+      "loss": 1.0105,
+      "step": 4119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444808833600835e-06,
+      "loss": 0.7734,
+      "step": 4120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.444010132158614e-06,
+      "loss": 0.8849,
+      "step": 4121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.443211263465362e-06,
+      "loss": 0.7512,
+      "step": 4122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.442412227559884e-06,
+      "loss": 0.4626,
+      "step": 4123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.441613024480978e-06,
+      "loss": 0.7836,
+      "step": 4124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440813654267455e-06,
+      "loss": 0.7197,
+      "step": 4125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.440014116958139e-06,
+      "loss": 1.0179,
+      "step": 4126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.439214412591851e-06,
+      "loss": 0.6266,
+      "step": 4127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.438414541207433e-06,
+      "loss": 0.9469,
+      "step": 4128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.437614502843727e-06,
+      "loss": 0.8845,
+      "step": 4129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436814297539586e-06,
+      "loss": 0.8235,
+      "step": 4130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.436013925333868e-06,
+      "loss": 0.7503,
+      "step": 4131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.435213386265443e-06,
+      "loss": 0.7231,
+      "step": 4132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.434412680373187e-06,
+      "loss": 0.6304,
+      "step": 4133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.433611807695985e-06,
+      "loss": 0.7944,
+      "step": 4134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43281076827273e-06,
+      "loss": 0.6086,
+      "step": 4135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43200956214232e-06,
+      "loss": 0.9184,
+      "step": 4136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43120818934367e-06,
+      "loss": 0.6355,
+      "step": 4137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.43040664991569e-06,
+      "loss": 1.0389,
+      "step": 4138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42960494389731e-06,
+      "loss": 0.9226,
+      "step": 4139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42880307132746e-06,
+      "loss": 0.7249,
+      "step": 4140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.428001032245081e-06,
+      "loss": 0.9077,
+      "step": 4141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.427198826689128e-06,
+      "loss": 1.0077,
+      "step": 4142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.426396454698552e-06,
+      "loss": 0.7169,
+      "step": 4143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42559391631232e-06,
+      "loss": 0.9266,
+      "step": 4144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.424791211569408e-06,
+      "loss": 0.3259,
+      "step": 4145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.423988340508793e-06,
+      "loss": 0.7534,
+      "step": 4146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.42318530316947e-06,
+      "loss": 0.5563,
+      "step": 4147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.422382099590434e-06,
+      "loss": 0.819,
+      "step": 4148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.421578729810693e-06,
+      "loss": 0.7163,
+      "step": 4149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.420775193869257e-06,
+      "loss": 0.9384,
+      "step": 4150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.41997149180515e-06,
+      "loss": 0.7669,
+      "step": 4151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.419167623657405e-06,
+      "loss": 0.7796,
+      "step": 4152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.418363589465055e-06,
+      "loss": 0.8736,
+      "step": 4153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.417559389267152e-06,
+      "loss": 0.8051,
+      "step": 4154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.416755023102743e-06,
+      "loss": 0.704,
+      "step": 4155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415950491010897e-06,
+      "loss": 0.6425,
+      "step": 4156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.415145793030682e-06,
+      "loss": 0.6608,
+      "step": 4157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.414340929201175e-06,
+      "loss": 0.8935,
+      "step": 4158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.413535899561464e-06,
+      "loss": 0.7646,
+      "step": 4159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.412730704150643e-06,
+      "loss": 0.8719,
+      "step": 4160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411925343007815e-06,
+      "loss": 0.8363,
+      "step": 4161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.411119816172091e-06,
+      "loss": 0.9259,
+      "step": 4162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.410314123682589e-06,
+      "loss": 0.855,
+      "step": 4163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.409508265578434e-06,
+      "loss": 0.7735,
+      "step": 4164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.408702241898766e-06,
+      "loss": 0.7124,
+      "step": 4165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407896052682722e-06,
+      "loss": 0.8898,
+      "step": 4166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.407089697969458e-06,
+      "loss": 0.7625,
+      "step": 4167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.406283177798128e-06,
+      "loss": 0.9638,
+      "step": 4168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.405476492207902e-06,
+      "loss": 0.8085,
+      "step": 4169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.404669641237952e-06,
+      "loss": 0.9521,
+      "step": 4170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.403862624927466e-06,
+      "loss": 0.8079,
+      "step": 4171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.403055443315633e-06,
+      "loss": 0.6786,
+      "step": 4172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.40224809644165e-06,
+      "loss": 0.9015,
+      "step": 4173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.401440584344725e-06,
+      "loss": 0.71,
+      "step": 4174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.400632907064074e-06,
+      "loss": 0.8949,
+      "step": 4175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39982506463892e-06,
+      "loss": 0.9994,
+      "step": 4176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.399017057108493e-06,
+      "loss": 0.9703,
+      "step": 4177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.398208884512033e-06,
+      "loss": 0.8754,
+      "step": 4178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.397400546888786e-06,
+      "loss": 0.9989,
+      "step": 4179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.396592044278012e-06,
+      "loss": 0.9009,
+      "step": 4180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.395783376718967e-06,
+      "loss": 0.6419,
+      "step": 4181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.394974544250926e-06,
+      "loss": 0.8321,
+      "step": 4182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39416554691317e-06,
+      "loss": 0.9737,
+      "step": 4183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.393356384744983e-06,
+      "loss": 0.7804,
+      "step": 4184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.392547057785662e-06,
+      "loss": 0.6492,
+      "step": 4185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.39173756607451e-06,
+      "loss": 0.7625,
+      "step": 4186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390927909650837e-06,
+      "loss": 0.6976,
+      "step": 4187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.390118088553963e-06,
+      "loss": 0.7535,
+      "step": 4188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.389308102823216e-06,
+      "loss": 0.9756,
+      "step": 4189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38849795249793e-06,
+      "loss": 0.7612,
+      "step": 4190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38768763761745e-06,
+      "loss": 0.4689,
+      "step": 4191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386877158221124e-06,
+      "loss": 0.9496,
+      "step": 4192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.386066514348317e-06,
+      "loss": 0.6656,
+      "step": 4193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.38525570603839e-06,
+      "loss": 0.9797,
+      "step": 4194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.384444733330722e-06,
+      "loss": 0.6605,
+      "step": 4195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.383633596264694e-06,
+      "loss": 0.8816,
+      "step": 4196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.382822294879698e-06,
+      "loss": 0.7572,
+      "step": 4197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.382010829215131e-06,
+      "loss": 0.7113,
+      "step": 4198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.381199199310406e-06,
+      "loss": 0.7786,
+      "step": 4199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.380387405204932e-06,
+      "loss": 0.7427,
+      "step": 4200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.379575446938136e-06,
+      "loss": 0.8849,
+      "step": 4201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.378763324549446e-06,
+      "loss": 0.7419,
+      "step": 4202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377951038078303e-06,
+      "loss": 0.7267,
+      "step": 4203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.377138587564152e-06,
+      "loss": 0.7326,
+      "step": 4204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37632597304645e-06,
+      "loss": 0.7016,
+      "step": 4205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.37551319456466e-06,
+      "loss": 0.7808,
+      "step": 4206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.374700252158252e-06,
+      "loss": 0.7746,
+      "step": 4207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373887145866704e-06,
+      "loss": 1.0231,
+      "step": 4208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.373073875729504e-06,
+      "loss": 0.5818,
+      "step": 4209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.372260441786146e-06,
+      "loss": 0.8631,
+      "step": 4210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.371446844076135e-06,
+      "loss": 0.784,
+      "step": 4211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.370633082638977e-06,
+      "loss": 0.7435,
+      "step": 4212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.369819157514195e-06,
+      "loss": 0.7259,
+      "step": 4213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.369005068741314e-06,
+      "loss": 0.9157,
+      "step": 4214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.368190816359867e-06,
+      "loss": 0.7426,
+      "step": 4215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3673764004094e-06,
+      "loss": 0.9047,
+      "step": 4216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.366561820929457e-06,
+      "loss": 0.7283,
+      "step": 4217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.365747077959605e-06,
+      "loss": 0.9181,
+      "step": 4218
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364932171539403e-06,
+      "loss": 0.7514,
+      "step": 4219
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.364117101708428e-06,
+      "loss": 0.7535,
+      "step": 4220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.363301868506264e-06,
+      "loss": 0.7552,
+      "step": 4221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.362486471972498e-06,
+      "loss": 0.8279,
+      "step": 4222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.361670912146728e-06,
+      "loss": 0.7095,
+      "step": 4223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360855189068562e-06,
+      "loss": 0.8333,
+      "step": 4224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.360039302777614e-06,
+      "loss": 0.704,
+      "step": 4225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.359223253313501e-06,
+      "loss": 0.7996,
+      "step": 4226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.358407040715861e-06,
+      "loss": 0.5944,
+      "step": 4227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.357590665024324e-06,
+      "loss": 0.5692,
+      "step": 4228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35677412627854e-06,
+      "loss": 0.8138,
+      "step": 4229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35595742451816e-06,
+      "loss": 1.0958,
+      "step": 4230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.355140559782847e-06,
+      "loss": 0.6744,
+      "step": 4231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35432353211227e-06,
+      "loss": 0.9004,
+      "step": 4232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.353506341546106e-06,
+      "loss": 0.691,
+      "step": 4233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35268898812404e-06,
+      "loss": 0.9569,
+      "step": 4234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.351871471885764e-06,
+      "loss": 0.7045,
+      "step": 4235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.35105379287098e-06,
+      "loss": 0.7309,
+      "step": 4236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.350235951119398e-06,
+      "loss": 0.8036,
+      "step": 4237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.349417946670732e-06,
+      "loss": 1.0012,
+      "step": 4238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.34859977956471e-06,
+      "loss": 0.6157,
+      "step": 4239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.347781449841063e-06,
+      "loss": 1.0372,
+      "step": 4240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.34696295753953e-06,
+      "loss": 0.9931,
+      "step": 4241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.346144302699861e-06,
+      "loss": 0.83,
+      "step": 4242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.345325485361811e-06,
+      "loss": 1.2426,
+      "step": 4243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.344506505565146e-06,
+      "loss": 0.8011,
+      "step": 4244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.343687363349637e-06,
+      "loss": 0.6191,
+      "step": 4245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342868058755062e-06,
+      "loss": 0.659,
+      "step": 4246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.342048591821213e-06,
+      "loss": 0.7649,
+      "step": 4247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.341228962587881e-06,
+      "loss": 0.7124,
+      "step": 4248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.340409171094874e-06,
+      "loss": 0.9972,
+      "step": 4249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.339589217382e-06,
+      "loss": 0.8691,
+      "step": 4250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.338769101489081e-06,
+      "loss": 0.8059,
+      "step": 4251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337948823455941e-06,
+      "loss": 0.8314,
+      "step": 4252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.337128383322418e-06,
+      "loss": 0.7316,
+      "step": 4253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.336307781128354e-06,
+      "loss": 0.8416,
+      "step": 4254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.335487016913597e-06,
+      "loss": 0.8287,
+      "step": 4255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.334666090718012e-06,
+      "loss": 0.8036,
+      "step": 4256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.33384500258146e-06,
+      "loss": 1.1051,
+      "step": 4257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.333023752543816e-06,
+      "loss": 0.6645,
+      "step": 4258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.332202340644964e-06,
+      "loss": 0.6904,
+      "step": 4259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.331380766924794e-06,
+      "loss": 0.8183,
+      "step": 4260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.330559031423204e-06,
+      "loss": 0.8887,
+      "step": 4261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.329737134180097e-06,
+      "loss": 0.6392,
+      "step": 4262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.328915075235393e-06,
+      "loss": 0.6642,
+      "step": 4263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.328092854629005e-06,
+      "loss": 0.6883,
+      "step": 4264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32727047240087e-06,
+      "loss": 0.7613,
+      "step": 4265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.326447928590921e-06,
+      "loss": 0.7726,
+      "step": 4266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.325625223239106e-06,
+      "loss": 0.8643,
+      "step": 4267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.324802356385375e-06,
+      "loss": 0.7809,
+      "step": 4268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.323979328069689e-06,
+      "loss": 0.8189,
+      "step": 4269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32315613833202e-06,
+      "loss": 0.7505,
+      "step": 4270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.32233278721234e-06,
+      "loss": 0.774,
+      "step": 4271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.321509274750637e-06,
+      "loss": 0.6313,
+      "step": 4272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.3206856009869e-06,
+      "loss": 0.945,
+      "step": 4273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319861765961132e-06,
+      "loss": 0.745,
+      "step": 4274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.319037769713338e-06,
+      "loss": 0.9753,
+      "step": 4275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.318213612283536e-06,
+      "loss": 0.8855,
+      "step": 4276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.317389293711747e-06,
+      "loss": 1.0914,
+      "step": 4277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.316564814038004e-06,
+      "loss": 0.7274,
+      "step": 4278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.315740173302346e-06,
+      "loss": 0.7194,
+      "step": 4279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314915371544822e-06,
+      "loss": 0.7496,
+      "step": 4280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.314090408805481e-06,
+      "loss": 0.6147,
+      "step": 4281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31326528512439e-06,
+      "loss": 0.9846,
+      "step": 4282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.31244000054162e-06,
+      "loss": 0.6757,
+      "step": 4283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.311614555097245e-06,
+      "loss": 0.9151,
+      "step": 4284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.310788948831355e-06,
+      "loss": 0.5397,
+      "step": 4285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309963181784042e-06,
+      "loss": 0.9606,
+      "step": 4286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.309137253995407e-06,
+      "loss": 0.605,
+      "step": 4287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.308311165505562e-06,
+      "loss": 0.6413,
+      "step": 4288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.307484916354623e-06,
+      "loss": 0.7623,
+      "step": 4289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.306658506582713e-06,
+      "loss": 0.7837,
+      "step": 4290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305831936229966e-06,
+      "loss": 0.6698,
+      "step": 4291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.305005205336526e-06,
+      "loss": 0.6751,
+      "step": 4292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.304178313942536e-06,
+      "loss": 0.8874,
+      "step": 4293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.303351262088155e-06,
+      "loss": 0.6544,
+      "step": 4294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.302524049813548e-06,
+      "loss": 0.8532,
+      "step": 4295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.301696677158885e-06,
+      "loss": 1.122,
+      "step": 4296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.300869144164346e-06,
+      "loss": 0.6144,
+      "step": 4297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.300041450870118e-06,
+      "loss": 1.0826,
+      "step": 4298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.299213597316398e-06,
+      "loss": 0.9065,
+      "step": 4299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.298385583543387e-06,
+      "loss": 0.9832,
+      "step": 4300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.297557409591297e-06,
+      "loss": 0.8523,
+      "step": 4301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.296729075500345e-06,
+      "loss": 1.1207,
+      "step": 4302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.295900581310758e-06,
+      "loss": 0.8024,
+      "step": 4303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.295071927062772e-06,
+      "loss": 0.5931,
+      "step": 4304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.294243112796625e-06,
+      "loss": 0.7715,
+      "step": 4305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29341413855257e-06,
+      "loss": 0.5714,
+      "step": 4306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.292585004370863e-06,
+      "loss": 0.5132,
+      "step": 4307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29175571029177e-06,
+      "loss": 0.7884,
+      "step": 4308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.29092625635556e-06,
+      "loss": 0.6462,
+      "step": 4309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.290096642602519e-06,
+      "loss": 0.8361,
+      "step": 4310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.289266869072933e-06,
+      "loss": 0.7823,
+      "step": 4311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.288436935807098e-06,
+      "loss": 0.6485,
+      "step": 4312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28760684284532e-06,
+      "loss": 0.6895,
+      "step": 4313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.286776590227907e-06,
+      "loss": 0.8787,
+      "step": 4314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.28594617799518e-06,
+      "loss": 0.7428,
+      "step": 4315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.285115606187468e-06,
+      "loss": 0.4862,
+      "step": 4316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.284284874845104e-06,
+      "loss": 0.9828,
+      "step": 4317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.283453984008433e-06,
+      "loss": 0.6834,
+      "step": 4318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.282622933717802e-06,
+      "loss": 0.807,
+      "step": 4319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.281791724013571e-06,
+      "loss": 0.8779,
+      "step": 4320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280960354936107e-06,
+      "loss": 1.0557,
+      "step": 4321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.280128826525782e-06,
+      "loss": 0.7624,
+      "step": 4322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.279297138822976e-06,
+      "loss": 0.8036,
+      "step": 4323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.278465291868083e-06,
+      "loss": 0.8638,
+      "step": 4324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.277633285701496e-06,
+      "loss": 0.7001,
+      "step": 4325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.276801120363622e-06,
+      "loss": 1.0777,
+      "step": 4326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.275968795894869e-06,
+      "loss": 0.615,
+      "step": 4327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27513631233566e-06,
+      "loss": 0.7705,
+      "step": 4328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.274303669726427e-06,
+      "loss": 0.8781,
+      "step": 4329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.273470868107599e-06,
+      "loss": 0.9043,
+      "step": 4330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.27263790751962e-06,
+      "loss": 0.8549,
+      "step": 4331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.271804788002944e-06,
+      "loss": 0.7219,
+      "step": 4332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270971509598026e-06,
+      "loss": 0.6915,
+      "step": 4333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.270138072345338e-06,
+      "loss": 0.6711,
+      "step": 4334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.269304476285349e-06,
+      "loss": 0.6757,
+      "step": 4335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.268470721458543e-06,
+      "loss": 0.8409,
+      "step": 4336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26763680790541e-06,
+      "loss": 0.8612,
+      "step": 4337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.266802735666443e-06,
+      "loss": 0.771,
+      "step": 4338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.265968504782152e-06,
+      "loss": 0.7613,
+      "step": 4339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.26513411529305e-06,
+      "loss": 0.9205,
+      "step": 4340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.264299567239652e-06,
+      "loss": 0.9012,
+      "step": 4341
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.263464860662491e-06,
+      "loss": 0.7458,
+      "step": 4342
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.2626299956021e-06,
+      "loss": 0.8245,
+      "step": 4343
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.261794972099023e-06,
+      "loss": 0.4957,
+      "step": 4344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.260959790193815e-06,
+      "loss": 0.6038,
+      "step": 4345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.260124449927028e-06,
+      "loss": 0.8312,
+      "step": 4346
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.259288951339233e-06,
+      "loss": 0.9184,
+      "step": 4347
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.258453294471005e-06,
+      "loss": 0.866,
+      "step": 4348
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.257617479362921e-06,
+      "loss": 0.9124,
+      "step": 4349
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.256781506055575e-06,
+      "loss": 1.0093,
+      "step": 4350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.255945374589563e-06,
+      "loss": 0.7121,
+      "step": 4351
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25510908500549e-06,
+      "loss": 0.7854,
+      "step": 4352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.254272637343968e-06,
+      "loss": 0.5462,
+      "step": 4353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.253436031645618e-06,
+      "loss": 0.7305,
+      "step": 4354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.252599267951068e-06,
+      "loss": 0.6693,
+      "step": 4355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.251762346300954e-06,
+      "loss": 0.7308,
+      "step": 4356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.25092526673592e-06,
+      "loss": 0.7242,
+      "step": 4357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.250088029296613e-06,
+      "loss": 0.8416,
+      "step": 4358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.249250634023694e-06,
+      "loss": 0.9505,
+      "step": 4359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.248413080957831e-06,
+      "loss": 0.5899,
+      "step": 4360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.247575370139695e-06,
+      "loss": 0.8542,
+      "step": 4361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.246737501609972e-06,
+      "loss": 0.5748,
+      "step": 4362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.245899475409348e-06,
+      "loss": 0.7629,
+      "step": 4363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24506129157852e-06,
+      "loss": 0.8132,
+      "step": 4364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.244222950158194e-06,
+      "loss": 0.7382,
+      "step": 4365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.24338445118908e-06,
+      "loss": 1.0812,
+      "step": 4366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.242545794711901e-06,
+      "loss": 0.9124,
+      "step": 4367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.241706980767382e-06,
+      "loss": 0.646,
+      "step": 4368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240868009396259e-06,
+      "loss": 1.0462,
+      "step": 4369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.240028880639277e-06,
+      "loss": 0.8156,
+      "step": 4370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.239189594537181e-06,
+      "loss": 0.9519,
+      "step": 4371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.238350151130734e-06,
+      "loss": 0.9143,
+      "step": 4372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.237510550460703e-06,
+      "loss": 0.8161,
+      "step": 4373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.236670792567856e-06,
+      "loss": 0.7652,
+      "step": 4374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.23583087749298e-06,
+      "loss": 0.7506,
+      "step": 4375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234990805276856e-06,
+      "loss": 0.8203,
+      "step": 4376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.234150575960288e-06,
+      "loss": 1.0766,
+      "step": 4377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.233310189584077e-06,
+      "loss": 0.6997,
+      "step": 4378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.232469646189033e-06,
+      "loss": 0.4352,
+      "step": 4379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.231628945815977e-06,
+      "loss": 0.8894,
+      "step": 4380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.230788088505737e-06,
+      "loss": 0.5829,
+      "step": 4381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.229947074299144e-06,
+      "loss": 0.906,
+      "step": 4382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.229105903237045e-06,
+      "loss": 0.7391,
+      "step": 4383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.228264575360284e-06,
+      "loss": 0.4736,
+      "step": 4384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.227423090709722e-06,
+      "loss": 1.0487,
+      "step": 4385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.226581449326223e-06,
+      "loss": 0.6315,
+      "step": 4386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22573965125066e-06,
+      "loss": 0.6612,
+      "step": 4387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.22489769652391e-06,
+      "loss": 0.5051,
+      "step": 4388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.224055585186866e-06,
+      "loss": 0.5658,
+      "step": 4389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.223213317280419e-06,
+      "loss": 0.8982,
+      "step": 4390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.222370892845475e-06,
+      "loss": 0.6737,
+      "step": 4391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.221528311922941e-06,
+      "loss": 0.7363,
+      "step": 4392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.220685574553739e-06,
+      "loss": 0.776,
+      "step": 4393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.219842680778793e-06,
+      "loss": 0.7476,
+      "step": 4394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.218999630639037e-06,
+      "loss": 0.6811,
+      "step": 4395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21815642417541e-06,
+      "loss": 0.9014,
+      "step": 4396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.217313061428863e-06,
+      "loss": 0.7525,
+      "step": 4397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.21646954244035e-06,
+      "loss": 0.7643,
+      "step": 4398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.215625867250836e-06,
+      "loss": 0.8575,
+      "step": 4399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.214782035901292e-06,
+      "loss": 1.0544,
+      "step": 4400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.213938048432697e-06,
+      "loss": 0.9196,
+      "step": 4401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.213093904886038e-06,
+      "loss": 1.0599,
+      "step": 4402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.212249605302307e-06,
+      "loss": 0.7187,
+      "step": 4403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.211405149722507e-06,
+      "loss": 0.7339,
+      "step": 4404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.210560538187648e-06,
+      "loss": 0.825,
+      "step": 4405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.209715770738746e-06,
+      "loss": 0.7155,
+      "step": 4406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208870847416824e-06,
+      "loss": 0.7497,
+      "step": 4407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.208025768262917e-06,
+      "loss": 0.7263,
+      "step": 4408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.207180533318061e-06,
+      "loss": 0.913,
+      "step": 4409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.206335142623305e-06,
+      "loss": 0.9457,
+      "step": 4410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.205489596219705e-06,
+      "loss": 0.8204,
+      "step": 4411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.204643894148318e-06,
+      "loss": 0.8116,
+      "step": 4412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.203798036450222e-06,
+      "loss": 0.6594,
+      "step": 4413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.202952023166485e-06,
+      "loss": 0.694,
+      "step": 4414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.202105854338197e-06,
+      "loss": 1.2451,
+      "step": 4415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.201259530006452e-06,
+      "loss": 0.745,
+      "step": 4416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.200413050212344e-06,
+      "loss": 0.9576,
+      "step": 4417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.199566414996986e-06,
+      "loss": 0.6154,
+      "step": 4418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.198719624401493e-06,
+      "loss": 0.8646,
+      "step": 4419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.197872678466982e-06,
+      "loss": 0.7257,
+      "step": 4420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.19702557723459e-06,
+      "loss": 0.8807,
+      "step": 4421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.196178320745449e-06,
+      "loss": 0.7734,
+      "step": 4422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.195330909040709e-06,
+      "loss": 0.6037,
+      "step": 4423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.194483342161518e-06,
+      "loss": 1.129,
+      "step": 4424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.193635620149041e-06,
+      "loss": 0.8683,
+      "step": 4425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.192787743044442e-06,
+      "loss": 0.7618,
+      "step": 4426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.1919397108889e-06,
+      "loss": 1.0441,
+      "step": 4427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.191091523723594e-06,
+      "loss": 0.7368,
+      "step": 4428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.190243181589716e-06,
+      "loss": 0.6371,
+      "step": 4429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.189394684528465e-06,
+      "loss": 0.7709,
+      "step": 4430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.188546032581043e-06,
+      "loss": 0.7824,
+      "step": 4431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18769722578867e-06,
+      "loss": 0.5721,
+      "step": 4432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18684826419256e-06,
+      "loss": 0.5744,
+      "step": 4433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185999147833943e-06,
+      "loss": 0.6027,
+      "step": 4434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.185149876754054e-06,
+      "loss": 0.9564,
+      "step": 4435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.184300450994137e-06,
+      "loss": 0.5719,
+      "step": 4436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.183450870595443e-06,
+      "loss": 1.0962,
+      "step": 4437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.182601135599227e-06,
+      "loss": 0.8051,
+      "step": 4438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.181751246046758e-06,
+      "loss": 0.6535,
+      "step": 4439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.18090120197931e-06,
+      "loss": 0.7027,
+      "step": 4440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.180051003438158e-06,
+      "loss": 0.8369,
+      "step": 4441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.179200650464596e-06,
+      "loss": 1.059,
+      "step": 4442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.178350143099916e-06,
+      "loss": 0.6554,
+      "step": 4443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.177499481385422e-06,
+      "loss": 0.7362,
+      "step": 4444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.176648665362426e-06,
+      "loss": 0.9928,
+      "step": 4445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.175797695072245e-06,
+      "loss": 0.9033,
+      "step": 4446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.174946570556204e-06,
+      "loss": 0.956,
+      "step": 4447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.174095291855636e-06,
+      "loss": 0.9656,
+      "step": 4448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.173243859011881e-06,
+      "loss": 0.9286,
+      "step": 4449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.17239227206629e-06,
+      "loss": 0.8498,
+      "step": 4450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.171540531060216e-06,
+      "loss": 0.708,
+      "step": 4451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.170688636035023e-06,
+      "loss": 0.7294,
+      "step": 4452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.169836587032079e-06,
+      "loss": 0.9233,
+      "step": 4453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168984384092766e-06,
+      "loss": 0.738,
+      "step": 4454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.168132027258467e-06,
+      "loss": 0.8109,
+      "step": 4455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.167279516570576e-06,
+      "loss": 0.6972,
+      "step": 4456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16642685207049e-06,
+      "loss": 0.8974,
+      "step": 4457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16557403379962e-06,
+      "loss": 0.608,
+      "step": 4458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.164721061799384e-06,
+      "loss": 0.6363,
+      "step": 4459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.1638679361112e-06,
+      "loss": 0.6422,
+      "step": 4460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.163014656776499e-06,
+      "loss": 0.9607,
+      "step": 4461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.16216122383672e-06,
+      "loss": 0.5868,
+      "step": 4462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.161307637333307e-06,
+      "loss": 0.8789,
+      "step": 4463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.160453897307714e-06,
+      "loss": 0.9891,
+      "step": 4464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.159600003801401e-06,
+      "loss": 0.5723,
+      "step": 4465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.158745956855834e-06,
+      "loss": 0.6332,
+      "step": 4466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157891756512488e-06,
+      "loss": 0.8247,
+      "step": 4467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.157037402812848e-06,
+      "loss": 0.7584,
+      "step": 4468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.156182895798402e-06,
+      "loss": 0.369,
+      "step": 4469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.155328235510646e-06,
+      "loss": 0.8279,
+      "step": 4470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.154473421991087e-06,
+      "loss": 1.0003,
+      "step": 4471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.153618455281236e-06,
+      "loss": 0.5433,
+      "step": 4472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.152763335422612e-06,
+      "loss": 0.663,
+      "step": 4473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151908062456746e-06,
+      "loss": 0.7146,
+      "step": 4474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.151052636425169e-06,
+      "loss": 0.6106,
+      "step": 4475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.150197057369422e-06,
+      "loss": 0.9209,
+      "step": 4476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.149341325331056e-06,
+      "loss": 0.8069,
+      "step": 4477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.148485440351628e-06,
+      "loss": 0.577,
+      "step": 4478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.147629402472703e-06,
+      "loss": 1.1098,
+      "step": 4479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.14677321173585e-06,
+      "loss": 1.0187,
+      "step": 4480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.145916868182652e-06,
+      "loss": 0.7468,
+      "step": 4481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.145060371854692e-06,
+      "loss": 0.8441,
+      "step": 4482
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.144203722793564e-06,
+      "loss": 0.6721,
+      "step": 4483
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.143346921040873e-06,
+      "loss": 0.832,
+      "step": 4484
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.142489966638223e-06,
+      "loss": 0.9015,
+      "step": 4485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.141632859627234e-06,
+      "loss": 0.7071,
+      "step": 4486
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.140775600049526e-06,
+      "loss": 0.7119,
+      "step": 4487
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139918187946734e-06,
+      "loss": 0.6955,
+      "step": 4488
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.139060623360494e-06,
+      "loss": 0.9389,
+      "step": 4489
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.138202906332452e-06,
+      "loss": 0.6002,
+      "step": 4490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.13734503690426e-06,
+      "loss": 0.7143,
+      "step": 4491
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.136487015117582e-06,
+      "loss": 0.6518,
+      "step": 4492
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.135628841014082e-06,
+      "loss": 0.8551,
+      "step": 4493
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.13477051463544e-06,
+      "loss": 0.7535,
+      "step": 4494
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.133912036023335e-06,
+      "loss": 0.9116,
+      "step": 4495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.133053405219457e-06,
+      "loss": 0.5349,
+      "step": 4496
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.132194622265508e-06,
+      "loss": 0.7063,
+      "step": 4497
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.131335687203187e-06,
+      "loss": 0.8949,
+      "step": 4498
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.13047660007421e-06,
+      "loss": 0.7855,
+      "step": 4499
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.129617360920297e-06,
+      "loss": 0.646,
+      "step": 4500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.128757969783171e-06,
+      "loss": 0.5085,
+      "step": 4501
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.127898426704575e-06,
+      "loss": 1.0224,
+      "step": 4502
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.127038731726241e-06,
+      "loss": 0.8301,
+      "step": 4503
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.126178884889924e-06,
+      "loss": 0.5165,
+      "step": 4504
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.125318886237382e-06,
+      "loss": 0.7075,
+      "step": 4505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.124458735810372e-06,
+      "loss": 0.9472,
+      "step": 4506
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123598433650673e-06,
+      "loss": 1.0073,
+      "step": 4507
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.12273797980006e-06,
+      "loss": 0.9502,
+      "step": 4508
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.121877374300318e-06,
+      "loss": 0.8986,
+      "step": 4509
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.121016617193243e-06,
+      "loss": 0.479,
+      "step": 4510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.120155708520637e-06,
+      "loss": 0.8908,
+      "step": 4511
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.119294648324302e-06,
+      "loss": 0.8709,
+      "step": 4512
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11843343664606e-06,
+      "loss": 1.0149,
+      "step": 4513
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.117572073527734e-06,
+      "loss": 0.8594,
+      "step": 4514
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.116710559011147e-06,
+      "loss": 0.652,
+      "step": 4515
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.115848893138146e-06,
+      "loss": 0.7821,
+      "step": 4516
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.114987075950568e-06,
+      "loss": 0.927,
+      "step": 4517
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.11412510749027e-06,
+      "loss": 0.7698,
+      "step": 4518
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.113262987799112e-06,
+      "loss": 0.5531,
+      "step": 4519
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.112400716918958e-06,
+      "loss": 0.8021,
+      "step": 4520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.111538294891684e-06,
+      "loss": 0.7737,
+      "step": 4521
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.110675721759172e-06,
+      "loss": 0.7037,
+      "step": 4522
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10981299756331e-06,
+      "loss": 0.6909,
+      "step": 4523
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.108950122345996e-06,
+      "loss": 0.9289,
+      "step": 4524
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.108087096149131e-06,
+      "loss": 0.7792,
+      "step": 4525
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10722391901463e-06,
+      "loss": 0.8349,
+      "step": 4526
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.106360590984406e-06,
+      "loss": 0.773,
+      "step": 4527
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10549711210039e-06,
+      "loss": 0.6762,
+      "step": 4528
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.10463348240451e-06,
+      "loss": 0.7701,
+      "step": 4529
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.103769701938711e-06,
+      "loss": 0.7526,
+      "step": 4530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.102905770744937e-06,
+      "loss": 0.7865,
+      "step": 4531
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.102041688865147e-06,
+      "loss": 0.6583,
+      "step": 4532
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.101177456341301e-06,
+      "loss": 0.8203,
+      "step": 4533
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.100313073215368e-06,
+      "loss": 0.9702,
+      "step": 4534
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.099448539529325e-06,
+      "loss": 0.6759,
+      "step": 4535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.098583855325157e-06,
+      "loss": 0.7972,
+      "step": 4536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.097719020644855e-06,
+      "loss": 0.9096,
+      "step": 4537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.09685403553042e-06,
+      "loss": 0.527,
+      "step": 4538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095988900023854e-06,
+      "loss": 0.6614,
+      "step": 4539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095123614167174e-06,
+      "loss": 0.64,
+      "step": 4540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.094258178002398e-06,
+      "loss": 1.047,
+      "step": 4541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.093392591571559e-06,
+      "loss": 0.5746,
+      "step": 4542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.092526854916686e-06,
+      "loss": 0.8148,
+      "step": 4543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.091660968079827e-06,
+      "loss": 0.7213,
+      "step": 4544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.090794931103026e-06,
+      "loss": 0.6245,
+      "step": 4545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.089928744028349e-06,
+      "loss": 0.844,
+      "step": 4546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08906240689785e-06,
+      "loss": 0.7716,
+      "step": 4547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08819591975361e-06,
+      "loss": 0.6033,
+      "step": 4548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.087329282637702e-06,
+      "loss": 0.978,
+      "step": 4549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.086462495592215e-06,
+      "loss": 0.7832,
+      "step": 4550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.085595558659244e-06,
+      "loss": 0.7467,
+      "step": 4551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.084728471880887e-06,
+      "loss": 0.9281,
+      "step": 4552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.083861235299253e-06,
+      "loss": 0.8672,
+      "step": 4553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.08299384895646e-06,
+      "loss": 0.8146,
+      "step": 4554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.082126312894628e-06,
+      "loss": 0.8425,
+      "step": 4555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.081258627155886e-06,
+      "loss": 0.9363,
+      "step": 4556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.080390791782375e-06,
+      "loss": 0.8894,
+      "step": 4557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.079522806816238e-06,
+      "loss": 0.9794,
+      "step": 4558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.078654672299625e-06,
+      "loss": 0.6231,
+      "step": 4559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0777863882747e-06,
+      "loss": 0.6779,
+      "step": 4560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076917954783623e-06,
+      "loss": 0.5692,
+      "step": 4561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.076049371868571e-06,
+      "loss": 0.9567,
+      "step": 4562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.075180639571726e-06,
+      "loss": 0.712,
+      "step": 4563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.074311757935274e-06,
+      "loss": 0.6032,
+      "step": 4564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07344272700141e-06,
+      "loss": 0.7621,
+      "step": 4565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.072573546812338e-06,
+      "loss": 0.9975,
+      "step": 4566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07170421741027e-06,
+      "loss": 0.8385,
+      "step": 4567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.07083473883742e-06,
+      "loss": 0.8463,
+      "step": 4568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.06996511113601e-06,
+      "loss": 0.726,
+      "step": 4569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.069095334348279e-06,
+      "loss": 0.8305,
+      "step": 4570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.068225408516459e-06,
+      "loss": 0.8967,
+      "step": 4571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.067355333682799e-06,
+      "loss": 0.7949,
+      "step": 4572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066485109889552e-06,
+      "loss": 0.9809,
+      "step": 4573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.065614737178979e-06,
+      "loss": 0.6842,
+      "step": 4574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.064744215593346e-06,
+      "loss": 0.6129,
+      "step": 4575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063873545174932e-06,
+      "loss": 0.6965,
+      "step": 4576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.063002725966014e-06,
+      "loss": 0.7774,
+      "step": 4577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.062131758008887e-06,
+      "loss": 0.7882,
+      "step": 4578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.061260641345842e-06,
+      "loss": 0.7783,
+      "step": 4579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.060389376019189e-06,
+      "loss": 0.7212,
+      "step": 4580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.059517962071234e-06,
+      "loss": 0.6084,
+      "step": 4581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.058646399544296e-06,
+      "loss": 0.7829,
+      "step": 4582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.057774688480704e-06,
+      "loss": 0.854,
+      "step": 4583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05690282892279e-06,
+      "loss": 0.9493,
+      "step": 4584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05603082091289e-06,
+      "loss": 0.5558,
+      "step": 4585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.055158664493355e-06,
+      "loss": 0.8285,
+      "step": 4586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.054286359706538e-06,
+      "loss": 0.6821,
+      "step": 4587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.0534139065948e-06,
+      "loss": 0.8981,
+      "step": 4588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.052541305200513e-06,
+      "loss": 0.5192,
+      "step": 4589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.05166855556605e-06,
+      "loss": 0.8797,
+      "step": 4590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.050795657733795e-06,
+      "loss": 0.9478,
+      "step": 4591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049922611746137e-06,
+      "loss": 0.6056,
+      "step": 4592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.049049417645478e-06,
+      "loss": 0.7568,
+      "step": 4593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.048176075474216e-06,
+      "loss": 0.6316,
+      "step": 4594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04730258527477e-06,
+      "loss": 0.8519,
+      "step": 4595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.046428947089556e-06,
+      "loss": 0.8533,
+      "step": 4596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.045555160961e-06,
+      "loss": 0.8182,
+      "step": 4597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.044681226931535e-06,
+      "loss": 0.9016,
+      "step": 4598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.043807145043604e-06,
+      "loss": 0.7716,
+      "step": 4599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.042932915339654e-06,
+      "loss": 0.517,
+      "step": 4600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.04205853786214e-06,
+      "loss": 0.9022,
+      "step": 4601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.041184012653522e-06,
+      "loss": 1.0466,
+      "step": 4602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.040309339756273e-06,
+      "loss": 0.7416,
+      "step": 4603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.039434519212868e-06,
+      "loss": 0.8024,
+      "step": 4604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.038559551065792e-06,
+      "loss": 0.8502,
+      "step": 4605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.037684435357534e-06,
+      "loss": 0.8053,
+      "step": 4606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.036809172130592e-06,
+      "loss": 0.9573,
+      "step": 4607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.035933761427475e-06,
+      "loss": 0.6289,
+      "step": 4608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.035058203290691e-06,
+      "loss": 0.7479,
+      "step": 4609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.034182497762761e-06,
+      "loss": 0.6444,
+      "step": 4610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.033306644886215e-06,
+      "loss": 0.7813,
+      "step": 4611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.032430644703581e-06,
+      "loss": 0.794,
+      "step": 4612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.031554497257406e-06,
+      "loss": 0.846,
+      "step": 4613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.030678202590232e-06,
+      "loss": 0.8319,
+      "step": 4614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.029801760744622e-06,
+      "loss": 0.9392,
+      "step": 4615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02892517176313e-06,
+      "loss": 0.7147,
+      "step": 4616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.028048435688333e-06,
+      "loss": 0.842,
+      "step": 4617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.027171552562804e-06,
+      "loss": 0.616,
+      "step": 4618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.02629452242913e-06,
+      "loss": 0.8992,
+      "step": 4619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.025417345329897e-06,
+      "loss": 0.581,
+      "step": 4620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024540021307709e-06,
+      "loss": 0.7408,
+      "step": 4621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.023662550405169e-06,
+      "loss": 0.8014,
+      "step": 4622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.022784932664885e-06,
+      "loss": 0.6326,
+      "step": 4623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021907168129483e-06,
+      "loss": 0.8954,
+      "step": 4624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.021029256841588e-06,
+      "loss": 0.8171,
+      "step": 4625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.020151198843833e-06,
+      "loss": 0.7577,
+      "step": 4626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.01927299417886e-06,
+      "loss": 0.4803,
+      "step": 4627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.018394642889316e-06,
+      "loss": 0.8101,
+      "step": 4628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.017516145017857e-06,
+      "loss": 0.7825,
+      "step": 4629
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.016637500607144e-06,
+      "loss": 0.8172,
+      "step": 4630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.01575870969985e-06,
+      "loss": 0.5396,
+      "step": 4631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014879772338649e-06,
+      "loss": 0.7274,
+      "step": 4632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.014000688566224e-06,
+      "loss": 0.7626,
+      "step": 4633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.013121458425267e-06,
+      "loss": 0.6125,
+      "step": 4634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.012242081958477e-06,
+      "loss": 0.6466,
+      "step": 4635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.011362559208557e-06,
+      "loss": 0.7841,
+      "step": 4636
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.010482890218219e-06,
+      "loss": 0.804,
+      "step": 4637
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.009603075030184e-06,
+      "loss": 0.6086,
+      "step": 4638
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.008723113687178e-06,
+      "loss": 0.8524,
+      "step": 4639
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.007843006231934e-06,
+      "loss": 0.5829,
+      "step": 4640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.006962752707193e-06,
+      "loss": 0.7285,
+      "step": 4641
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.0060823531557e-06,
+      "loss": 0.599,
+      "step": 4642
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.005201807620215e-06,
+      "loss": 0.8052,
+      "step": 4643
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.004321116143496e-06,
+      "loss": 0.6522,
+      "step": 4644
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00344027876831e-06,
+      "loss": 0.6148,
+      "step": 4645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.00255929553744e-06,
+      "loss": 0.8492,
+      "step": 4646
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.001678166493663e-06,
+      "loss": 0.6685,
+      "step": 4647
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.000796891679771e-06,
+      "loss": 0.5767,
+      "step": 4648
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999915471138562e-06,
+      "loss": 0.7486,
+      "step": 4649
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.999033904912837e-06,
+      "loss": 0.6774,
+      "step": 4650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.998152193045412e-06,
+      "loss": 0.8223,
+      "step": 4651
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.997270335579102e-06,
+      "loss": 0.8073,
+      "step": 4652
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.996388332556735e-06,
+      "loss": 1.0251,
+      "step": 4653
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.995506184021143e-06,
+      "loss": 0.8343,
+      "step": 4654
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.994623890015164e-06,
+      "loss": 0.733,
+      "step": 4655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.993741450581646e-06,
+      "loss": 0.7104,
+      "step": 4656
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.992858865763444e-06,
+      "loss": 0.7541,
+      "step": 4657
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991976135603417e-06,
+      "loss": 0.8873,
+      "step": 4658
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.991093260144431e-06,
+      "loss": 0.8111,
+      "step": 4659
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.990210239429366e-06,
+      "loss": 0.7696,
+      "step": 4660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.989327073501101e-06,
+      "loss": 0.6559,
+      "step": 4661
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.988443762402525e-06,
+      "loss": 0.794,
+      "step": 4662
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.987560306176535e-06,
+      "loss": 0.6523,
+      "step": 4663
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.986676704866035e-06,
+      "loss": 0.6438,
+      "step": 4664
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.985792958513932e-06,
+      "loss": 0.4297,
+      "step": 4665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.984909067163147e-06,
+      "loss": 0.6192,
+      "step": 4666
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9840250308566e-06,
+      "loss": 0.8606,
+      "step": 4667
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.983140849637227e-06,
+      "loss": 0.7823,
+      "step": 4668
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.982256523547964e-06,
+      "loss": 0.7453,
+      "step": 4669
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.981372052631757e-06,
+      "loss": 0.819,
+      "step": 4670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.980487436931558e-06,
+      "loss": 0.8538,
+      "step": 4671
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.979602676490325e-06,
+      "loss": 0.5461,
+      "step": 4672
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97871777135103e-06,
+      "loss": 0.7818,
+      "step": 4673
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.97783272155664e-06,
+      "loss": 0.9359,
+      "step": 4674
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976947527150137e-06,
+      "loss": 0.9971,
+      "step": 4675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.976062188174513e-06,
+      "loss": 0.6862,
+      "step": 4676
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.975176704672757e-06,
+      "loss": 0.7988,
+      "step": 4677
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.974291076687874e-06,
+      "loss": 0.4702,
+      "step": 4678
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.973405304262874e-06,
+      "loss": 0.7863,
+      "step": 4679
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.972519387440767e-06,
+      "loss": 0.6236,
+      "step": 4680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.971633326264581e-06,
+      "loss": 0.8801,
+      "step": 4681
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.970747120777342e-06,
+      "loss": 0.8759,
+      "step": 4682
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.96986077102209e-06,
+      "loss": 0.621,
+      "step": 4683
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.968974277041865e-06,
+      "loss": 0.8613,
+      "step": 4684
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.968087638879722e-06,
+      "loss": 0.6746,
+      "step": 4685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.967200856578714e-06,
+      "loss": 0.8906,
+      "step": 4686
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.966313930181912e-06,
+      "loss": 0.8808,
+      "step": 4687
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.965426859732381e-06,
+      "loss": 0.5983,
+      "step": 4688
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.964539645273204e-06,
+      "loss": 0.8553,
+      "step": 4689
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.963652286847464e-06,
+      "loss": 0.7682,
+      "step": 4690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.962764784498254e-06,
+      "loss": 0.649,
+      "step": 4691
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.961877138268677e-06,
+      "loss": 0.7983,
+      "step": 4692
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960989348201837e-06,
+      "loss": 0.7002,
+      "step": 4693
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.960101414340847e-06,
+      "loss": 0.7793,
+      "step": 4694
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.959213336728828e-06,
+      "loss": 0.658,
+      "step": 4695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.95832511540891e-06,
+      "loss": 1.0912,
+      "step": 4696
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.957436750424223e-06,
+      "loss": 0.6396,
+      "step": 4697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.956548241817914e-06,
+      "loss": 0.9055,
+      "step": 4698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.955659589633126e-06,
+      "loss": 0.7494,
+      "step": 4699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.954770793913018e-06,
+      "loss": 0.7089,
+      "step": 4700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.953881854700754e-06,
+      "loss": 0.5153,
+      "step": 4701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9529927720395e-06,
+      "loss": 0.98,
+      "step": 4702
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.952103545972431e-06,
+      "loss": 0.6709,
+      "step": 4703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.951214176542735e-06,
+      "loss": 0.5946,
+      "step": 4704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9503246637936e-06,
+      "loss": 0.8149,
+      "step": 4705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.949435007768225e-06,
+      "loss": 0.8313,
+      "step": 4706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.948545208509811e-06,
+      "loss": 0.6847,
+      "step": 4707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.947655266061572e-06,
+      "loss": 0.8669,
+      "step": 4708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.946765180466725e-06,
+      "loss": 0.5057,
+      "step": 4709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.945874951768495e-06,
+      "loss": 0.9442,
+      "step": 4710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944984580010116e-06,
+      "loss": 1.0194,
+      "step": 4711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.944094065234823e-06,
+      "loss": 0.7048,
+      "step": 4712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.943203407485864e-06,
+      "loss": 0.6619,
+      "step": 4713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.942312606806495e-06,
+      "loss": 0.9722,
+      "step": 4714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.941421663239971e-06,
+      "loss": 0.613,
+      "step": 4715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.940530576829562e-06,
+      "loss": 0.7098,
+      "step": 4716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.939639347618541e-06,
+      "loss": 0.8921,
+      "step": 4717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.938747975650187e-06,
+      "loss": 0.794,
+      "step": 4718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93785646096779e-06,
+      "loss": 0.8209,
+      "step": 4719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.936964803614642e-06,
+      "loss": 0.9501,
+      "step": 4720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.936073003634045e-06,
+      "loss": 0.756,
+      "step": 4721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93518106106931e-06,
+      "loss": 0.7392,
+      "step": 4722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.93428897596375e-06,
+      "loss": 0.8633,
+      "step": 4723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.933396748360688e-06,
+      "loss": 1.0391,
+      "step": 4724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.932504378303452e-06,
+      "loss": 0.7234,
+      "step": 4725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.931611865835378e-06,
+      "loss": 0.9716,
+      "step": 4726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.930719210999812e-06,
+      "loss": 0.7621,
+      "step": 4727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.9298264138401e-06,
+      "loss": 0.5718,
+      "step": 4728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928933474399601e-06,
+      "loss": 0.7036,
+      "step": 4729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.928040392721678e-06,
+      "loss": 0.8063,
+      "step": 4730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.927147168849704e-06,
+      "loss": 0.7748,
+      "step": 4731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.92625380282705e-06,
+      "loss": 0.7359,
+      "step": 4732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.925360294697109e-06,
+      "loss": 0.7551,
+      "step": 4733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.924466644503265e-06,
+      "loss": 0.8221,
+      "step": 4734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.923572852288922e-06,
+      "loss": 0.6152,
+      "step": 4735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.922678918097481e-06,
+      "loss": 0.605,
+      "step": 4736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.921784841972355e-06,
+      "loss": 0.8242,
+      "step": 4737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.920890623956965e-06,
+      "loss": 0.7387,
+      "step": 4738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.919996264094732e-06,
+      "loss": 1.0117,
+      "step": 4739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.919101762429094e-06,
+      "loss": 0.9477,
+      "step": 4740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.918207119003488e-06,
+      "loss": 0.6861,
+      "step": 4741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91731233386136e-06,
+      "loss": 0.9652,
+      "step": 4742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.916417407046166e-06,
+      "loss": 0.9129,
+      "step": 4743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.915522338601362e-06,
+      "loss": 0.722,
+      "step": 4744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91462712857042e-06,
+      "loss": 0.7483,
+      "step": 4745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.91373177699681e-06,
+      "loss": 0.7157,
+      "step": 4746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.912836283924014e-06,
+      "loss": 0.5705,
+      "step": 4747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911940649395521e-06,
+      "loss": 0.7863,
+      "step": 4748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.911044873454823e-06,
+      "loss": 1.1284,
+      "step": 4749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.910148956145425e-06,
+      "loss": 0.9308,
+      "step": 4750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.909252897510833e-06,
+      "loss": 0.778,
+      "step": 4751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.908356697594562e-06,
+      "loss": 0.6201,
+      "step": 4752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.907460356440133e-06,
+      "loss": 1.056,
+      "step": 4753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90656387409108e-06,
+      "loss": 0.6248,
+      "step": 4754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.905667250590932e-06,
+      "loss": 0.5503,
+      "step": 4755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.904770485983236e-06,
+      "loss": 0.869,
+      "step": 4756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.90387358031154e-06,
+      "loss": 0.9011,
+      "step": 4757
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902976533619401e-06,
+      "loss": 0.8277,
+      "step": 4758
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.902079345950385e-06,
+      "loss": 0.6372,
+      "step": 4759
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.901182017348055e-06,
+      "loss": 0.67,
+      "step": 4760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.900284547855992e-06,
+      "loss": 0.9573,
+      "step": 4761
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.899386937517778e-06,
+      "loss": 0.6887,
+      "step": 4762
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.898489186377009e-06,
+      "loss": 0.8135,
+      "step": 4763
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.897591294477276e-06,
+      "loss": 0.7636,
+      "step": 4764
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.896693261862185e-06,
+      "loss": 0.6934,
+      "step": 4765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.895795088575348e-06,
+      "loss": 0.8921,
+      "step": 4766
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.894896774660382e-06,
+      "loss": 0.9207,
+      "step": 4767
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.893998320160913e-06,
+      "loss": 0.7232,
+      "step": 4768
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89309972512057e-06,
+      "loss": 0.908,
+      "step": 4769
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.892200989582994e-06,
+      "loss": 0.9619,
+      "step": 4770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.89130211359183e-06,
+      "loss": 0.987,
+      "step": 4771
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.890403097190729e-06,
+      "loss": 0.8549,
+      "step": 4772
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.88950394042335e-06,
+      "loss": 0.5622,
+      "step": 4773
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.888604643333357e-06,
+      "loss": 0.7658,
+      "step": 4774
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.887705205964426e-06,
+      "loss": 0.6614,
+      "step": 4775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.886805628360234e-06,
+      "loss": 0.6811,
+      "step": 4776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.885905910564466e-06,
+      "loss": 0.9146,
+      "step": 4777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.885006052620819e-06,
+      "loss": 0.7805,
+      "step": 4778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.884106054572987e-06,
+      "loss": 0.8623,
+      "step": 4779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.88320591646468e-06,
+      "loss": 0.6467,
+      "step": 4780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.882305638339612e-06,
+      "loss": 0.9547,
+      "step": 4781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.881405220241503e-06,
+      "loss": 0.7337,
+      "step": 4782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.880504662214076e-06,
+      "loss": 0.5726,
+      "step": 4783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.879603964301069e-06,
+      "loss": 0.658,
+      "step": 4784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.87870312654622e-06,
+      "loss": 0.7879,
+      "step": 4785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.877802148993278e-06,
+      "loss": 0.7547,
+      "step": 4786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.876901031685997e-06,
+      "loss": 0.4742,
+      "step": 4787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.875999774668135e-06,
+      "loss": 0.5972,
+      "step": 4788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.875098377983462e-06,
+      "loss": 0.6628,
+      "step": 4789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.874196841675754e-06,
+      "loss": 0.849,
+      "step": 4790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.873295165788787e-06,
+      "loss": 0.6663,
+      "step": 4791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.872393350366356e-06,
+      "loss": 0.7266,
+      "step": 4792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.87149139545225e-06,
+      "loss": 0.5769,
+      "step": 4793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.870589301090274e-06,
+      "loss": 0.5474,
+      "step": 4794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.869687067324235e-06,
+      "loss": 0.9086,
+      "step": 4795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.868784694197947e-06,
+      "loss": 0.6839,
+      "step": 4796
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86788218175523e-06,
+      "loss": 0.8597,
+      "step": 4797
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86697953003992e-06,
+      "loss": 0.6906,
+      "step": 4798
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.866076739095846e-06,
+      "loss": 0.716,
+      "step": 4799
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.865173808966851e-06,
+      "loss": 0.5714,
+      "step": 4800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.864270739696785e-06,
+      "loss": 0.6855,
+      "step": 4801
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.863367531329503e-06,
+      "loss": 0.639,
+      "step": 4802
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.862464183908869e-06,
+      "loss": 0.7052,
+      "step": 4803
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.86156069747875e-06,
+      "loss": 1.1251,
+      "step": 4804
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.860657072083022e-06,
+      "loss": 0.8605,
+      "step": 4805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.859753307765571e-06,
+      "loss": 0.6675,
+      "step": 4806
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.858849404570282e-06,
+      "loss": 0.6228,
+      "step": 4807
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857945362541053e-06,
+      "loss": 0.795,
+      "step": 4808
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.857041181721788e-06,
+      "loss": 0.6974,
+      "step": 4809
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.856136862156395e-06,
+      "loss": 0.6744,
+      "step": 4810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.855232403888789e-06,
+      "loss": 0.9701,
+      "step": 4811
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.854327806962896e-06,
+      "loss": 0.6719,
+      "step": 4812
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.853423071422644e-06,
+      "loss": 1.0354,
+      "step": 4813
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.852518197311972e-06,
+      "loss": 0.7331,
+      "step": 4814
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.851613184674821e-06,
+      "loss": 0.6916,
+      "step": 4815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.850708033555142e-06,
+      "loss": 0.8126,
+      "step": 4816
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.849802743996891e-06,
+      "loss": 0.7149,
+      "step": 4817
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.84889731604403e-06,
+      "loss": 0.9407,
+      "step": 4818
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.847991749740533e-06,
+      "loss": 0.6051,
+      "step": 4819
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.847086045130374e-06,
+      "loss": 0.8816,
+      "step": 4820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.846180202257539e-06,
+      "loss": 0.7812,
+      "step": 4821
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.845274221166015e-06,
+      "loss": 1.0128,
+      "step": 4822
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.844368101899801e-06,
+      "loss": 0.7738,
+      "step": 4823
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.843461844502903e-06,
+      "loss": 0.9808,
+      "step": 4824
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.842555449019326e-06,
+      "loss": 0.8205,
+      "step": 4825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.841648915493092e-06,
+      "loss": 0.6107,
+      "step": 4826
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.840742243968222e-06,
+      "loss": 0.6255,
+      "step": 4827
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.839835434488748e-06,
+      "loss": 0.8649,
+      "step": 4828
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.838928487098707e-06,
+      "loss": 0.8069,
+      "step": 4829
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.838021401842145e-06,
+      "loss": 0.6563,
+      "step": 4830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.837114178763108e-06,
+      "loss": 0.8003,
+      "step": 4831
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.836206817905655e-06,
+      "loss": 1.0281,
+      "step": 4832
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.835299319313854e-06,
+      "loss": 0.8938,
+      "step": 4833
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.834391683031773e-06,
+      "loss": 0.8298,
+      "step": 4834
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.833483909103487e-06,
+      "loss": 0.9072,
+      "step": 4835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.832575997573082e-06,
+      "loss": 0.7033,
+      "step": 4836
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.831667948484651e-06,
+      "loss": 0.8779,
+      "step": 4837
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.83075976188229e-06,
+      "loss": 0.7842,
+      "step": 4838
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.829851437810099e-06,
+      "loss": 0.9074,
+      "step": 4839
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.828942976312197e-06,
+      "loss": 0.7292,
+      "step": 4840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.828034377432694e-06,
+      "loss": 0.6061,
+      "step": 4841
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.827125641215718e-06,
+      "loss": 0.7248,
+      "step": 4842
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.8262167677054e-06,
+      "loss": 0.9962,
+      "step": 4843
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.825307756945874e-06,
+      "loss": 0.6861,
+      "step": 4844
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.82439860898129e-06,
+      "loss": 0.8565,
+      "step": 4845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.823489323855794e-06,
+      "loss": 0.8335,
+      "step": 4846
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.822579901613546e-06,
+      "loss": 0.8066,
+      "step": 4847
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.821670342298706e-06,
+      "loss": 0.718,
+      "step": 4848
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.820760645955452e-06,
+      "loss": 0.9073,
+      "step": 4849
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.819850812627956e-06,
+      "loss": 0.8344,
+      "step": 4850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.818940842360404e-06,
+      "loss": 0.5193,
+      "step": 4851
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.818030735196984e-06,
+      "loss": 0.8215,
+      "step": 4852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.817120491181897e-06,
+      "loss": 0.7198,
+      "step": 4853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.816210110359347e-06,
+      "loss": 0.7484,
+      "step": 4854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.815299592773545e-06,
+      "loss": 0.9153,
+      "step": 4855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.814388938468704e-06,
+      "loss": 0.8063,
+      "step": 4856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.813478147489052e-06,
+      "loss": 0.6581,
+      "step": 4857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.812567219878818e-06,
+      "loss": 0.7423,
+      "step": 4858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.811656155682241e-06,
+      "loss": 0.8281,
+      "step": 4859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.810744954943564e-06,
+      "loss": 0.9144,
+      "step": 4860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.809833617707035e-06,
+      "loss": 1.023,
+      "step": 4861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.808922144016915e-06,
+      "loss": 0.9133,
+      "step": 4862
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.808010533917465e-06,
+      "loss": 0.8181,
+      "step": 4863
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.807098787452958e-06,
+      "loss": 0.7671,
+      "step": 4864
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80618690466767e-06,
+      "loss": 0.7099,
+      "step": 4865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.805274885605883e-06,
+      "loss": 1.0023,
+      "step": 4866
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.804362730311887e-06,
+      "loss": 0.6587,
+      "step": 4867
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.803450438829983e-06,
+      "loss": 0.7427,
+      "step": 4868
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.80253801120447e-06,
+      "loss": 0.9583,
+      "step": 4869
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.801625447479661e-06,
+      "loss": 0.8312,
+      "step": 4870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.800712747699873e-06,
+      "loss": 0.8117,
+      "step": 4871
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.799799911909427e-06,
+      "loss": 0.5947,
+      "step": 4872
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.798886940152654e-06,
+      "loss": 0.8052,
+      "step": 4873
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79797383247389e-06,
+      "loss": 0.5967,
+      "step": 4874
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.797060588917477e-06,
+      "loss": 0.8533,
+      "step": 4875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.796147209527769e-06,
+      "loss": 0.8985,
+      "step": 4876
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.795233694349119e-06,
+      "loss": 0.8198,
+      "step": 4877
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.79432004342589e-06,
+      "loss": 0.7529,
+      "step": 4878
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.793406256802453e-06,
+      "loss": 0.6908,
+      "step": 4879
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.792492334523182e-06,
+      "loss": 0.6187,
+      "step": 4880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.791578276632461e-06,
+      "loss": 0.8785,
+      "step": 4881
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.790664083174681e-06,
+      "loss": 0.6323,
+      "step": 4882
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.789749754194237e-06,
+      "loss": 0.6631,
+      "step": 4883
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.788835289735527e-06,
+      "loss": 0.9992,
+      "step": 4884
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.787920689842965e-06,
+      "loss": 0.6973,
+      "step": 4885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.787005954560965e-06,
+      "loss": 0.9274,
+      "step": 4886
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.78609108393395e-06,
+      "loss": 0.8245,
+      "step": 4887
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.785176078006346e-06,
+      "loss": 0.6166,
+      "step": 4888
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.784260936822592e-06,
+      "loss": 0.9866,
+      "step": 4889
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.783345660427129e-06,
+      "loss": 0.614,
+      "step": 4890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.782430248864403e-06,
+      "loss": 0.7093,
+      "step": 4891
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.781514702178871e-06,
+      "loss": 0.7558,
+      "step": 4892
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.780599020414997e-06,
+      "loss": 0.7576,
+      "step": 4893
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.779683203617242e-06,
+      "loss": 0.8672,
+      "step": 4894
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.778767251830089e-06,
+      "loss": 0.9933,
+      "step": 4895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.777851165098012e-06,
+      "loss": 0.6333,
+      "step": 4896
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.776934943465503e-06,
+      "loss": 0.6421,
+      "step": 4897
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.776018586977057e-06,
+      "loss": 0.6856,
+      "step": 4898
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.775102095677172e-06,
+      "loss": 0.8554,
+      "step": 4899
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.774185469610358e-06,
+      "loss": 0.6908,
+      "step": 4900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.773268708821127e-06,
+      "loss": 0.6401,
+      "step": 4901
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.772351813354e-06,
+      "loss": 0.7679,
+      "step": 4902
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.771434783253507e-06,
+      "loss": 0.6294,
+      "step": 4903
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.770517618564176e-06,
+      "loss": 0.5455,
+      "step": 4904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.769600319330553e-06,
+      "loss": 0.7645,
+      "step": 4905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76868288559718e-06,
+      "loss": 0.6235,
+      "step": 4906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.767765317408614e-06,
+      "loss": 0.9007,
+      "step": 4907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76684761480941e-06,
+      "loss": 0.8461,
+      "step": 4908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765929777844139e-06,
+      "loss": 0.4142,
+      "step": 4909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.765011806557372e-06,
+      "loss": 0.564,
+      "step": 4910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.764093700993688e-06,
+      "loss": 0.5961,
+      "step": 4911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.763175461197673e-06,
+      "loss": 0.7222,
+      "step": 4912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.76225708721392e-06,
+      "loss": 0.5112,
+      "step": 4913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.761338579087026e-06,
+      "loss": 0.8123,
+      "step": 4914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.760419936861599e-06,
+      "loss": 0.8014,
+      "step": 4915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.75950116058225e-06,
+      "loss": 0.697,
+      "step": 4916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.758582250293596e-06,
+      "loss": 0.8262,
+      "step": 4917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.757663206040264e-06,
+      "loss": 0.7416,
+      "step": 4918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.756744027866884e-06,
+      "loss": 0.952,
+      "step": 4919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.755824715818094e-06,
+      "loss": 0.8875,
+      "step": 4920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.75490526993854e-06,
+      "loss": 0.5228,
+      "step": 4921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.753985690272871e-06,
+      "loss": 0.6284,
+      "step": 4922
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.753065976865745e-06,
+      "loss": 0.3453,
+      "step": 4923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.752146129761826e-06,
+      "loss": 0.7935,
+      "step": 4924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.751226149005785e-06,
+      "loss": 0.7827,
+      "step": 4925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.750306034642298e-06,
+      "loss": 0.814,
+      "step": 4926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.749385786716048e-06,
+      "loss": 0.5518,
+      "step": 4927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.748465405271728e-06,
+      "loss": 0.9052,
+      "step": 4928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.747544890354031e-06,
+      "loss": 0.8163,
+      "step": 4929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.74662424200766e-06,
+      "loss": 0.964,
+      "step": 4930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.745703460277327e-06,
+      "loss": 0.9677,
+      "step": 4931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.744782545207745e-06,
+      "loss": 0.6555,
+      "step": 4932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.743861496843637e-06,
+      "loss": 0.8494,
+      "step": 4933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.742940315229733e-06,
+      "loss": 0.6181,
+      "step": 4934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.742019000410767e-06,
+      "loss": 0.6444,
+      "step": 4935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.741097552431481e-06,
+      "loss": 0.7288,
+      "step": 4936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.740175971336624e-06,
+      "loss": 0.8102,
+      "step": 4937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.739254257170947e-06,
+      "loss": 0.7254,
+      "step": 4938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.738332409979216e-06,
+      "loss": 0.6953,
+      "step": 4939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.737410429806197e-06,
+      "loss": 0.811,
+      "step": 4940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.736488316696663e-06,
+      "loss": 0.753,
+      "step": 4941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.735566070695394e-06,
+      "loss": 0.8536,
+      "step": 4942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.734643691847178e-06,
+      "loss": 0.726,
+      "step": 4943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.73372118019681e-06,
+      "loss": 0.6398,
+      "step": 4944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.732798535789086e-06,
+      "loss": 0.7031,
+      "step": 4945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.731875758668815e-06,
+      "loss": 0.8628,
+      "step": 4946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730952848880809e-06,
+      "loss": 0.8842,
+      "step": 4947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.730029806469886e-06,
+      "loss": 0.8883,
+      "step": 4948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.729106631480874e-06,
+      "loss": 0.9889,
+      "step": 4949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.728183323958603e-06,
+      "loss": 0.7578,
+      "step": 4950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.727259883947913e-06,
+      "loss": 0.5768,
+      "step": 4951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.726336311493647e-06,
+      "loss": 0.7966,
+      "step": 4952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.725412606640658e-06,
+      "loss": 0.8809,
+      "step": 4953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.724488769433805e-06,
+      "loss": 0.9376,
+      "step": 4954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.723564799917949e-06,
+      "loss": 0.7244,
+      "step": 4955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.722640698137961e-06,
+      "loss": 0.6978,
+      "step": 4956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.72171646413872e-06,
+      "loss": 0.8785,
+      "step": 4957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.720792097965109e-06,
+      "loss": 0.8111,
+      "step": 4958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.719867599662017e-06,
+      "loss": 0.5374,
+      "step": 4959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71894296927434e-06,
+      "loss": 0.9343,
+      "step": 4960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.718018206846981e-06,
+      "loss": 0.8484,
+      "step": 4961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71709331242485e-06,
+      "loss": 0.7376,
+      "step": 4962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.716168286052864e-06,
+      "loss": 0.7244,
+      "step": 4963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71524312777594e-06,
+      "loss": 0.8853,
+      "step": 4964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.714317837639011e-06,
+      "loss": 0.4048,
+      "step": 4965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.71339241568701e-06,
+      "loss": 0.6247,
+      "step": 4966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.712466861964877e-06,
+      "loss": 0.6559,
+      "step": 4967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.711541176517563e-06,
+      "loss": 0.8835,
+      "step": 4968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.710615359390018e-06,
+      "loss": 0.7999,
+      "step": 4969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.709689410627204e-06,
+      "loss": 0.6704,
+      "step": 4970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.708763330274088e-06,
+      "loss": 0.7886,
+      "step": 4971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.707837118375643e-06,
+      "loss": 0.8358,
+      "step": 4972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.706910774976849e-06,
+      "loss": 0.7636,
+      "step": 4973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70598430012269e-06,
+      "loss": 0.8747,
+      "step": 4974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.70505769385816e-06,
+      "loss": 0.5973,
+      "step": 4975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.704130956228259e-06,
+      "loss": 0.7645,
+      "step": 4976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.703204087277989e-06,
+      "loss": 0.9497,
+      "step": 4977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.702277087052362e-06,
+      "loss": 0.8042,
+      "step": 4978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.701349955596397e-06,
+      "loss": 0.8416,
+      "step": 4979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.700422692955118e-06,
+      "loss": 0.7031,
+      "step": 4980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.699495299173554e-06,
+      "loss": 1.0981,
+      "step": 4981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.698567774296745e-06,
+      "loss": 0.689,
+      "step": 4982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69764011836973e-06,
+      "loss": 0.9104,
+      "step": 4983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.696712331437565e-06,
+      "loss": 0.8086,
+      "step": 4984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6957844135453e-06,
+      "loss": 0.6499,
+      "step": 4985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.694856364737997e-06,
+      "loss": 0.8269,
+      "step": 4986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69392818506073e-06,
+      "loss": 0.7701,
+      "step": 4987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.69299987455857e-06,
+      "loss": 0.6883,
+      "step": 4988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6920714332766e-06,
+      "loss": 0.7205,
+      "step": 4989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.691142861259909e-06,
+      "loss": 0.9373,
+      "step": 4990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.690214158553586e-06,
+      "loss": 0.7253,
+      "step": 4991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.689285325202735e-06,
+      "loss": 0.9155,
+      "step": 4992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.688356361252466e-06,
+      "loss": 0.5142,
+      "step": 4993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.687427266747886e-06,
+      "loss": 0.9313,
+      "step": 4994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.686498041734121e-06,
+      "loss": 0.8265,
+      "step": 4995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.685568686256291e-06,
+      "loss": 0.7463,
+      "step": 4996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.684639200359529e-06,
+      "loss": 0.7921,
+      "step": 4997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.683709584088976e-06,
+      "loss": 0.9222,
+      "step": 4998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.682779837489777e-06,
+      "loss": 0.8215,
+      "step": 4999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.681849960607081e-06,
+      "loss": 0.807,
+      "step": 5000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.680919953486047e-06,
+      "loss": 0.864,
+      "step": 5001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.679989816171838e-06,
+      "loss": 0.7861,
+      "step": 5002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.679059548709624e-06,
+      "loss": 0.813,
+      "step": 5003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.678129151144582e-06,
+      "loss": 0.8587,
+      "step": 5004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.677198623521894e-06,
+      "loss": 0.705,
+      "step": 5005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.676267965886753e-06,
+      "loss": 0.6254,
+      "step": 5006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.675337178284347e-06,
+      "loss": 0.8794,
+      "step": 5007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.674406260759884e-06,
+      "loss": 0.6688,
+      "step": 5008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67347521335857e-06,
+      "loss": 0.6905,
+      "step": 5009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.672544036125618e-06,
+      "loss": 0.7303,
+      "step": 5010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.67161272910625e-06,
+      "loss": 0.9397,
+      "step": 5011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.670681292345694e-06,
+      "loss": 0.5811,
+      "step": 5012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.669749725889182e-06,
+      "loss": 0.6197,
+      "step": 5013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.668818029781952e-06,
+      "loss": 0.6021,
+      "step": 5014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.667886204069253e-06,
+      "loss": 0.9616,
+      "step": 5015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666954248796335e-06,
+      "loss": 0.8168,
+      "step": 5016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.666022164008458e-06,
+      "loss": 0.8826,
+      "step": 5017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.665089949750883e-06,
+      "loss": 0.6141,
+      "step": 5018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.664157606068885e-06,
+      "loss": 0.8633,
+      "step": 5019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.66322513300774e-06,
+      "loss": 0.8681,
+      "step": 5020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.662292530612734e-06,
+      "loss": 0.7502,
+      "step": 5021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.661359798929152e-06,
+      "loss": 0.7839,
+      "step": 5022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.660426938002293e-06,
+      "loss": 0.3887,
+      "step": 5023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.659493947877459e-06,
+      "loss": 0.5399,
+      "step": 5024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65856082859996e-06,
+      "loss": 0.6073,
+      "step": 5025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.657627580215107e-06,
+      "loss": 0.7443,
+      "step": 5026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.656694202768225e-06,
+      "loss": 0.6587,
+      "step": 5027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.655760696304642e-06,
+      "loss": 0.9395,
+      "step": 5028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.654827060869687e-06,
+      "loss": 0.5275,
+      "step": 5029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.653893296508705e-06,
+      "loss": 0.625,
+      "step": 5030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.65295940326704e-06,
+      "loss": 0.82,
+      "step": 5031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.652025381190045e-06,
+      "loss": 0.6692,
+      "step": 5032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.651091230323079e-06,
+      "loss": 0.7892,
+      "step": 5033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.650156950711506e-06,
+      "loss": 0.6575,
+      "step": 5034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.649222542400698e-06,
+      "loss": 0.5473,
+      "step": 5035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.648288005436034e-06,
+      "loss": 0.9647,
+      "step": 5036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.647353339862895e-06,
+      "loss": 0.915,
+      "step": 5037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.646418545726673e-06,
+      "loss": 1.0301,
+      "step": 5038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.645483623072764e-06,
+      "loss": 0.8235,
+      "step": 5039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.644548571946569e-06,
+      "loss": 0.8144,
+      "step": 5040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.643613392393498e-06,
+      "loss": 0.778,
+      "step": 5041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.642678084458968e-06,
+      "loss": 0.9328,
+      "step": 5042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.641742648188397e-06,
+      "loss": 0.7801,
+      "step": 5043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.640807083627214e-06,
+      "loss": 0.669,
+      "step": 5044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.639871390820853e-06,
+      "loss": 0.6646,
+      "step": 5045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.638935569814752e-06,
+      "loss": 0.9052,
+      "step": 5046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.637999620654359e-06,
+      "loss": 0.6178,
+      "step": 5047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.637063543385128e-06,
+      "loss": 0.9546,
+      "step": 5048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.636127338052513e-06,
+      "loss": 0.7924,
+      "step": 5049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.635191004701981e-06,
+      "loss": 0.6635,
+      "step": 5050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.634254543379005e-06,
+      "loss": 0.564,
+      "step": 5051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.63331795412906e-06,
+      "loss": 0.6064,
+      "step": 5052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.632381236997633e-06,
+      "loss": 0.7024,
+      "step": 5053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.631444392030206e-06,
+      "loss": 0.7518,
+      "step": 5054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.630507419272281e-06,
+      "loss": 0.8025,
+      "step": 5055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.62957031876936e-06,
+      "loss": 0.7385,
+      "step": 5056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.628633090566949e-06,
+      "loss": 0.7555,
+      "step": 5057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.627695734710565e-06,
+      "loss": 0.7476,
+      "step": 5058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6267582512457275e-06,
+      "loss": 0.7793,
+      "step": 5059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6258206402179604e-06,
+      "loss": 0.8153,
+      "step": 5060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.624882901672801e-06,
+      "loss": 0.8633,
+      "step": 5061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.623945035655787e-06,
+      "loss": 0.7444,
+      "step": 5062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.623007042212466e-06,
+      "loss": 0.7417,
+      "step": 5063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.622068921388384e-06,
+      "loss": 0.653,
+      "step": 5064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.621130673229105e-06,
+      "loss": 0.7336,
+      "step": 5065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.620192297780189e-06,
+      "loss": 0.8137,
+      "step": 5066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.619253795087209e-06,
+      "loss": 0.6509,
+      "step": 5067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.61831516519574e-06,
+      "loss": 0.8519,
+      "step": 5068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.617376408151363e-06,
+      "loss": 0.6915,
+      "step": 5069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.6164375239996694e-06,
+      "loss": 0.7119,
+      "step": 5070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.615498512786252e-06,
+      "loss": 0.8594,
+      "step": 5071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.614559374556716e-06,
+      "loss": 0.7006,
+      "step": 5072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.613620109356663e-06,
+      "loss": 0.6797,
+      "step": 5073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.612680717231709e-06,
+      "loss": 0.666,
+      "step": 5074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.611741198227475e-06,
+      "loss": 0.9733,
+      "step": 5075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.610801552389584e-06,
+      "loss": 0.7424,
+      "step": 5076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.609861779763671e-06,
+      "loss": 0.9987,
+      "step": 5077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60892188039537e-06,
+      "loss": 0.6403,
+      "step": 5078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60798185433033e-06,
+      "loss": 0.696,
+      "step": 5079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.607041701614198e-06,
+      "loss": 0.6179,
+      "step": 5080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.606101422292629e-06,
+      "loss": 0.8067,
+      "step": 5081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.605161016411291e-06,
+      "loss": 1.006,
+      "step": 5082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.60422048401585e-06,
+      "loss": 0.6117,
+      "step": 5083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.603279825151979e-06,
+      "loss": 0.8111,
+      "step": 5084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.602339039865362e-06,
+      "loss": 0.6948,
+      "step": 5085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.601398128201685e-06,
+      "loss": 0.8428,
+      "step": 5086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.6004570902066425e-06,
+      "loss": 0.7007,
+      "step": 5087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.599515925925931e-06,
+      "loss": 0.8539,
+      "step": 5088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5985746354052595e-06,
+      "loss": 0.9391,
+      "step": 5089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.597633218690339e-06,
+      "loss": 0.7791,
+      "step": 5090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.596691675826886e-06,
+      "loss": 0.6714,
+      "step": 5091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.595750006860625e-06,
+      "loss": 0.7414,
+      "step": 5092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.594808211837288e-06,
+      "loss": 0.7621,
+      "step": 5093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.593866290802608e-06,
+      "loss": 0.7313,
+      "step": 5094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.592924243802329e-06,
+      "loss": 0.7985,
+      "step": 5095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5919820708822025e-06,
+      "loss": 0.7658,
+      "step": 5096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5910397720879785e-06,
+      "loss": 0.9354,
+      "step": 5097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.590097347465419e-06,
+      "loss": 0.6762,
+      "step": 5098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.589154797060292e-06,
+      "loss": 0.6796,
+      "step": 5099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5882121209183704e-06,
+      "loss": 0.7125,
+      "step": 5100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.587269319085433e-06,
+      "loss": 0.7614,
+      "step": 5101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5863263916072636e-06,
+      "loss": 0.6526,
+      "step": 5102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5853833385296545e-06,
+      "loss": 0.8067,
+      "step": 5103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584440159898405e-06,
+      "loss": 0.9494,
+      "step": 5104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5834968557593155e-06,
+      "loss": 1.0059,
+      "step": 5105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.582553426158196e-06,
+      "loss": 0.7362,
+      "step": 5106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5816098711408644e-06,
+      "loss": 0.73,
+      "step": 5107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.580666190753141e-06,
+      "loss": 0.5355,
+      "step": 5108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.579722385040852e-06,
+      "loss": 0.6754,
+      "step": 5109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.578778454049834e-06,
+      "loss": 0.8554,
+      "step": 5110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.577834397825925e-06,
+      "loss": 0.7224,
+      "step": 5111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.576890216414973e-06,
+      "loss": 0.5296,
+      "step": 5112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.575945909862829e-06,
+      "loss": 0.7637,
+      "step": 5113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.575001478215351e-06,
+      "loss": 0.9581,
+      "step": 5114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5740569215184044e-06,
+      "loss": 0.6561,
+      "step": 5115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.573112239817857e-06,
+      "loss": 0.625,
+      "step": 5116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5721674331595895e-06,
+      "loss": 0.8347,
+      "step": 5117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.571222501589481e-06,
+      "loss": 0.7531,
+      "step": 5118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.570277445153422e-06,
+      "loss": 0.7447,
+      "step": 5119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.569332263897305e-06,
+      "loss": 0.541,
+      "step": 5120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.568386957867033e-06,
+      "loss": 0.6174,
+      "step": 5121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.567441527108511e-06,
+      "loss": 0.7271,
+      "step": 5122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.566495971667654e-06,
+      "loss": 0.6534,
+      "step": 5123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.565550291590378e-06,
+      "loss": 0.8348,
+      "step": 5124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.564604486922611e-06,
+      "loss": 0.8592,
+      "step": 5125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.563658557710281e-06,
+      "loss": 0.9638,
+      "step": 5126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.562712503999327e-06,
+      "loss": 0.6923,
+      "step": 5127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.561766325835692e-06,
+      "loss": 0.6658,
+      "step": 5128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5608200232653254e-06,
+      "loss": 0.6262,
+      "step": 5129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.559873596334179e-06,
+      "loss": 0.4496,
+      "step": 5130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.558927045088219e-06,
+      "loss": 0.812,
+      "step": 5131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.55798036957341e-06,
+      "loss": 0.6898,
+      "step": 5132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.557033569835726e-06,
+      "loss": 0.8707,
+      "step": 5133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.556086645921144e-06,
+      "loss": 0.6404,
+      "step": 5134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.555139597875654e-06,
+      "loss": 0.7675,
+      "step": 5135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.554192425745243e-06,
+      "loss": 0.9242,
+      "step": 5136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.55324512957591e-06,
+      "loss": 1.0813,
+      "step": 5137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5522977094136584e-06,
+      "loss": 0.8034,
+      "step": 5138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5513501653045e-06,
+      "loss": 0.6968,
+      "step": 5139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5504024972944445e-06,
+      "loss": 0.609,
+      "step": 5140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.549454705429518e-06,
+      "loss": 1.0148,
+      "step": 5141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.54850678975575e-06,
+      "loss": 0.8269,
+      "step": 5142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.547558750319168e-06,
+      "loss": 0.5065,
+      "step": 5143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.546610587165815e-06,
+      "loss": 0.7718,
+      "step": 5144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.545662300341736e-06,
+      "loss": 0.9522,
+      "step": 5145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.544713889892982e-06,
+      "loss": 0.9752,
+      "step": 5146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.543765355865611e-06,
+      "loss": 0.8575,
+      "step": 5147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.542816698305686e-06,
+      "loss": 0.6789,
+      "step": 5148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.541867917259278e-06,
+      "loss": 0.8017,
+      "step": 5149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.540919012772462e-06,
+      "loss": 0.6264,
+      "step": 5150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5399699848913186e-06,
+      "loss": 0.6524,
+      "step": 5151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.539020833661937e-06,
+      "loss": 0.8175,
+      "step": 5152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.538071559130409e-06,
+      "loss": 0.6645,
+      "step": 5153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.537122161342833e-06,
+      "loss": 0.7791,
+      "step": 5154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.536172640345318e-06,
+      "loss": 0.6445,
+      "step": 5155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.535222996183973e-06,
+      "loss": 0.8237,
+      "step": 5156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.534273228904916e-06,
+      "loss": 0.8211,
+      "step": 5157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5333233385542705e-06,
+      "loss": 0.6161,
+      "step": 5158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.532373325178166e-06,
+      "loss": 0.9075,
+      "step": 5159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.531423188822738e-06,
+      "loss": 0.6721,
+      "step": 5160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.530472929534126e-06,
+      "loss": 0.6018,
+      "step": 5161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529522547358481e-06,
+      "loss": 0.6173,
+      "step": 5162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.528572042341954e-06,
+      "loss": 0.6589,
+      "step": 5163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.527621414530704e-06,
+      "loss": 0.7594,
+      "step": 5164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5266706639708965e-06,
+      "loss": 0.6256,
+      "step": 5165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.525719790708703e-06,
+      "loss": 0.7928,
+      "step": 5166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5247687947903e-06,
+      "loss": 0.7957,
+      "step": 5167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.523817676261871e-06,
+      "loss": 0.8855,
+      "step": 5168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.522866435169606e-06,
+      "loss": 0.6789,
+      "step": 5169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5219150715597e-06,
+      "loss": 0.6902,
+      "step": 5170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.520963585478353e-06,
+      "loss": 1.0054,
+      "step": 5171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.520011976971771e-06,
+      "loss": 0.6979,
+      "step": 5172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.519060246086168e-06,
+      "loss": 0.8029,
+      "step": 5173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.518108392867762e-06,
+      "loss": 0.8313,
+      "step": 5174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5171564173627795e-06,
+      "loss": 1.0572,
+      "step": 5175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.516204319617451e-06,
+      "loss": 0.7178,
+      "step": 5176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.515252099678011e-06,
+      "loss": 1.0493,
+      "step": 5177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.514299757590701e-06,
+      "loss": 0.8977,
+      "step": 5178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.513347293401774e-06,
+      "loss": 0.584,
+      "step": 5179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5123947071574836e-06,
+      "loss": 0.8076,
+      "step": 5180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5114419989040856e-06,
+      "loss": 0.7288,
+      "step": 5181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.510489168687851e-06,
+      "loss": 0.855,
+      "step": 5182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.509536216555049e-06,
+      "loss": 0.8696,
+      "step": 5183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.508583142551959e-06,
+      "loss": 0.7795,
+      "step": 5184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.507629946724866e-06,
+      "loss": 1.0367,
+      "step": 5185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.506676629120058e-06,
+      "loss": 0.6361,
+      "step": 5186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.505723189783831e-06,
+      "loss": 0.7231,
+      "step": 5187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.504769628762488e-06,
+      "loss": 0.7938,
+      "step": 5188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.503815946102338e-06,
+      "loss": 0.787,
+      "step": 5189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.5028621418496905e-06,
+      "loss": 0.6152,
+      "step": 5190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.501908216050867e-06,
+      "loss": 0.7051,
+      "step": 5191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500954168752193e-06,
+      "loss": 0.8824,
+      "step": 5192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.9019,
+      "step": 5193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.499045709840626e-06,
+      "loss": 0.6259,
+      "step": 5194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.498091298320412e-06,
+      "loss": 0.6174,
+      "step": 5195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.497136765485708e-06,
+      "loss": 0.6352,
+      "step": 5196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.49618211138287e-06,
+      "loss": 0.6983,
+      "step": 5197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.495227336058256e-06,
+      "loss": 0.7555,
+      "step": 5198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.494272439558235e-06,
+      "loss": 0.7079,
+      "step": 5199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.493317421929179e-06,
+      "loss": 1.0292,
+      "step": 5200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.492362283217467e-06,
+      "loss": 0.8883,
+      "step": 5201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4914070234694804e-06,
+      "loss": 0.8459,
+      "step": 5202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.490451642731612e-06,
+      "loss": 0.8223,
+      "step": 5203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.489496141050259e-06,
+      "loss": 0.6759,
+      "step": 5204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.488540518471819e-06,
+      "loss": 0.7034,
+      "step": 5205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.487584775042702e-06,
+      "loss": 0.625,
+      "step": 5206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4866289108093235e-06,
+      "loss": 0.6219,
+      "step": 5207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.485672925818102e-06,
+      "loss": 0.9086,
+      "step": 5208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.484716820115461e-06,
+      "loss": 0.6299,
+      "step": 5209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.483760593747834e-06,
+      "loss": 0.8854,
+      "step": 5210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.482804246761657e-06,
+      "loss": 0.6664,
+      "step": 5211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4818477792033725e-06,
+      "loss": 0.8947,
+      "step": 5212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.4808911911194305e-06,
+      "loss": 0.89,
+      "step": 5213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.479934482556284e-06,
+      "loss": 0.8498,
+      "step": 5214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.478977653560398e-06,
+      "loss": 0.7717,
+      "step": 5215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.47802070417823e-06,
+      "loss": 0.6089,
+      "step": 5216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.477063634456263e-06,
+      "loss": 0.5539,
+      "step": 5217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.476106444440967e-06,
+      "loss": 0.6795,
+      "step": 5218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.475149134178829e-06,
+      "loss": 0.8078,
+      "step": 5219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.474191703716339e-06,
+      "loss": 0.9287,
+      "step": 5220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.473234153099991e-06,
+      "loss": 0.6735,
+      "step": 5221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.472276482376288e-06,
+      "loss": 0.7748,
+      "step": 5222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4713186915917365e-06,
+      "loss": 0.8883,
+      "step": 5223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.47036078079285e-06,
+      "loss": 0.6351,
+      "step": 5224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.469402750026147e-06,
+      "loss": 0.9876,
+      "step": 5225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.468444599338152e-06,
+      "loss": 0.7617,
+      "step": 5226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.467486328775395e-06,
+      "loss": 0.7438,
+      "step": 5227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.466527938384416e-06,
+      "loss": 0.4854,
+      "step": 5228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.465569428211752e-06,
+      "loss": 0.7271,
+      "step": 5229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.464610798303953e-06,
+      "loss": 0.771,
+      "step": 5230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.463652048707576e-06,
+      "loss": 1.0846,
+      "step": 5231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.462693179469176e-06,
+      "loss": 0.6972,
+      "step": 5232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.46173419063532e-06,
+      "loss": 0.5495,
+      "step": 5233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.460775082252581e-06,
+      "loss": 0.5519,
+      "step": 5234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.459815854367533e-06,
+      "loss": 0.4938,
+      "step": 5235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.458856507026762e-06,
+      "loss": 0.6596,
+      "step": 5236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.457897040276853e-06,
+      "loss": 0.8081,
+      "step": 5237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.456937454164404e-06,
+      "loss": 0.8403,
+      "step": 5238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.455977748736015e-06,
+      "loss": 0.5925,
+      "step": 5239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.455017924038289e-06,
+      "loss": 1.0273,
+      "step": 5240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.454057980117842e-06,
+      "loss": 0.7116,
+      "step": 5241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.453097917021288e-06,
+      "loss": 0.8241,
+      "step": 5242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.452137734795251e-06,
+      "loss": 0.8747,
+      "step": 5243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.451177433486362e-06,
+      "loss": 0.7344,
+      "step": 5244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.450217013141256e-06,
+      "loss": 0.7069,
+      "step": 5245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.449256473806572e-06,
+      "loss": 0.6336,
+      "step": 5246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.448295815528956e-06,
+      "loss": 0.6732,
+      "step": 5247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.447335038355064e-06,
+      "loss": 0.7421,
+      "step": 5248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.446374142331551e-06,
+      "loss": 1.0397,
+      "step": 5249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.445413127505081e-06,
+      "loss": 0.879,
+      "step": 5250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.444451993922325e-06,
+      "loss": 0.8121,
+      "step": 5251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.443490741629957e-06,
+      "loss": 0.9638,
+      "step": 5252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.442529370674661e-06,
+      "loss": 0.6344,
+      "step": 5253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.441567881103118e-06,
+      "loss": 0.9191,
+      "step": 5254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.440606272962027e-06,
+      "loss": 0.7831,
+      "step": 5255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.439644546298084e-06,
+      "loss": 0.767,
+      "step": 5256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.438682701157993e-06,
+      "loss": 0.6799,
+      "step": 5257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.437720737588464e-06,
+      "loss": 0.9373,
+      "step": 5258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4367586556362125e-06,
+      "loss": 0.6329,
+      "step": 5259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4357964553479615e-06,
+      "loss": 0.9578,
+      "step": 5260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.434834136770435e-06,
+      "loss": 0.7367,
+      "step": 5261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.433871699950369e-06,
+      "loss": 0.7954,
+      "step": 5262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.432909144934502e-06,
+      "loss": 0.9472,
+      "step": 5263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.431946471769576e-06,
+      "loss": 0.6453,
+      "step": 5264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.430983680502344e-06,
+      "loss": 0.8239,
+      "step": 5265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.430020771179562e-06,
+      "loss": 0.5781,
+      "step": 5266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.429057743847987e-06,
+      "loss": 0.8228,
+      "step": 5267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.428094598554392e-06,
+      "loss": 0.6526,
+      "step": 5268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.427131335345547e-06,
+      "loss": 0.7398,
+      "step": 5269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.426167954268232e-06,
+      "loss": 0.9939,
+      "step": 5270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.425204455369231e-06,
+      "loss": 1.0279,
+      "step": 5271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4242408386953346e-06,
+      "loss": 0.8694,
+      "step": 5272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.423277104293338e-06,
+      "loss": 0.6355,
+      "step": 5273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.422313252210045e-06,
+      "loss": 0.505,
+      "step": 5274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.42134928249226e-06,
+      "loss": 0.7461,
+      "step": 5275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.420385195186799e-06,
+      "loss": 0.7602,
+      "step": 5276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.41942099034048e-06,
+      "loss": 0.6026,
+      "step": 5277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4184566680001255e-06,
+      "loss": 0.7789,
+      "step": 5278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.417492228212568e-06,
+      "loss": 0.7234,
+      "step": 5279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.416527671024645e-06,
+      "loss": 0.8229,
+      "step": 5280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.415562996483193e-06,
+      "loss": 0.9053,
+      "step": 5281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.414598204635064e-06,
+      "loss": 0.8628,
+      "step": 5282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.413633295527109e-06,
+      "loss": 0.6671,
+      "step": 5283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.412668269206187e-06,
+      "loss": 0.7137,
+      "step": 5284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.411703125719163e-06,
+      "loss": 0.8569,
+      "step": 5285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.410737865112907e-06,
+      "loss": 0.6917,
+      "step": 5286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.4097724874342946e-06,
+      "loss": 0.8486,
+      "step": 5287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.408806992730208e-06,
+      "loss": 0.7413,
+      "step": 5288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.407841381047533e-06,
+      "loss": 0.7233,
+      "step": 5289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.406875652433162e-06,
+      "loss": 0.8855,
+      "step": 5290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.405909806933998e-06,
+      "loss": 0.6974,
+      "step": 5291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.404943844596939e-06,
+      "loss": 0.5652,
+      "step": 5292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.403977765468899e-06,
+      "loss": 0.8218,
+      "step": 5293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.403011569596793e-06,
+      "loss": 0.589,
+      "step": 5294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.402045257027541e-06,
+      "loss": 0.9292,
+      "step": 5295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.401078827808072e-06,
+      "loss": 0.6835,
+      "step": 5296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.400112281985316e-06,
+      "loss": 0.8938,
+      "step": 5297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.399145619606212e-06,
+      "loss": 0.8074,
+      "step": 5298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.398178840717705e-06,
+      "loss": 0.6829,
+      "step": 5299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.397211945366745e-06,
+      "loss": 0.7128,
+      "step": 5300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.396244933600285e-06,
+      "loss": 0.7637,
+      "step": 5301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.395277805465287e-06,
+      "loss": 0.7858,
+      "step": 5302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.394310561008717e-06,
+      "loss": 0.6712,
+      "step": 5303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.393343200277551e-06,
+      "loss": 0.7423,
+      "step": 5304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.392375723318761e-06,
+      "loss": 0.9495,
+      "step": 5305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.391408130179332e-06,
+      "loss": 0.7521,
+      "step": 5306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.390440420906257e-06,
+      "loss": 0.8665,
+      "step": 5307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.389472595546525e-06,
+      "loss": 0.8159,
+      "step": 5308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3885046541471404e-06,
+      "loss": 0.7736,
+      "step": 5309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.387536596755108e-06,
+      "loss": 0.6695,
+      "step": 5310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.38656842341744e-06,
+      "loss": 0.738,
+      "step": 5311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.385600134181152e-06,
+      "loss": 0.888,
+      "step": 5312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.38463172909327e-06,
+      "loss": 0.7578,
+      "step": 5313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.383663208200819e-06,
+      "loss": 0.7984,
+      "step": 5314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.382694571550837e-06,
+      "loss": 0.8131,
+      "step": 5315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3817258191903585e-06,
+      "loss": 0.6677,
+      "step": 5316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.380756951166433e-06,
+      "loss": 0.8868,
+      "step": 5317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.379787967526112e-06,
+      "loss": 0.7449,
+      "step": 5318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.378818868316449e-06,
+      "loss": 0.879,
+      "step": 5319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.377849653584508e-06,
+      "loss": 0.7284,
+      "step": 5320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.376880323377357e-06,
+      "loss": 0.7427,
+      "step": 5321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37591087774207e-06,
+      "loss": 0.5801,
+      "step": 5322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.374941316725723e-06,
+      "loss": 0.6183,
+      "step": 5323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373971640375405e-06,
+      "loss": 0.7775,
+      "step": 5324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.373001848738203e-06,
+      "loss": 0.5529,
+      "step": 5325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3720319418612144e-06,
+      "loss": 0.6655,
+      "step": 5326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.37106191979154e-06,
+      "loss": 0.6176,
+      "step": 5327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.370091782576287e-06,
+      "loss": 0.7865,
+      "step": 5328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3691215302625695e-06,
+      "loss": 0.8682,
+      "step": 5329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.368151162897502e-06,
+      "loss": 0.8259,
+      "step": 5330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.367180680528215e-06,
+      "loss": 0.8929,
+      "step": 5331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.36621008320183e-06,
+      "loss": 0.7981,
+      "step": 5332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.365239370965487e-06,
+      "loss": 0.9343,
+      "step": 5333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.364268543866324e-06,
+      "loss": 0.7801,
+      "step": 5334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3632976019514894e-06,
+      "loss": 0.6579,
+      "step": 5335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.362326545268133e-06,
+      "loss": 0.7469,
+      "step": 5336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.361355373863415e-06,
+      "loss": 0.9505,
+      "step": 5337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.360384087784494e-06,
+      "loss": 0.8432,
+      "step": 5338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.359412687078543e-06,
+      "loss": 0.9171,
+      "step": 5339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3584411717927315e-06,
+      "loss": 0.9326,
+      "step": 5340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.357469541974243e-06,
+      "loss": 0.6036,
+      "step": 5341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3564977976702616e-06,
+      "loss": 0.5965,
+      "step": 5342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.355525938927975e-06,
+      "loss": 0.753,
+      "step": 5343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.354553965794582e-06,
+      "loss": 0.8029,
+      "step": 5344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.353581878317287e-06,
+      "loss": 0.7119,
+      "step": 5345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.352609676543292e-06,
+      "loss": 0.9489,
+      "step": 5346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.351637360519813e-06,
+      "loss": 1.0089,
+      "step": 5347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.350664930294068e-06,
+      "loss": 0.7757,
+      "step": 5348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.349692385913281e-06,
+      "loss": 0.9386,
+      "step": 5349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3487197274246826e-06,
+      "loss": 0.6766,
+      "step": 5350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.347746954875506e-06,
+      "loss": 0.8188,
+      "step": 5351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3467740683129914e-06,
+      "loss": 0.6351,
+      "step": 5352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.345801067784388e-06,
+      "loss": 0.9,
+      "step": 5353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.344827953336943e-06,
+      "loss": 0.9105,
+      "step": 5354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.343854725017919e-06,
+      "loss": 0.6483,
+      "step": 5355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.342881382874576e-06,
+      "loss": 0.8339,
+      "step": 5356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.341907926954182e-06,
+      "loss": 0.9455,
+      "step": 5357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.340934357304011e-06,
+      "loss": 0.6301,
+      "step": 5358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.339960673971343e-06,
+      "loss": 0.7224,
+      "step": 5359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.338986877003461e-06,
+      "loss": 0.7752,
+      "step": 5360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3380129664476574e-06,
+      "loss": 0.7791,
+      "step": 5361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3370389423512275e-06,
+      "loss": 0.9716,
+      "step": 5362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.336064804761474e-06,
+      "loss": 0.8674,
+      "step": 5363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.3350905537257e-06,
+      "loss": 0.891,
+      "step": 5364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.334116189291223e-06,
+      "loss": 0.7527,
+      "step": 5365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.333141711505357e-06,
+      "loss": 0.8598,
+      "step": 5366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3321671204154265e-06,
+      "loss": 0.5176,
+      "step": 5367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.331192416068761e-06,
+      "loss": 0.5259,
+      "step": 5368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.330217598512696e-06,
+      "loss": 0.9429,
+      "step": 5369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3292426677945685e-06,
+      "loss": 1.0362,
+      "step": 5370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.328267623961726e-06,
+      "loss": 0.8696,
+      "step": 5371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.327292467061518e-06,
+      "loss": 0.847,
+      "step": 5372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.326317197141304e-06,
+      "loss": 0.8204,
+      "step": 5373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.325341814248443e-06,
+      "loss": 0.7549,
+      "step": 5374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.324366318430302e-06,
+      "loss": 0.5314,
+      "step": 5375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.323390709734256e-06,
+      "loss": 0.9271,
+      "step": 5376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.322414988207683e-06,
+      "loss": 0.7426,
+      "step": 5377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.321439153897964e-06,
+      "loss": 0.7897,
+      "step": 5378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.320463206852493e-06,
+      "loss": 0.8769,
+      "step": 5379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.319487147118663e-06,
+      "loss": 0.9537,
+      "step": 5380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3185109747438725e-06,
+      "loss": 0.5236,
+      "step": 5381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.317534689775527e-06,
+      "loss": 0.6465,
+      "step": 5382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.316558292261042e-06,
+      "loss": 0.5828,
+      "step": 5383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.315581782247829e-06,
+      "loss": 0.9466,
+      "step": 5384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.314605159783313e-06,
+      "loss": 0.9535,
+      "step": 5385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.313628424914921e-06,
+      "loss": 0.8773,
+      "step": 5386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3126515776900865e-06,
+      "loss": 0.883,
+      "step": 5387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.311674618156248e-06,
+      "loss": 1.0235,
+      "step": 5388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.310697546360847e-06,
+      "loss": 0.6658,
+      "step": 5389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.309720362351337e-06,
+      "loss": 0.6457,
+      "step": 5390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.308743066175172e-06,
+      "loss": 1.0472,
+      "step": 5391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.307765657879807e-06,
+      "loss": 0.6232,
+      "step": 5392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.306788137512715e-06,
+      "loss": 0.6657,
+      "step": 5393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.305810505121363e-06,
+      "loss": 0.8355,
+      "step": 5394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.3048327607532285e-06,
+      "loss": 0.603,
+      "step": 5395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.303854904455794e-06,
+      "loss": 0.684,
+      "step": 5396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.302876936276547e-06,
+      "loss": 0.9129,
+      "step": 5397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.30189885626298e-06,
+      "loss": 0.9262,
+      "step": 5398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.300920664462592e-06,
+      "loss": 0.6969,
+      "step": 5399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.299942360922886e-06,
+      "loss": 0.6866,
+      "step": 5400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.298963945691371e-06,
+      "loss": 0.8327,
+      "step": 5401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.297985418815563e-06,
+      "loss": 0.4549,
+      "step": 5402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.297006780342981e-06,
+      "loss": 0.9233,
+      "step": 5403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.296028030321153e-06,
+      "loss": 0.9745,
+      "step": 5404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.295049168797605e-06,
+      "loss": 0.9553,
+      "step": 5405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.294070195819877e-06,
+      "loss": 0.9203,
+      "step": 5406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.293091111435511e-06,
+      "loss": 0.8199,
+      "step": 5407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2921119156920515e-06,
+      "loss": 0.7719,
+      "step": 5408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.291132608637053e-06,
+      "loss": 0.7575,
+      "step": 5409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.290153190318073e-06,
+      "loss": 0.7557,
+      "step": 5410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.289173660782675e-06,
+      "loss": 0.7647,
+      "step": 5411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2881940200784275e-06,
+      "loss": 0.5597,
+      "step": 5412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2872142682529045e-06,
+      "loss": 0.7512,
+      "step": 5413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.286234405353686e-06,
+      "loss": 0.6164,
+      "step": 5414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.285254431428358e-06,
+      "loss": 0.752,
+      "step": 5415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.284274346524508e-06,
+      "loss": 0.7947,
+      "step": 5416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.283294150689735e-06,
+      "loss": 1.0104,
+      "step": 5417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2823138439716375e-06,
+      "loss": 0.6581,
+      "step": 5418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2813334264178235e-06,
+      "loss": 0.7574,
+      "step": 5419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.280352898075903e-06,
+      "loss": 0.7949,
+      "step": 5420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.279372258993498e-06,
+      "loss": 0.7898,
+      "step": 5421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.278391509218225e-06,
+      "loss": 0.8876,
+      "step": 5422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.277410648797716e-06,
+      "loss": 0.6746,
+      "step": 5423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.276429677779603e-06,
+      "loss": 0.5254,
+      "step": 5424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.275448596211525e-06,
+      "loss": 0.7366,
+      "step": 5425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.274467404141127e-06,
+      "loss": 0.8965,
+      "step": 5426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.273486101616057e-06,
+      "loss": 0.7016,
+      "step": 5427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.27250468868397e-06,
+      "loss": 1.0565,
+      "step": 5428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.271523165392529e-06,
+      "loss": 0.9096,
+      "step": 5429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.270541531789395e-06,
+      "loss": 0.9365,
+      "step": 5430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2695597879222424e-06,
+      "loss": 0.6976,
+      "step": 5431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.268577933838748e-06,
+      "loss": 0.6766,
+      "step": 5432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2675959695865896e-06,
+      "loss": 0.7793,
+      "step": 5433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.266613895213457e-06,
+      "loss": 0.5172,
+      "step": 5434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.265631710767041e-06,
+      "loss": 0.7606,
+      "step": 5435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.264649416295042e-06,
+      "loss": 0.8413,
+      "step": 5436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.263667011845161e-06,
+      "loss": 0.8326,
+      "step": 5437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.262684497465108e-06,
+      "loss": 0.9786,
+      "step": 5438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.261701873202595e-06,
+      "loss": 0.4989,
+      "step": 5439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.26071913910534e-06,
+      "loss": 0.951,
+      "step": 5440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.259736295221069e-06,
+      "loss": 0.681,
+      "step": 5441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.258753341597514e-06,
+      "loss": 0.7998,
+      "step": 5442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.257770278282407e-06,
+      "loss": 0.5767,
+      "step": 5443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.256787105323488e-06,
+      "loss": 0.7212,
+      "step": 5444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.255803822768504e-06,
+      "loss": 1.0152,
+      "step": 5445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.254820430665207e-06,
+      "loss": 1.0506,
+      "step": 5446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.253836929061351e-06,
+      "loss": 0.6489,
+      "step": 5447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2528533180047e-06,
+      "loss": 0.8512,
+      "step": 5448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.251869597543019e-06,
+      "loss": 0.7495,
+      "step": 5449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.250885767724081e-06,
+      "loss": 0.7407,
+      "step": 5450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.249901828595663e-06,
+      "loss": 0.719,
+      "step": 5451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24891778020555e-06,
+      "loss": 0.7768,
+      "step": 5452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.24793362260153e-06,
+      "loss": 0.7869,
+      "step": 5453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.246949355831392e-06,
+      "loss": 0.6365,
+      "step": 5454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.245964979942941e-06,
+      "loss": 0.7972,
+      "step": 5455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.244980494983978e-06,
+      "loss": 0.7694,
+      "step": 5456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.243995901002312e-06,
+      "loss": 0.9233,
+      "step": 5457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2430111980457585e-06,
+      "loss": 0.8275,
+      "step": 5458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2420263861621374e-06,
+      "loss": 0.8875,
+      "step": 5459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.241041465399274e-06,
+      "loss": 0.95,
+      "step": 5460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.240056435805e-06,
+      "loss": 0.5392,
+      "step": 5461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.23907129742715e-06,
+      "loss": 0.7429,
+      "step": 5462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.238086050313563e-06,
+      "loss": 0.7429,
+      "step": 5463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.237100694512089e-06,
+      "loss": 0.6238,
+      "step": 5464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2361152300705795e-06,
+      "loss": 0.8386,
+      "step": 5465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.23512965703689e-06,
+      "loss": 0.9496,
+      "step": 5466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.234143975458885e-06,
+      "loss": 0.7176,
+      "step": 5467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.233158185384427e-06,
+      "loss": 0.8355,
+      "step": 5468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.232172286861394e-06,
+      "loss": 0.9296,
+      "step": 5469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.231186279937662e-06,
+      "loss": 0.7813,
+      "step": 5470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.230200164661114e-06,
+      "loss": 0.7454,
+      "step": 5471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2292139410796395e-06,
+      "loss": 0.5454,
+      "step": 5472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.228227609241131e-06,
+      "loss": 0.8224,
+      "step": 5473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.227241169193488e-06,
+      "loss": 0.6484,
+      "step": 5474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.226254620984617e-06,
+      "loss": 0.6766,
+      "step": 5475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.225267964662426e-06,
+      "loss": 0.5836,
+      "step": 5476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.22428120027483e-06,
+      "loss": 0.876,
+      "step": 5477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.223294327869746e-06,
+      "loss": 0.8625,
+      "step": 5478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.222307347495106e-06,
+      "loss": 0.8508,
+      "step": 5479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2213202591988364e-06,
+      "loss": 0.8246,
+      "step": 5480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2203330630288714e-06,
+      "loss": 0.7535,
+      "step": 5481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.219345759033155e-06,
+      "loss": 0.6445,
+      "step": 5482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.218358347259634e-06,
+      "loss": 0.7223,
+      "step": 5483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.217370827756258e-06,
+      "loss": 0.8931,
+      "step": 5484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.216383200570984e-06,
+      "loss": 0.7915,
+      "step": 5485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.215395465751775e-06,
+      "loss": 0.7234,
+      "step": 5486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2144076233465976e-06,
+      "loss": 0.5959,
+      "step": 5487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.213419673403425e-06,
+      "loss": 0.765,
+      "step": 5488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.212431615970234e-06,
+      "loss": 0.7959,
+      "step": 5489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.211443451095007e-06,
+      "loss": 0.8169,
+      "step": 5490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.210455178825735e-06,
+      "loss": 0.7505,
+      "step": 5491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.209466799210406e-06,
+      "loss": 0.8724,
+      "step": 5492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.208478312297024e-06,
+      "loss": 0.526,
+      "step": 5493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.207489718133591e-06,
+      "loss": 0.8337,
+      "step": 5494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.206501016768115e-06,
+      "loss": 0.9113,
+      "step": 5495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.205512208248611e-06,
+      "loss": 0.8704,
+      "step": 5496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2045232926230965e-06,
+      "loss": 0.7511,
+      "step": 5497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2035342699395995e-06,
+      "loss": 0.6532,
+      "step": 5498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.202545140246148e-06,
+      "loss": 0.974,
+      "step": 5499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.2015559035907754e-06,
+      "loss": 0.7113,
+      "step": 5500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.200566560021525e-06,
+      "loss": 0.6384,
+      "step": 5501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.199577109586439e-06,
+      "loss": 0.7695,
+      "step": 5502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.1985875523335715e-06,
+      "loss": 0.9244,
+      "step": 5503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.197597888310977e-06,
+      "loss": 0.6574,
+      "step": 5504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.196608117566714e-06,
+      "loss": 1.0465,
+      "step": 5505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.195618240148849e-06,
+      "loss": 0.7827,
+      "step": 5506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.194628256105457e-06,
+      "loss": 0.7585,
+      "step": 5507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.193638165484611e-06,
+      "loss": 0.7939,
+      "step": 5508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.192647968334394e-06,
+      "loss": 0.7166,
+      "step": 5509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.191657664702893e-06,
+      "loss": 0.767,
+      "step": 5510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.190667254638199e-06,
+      "loss": 0.638,
+      "step": 5511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.189676738188411e-06,
+      "loss": 0.707,
+      "step": 5512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.188686115401628e-06,
+      "loss": 0.716,
+      "step": 5513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.187695386325961e-06,
+      "loss": 0.6557,
+      "step": 5514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.186704551009521e-06,
+      "loss": 0.4424,
+      "step": 5515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.185713609500426e-06,
+      "loss": 0.8286,
+      "step": 5516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1847225618467975e-06,
+      "loss": 0.8951,
+      "step": 5517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1837314080967675e-06,
+      "loss": 0.6183,
+      "step": 5518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.182740148298466e-06,
+      "loss": 0.4976,
+      "step": 5519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.181748782500031e-06,
+      "loss": 0.5742,
+      "step": 5520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.180757310749608e-06,
+      "loss": 0.7531,
+      "step": 5521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.179765733095347e-06,
+      "loss": 1.0696,
+      "step": 5522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.178774049585397e-06,
+      "loss": 0.8344,
+      "step": 5523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.177782260267921e-06,
+      "loss": 0.5412,
+      "step": 5524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.176790365191083e-06,
+      "loss": 0.7987,
+      "step": 5525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.175798364403051e-06,
+      "loss": 0.762,
+      "step": 5526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.174806257951999e-06,
+      "loss": 0.7809,
+      "step": 5527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.173814045886107e-06,
+      "loss": 0.5229,
+      "step": 5528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.172821728253563e-06,
+      "loss": 0.7817,
+      "step": 5529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.17182930510255e-06,
+      "loss": 0.7407,
+      "step": 5530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.170836776481267e-06,
+      "loss": 0.8411,
+      "step": 5531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.169844142437916e-06,
+      "loss": 0.7207,
+      "step": 5532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.168851403020699e-06,
+      "loss": 0.6129,
+      "step": 5533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.167858558277827e-06,
+      "loss": 0.7963,
+      "step": 5534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.166865608257515e-06,
+      "loss": 1.0023,
+      "step": 5535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.165872553007984e-06,
+      "loss": 0.8115,
+      "step": 5536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.164879392577461e-06,
+      "loss": 0.6079,
+      "step": 5537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1638861270141746e-06,
+      "loss": 0.7194,
+      "step": 5538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1628927563663604e-06,
+      "loss": 0.7026,
+      "step": 5539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.161899280682262e-06,
+      "loss": 0.7743,
+      "step": 5540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.160905700010122e-06,
+      "loss": 0.8098,
+      "step": 5541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.159912014398196e-06,
+      "loss": 0.7263,
+      "step": 5542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.158918223894735e-06,
+      "loss": 0.7579,
+      "step": 5543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.157924328548003e-06,
+      "loss": 0.9133,
+      "step": 5544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.156930328406268e-06,
+      "loss": 0.682,
+      "step": 5545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.155936223517799e-06,
+      "loss": 0.7309,
+      "step": 5546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.154942013930874e-06,
+      "loss": 0.8474,
+      "step": 5547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.153947699693773e-06,
+      "loss": 0.8081,
+      "step": 5548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.152953280854784e-06,
+      "loss": 0.9666,
+      "step": 5549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1519587574621984e-06,
+      "loss": 1.0325,
+      "step": 5550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1509641295643124e-06,
+      "loss": 0.7126,
+      "step": 5551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.149969397209431e-06,
+      "loss": 0.655,
+      "step": 5552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.148974560445859e-06,
+      "loss": 0.9218,
+      "step": 5553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.147979619321907e-06,
+      "loss": 0.5137,
+      "step": 5554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.146984573885896e-06,
+      "loss": 0.9172,
+      "step": 5555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.145989424186146e-06,
+      "loss": 0.8883,
+      "step": 5556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.144994170270984e-06,
+      "loss": 1.1378,
+      "step": 5557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.143998812188742e-06,
+      "loss": 0.7213,
+      "step": 5558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.143003349987759e-06,
+      "loss": 0.7716,
+      "step": 5559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.142007783716377e-06,
+      "loss": 0.637,
+      "step": 5560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.141012113422942e-06,
+      "loss": 0.8033,
+      "step": 5561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.140016339155811e-06,
+      "loss": 0.5275,
+      "step": 5562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.139020460963337e-06,
+      "loss": 0.83,
+      "step": 5563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.138024478893883e-06,
+      "loss": 0.8522,
+      "step": 5564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13702839299582e-06,
+      "loss": 0.663,
+      "step": 5565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.136032203317519e-06,
+      "loss": 0.9459,
+      "step": 5566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.135035909907359e-06,
+      "loss": 0.722,
+      "step": 5567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.134039512813718e-06,
+      "loss": 0.5763,
+      "step": 5568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.13304301208499e-06,
+      "loss": 0.5286,
+      "step": 5569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.132046407769566e-06,
+      "loss": 0.4752,
+      "step": 5570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.131049699915842e-06,
+      "loss": 0.8899,
+      "step": 5571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.130052888572223e-06,
+      "loss": 0.6851,
+      "step": 5572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.129055973787117e-06,
+      "loss": 0.5153,
+      "step": 5573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.128058955608935e-06,
+      "loss": 0.7945,
+      "step": 5574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.127061834086098e-06,
+      "loss": 0.6885,
+      "step": 5575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.126064609267027e-06,
+      "loss": 0.7755,
+      "step": 5576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1250672812001505e-06,
+      "loss": 0.998,
+      "step": 5577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.124069849933903e-06,
+      "loss": 0.8445,
+      "step": 5578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.123072315516721e-06,
+      "loss": 0.9596,
+      "step": 5579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.122074677997049e-06,
+      "loss": 0.8682,
+      "step": 5580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.121076937423333e-06,
+      "loss": 0.7559,
+      "step": 5581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.120079093844027e-06,
+      "loss": 0.8136,
+      "step": 5582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.119081147307591e-06,
+      "loss": 0.5982,
+      "step": 5583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.118083097862486e-06,
+      "loss": 0.7922,
+      "step": 5584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.117084945557181e-06,
+      "loss": 0.7591,
+      "step": 5585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.116086690440149e-06,
+      "loss": 0.7917,
+      "step": 5586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.115088332559869e-06,
+      "loss": 0.9682,
+      "step": 5587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.114089871964822e-06,
+      "loss": 0.6429,
+      "step": 5588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.113091308703498e-06,
+      "loss": 0.9699,
+      "step": 5589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.112092642824389e-06,
+      "loss": 0.7264,
+      "step": 5590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.111093874375995e-06,
+      "loss": 0.4906,
+      "step": 5591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.110095003406816e-06,
+      "loss": 0.7293,
+      "step": 5592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.109096029965362e-06,
+      "loss": 0.5135,
+      "step": 5593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.108096954100148e-06,
+      "loss": 1.0169,
+      "step": 5594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.107097775859688e-06,
+      "loss": 0.5231,
+      "step": 5595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.106098495292505e-06,
+      "loss": 0.7042,
+      "step": 5596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.105099112447131e-06,
+      "loss": 0.5249,
+      "step": 5597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.104099627372096e-06,
+      "loss": 1.0051,
+      "step": 5598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.103100040115939e-06,
+      "loss": 0.6337,
+      "step": 5599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1021003507272014e-06,
+      "loss": 0.8528,
+      "step": 5600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1011005592544314e-06,
+      "loss": 0.7691,
+      "step": 5601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.100100665746184e-06,
+      "loss": 0.8201,
+      "step": 5602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.099100670251013e-06,
+      "loss": 0.8073,
+      "step": 5603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.098100572817483e-06,
+      "loss": 0.6972,
+      "step": 5604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.097100373494164e-06,
+      "loss": 0.9096,
+      "step": 5605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096100072329624e-06,
+      "loss": 0.7138,
+      "step": 5606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.095099669372444e-06,
+      "loss": 0.971,
+      "step": 5607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0940991646712045e-06,
+      "loss": 0.9725,
+      "step": 5608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.093098558274494e-06,
+      "loss": 0.9486,
+      "step": 5609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0920978502309044e-06,
+      "loss": 0.7696,
+      "step": 5610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.091097040589033e-06,
+      "loss": 0.9972,
+      "step": 5611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.090096129397481e-06,
+      "loss": 0.6616,
+      "step": 5612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0890951167048586e-06,
+      "loss": 0.6766,
+      "step": 5613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.088094002559774e-06,
+      "loss": 0.8036,
+      "step": 5614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.087092787010847e-06,
+      "loss": 0.6419,
+      "step": 5615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.086091470106698e-06,
+      "loss": 0.8037,
+      "step": 5616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.085090051895953e-06,
+      "loss": 0.8162,
+      "step": 5617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.084088532427247e-06,
+      "loss": 0.7132,
+      "step": 5618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.083086911749214e-06,
+      "loss": 0.7968,
+      "step": 5619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.082085189910495e-06,
+      "loss": 0.7363,
+      "step": 5620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.081083366959739e-06,
+      "loss": 0.7848,
+      "step": 5621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.080081442945597e-06,
+      "loss": 0.5991,
+      "step": 5622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0790794179167225e-06,
+      "loss": 0.5167,
+      "step": 5623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0780772919217785e-06,
+      "loss": 0.7522,
+      "step": 5624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0770750650094335e-06,
+      "loss": 0.7209,
+      "step": 5625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.076072737228355e-06,
+      "loss": 0.8779,
+      "step": 5626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.075070308627219e-06,
+      "loss": 0.7344,
+      "step": 5627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0740677792547094e-06,
+      "loss": 0.8314,
+      "step": 5628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0730651491595105e-06,
+      "loss": 0.5215,
+      "step": 5629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.07206241839031e-06,
+      "loss": 0.7076,
+      "step": 5630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.071059586995806e-06,
+      "loss": 0.8573,
+      "step": 5631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0700566550247015e-06,
+      "loss": 0.578,
+      "step": 5632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.069053622525697e-06,
+      "loss": 0.8622,
+      "step": 5633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.068050489547505e-06,
+      "loss": 0.6866,
+      "step": 5634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0670472561388395e-06,
+      "loss": 0.7551,
+      "step": 5635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0660439223484216e-06,
+      "loss": 0.774,
+      "step": 5636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.065040488224975e-06,
+      "loss": 0.8063,
+      "step": 5637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.064036953817229e-06,
+      "loss": 0.8358,
+      "step": 5638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.06303331917392e-06,
+      "loss": 0.6347,
+      "step": 5639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.062029584343788e-06,
+      "loss": 0.7806,
+      "step": 5640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.061025749375572e-06,
+      "loss": 0.6514,
+      "step": 5641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.060021814318027e-06,
+      "loss": 0.6447,
+      "step": 5642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.059017779219904e-06,
+      "loss": 1.0316,
+      "step": 5643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.058013644129962e-06,
+      "loss": 0.9775,
+      "step": 5644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.057009409096967e-06,
+      "loss": 0.852,
+      "step": 5645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.056005074169684e-06,
+      "loss": 0.755,
+      "step": 5646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.055000639396888e-06,
+      "loss": 1.0529,
+      "step": 5647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.053996104827357e-06,
+      "loss": 0.8116,
+      "step": 5648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0529914705098755e-06,
+      "loss": 0.8806,
+      "step": 5649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.05198673649323e-06,
+      "loss": 0.7855,
+      "step": 5650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.050981902826214e-06,
+      "loss": 0.6813,
+      "step": 5651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.049976969557624e-06,
+      "loss": 0.5125,
+      "step": 5652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.048971936736264e-06,
+      "loss": 0.6173,
+      "step": 5653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.0479668044109384e-06,
+      "loss": 0.9732,
+      "step": 5654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.046961572630463e-06,
+      "loss": 0.7608,
+      "step": 5655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.045956241443655e-06,
+      "loss": 0.6953,
+      "step": 5656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.044950810899332e-06,
+      "loss": 0.505,
+      "step": 5657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.043945281046321e-06,
+      "loss": 0.69,
+      "step": 5658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.042939651933458e-06,
+      "loss": 0.7959,
+      "step": 5659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0419339236095755e-06,
+      "loss": 0.7426,
+      "step": 5660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.040928096123516e-06,
+      "loss": 0.7771,
+      "step": 5661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.039922169524125e-06,
+      "loss": 0.6849,
+      "step": 5662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.038916143860251e-06,
+      "loss": 0.806,
+      "step": 5663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.037910019180756e-06,
+      "loss": 0.5928,
+      "step": 5664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.036903795534491e-06,
+      "loss": 0.667,
+      "step": 5665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.035897472970329e-06,
+      "loss": 1.0036,
+      "step": 5666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0348910515371375e-06,
+      "loss": 0.8126,
+      "step": 5667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.033884531283788e-06,
+      "loss": 0.91,
+      "step": 5668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.032877912259164e-06,
+      "loss": 0.7083,
+      "step": 5669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.03187119451215e-06,
+      "loss": 0.6174,
+      "step": 5670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.030864378091632e-06,
+      "loss": 0.6109,
+      "step": 5671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.029857463046505e-06,
+      "loss": 0.6466,
+      "step": 5672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.02885044942567e-06,
+      "loss": 0.7437,
+      "step": 5673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.027843337278027e-06,
+      "loss": 0.8889,
+      "step": 5674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.026836126652487e-06,
+      "loss": 0.6555,
+      "step": 5675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.025828817597962e-06,
+      "loss": 0.5796,
+      "step": 5676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0248214101633685e-06,
+      "loss": 0.6996,
+      "step": 5677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.023813904397632e-06,
+      "loss": 0.6122,
+      "step": 5678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.022806300349676e-06,
+      "loss": 0.7593,
+      "step": 5679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0217985980684365e-06,
+      "loss": 0.6161,
+      "step": 5680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.02079079760285e-06,
+      "loss": 0.8193,
+      "step": 5681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.019782899001855e-06,
+      "loss": 0.7412,
+      "step": 5682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0187749023144e-06,
+      "loss": 0.7818,
+      "step": 5683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.017766807589438e-06,
+      "loss": 1.0237,
+      "step": 5684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.016758614875921e-06,
+      "loss": 0.7794,
+      "step": 5685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.015750324222813e-06,
+      "loss": 0.7465,
+      "step": 5686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.014741935679078e-06,
+      "loss": 0.7046,
+      "step": 5687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0137334492936875e-06,
+      "loss": 0.6269,
+      "step": 5688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.012724865115615e-06,
+      "loss": 0.5542,
+      "step": 5689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.011716183193839e-06,
+      "loss": 0.6245,
+      "step": 5690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.010707403577349e-06,
+      "loss": 0.7819,
+      "step": 5691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.009698526315129e-06,
+      "loss": 0.8296,
+      "step": 5692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.008689551456176e-06,
+      "loss": 0.7504,
+      "step": 5693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.007680479049488e-06,
+      "loss": 0.8533,
+      "step": 5694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.006671309144068e-06,
+      "loss": 0.552,
+      "step": 5695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.0056620417889235e-06,
+      "loss": 0.7719,
+      "step": 5696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.004652677033069e-06,
+      "loss": 0.8918,
+      "step": 5697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.003643214925522e-06,
+      "loss": 0.7127,
+      "step": 5698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.002633655515304e-06,
+      "loss": 0.9576,
+      "step": 5699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.001623998851443e-06,
+      "loss": 0.5795,
+      "step": 5700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.000614244982968e-06,
+      "loss": 0.7676,
+      "step": 5701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.999604393958921e-06,
+      "loss": 0.6837,
+      "step": 5702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9985944458283375e-06,
+      "loss": 0.7683,
+      "step": 5703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997584400640267e-06,
+      "loss": 0.6483,
+      "step": 5704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.996574258443761e-06,
+      "loss": 0.8177,
+      "step": 5705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9955640192878705e-06,
+      "loss": 0.8513,
+      "step": 5706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.99455368322166e-06,
+      "loss": 0.7115,
+      "step": 5707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993543250294193e-06,
+      "loss": 0.6829,
+      "step": 5708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.992532720554538e-06,
+      "loss": 0.601,
+      "step": 5709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9915220940517704e-06,
+      "loss": 0.7403,
+      "step": 5710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9905113708349684e-06,
+      "loss": 0.8805,
+      "step": 5711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.989500550953215e-06,
+      "loss": 0.8616,
+      "step": 5712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9884896344556e-06,
+      "loss": 0.5452,
+      "step": 5713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.987478621391215e-06,
+      "loss": 0.6798,
+      "step": 5714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.98646751180916e-06,
+      "loss": 0.7608,
+      "step": 5715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9854563057585354e-06,
+      "loss": 0.7277,
+      "step": 5716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.984445003288447e-06,
+      "loss": 0.7658,
+      "step": 5717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.983433604448011e-06,
+      "loss": 0.9222,
+      "step": 5718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.982422109286339e-06,
+      "loss": 0.6615,
+      "step": 5719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9814105178525545e-06,
+      "loss": 0.6654,
+      "step": 5720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.980398830195785e-06,
+      "loss": 0.7109,
+      "step": 5721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.979387046365158e-06,
+      "loss": 0.7814,
+      "step": 5722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.978375166409809e-06,
+      "loss": 0.7318,
+      "step": 5723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.977363190378879e-06,
+      "loss": 0.7319,
+      "step": 5724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9763511183215106e-06,
+      "loss": 0.917,
+      "step": 5725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.975338950286855e-06,
+      "loss": 0.6416,
+      "step": 5726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.974326686324065e-06,
+      "loss": 0.7774,
+      "step": 5727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9733143264823e-06,
+      "loss": 0.886,
+      "step": 5728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.972301870810723e-06,
+      "loss": 0.7492,
+      "step": 5729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.971289319358499e-06,
+      "loss": 0.9448,
+      "step": 5730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.970276672174803e-06,
+      "loss": 0.5287,
+      "step": 5731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.969263929308812e-06,
+      "loss": 0.6129,
+      "step": 5732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.968251090809708e-06,
+      "loss": 0.9173,
+      "step": 5733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9672381567266745e-06,
+      "loss": 0.8056,
+      "step": 5734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.966225127108907e-06,
+      "loss": 0.9769,
+      "step": 5735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.965212002005597e-06,
+      "loss": 0.7537,
+      "step": 5736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.964198781465948e-06,
+      "loss": 0.6218,
+      "step": 5737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.963185465539162e-06,
+      "loss": 0.8795,
+      "step": 5738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.962172054274449e-06,
+      "loss": 0.8531,
+      "step": 5739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9611585477210275e-06,
+      "loss": 0.8556,
+      "step": 5740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9601449459281105e-06,
+      "loss": 0.7014,
+      "step": 5741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9591312489449235e-06,
+      "loss": 0.7976,
+      "step": 5742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.958117456820696e-06,
+      "loss": 0.9865,
+      "step": 5743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.957103569604658e-06,
+      "loss": 0.6337,
+      "step": 5744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.95608958734605e-06,
+      "loss": 0.6145,
+      "step": 5745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.955075510094111e-06,
+      "loss": 0.9001,
+      "step": 5746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.95406133789809e-06,
+      "loss": 0.664,
+      "step": 5747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.953047070807235e-06,
+      "loss": 0.6769,
+      "step": 5748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.952032708870805e-06,
+      "loss": 0.9315,
+      "step": 5749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.951018252138058e-06,
+      "loss": 0.6619,
+      "step": 5750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.95000370065826e-06,
+      "loss": 0.7644,
+      "step": 5751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9489890544806795e-06,
+      "loss": 0.4206,
+      "step": 5752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.947974313654592e-06,
+      "loss": 0.6787,
+      "step": 5753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.946959478229277e-06,
+      "loss": 0.7648,
+      "step": 5754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.945944548254013e-06,
+      "loss": 0.7168,
+      "step": 5755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.944929523778095e-06,
+      "loss": 0.793,
+      "step": 5756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.943914404850811e-06,
+      "loss": 0.8922,
+      "step": 5757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.942899191521457e-06,
+      "loss": 0.795,
+      "step": 5758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.94188388383934e-06,
+      "loss": 1.0378,
+      "step": 5759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9408684818537595e-06,
+      "loss": 0.8588,
+      "step": 5760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.939852985614031e-06,
+      "loss": 0.689,
+      "step": 5761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.93883739516947e-06,
+      "loss": 0.5797,
+      "step": 5762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.937821710569394e-06,
+      "loss": 0.9819,
+      "step": 5763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.93680593186313e-06,
+      "loss": 0.6931,
+      "step": 5764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9357900591000034e-06,
+      "loss": 0.8021,
+      "step": 5765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9347740923293515e-06,
+      "loss": 0.8569,
+      "step": 5766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.933758031600513e-06,
+      "loss": 0.7555,
+      "step": 5767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.932741876962826e-06,
+      "loss": 0.9399,
+      "step": 5768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.931725628465643e-06,
+      "loss": 0.8515,
+      "step": 5769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.930709286158315e-06,
+      "loss": 0.732,
+      "step": 5770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.929692850090196e-06,
+      "loss": 0.7297,
+      "step": 5771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9286763203106475e-06,
+      "loss": 0.7242,
+      "step": 5772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.927659696869039e-06,
+      "loss": 1.034,
+      "step": 5773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9266429798147365e-06,
+      "loss": 0.5717,
+      "step": 5774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.925626169197115e-06,
+      "loss": 0.6302,
+      "step": 5775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.924609265065556e-06,
+      "loss": 0.9304,
+      "step": 5776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.923592267469441e-06,
+      "loss": 0.5999,
+      "step": 5777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.922575176458162e-06,
+      "loss": 0.7628,
+      "step": 5778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.921557992081105e-06,
+      "loss": 0.7865,
+      "step": 5779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.920540714387674e-06,
+      "loss": 0.688,
+      "step": 5780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.919523343427269e-06,
+      "loss": 0.6978,
+      "step": 5781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.918505879249294e-06,
+      "loss": 0.8418,
+      "step": 5782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9174883219031655e-06,
+      "loss": 0.7593,
+      "step": 5783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.916470671438292e-06,
+      "loss": 0.7466,
+      "step": 5784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9154529279040985e-06,
+      "loss": 0.9064,
+      "step": 5785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.914435091350008e-06,
+      "loss": 0.8027,
+      "step": 5786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.913417161825449e-06,
+      "loss": 0.6015,
+      "step": 5787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9123991393798575e-06,
+      "loss": 0.6832,
+      "step": 5788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.911381024062668e-06,
+      "loss": 0.7371,
+      "step": 5789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.910362815923325e-06,
+      "loss": 0.8084,
+      "step": 5790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9093445150112756e-06,
+      "loss": 0.8616,
+      "step": 5791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.908326121375973e-06,
+      "loss": 0.6483,
+      "step": 5792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.90730763506687e-06,
+      "loss": 0.5447,
+      "step": 5793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.906289056133431e-06,
+      "loss": 0.6717,
+      "step": 5794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.905270384625115e-06,
+      "loss": 0.8964,
+      "step": 5795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.904251620591398e-06,
+      "loss": 0.914,
+      "step": 5796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.903232764081753e-06,
+      "loss": 0.7605,
+      "step": 5797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.902213815145655e-06,
+      "loss": 0.7928,
+      "step": 5798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.9011947738325915e-06,
+      "loss": 0.6556,
+      "step": 5799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.900175640192046e-06,
+      "loss": 0.7695,
+      "step": 5800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.899156414273514e-06,
+      "loss": 0.7619,
+      "step": 5801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.898137096126492e-06,
+      "loss": 0.6571,
+      "step": 5802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.897117685800476e-06,
+      "loss": 0.725,
+      "step": 5803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.896098183344977e-06,
+      "loss": 0.9331,
+      "step": 5804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.895078588809503e-06,
+      "loss": 0.6125,
+      "step": 5805
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8940589022435685e-06,
+      "loss": 0.6127,
+      "step": 5806
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.893039123696692e-06,
+      "loss": 0.6106,
+      "step": 5807
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.892019253218399e-06,
+      "loss": 0.4651,
+      "step": 5808
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.890999290858213e-06,
+      "loss": 0.9589,
+      "step": 5809
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.889979236665669e-06,
+      "loss": 0.9773,
+      "step": 5810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.888959090690306e-06,
+      "loss": 0.601,
+      "step": 5811
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.887938852981662e-06,
+      "loss": 0.5052,
+      "step": 5812
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.886918523589283e-06,
+      "loss": 0.7706,
+      "step": 5813
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.88589810256272e-06,
+      "loss": 0.6684,
+      "step": 5814
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.884877589951529e-06,
+      "loss": 0.6242,
+      "step": 5815
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.883856985805266e-06,
+      "loss": 0.809,
+      "step": 5816
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.882836290173493e-06,
+      "loss": 0.607,
+      "step": 5817
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.881815503105784e-06,
+      "loss": 0.9442,
+      "step": 5818
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.880794624651709e-06,
+      "loss": 0.8558,
+      "step": 5819
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8797736548608405e-06,
+      "loss": 0.8266,
+      "step": 5820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8787525937827656e-06,
+      "loss": 1.0987,
+      "step": 5821
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.877731441467067e-06,
+      "loss": 0.899,
+      "step": 5822
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.876710197963334e-06,
+      "loss": 0.8114,
+      "step": 5823
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.875688863321164e-06,
+      "loss": 0.8931,
+      "step": 5824
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.874667437590152e-06,
+      "loss": 1.0903,
+      "step": 5825
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.873645920819906e-06,
+      "loss": 0.3309,
+      "step": 5826
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.87262431306003e-06,
+      "loss": 0.6959,
+      "step": 5827
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.871602614360138e-06,
+      "loss": 0.8068,
+      "step": 5828
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.870580824769848e-06,
+      "loss": 0.8677,
+      "step": 5829
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.869558944338776e-06,
+      "loss": 0.4802,
+      "step": 5830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.868536973116553e-06,
+      "loss": 0.6457,
+      "step": 5831
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.867514911152808e-06,
+      "loss": 0.8562,
+      "step": 5832
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.866492758497171e-06,
+      "loss": 0.6954,
+      "step": 5833
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.865470515199283e-06,
+      "loss": 0.7482,
+      "step": 5834
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.86444818130879e-06,
+      "loss": 0.7545,
+      "step": 5835
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8634257568753375e-06,
+      "loss": 0.7449,
+      "step": 5836
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.862403241948574e-06,
+      "loss": 0.7287,
+      "step": 5837
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.861380636578161e-06,
+      "loss": 0.5556,
+      "step": 5838
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.860357940813756e-06,
+      "loss": 0.687,
+      "step": 5839
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.859335154705027e-06,
+      "loss": 0.7312,
+      "step": 5840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.858312278301638e-06,
+      "loss": 0.7909,
+      "step": 5841
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.857289311653268e-06,
+      "loss": 0.9105,
+      "step": 5842
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.856266254809594e-06,
+      "loss": 0.7119,
+      "step": 5843
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.855243107820298e-06,
+      "loss": 0.8121,
+      "step": 5844
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8542198707350635e-06,
+      "loss": 0.9121,
+      "step": 5845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.853196543603589e-06,
+      "loss": 0.9281,
+      "step": 5846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.852173126475565e-06,
+      "loss": 0.7815,
+      "step": 5847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.851149619400693e-06,
+      "loss": 0.7779,
+      "step": 5848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.850126022428678e-06,
+      "loss": 0.6296,
+      "step": 5849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8491023356092275e-06,
+      "loss": 0.6178,
+      "step": 5850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.848078558992056e-06,
+      "loss": 0.9478,
+      "step": 5851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.84705469262688e-06,
+      "loss": 0.7373,
+      "step": 5852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8460307365634225e-06,
+      "loss": 0.8942,
+      "step": 5853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8450066908514104e-06,
+      "loss": 0.9211,
+      "step": 5854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.84398255554057e-06,
+      "loss": 0.5188,
+      "step": 5855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.84295833068064e-06,
+      "loss": 1.2166,
+      "step": 5856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.841934016321361e-06,
+      "loss": 0.9233,
+      "step": 5857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.840909612512473e-06,
+      "loss": 0.6628,
+      "step": 5858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.839885119303726e-06,
+      "loss": 0.8947,
+      "step": 5859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.838860536744872e-06,
+      "loss": 0.4688,
+      "step": 5860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.837835864885668e-06,
+      "loss": 0.963,
+      "step": 5861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.836811103775874e-06,
+      "loss": 0.5042,
+      "step": 5862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.835786253465256e-06,
+      "loss": 0.5766,
+      "step": 5863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.834761314003584e-06,
+      "loss": 0.7556,
+      "step": 5864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.833736285440632e-06,
+      "loss": 0.7655,
+      "step": 5865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.832711167826177e-06,
+      "loss": 0.4525,
+      "step": 5866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8316859612100036e-06,
+      "loss": 0.6736,
+      "step": 5867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.830660665641898e-06,
+      "loss": 0.8281,
+      "step": 5868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82963528117165e-06,
+      "loss": 0.746,
+      "step": 5869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.828609807849058e-06,
+      "loss": 0.7045,
+      "step": 5870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.827584245723919e-06,
+      "loss": 1.0678,
+      "step": 5871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.826558594846039e-06,
+      "loss": 0.6969,
+      "step": 5872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.825532855265227e-06,
+      "loss": 0.8126,
+      "step": 5873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.824507027031295e-06,
+      "loss": 0.7169,
+      "step": 5874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.82348111019406e-06,
+      "loss": 0.9603,
+      "step": 5875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.822455104803344e-06,
+      "loss": 0.6941,
+      "step": 5876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.821429010908972e-06,
+      "loss": 1.0844,
+      "step": 5877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.820402828560776e-06,
+      "loss": 0.82,
+      "step": 5878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.819376557808587e-06,
+      "loss": 0.7351,
+      "step": 5879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.818350198702246e-06,
+      "loss": 0.7095,
+      "step": 5880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.817323751291598e-06,
+      "loss": 0.7311,
+      "step": 5881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.816297215626485e-06,
+      "loss": 0.8344,
+      "step": 5882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.815270591756762e-06,
+      "loss": 0.7871,
+      "step": 5883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.814243879732285e-06,
+      "loss": 0.5799,
+      "step": 5884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.813217079602912e-06,
+      "loss": 0.8295,
+      "step": 5885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.812190191418508e-06,
+      "loss": 0.6487,
+      "step": 5886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.811163215228944e-06,
+      "loss": 0.9404,
+      "step": 5887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.81013615108409e-06,
+      "loss": 0.764,
+      "step": 5888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.809108999033825e-06,
+      "loss": 0.7711,
+      "step": 5889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80808175912803e-06,
+      "loss": 0.8715,
+      "step": 5890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8070544314165894e-06,
+      "loss": 0.6471,
+      "step": 5891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.806027015949397e-06,
+      "loss": 0.7947,
+      "step": 5892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80499951277634e-06,
+      "loss": 0.5704,
+      "step": 5893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.8039719219473256e-06,
+      "loss": 0.7423,
+      "step": 5894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.802944243512249e-06,
+      "loss": 0.9138,
+      "step": 5895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.80191647752102e-06,
+      "loss": 0.7673,
+      "step": 5896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.800888624023552e-06,
+      "loss": 0.6669,
+      "step": 5897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.799860683069759e-06,
+      "loss": 0.7935,
+      "step": 5898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7988326547095605e-06,
+      "loss": 0.7464,
+      "step": 5899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.797804538992881e-06,
+      "loss": 0.7973,
+      "step": 5900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.796776335969646e-06,
+      "loss": 0.8873,
+      "step": 5901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.795748045689792e-06,
+      "loss": 1.0022,
+      "step": 5902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.794719668203253e-06,
+      "loss": 0.8285,
+      "step": 5903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.793691203559972e-06,
+      "loss": 0.6582,
+      "step": 5904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.792662651809895e-06,
+      "loss": 0.8338,
+      "step": 5905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.791634013002967e-06,
+      "loss": 0.657,
+      "step": 5906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.790605287189143e-06,
+      "loss": 0.6962,
+      "step": 5907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.789576474418385e-06,
+      "loss": 0.8237,
+      "step": 5908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.788547574740653e-06,
+      "loss": 0.9742,
+      "step": 5909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.787518588205909e-06,
+      "loss": 0.8217,
+      "step": 5910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7864895148641295e-06,
+      "loss": 0.975,
+      "step": 5911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.785460354765287e-06,
+      "loss": 0.5951,
+      "step": 5912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78443110795936e-06,
+      "loss": 0.7243,
+      "step": 5913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.783401774496332e-06,
+      "loss": 0.8438,
+      "step": 5914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.78237235442619e-06,
+      "loss": 1.0585,
+      "step": 5915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.781342847798927e-06,
+      "loss": 0.5943,
+      "step": 5916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7803132546645364e-06,
+      "loss": 0.8248,
+      "step": 5917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.779283575073021e-06,
+      "loss": 0.7706,
+      "step": 5918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.778253809074384e-06,
+      "loss": 0.8608,
+      "step": 5919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.777223956718632e-06,
+      "loss": 0.8983,
+      "step": 5920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7761940180557785e-06,
+      "loss": 0.6217,
+      "step": 5921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.775163993135843e-06,
+      "loss": 0.8355,
+      "step": 5922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.774133882008843e-06,
+      "loss": 1.0002,
+      "step": 5923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.773103684724806e-06,
+      "loss": 0.9218,
+      "step": 5924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.772073401333758e-06,
+      "loss": 0.7953,
+      "step": 5925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.771043031885736e-06,
+      "loss": 0.6109,
+      "step": 5926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.770012576430777e-06,
+      "loss": 0.6382,
+      "step": 5927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.768982035018922e-06,
+      "loss": 0.5345,
+      "step": 5928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.767951407700217e-06,
+      "loss": 0.7905,
+      "step": 5929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.766920694524714e-06,
+      "loss": 0.4549,
+      "step": 5930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.765889895542465e-06,
+      "loss": 0.6928,
+      "step": 5931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76485901080353e-06,
+      "loss": 0.6942,
+      "step": 5932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.76382804035797e-06,
+      "loss": 0.7947,
+      "step": 5933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.762796984255854e-06,
+      "loss": 0.9877,
+      "step": 5934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.761765842547253e-06,
+      "loss": 0.6933,
+      "step": 5935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.760734615282241e-06,
+      "loss": 0.7044,
+      "step": 5936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.759703302510898e-06,
+      "loss": 1.059,
+      "step": 5937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.758671904283307e-06,
+      "loss": 0.587,
+      "step": 5938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.757640420649556e-06,
+      "loss": 0.7179,
+      "step": 5939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.756608851659736e-06,
+      "loss": 0.7183,
+      "step": 5940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.755577197363945e-06,
+      "loss": 0.7557,
+      "step": 5941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.75454545781228e-06,
+      "loss": 0.9632,
+      "step": 5942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.75351363305485e-06,
+      "loss": 0.7163,
+      "step": 5943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7524817231417595e-06,
+      "loss": 0.7435,
+      "step": 5944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.75144972812312e-06,
+      "loss": 1.0145,
+      "step": 5945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.750417648049053e-06,
+      "loss": 0.8805,
+      "step": 5946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.749385482969675e-06,
+      "loss": 0.6966,
+      "step": 5947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.74835323293511e-06,
+      "loss": 0.7574,
+      "step": 5948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.747320897995493e-06,
+      "loss": 0.7753,
+      "step": 5949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.746288478200951e-06,
+      "loss": 0.7232,
+      "step": 5950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.745255973601625e-06,
+      "loss": 0.8634,
+      "step": 5951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.7442233842476545e-06,
+      "loss": 0.6909,
+      "step": 5952
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.743190710189186e-06,
+      "loss": 0.6354,
+      "step": 5953
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.742157951476369e-06,
+      "loss": 0.8126,
+      "step": 5954
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.741125108159356e-06,
+      "loss": 0.8961,
+      "step": 5955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.740092180288307e-06,
+      "loss": 0.7899,
+      "step": 5956
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.739059167913382e-06,
+      "loss": 0.8249,
+      "step": 5957
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.738026071084749e-06,
+      "loss": 0.6387,
+      "step": 5958
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.736992889852575e-06,
+      "loss": 0.5304,
+      "step": 5959
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.735959624267037e-06,
+      "loss": 0.7004,
+      "step": 5960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.734926274378313e-06,
+      "loss": 0.7152,
+      "step": 5961
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.733892840236584e-06,
+      "loss": 0.8068,
+      "step": 5962
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732859321892038e-06,
+      "loss": 0.723,
+      "step": 5963
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.731825719394866e-06,
+      "loss": 0.7388,
+      "step": 5964
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.730792032795261e-06,
+      "loss": 0.5912,
+      "step": 5965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.729758262143422e-06,
+      "loss": 0.8453,
+      "step": 5966
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.728724407489554e-06,
+      "loss": 0.7883,
+      "step": 5967
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7276904688838625e-06,
+      "loss": 0.8883,
+      "step": 5968
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.726656446376557e-06,
+      "loss": 0.6993,
+      "step": 5969
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.725622340017855e-06,
+      "loss": 0.7437,
+      "step": 5970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.724588149857976e-06,
+      "loss": 0.9707,
+      "step": 5971
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.723553875947139e-06,
+      "loss": 0.5858,
+      "step": 5972
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.722519518335576e-06,
+      "loss": 0.8085,
+      "step": 5973
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.721485077073517e-06,
+      "loss": 0.7425,
+      "step": 5974
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.720450552211197e-06,
+      "loss": 0.6329,
+      "step": 5975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.719415943798855e-06,
+      "loss": 0.6212,
+      "step": 5976
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.7183812518867365e-06,
+      "loss": 0.616,
+      "step": 5977
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.717346476525087e-06,
+      "loss": 0.6247,
+      "step": 5978
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.716311617764159e-06,
+      "loss": 1.0037,
+      "step": 5979
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.715276675654208e-06,
+      "loss": 1.0317,
+      "step": 5980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.714241650245494e-06,
+      "loss": 0.8092,
+      "step": 5981
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.713206541588281e-06,
+      "loss": 0.8309,
+      "step": 5982
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.712171349732836e-06,
+      "loss": 0.9833,
+      "step": 5983
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.711136074729432e-06,
+      "loss": 0.6983,
+      "step": 5984
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.710100716628345e-06,
+      "loss": 0.9746,
+      "step": 5985
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.709065275479853e-06,
+      "loss": 0.9872,
+      "step": 5986
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.708029751334242e-06,
+      "loss": 0.499,
+      "step": 5987
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.706994144241799e-06,
+      "loss": 0.7461,
+      "step": 5988
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.705958454252817e-06,
+      "loss": 0.8627,
+      "step": 5989
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.704922681417593e-06,
+      "loss": 0.5856,
+      "step": 5990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.703886825786423e-06,
+      "loss": 0.817,
+      "step": 5991
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.702850887409617e-06,
+      "loss": 0.7423,
+      "step": 5992
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.701814866337477e-06,
+      "loss": 0.7995,
+      "step": 5993
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.700778762620319e-06,
+      "loss": 0.7815,
+      "step": 5994
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.699742576308459e-06,
+      "loss": 0.8219,
+      "step": 5995
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.698706307452215e-06,
+      "loss": 0.5869,
+      "step": 5996
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.697669956101914e-06,
+      "loss": 0.9415,
+      "step": 5997
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.696633522307883e-06,
+      "loss": 0.8354,
+      "step": 5998
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.695597006120453e-06,
+      "loss": 0.7718,
+      "step": 5999
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6945604075899604e-06,
+      "loss": 0.8965,
+      "step": 6000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.693523726766747e-06,
+      "loss": 0.775,
+      "step": 6001
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.692486963701157e-06,
+      "loss": 0.7829,
+      "step": 6002
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.691450118443538e-06,
+      "loss": 0.7426,
+      "step": 6003
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.69041319104424e-06,
+      "loss": 0.8631,
+      "step": 6004
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.689376181553624e-06,
+      "loss": 0.9523,
+      "step": 6005
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.688339090022047e-06,
+      "loss": 0.6509,
+      "step": 6006
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.687301916499871e-06,
+      "loss": 0.7483,
+      "step": 6007
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68626466103747e-06,
+      "loss": 0.8123,
+      "step": 6008
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.685227323685209e-06,
+      "loss": 0.5957,
+      "step": 6009
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68418990449347e-06,
+      "loss": 0.7768,
+      "step": 6010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.68315240351263e-06,
+      "loss": 0.8194,
+      "step": 6011
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.682114820793075e-06,
+      "loss": 0.8931,
+      "step": 6012
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.681077156385192e-06,
+      "loss": 0.9722,
+      "step": 6013
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.680039410339371e-06,
+      "loss": 0.7174,
+      "step": 6014
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.67900158270601e-06,
+      "loss": 0.6897,
+      "step": 6015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.677963673535512e-06,
+      "loss": 0.772,
+      "step": 6016
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.676925682878273e-06,
+      "loss": 0.5045,
+      "step": 6017
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.675887610784708e-06,
+      "loss": 0.5413,
+      "step": 6018
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.674849457305227e-06,
+      "loss": 0.6752,
+      "step": 6019
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.673811222490243e-06,
+      "loss": 0.5319,
+      "step": 6020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.672772906390177e-06,
+      "loss": 0.5314,
+      "step": 6021
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.671734509055455e-06,
+      "loss": 1.0502,
+      "step": 6022
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.670696030536501e-06,
+      "loss": 0.7256,
+      "step": 6023
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.669657470883749e-06,
+      "loss": 0.8343,
+      "step": 6024
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.668618830147634e-06,
+      "loss": 0.5216,
+      "step": 6025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.667580108378594e-06,
+      "loss": 0.6653,
+      "step": 6026
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.666541305627074e-06,
+      "loss": 0.8699,
+      "step": 6027
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6655024219435195e-06,
+      "loss": 0.7995,
+      "step": 6028
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6644634573783825e-06,
+      "loss": 0.9218,
+      "step": 6029
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.663424411982121e-06,
+      "loss": 0.8266,
+      "step": 6030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6623852858051885e-06,
+      "loss": 0.6639,
+      "step": 6031
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.661346078898052e-06,
+      "loss": 1.1646,
+      "step": 6032
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.660306791311177e-06,
+      "loss": 0.7066,
+      "step": 6033
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6592674230950346e-06,
+      "loss": 0.6438,
+      "step": 6034
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6582279743001e-06,
+      "loss": 0.8468,
+      "step": 6035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65718844497685e-06,
+      "loss": 0.8845,
+      "step": 6036
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6561488351757686e-06,
+      "loss": 0.8161,
+      "step": 6037
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.655109144947341e-06,
+      "loss": 0.5974,
+      "step": 6038
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.65406937434206e-06,
+      "loss": 0.7496,
+      "step": 6039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.653029523410418e-06,
+      "loss": 0.6878,
+      "step": 6040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.651989592202913e-06,
+      "loss": 0.6942,
+      "step": 6041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.650949580770048e-06,
+      "loss": 0.8683,
+      "step": 6042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.649909489162329e-06,
+      "loss": 0.8371,
+      "step": 6043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.648869317430265e-06,
+      "loss": 0.831,
+      "step": 6044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.647829065624368e-06,
+      "loss": 1.0226,
+      "step": 6045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.646788733795161e-06,
+      "loss": 0.874,
+      "step": 6046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6457483219931605e-06,
+      "loss": 0.8159,
+      "step": 6047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.644707830268894e-06,
+      "loss": 0.7306,
+      "step": 6048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.643667258672889e-06,
+      "loss": 0.8898,
+      "step": 6049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.642626607255681e-06,
+      "loss": 0.7535,
+      "step": 6050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.641585876067807e-06,
+      "loss": 0.7837,
+      "step": 6051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.640545065159804e-06,
+      "loss": 0.7362,
+      "step": 6052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.639504174582222e-06,
+      "loss": 0.8084,
+      "step": 6053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6384632043856075e-06,
+      "loss": 0.6555,
+      "step": 6054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.637422154620511e-06,
+      "loss": 1.0156,
+      "step": 6055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.636381025337492e-06,
+      "loss": 0.6616,
+      "step": 6056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.635339816587109e-06,
+      "loss": 0.7411,
+      "step": 6057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.634298528419927e-06,
+      "loss": 0.7425,
+      "step": 6058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6332571608865116e-06,
+      "loss": 0.6631,
+      "step": 6059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.632215714037439e-06,
+      "loss": 0.5736,
+      "step": 6060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.63117418792328e-06,
+      "loss": 0.9782,
+      "step": 6061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.630132582594617e-06,
+      "loss": 0.5795,
+      "step": 6062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.629090898102034e-06,
+      "loss": 0.5564,
+      "step": 6063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.628049134496116e-06,
+      "loss": 0.5508,
+      "step": 6064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.627007291827454e-06,
+      "loss": 0.585,
+      "step": 6065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.625965370146645e-06,
+      "loss": 0.6527,
+      "step": 6066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6249233695042856e-06,
+      "loss": 0.795,
+      "step": 6067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623881289950982e-06,
+      "loss": 0.6184,
+      "step": 6068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.622839131537334e-06,
+      "loss": 0.5787,
+      "step": 6069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.621796894313958e-06,
+      "loss": 0.7933,
+      "step": 6070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6207545783314656e-06,
+      "loss": 0.8728,
+      "step": 6071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.619712183640474e-06,
+      "loss": 0.8701,
+      "step": 6072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.618669710291607e-06,
+      "loss": 0.621,
+      "step": 6073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6176271583354865e-06,
+      "loss": 0.8077,
+      "step": 6074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.616584527822745e-06,
+      "loss": 0.7167,
+      "step": 6075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.615541818804014e-06,
+      "loss": 0.5561,
+      "step": 6076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.614499031329933e-06,
+      "loss": 0.7378,
+      "step": 6077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6134561654511395e-06,
+      "loss": 0.726,
+      "step": 6078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.612413221218279e-06,
+      "loss": 0.6188,
+      "step": 6079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.611370198682001e-06,
+      "loss": 0.6169,
+      "step": 6080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.610327097892959e-06,
+      "loss": 0.8683,
+      "step": 6081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.6092839189018045e-06,
+      "loss": 0.6447,
+      "step": 6082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.608240661759199e-06,
+      "loss": 1.147,
+      "step": 6083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.607197326515808e-06,
+      "loss": 0.8762,
+      "step": 6084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.606153913222298e-06,
+      "loss": 0.8193,
+      "step": 6085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.605110421929339e-06,
+      "loss": 0.5869,
+      "step": 6086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.604066852687607e-06,
+      "loss": 0.6947,
+      "step": 6087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.603023205547782e-06,
+      "loss": 0.6895,
+      "step": 6088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.601979480560543e-06,
+      "loss": 0.7951,
+      "step": 6089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.600935677776579e-06,
+      "loss": 0.9633,
+      "step": 6090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.599891797246578e-06,
+      "loss": 0.6267,
+      "step": 6091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5988478390212375e-06,
+      "loss": 0.7779,
+      "step": 6092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5978038031512505e-06,
+      "loss": 0.8669,
+      "step": 6093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5967596896873205e-06,
+      "loss": 0.8032,
+      "step": 6094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.595715498680155e-06,
+      "loss": 0.7457,
+      "step": 6095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.59467123018046e-06,
+      "loss": 0.8487,
+      "step": 6096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.5936268842389475e-06,
+      "loss": 0.4437,
+      "step": 6097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.592582460906337e-06,
+      "loss": 0.905,
+      "step": 6098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.591537960233345e-06,
+      "loss": 0.8905,
+      "step": 6099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5904933822707e-06,
+      "loss": 0.6232,
+      "step": 6100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.589448727069126e-06,
+      "loss": 0.8571,
+      "step": 6101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5884039946793545e-06,
+      "loss": 0.958,
+      "step": 6102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.587359185152122e-06,
+      "loss": 0.6969,
+      "step": 6103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.586314298538168e-06,
+      "loss": 0.6101,
+      "step": 6104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5852693348882345e-06,
+      "loss": 0.9625,
+      "step": 6105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.584224294253069e-06,
+      "loss": 0.8203,
+      "step": 6106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.583179176683418e-06,
+      "loss": 0.8716,
+      "step": 6107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5821339822300405e-06,
+      "loss": 0.9548,
+      "step": 6108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.581088710943691e-06,
+      "loss": 0.6169,
+      "step": 6109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.58004336287513e-06,
+      "loss": 0.8491,
+      "step": 6110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.578997938075126e-06,
+      "loss": 0.5811,
+      "step": 6111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5779524365944445e-06,
+      "loss": 0.8394,
+      "step": 6112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5769068584838606e-06,
+      "loss": 0.8092,
+      "step": 6113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5758612037941484e-06,
+      "loss": 0.4952,
+      "step": 6114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5748154725760895e-06,
+      "loss": 0.8268,
+      "step": 6115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.573769664880468e-06,
+      "loss": 0.7383,
+      "step": 6116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.572723780758069e-06,
+      "loss": 0.8493,
+      "step": 6117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.571677820259686e-06,
+      "loss": 0.6871,
+      "step": 6118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5706317834361136e-06,
+      "loss": 1.0051,
+      "step": 6119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.569585670338148e-06,
+      "loss": 0.7843,
+      "step": 6120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.568539481016593e-06,
+      "loss": 0.8317,
+      "step": 6121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.567493215522256e-06,
+      "loss": 0.7795,
+      "step": 6122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.566446873905945e-06,
+      "loss": 0.6862,
+      "step": 6123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.565400456218473e-06,
+      "loss": 0.7649,
+      "step": 6124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.564353962510658e-06,
+      "loss": 0.386,
+      "step": 6125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.56330739283332e-06,
+      "loss": 0.7246,
+      "step": 6126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.562260747237286e-06,
+      "loss": 0.5641,
+      "step": 6127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5612140257733795e-06,
+      "loss": 0.5453,
+      "step": 6128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.560167228492436e-06,
+      "loss": 0.8995,
+      "step": 6129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.559120355445291e-06,
+      "loss": 0.7974,
+      "step": 6130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.55807340668278e-06,
+      "loss": 0.6296,
+      "step": 6131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.557026382255749e-06,
+      "loss": 0.9328,
+      "step": 6132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.555979282215046e-06,
+      "loss": 0.8719,
+      "step": 6133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.554932106611517e-06,
+      "loss": 0.6479,
+      "step": 6134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.553884855496016e-06,
+      "loss": 0.7619,
+      "step": 6135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.552837528919407e-06,
+      "loss": 0.9847,
+      "step": 6136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.551790126932543e-06,
+      "loss": 0.9508,
+      "step": 6137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.550742649586293e-06,
+      "loss": 0.7777,
+      "step": 6138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.549695096931526e-06,
+      "loss": 0.767,
+      "step": 6139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.548647469019113e-06,
+      "loss": 0.6809,
+      "step": 6140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.547599765899929e-06,
+      "loss": 0.7338,
+      "step": 6141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.546551987624855e-06,
+      "loss": 0.81,
+      "step": 6142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.545504134244773e-06,
+      "loss": 0.5217,
+      "step": 6143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5444562058105725e-06,
+      "loss": 0.6993,
+      "step": 6144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.543408202373138e-06,
+      "loss": 0.7148,
+      "step": 6145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.542360123983371e-06,
+      "loss": 0.5114,
+      "step": 6146
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5413119706921635e-06,
+      "loss": 0.9432,
+      "step": 6147
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.540263742550419e-06,
+      "loss": 0.7962,
+      "step": 6148
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.539215439609043e-06,
+      "loss": 0.5867,
+      "step": 6149
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.538167061918942e-06,
+      "loss": 0.8794,
+      "step": 6150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53711860953103e-06,
+      "loss": 0.6194,
+      "step": 6151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.536070082496222e-06,
+      "loss": 0.6795,
+      "step": 6152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.535021480865439e-06,
+      "loss": 0.7662,
+      "step": 6153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.533972804689603e-06,
+      "loss": 0.7677,
+      "step": 6154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.53292405401964e-06,
+      "loss": 0.8178,
+      "step": 6155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5318752289064815e-06,
+      "loss": 0.7555,
+      "step": 6156
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.530826329401062e-06,
+      "loss": 0.9138,
+      "step": 6157
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.529777355554317e-06,
+      "loss": 1.1664,
+      "step": 6158
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5287283074171884e-06,
+      "loss": 1.1173,
+      "step": 6159
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.527679185040623e-06,
+      "loss": 0.6093,
+      "step": 6160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.526629988475567e-06,
+      "loss": 0.8414,
+      "step": 6161
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.525580717772973e-06,
+      "loss": 0.7041,
+      "step": 6162
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5245313729837956e-06,
+      "loss": 0.5739,
+      "step": 6163
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5234819541589956e-06,
+      "loss": 0.7255,
+      "step": 6164
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.522432461349536e-06,
+      "loss": 0.893,
+      "step": 6165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5213828946063815e-06,
+      "loss": 0.8503,
+      "step": 6166
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.520333253980503e-06,
+      "loss": 0.7619,
+      "step": 6167
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.519283539522876e-06,
+      "loss": 0.7038,
+      "step": 6168
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5182337512844725e-06,
+      "loss": 0.5835,
+      "step": 6169
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.517183889316278e-06,
+      "loss": 0.7112,
+      "step": 6170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.516133953669277e-06,
+      "loss": 0.8106,
+      "step": 6171
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.515083944394454e-06,
+      "loss": 0.7718,
+      "step": 6172
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.514033861542801e-06,
+      "loss": 0.6786,
+      "step": 6173
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.512983705165315e-06,
+      "loss": 0.8468,
+      "step": 6174
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.511933475312995e-06,
+      "loss": 0.8633,
+      "step": 6175
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.51088317203684e-06,
+      "loss": 0.7061,
+      "step": 6176
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5098327953878585e-06,
+      "loss": 1.0315,
+      "step": 6177
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.50878234541706e-06,
+      "loss": 0.9735,
+      "step": 6178
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5077318221754536e-06,
+      "loss": 0.6532,
+      "step": 6179
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.506681225714061e-06,
+      "loss": 0.9323,
+      "step": 6180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.505630556083898e-06,
+      "loss": 0.5682,
+      "step": 6181
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.5045798133359924e-06,
+      "loss": 0.6839,
+      "step": 6182
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.503528997521365e-06,
+      "loss": 0.6233,
+      "step": 6183
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.502478108691054e-06,
+      "loss": 0.576,
+      "step": 6184
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.501427146896087e-06,
+      "loss": 0.5271,
+      "step": 6185
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.500376112187505e-06,
+      "loss": 0.5909,
+      "step": 6186
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.499325004616349e-06,
+      "loss": 0.7717,
+      "step": 6187
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.498273824233664e-06,
+      "loss": 0.8379,
+      "step": 6188
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.497222571090497e-06,
+      "loss": 0.7574,
+      "step": 6189
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.496171245237901e-06,
+      "loss": 1.0838,
+      "step": 6190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4951198467269305e-06,
+      "loss": 1.05,
+      "step": 6191
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.494068375608646e-06,
+      "loss": 0.8458,
+      "step": 6192
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.493016831934108e-06,
+      "loss": 0.8817,
+      "step": 6193
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.491965215754383e-06,
+      "loss": 0.656,
+      "step": 6194
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.490913527120543e-06,
+      "loss": 0.8587,
+      "step": 6195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.489861766083658e-06,
+      "loss": 0.769,
+      "step": 6196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.488809932694805e-06,
+      "loss": 0.8312,
+      "step": 6197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.487758027005066e-06,
+      "loss": 0.7445,
+      "step": 6198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.486706049065523e-06,
+      "loss": 0.4948,
+      "step": 6199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.485653998927261e-06,
+      "loss": 0.5979,
+      "step": 6200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.484601876641375e-06,
+      "loss": 0.5969,
+      "step": 6201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.483549682258956e-06,
+      "loss": 0.5872,
+      "step": 6202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.482497415831103e-06,
+      "loss": 0.912,
+      "step": 6203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4814450774089156e-06,
+      "loss": 0.9734,
+      "step": 6204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4803926670435e-06,
+      "loss": 1.0051,
+      "step": 6205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.479340184785965e-06,
+      "loss": 0.7916,
+      "step": 6206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.478287630687418e-06,
+      "loss": 0.7841,
+      "step": 6207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.477235004798977e-06,
+      "loss": 0.5884,
+      "step": 6208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.476182307171763e-06,
+      "loss": 0.9333,
+      "step": 6209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.475129537856893e-06,
+      "loss": 0.578,
+      "step": 6210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.474076696905496e-06,
+      "loss": 0.7522,
+      "step": 6211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4730237843686995e-06,
+      "loss": 0.9127,
+      "step": 6212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4719708002976355e-06,
+      "loss": 0.6423,
+      "step": 6213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.470917744743441e-06,
+      "loss": 0.742,
+      "step": 6214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.469864617757256e-06,
+      "loss": 0.8647,
+      "step": 6215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.468811419390222e-06,
+      "loss": 0.8304,
+      "step": 6216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.467758149693486e-06,
+      "loss": 0.7006,
+      "step": 6217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.466704808718198e-06,
+      "loss": 0.7496,
+      "step": 6218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.465651396515511e-06,
+      "loss": 0.9503,
+      "step": 6219
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.464597913136581e-06,
+      "loss": 0.6779,
+      "step": 6220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.463544358632569e-06,
+      "loss": 1.0167,
+      "step": 6221
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.462490733054638e-06,
+      "loss": 0.4509,
+      "step": 6222
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.461437036453957e-06,
+      "loss": 0.7862,
+      "step": 6223
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.460383268881693e-06,
+      "loss": 0.6806,
+      "step": 6224
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.459329430389023e-06,
+      "loss": 1.0634,
+      "step": 6225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.458275521027124e-06,
+      "loss": 0.8281,
+      "step": 6226
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4572215408471765e-06,
+      "loss": 0.5559,
+      "step": 6227
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.456167489900363e-06,
+      "loss": 0.8475,
+      "step": 6228
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.455113368237874e-06,
+      "loss": 0.556,
+      "step": 6229
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4540591759109005e-06,
+      "loss": 0.7889,
+      "step": 6230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4530049129706344e-06,
+      "loss": 0.7835,
+      "step": 6231
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.451950579468276e-06,
+      "loss": 0.7901,
+      "step": 6232
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.450896175455027e-06,
+      "loss": 0.73,
+      "step": 6233
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.449841700982091e-06,
+      "loss": 0.9557,
+      "step": 6234
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.448787156100675e-06,
+      "loss": 0.6919,
+      "step": 6235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.447732540861996e-06,
+      "loss": 0.849,
+      "step": 6236
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.446677855317265e-06,
+      "loss": 0.6936,
+      "step": 6237
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4456230995177e-06,
+      "loss": 0.6089,
+      "step": 6238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.444568273514524e-06,
+      "loss": 0.6384,
+      "step": 6239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.443513377358964e-06,
+      "loss": 0.8208,
+      "step": 6240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.442458411102247e-06,
+      "loss": 0.3728,
+      "step": 6241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.441403374795606e-06,
+      "loss": 0.7543,
+      "step": 6242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4403482684902755e-06,
+      "loss": 0.959,
+      "step": 6243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.4392930922374964e-06,
+      "loss": 0.705,
+      "step": 6244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.438237846088509e-06,
+      "loss": 0.6655,
+      "step": 6245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.437182530094559e-06,
+      "loss": 0.6109,
+      "step": 6246
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.436127144306898e-06,
+      "loss": 0.76,
+      "step": 6247
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.435071688776777e-06,
+      "loss": 0.677,
+      "step": 6248
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.434016163555452e-06,
+      "loss": 0.5672,
+      "step": 6249
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.432960568694182e-06,
+      "loss": 0.7931,
+      "step": 6250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.43190490424423e-06,
+      "loss": 1.0878,
+      "step": 6251
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.430849170256863e-06,
+      "loss": 0.9173,
+      "step": 6252
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.42979336678335e-06,
+      "loss": 0.7305,
+      "step": 6253
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4287374938749625e-06,
+      "loss": 0.5952,
+      "step": 6254
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.427681551582978e-06,
+      "loss": 0.7439,
+      "step": 6255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.426625539958676e-06,
+      "loss": 0.789,
+      "step": 6256
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.425569459053341e-06,
+      "loss": 0.8449,
+      "step": 6257
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.424513308918256e-06,
+      "loss": 0.8773,
+      "step": 6258
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.423457089604712e-06,
+      "loss": 0.5271,
+      "step": 6259
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.422400801164004e-06,
+      "loss": 0.8257,
+      "step": 6260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.421344443647425e-06,
+      "loss": 0.8883,
+      "step": 6261
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4202880171062774e-06,
+      "loss": 0.7029,
+      "step": 6262
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.419231521591863e-06,
+      "loss": 0.3911,
+      "step": 6263
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.418174957155489e-06,
+      "loss": 0.8451,
+      "step": 6264
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.417118323848465e-06,
+      "loss": 0.7767,
+      "step": 6265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.416061621722104e-06,
+      "loss": 0.5993,
+      "step": 6266
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4150048508277215e-06,
+      "loss": 0.6035,
+      "step": 6267
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.413948011216642e-06,
+      "loss": 0.5519,
+      "step": 6268
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4128911029401805e-06,
+      "loss": 0.7093,
+      "step": 6269
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4118341260496705e-06,
+      "loss": 0.7129,
+      "step": 6270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4107770805964395e-06,
+      "loss": 0.816,
+      "step": 6271
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.40971996663182e-06,
+      "loss": 0.6985,
+      "step": 6272
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.408662784207149e-06,
+      "loss": 0.8539,
+      "step": 6273
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.407605533373767e-06,
+      "loss": 0.4858,
+      "step": 6274
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.406548214183016e-06,
+      "loss": 0.654,
+      "step": 6275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.405490826686243e-06,
+      "loss": 0.7457,
+      "step": 6276
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.404433370934797e-06,
+      "loss": 0.6961,
+      "step": 6277
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.4033758469800334e-06,
+      "loss": 0.9222,
+      "step": 6278
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.402318254873305e-06,
+      "loss": 0.8181,
+      "step": 6279
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.401260594665975e-06,
+      "loss": 0.854,
+      "step": 6280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.400202866409405e-06,
+      "loss": 0.7192,
+      "step": 6281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.399145070154962e-06,
+      "loss": 0.864,
+      "step": 6282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.398087205954012e-06,
+      "loss": 0.8505,
+      "step": 6283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.397029273857934e-06,
+      "loss": 0.9964,
+      "step": 6284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3959712739181e-06,
+      "loss": 0.9802,
+      "step": 6285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.39491320618589e-06,
+      "loss": 0.7941,
+      "step": 6286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3938550707126865e-06,
+      "loss": 0.6007,
+      "step": 6287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.392796867549878e-06,
+      "loss": 0.5719,
+      "step": 6288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3917385967488525e-06,
+      "loss": 0.8088,
+      "step": 6289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.390680258361002e-06,
+      "loss": 0.9308,
+      "step": 6290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.389621852437723e-06,
+      "loss": 0.9643,
+      "step": 6291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.388563379030415e-06,
+      "loss": 0.921,
+      "step": 6292
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.387504838190479e-06,
+      "loss": 0.4388,
+      "step": 6293
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.386446229969323e-06,
+      "loss": 0.5841,
+      "step": 6294
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.385387554418357e-06,
+      "loss": 0.5228,
+      "step": 6295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.384328811588989e-06,
+      "loss": 0.5453,
+      "step": 6296
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.383270001532636e-06,
+      "loss": 0.6579,
+      "step": 6297
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38221112430072e-06,
+      "loss": 0.6996,
+      "step": 6298
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.38115217994466e-06,
+      "loss": 0.7,
+      "step": 6299
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.380093168515882e-06,
+      "loss": 0.919,
+      "step": 6300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.379034090065816e-06,
+      "loss": 0.8125,
+      "step": 6301
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.377974944645893e-06,
+      "loss": 0.6827,
+      "step": 6302
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.376915732307547e-06,
+      "loss": 0.7387,
+      "step": 6303
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.375856453102217e-06,
+      "loss": 0.8056,
+      "step": 6304
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.374797107081347e-06,
+      "loss": 0.5647,
+      "step": 6305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.373737694296379e-06,
+      "loss": 0.6159,
+      "step": 6306
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.372678214798761e-06,
+      "loss": 0.6712,
+      "step": 6307
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.371618668639946e-06,
+      "loss": 0.9184,
+      "step": 6308
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.370559055871389e-06,
+      "loss": 0.7683,
+      "step": 6309
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.369499376544546e-06,
+      "loss": 0.7416,
+      "step": 6310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.368439630710879e-06,
+      "loss": 0.7332,
+      "step": 6311
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3673798184218525e-06,
+      "loss": 0.7588,
+      "step": 6312
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.366319939728934e-06,
+      "loss": 0.9838,
+      "step": 6313
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.365259994683594e-06,
+      "loss": 0.623,
+      "step": 6314
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.364199983337306e-06,
+      "loss": 0.7596,
+      "step": 6315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.363139905741548e-06,
+      "loss": 0.5817,
+      "step": 6316
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3620797619478e-06,
+      "loss": 0.6434,
+      "step": 6317
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.361019552007546e-06,
+      "loss": 0.8963,
+      "step": 6318
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.359959275972272e-06,
+      "loss": 0.9683,
+      "step": 6319
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3588989338934695e-06,
+      "loss": 0.6678,
+      "step": 6320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.357838525822629e-06,
+      "loss": 0.754,
+      "step": 6321
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.356778051811252e-06,
+      "loss": 0.8596,
+      "step": 6322
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.355717511910833e-06,
+      "loss": 1.0279,
+      "step": 6323
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.354656906172878e-06,
+      "loss": 0.7916,
+      "step": 6324
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.353596234648891e-06,
+      "loss": 1.0093,
+      "step": 6325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3525354973903815e-06,
+      "loss": 0.4335,
+      "step": 6326
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.351474694448865e-06,
+      "loss": 0.9154,
+      "step": 6327
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.350413825875852e-06,
+      "loss": 0.697,
+      "step": 6328
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3493528917228664e-06,
+      "loss": 0.8433,
+      "step": 6329
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.348291892041427e-06,
+      "loss": 0.7312,
+      "step": 6330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.34723082688306e-06,
+      "loss": 0.618,
+      "step": 6331
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.346169696299294e-06,
+      "loss": 1.0115,
+      "step": 6332
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3451085003416604e-06,
+      "loss": 0.5473,
+      "step": 6333
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3440472390616925e-06,
+      "loss": 0.6627,
+      "step": 6334
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.342985912510929e-06,
+      "loss": 0.5446,
+      "step": 6335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.341924520740914e-06,
+      "loss": 0.6961,
+      "step": 6336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.340863063803187e-06,
+      "loss": 0.6359,
+      "step": 6337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3398015417493e-06,
+      "loss": 0.6083,
+      "step": 6338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3387399546308e-06,
+      "loss": 0.7645,
+      "step": 6339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.337678302499239e-06,
+      "loss": 0.8401,
+      "step": 6340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.33661658540618e-06,
+      "loss": 0.668,
+      "step": 6341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.335554803403179e-06,
+      "loss": 0.717,
+      "step": 6342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3344929565418e-06,
+      "loss": 1.0715,
+      "step": 6343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.333431044873609e-06,
+      "loss": 0.6997,
+      "step": 6344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.332369068450175e-06,
+      "loss": 0.6875,
+      "step": 6345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.331307027323073e-06,
+      "loss": 0.5394,
+      "step": 6346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.330244921543876e-06,
+      "loss": 0.8633,
+      "step": 6347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.329182751164164e-06,
+      "loss": 0.6159,
+      "step": 6348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.328120516235521e-06,
+      "loss": 0.8185,
+      "step": 6349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.327058216809529e-06,
+      "loss": 0.6181,
+      "step": 6350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.325995852937776e-06,
+      "loss": 0.7501,
+      "step": 6351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3249334246718575e-06,
+      "loss": 0.7702,
+      "step": 6352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.323870932063364e-06,
+      "loss": 0.8205,
+      "step": 6353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.322808375163896e-06,
+      "loss": 0.5059,
+      "step": 6354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3217457540250516e-06,
+      "loss": 0.9864,
+      "step": 6355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3206830686984366e-06,
+      "loss": 0.621,
+      "step": 6356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.319620319235658e-06,
+      "loss": 0.8313,
+      "step": 6357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.318557505688327e-06,
+      "loss": 0.9566,
+      "step": 6358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.317494628108054e-06,
+      "loss": 0.6336,
+      "step": 6359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.316431686546458e-06,
+      "loss": 0.8941,
+      "step": 6360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.315368681055157e-06,
+      "loss": 0.9444,
+      "step": 6361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.314305611685774e-06,
+      "loss": 0.6291,
+      "step": 6362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.313242478489934e-06,
+      "loss": 0.6875,
+      "step": 6363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.312179281519268e-06,
+      "loss": 0.9125,
+      "step": 6364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.311116020825405e-06,
+      "loss": 1.0869,
+      "step": 6365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3100526964599815e-06,
+      "loss": 0.6384,
+      "step": 6366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.308989308474636e-06,
+      "loss": 0.7996,
+      "step": 6367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.30792585692101e-06,
+      "loss": 0.5724,
+      "step": 6368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.306862341850746e-06,
+      "loss": 0.7798,
+      "step": 6369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3057987633154915e-06,
+      "loss": 0.8157,
+      "step": 6370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3047351213669006e-06,
+      "loss": 0.7424,
+      "step": 6371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.3036714160566216e-06,
+      "loss": 0.7857,
+      "step": 6372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.302607647436312e-06,
+      "loss": 0.7741,
+      "step": 6373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.301543815557636e-06,
+      "loss": 1.0886,
+      "step": 6374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.300479920472252e-06,
+      "loss": 0.7751,
+      "step": 6375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.299415962231827e-06,
+      "loss": 0.5069,
+      "step": 6376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.29835194088803e-06,
+      "loss": 0.6952,
+      "step": 6377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2972878564925325e-06,
+      "loss": 1.0343,
+      "step": 6378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.296223709097009e-06,
+      "loss": 0.7019,
+      "step": 6379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.295159498753141e-06,
+      "loss": 0.9095,
+      "step": 6380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.294095225512604e-06,
+      "loss": 0.8872,
+      "step": 6381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.293030889427088e-06,
+      "loss": 0.6873,
+      "step": 6382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2919664905482745e-06,
+      "loss": 0.7465,
+      "step": 6383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.290902028927858e-06,
+      "loss": 0.6835,
+      "step": 6384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.289837504617531e-06,
+      "loss": 0.853,
+      "step": 6385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.28877291766899e-06,
+      "loss": 0.7831,
+      "step": 6386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.287708268133932e-06,
+      "loss": 0.9107,
+      "step": 6387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.28664355606406e-06,
+      "loss": 0.871,
+      "step": 6388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.285578781511083e-06,
+      "loss": 0.6911,
+      "step": 6389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.284513944526707e-06,
+      "loss": 0.5974,
+      "step": 6390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.283449045162642e-06,
+      "loss": 0.785,
+      "step": 6391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.282384083470605e-06,
+      "loss": 1.0211,
+      "step": 6392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.2813190595023135e-06,
+      "loss": 0.7341,
+      "step": 6393
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.280253973309487e-06,
+      "loss": 1.0983,
+      "step": 6394
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.279188824943851e-06,
+      "loss": 0.6979,
+      "step": 6395
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27812361445713e-06,
+      "loss": 0.8564,
+      "step": 6396
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2770583419010525e-06,
+      "loss": 0.777,
+      "step": 6397
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.275993007327356e-06,
+      "loss": 0.7151,
+      "step": 6398
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.274927610787771e-06,
+      "loss": 1.0836,
+      "step": 6399
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.27386215233404e-06,
+      "loss": 0.7658,
+      "step": 6400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.272796632017902e-06,
+      "loss": 0.5836,
+      "step": 6401
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.271731049891104e-06,
+      "loss": 0.8584,
+      "step": 6402
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.270665406005393e-06,
+      "loss": 0.4427,
+      "step": 6403
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.26959970041252e-06,
+      "loss": 0.6926,
+      "step": 6404
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.268533933164237e-06,
+      "loss": 0.53,
+      "step": 6405
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.267468104312303e-06,
+      "loss": 0.849,
+      "step": 6406
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.266402213908474e-06,
+      "loss": 0.6271,
+      "step": 6407
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.265336262004518e-06,
+      "loss": 0.8083,
+      "step": 6408
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.264270248652199e-06,
+      "loss": 0.7904,
+      "step": 6409
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.263204173903282e-06,
+      "loss": 0.7846,
+      "step": 6410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.262138037809542e-06,
+      "loss": 0.5522,
+      "step": 6411
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2610718404227546e-06,
+      "loss": 0.9341,
+      "step": 6412
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.260005581794695e-06,
+      "loss": 0.7207,
+      "step": 6413
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2589392619771435e-06,
+      "loss": 0.7612,
+      "step": 6414
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.257872881021886e-06,
+      "loss": 0.7031,
+      "step": 6415
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.256806438980707e-06,
+      "loss": 0.8063,
+      "step": 6416
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.255739935905396e-06,
+      "loss": 0.6669,
+      "step": 6417
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.254673371847748e-06,
+      "loss": 0.5289,
+      "step": 6418
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.253606746859555e-06,
+      "loss": 0.7916,
+      "step": 6419
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.252540060992619e-06,
+      "loss": 0.8885,
+      "step": 6420
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.251473314298737e-06,
+      "loss": 0.7087,
+      "step": 6421
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.250406506829717e-06,
+      "loss": 0.6393,
+      "step": 6422
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.249339638637365e-06,
+      "loss": 0.856,
+      "step": 6423
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.248272709773491e-06,
+      "loss": 0.714,
+      "step": 6424
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.247205720289907e-06,
+      "loss": 0.7585,
+      "step": 6425
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2461386702384306e-06,
+      "loss": 0.5375,
+      "step": 6426
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.245071559670881e-06,
+      "loss": 0.6691,
+      "step": 6427
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.24400438863908e-06,
+      "loss": 0.755,
+      "step": 6428
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.242937157194851e-06,
+      "loss": 1.0376,
+      "step": 6429
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.241869865390024e-06,
+      "loss": 0.614,
+      "step": 6430
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.240802513276427e-06,
+      "loss": 0.7329,
+      "step": 6431
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.239735100905897e-06,
+      "loss": 0.7819,
+      "step": 6432
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.238667628330269e-06,
+      "loss": 1.0911,
+      "step": 6433
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.237600095601382e-06,
+      "loss": 0.7194,
+      "step": 6434
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.236532502771078e-06,
+      "loss": 0.648,
+      "step": 6435
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.235464849891205e-06,
+      "loss": 0.5194,
+      "step": 6436
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.234397137013608e-06,
+      "loss": 0.553,
+      "step": 6437
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2333293641901414e-06,
+      "loss": 0.5425,
+      "step": 6438
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.232261531472656e-06,
+      "loss": 0.7689,
+      "step": 6439
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.231193638913012e-06,
+      "loss": 0.9868,
+      "step": 6440
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.230125686563068e-06,
+      "loss": 0.8181,
+      "step": 6441
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2290576744746855e-06,
+      "loss": 1.0289,
+      "step": 6442
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.227989602699732e-06,
+      "loss": 0.5704,
+      "step": 6443
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2269214712900776e-06,
+      "loss": 0.6919,
+      "step": 6444
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.225853280297588e-06,
+      "loss": 0.5447,
+      "step": 6445
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.224785029774143e-06,
+      "loss": 0.805,
+      "step": 6446
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2237167197716195e-06,
+      "loss": 0.6123,
+      "step": 6447
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.222648350341897e-06,
+      "loss": 1.0211,
+      "step": 6448
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.221579921536856e-06,
+      "loss": 0.967,
+      "step": 6449
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.220511433408386e-06,
+      "loss": 0.7001,
+      "step": 6450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.219442886008375e-06,
+      "loss": 0.5883,
+      "step": 6451
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.218374279388713e-06,
+      "loss": 0.6906,
+      "step": 6452
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.217305613601296e-06,
+      "loss": 0.7633,
+      "step": 6453
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.216236888698021e-06,
+      "loss": 0.9152,
+      "step": 6454
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.215168104730788e-06,
+      "loss": 0.783,
+      "step": 6455
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.214099261751503e-06,
+      "loss": 0.6008,
+      "step": 6456
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.213030359812069e-06,
+      "loss": 0.7602,
+      "step": 6457
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.211961398964397e-06,
+      "loss": 0.8006,
+      "step": 6458
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2108923792603945e-06,
+      "loss": 0.7812,
+      "step": 6459
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.209823300751983e-06,
+      "loss": 0.7074,
+      "step": 6460
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.208754163491075e-06,
+      "loss": 0.6191,
+      "step": 6461
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.207684967529592e-06,
+      "loss": 0.7356,
+      "step": 6462
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.206615712919458e-06,
+      "loss": 0.8401,
+      "step": 6463
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.205546399712599e-06,
+      "loss": 0.733,
+      "step": 6464
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.204477027960942e-06,
+      "loss": 0.5745,
+      "step": 6465
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.203407597716421e-06,
+      "loss": 0.8018,
+      "step": 6466
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.20233810903097e-06,
+      "loss": 0.6075,
+      "step": 6467
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.201268561956526e-06,
+      "loss": 0.819,
+      "step": 6468
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.2001989565450305e-06,
+      "loss": 0.8192,
+      "step": 6469
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.199129292848425e-06,
+      "loss": 0.6692,
+      "step": 6470
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.198059570918656e-06,
+      "loss": 0.6706,
+      "step": 6471
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.196989790807674e-06,
+      "loss": 0.9641,
+      "step": 6472
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.195919952567426e-06,
+      "loss": 0.8272,
+      "step": 6473
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.194850056249873e-06,
+      "loss": 0.7517,
+      "step": 6474
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.193780101906967e-06,
+      "loss": 0.7122,
+      "step": 6475
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.192710089590669e-06,
+      "loss": 0.9842,
+      "step": 6476
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.191640019352942e-06,
+      "loss": 0.9487,
+      "step": 6477
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1905698912457546e-06,
+      "loss": 0.729,
+      "step": 6478
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.189499705321073e-06,
+      "loss": 0.7684,
+      "step": 6479
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.188429461630866e-06,
+      "loss": 0.8886,
+      "step": 6480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.187359160227111e-06,
+      "loss": 0.7174,
+      "step": 6481
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.186288801161786e-06,
+      "loss": 0.6443,
+      "step": 6482
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.185218384486866e-06,
+      "loss": 0.6454,
+      "step": 6483
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.184147910254339e-06,
+      "loss": 0.8527,
+      "step": 6484
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.183077378516187e-06,
+      "loss": 0.9292,
+      "step": 6485
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.182006789324399e-06,
+      "loss": 0.7198,
+      "step": 6486
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.180936142730964e-06,
+      "loss": 0.73,
+      "step": 6487
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.179865438787878e-06,
+      "loss": 0.8238,
+      "step": 6488
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.178794677547138e-06,
+      "loss": 0.8519,
+      "step": 6489
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.17772385906074e-06,
+      "loss": 0.835,
+      "step": 6490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.176652983380689e-06,
+      "loss": 0.9483,
+      "step": 6491
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.175582050558988e-06,
+      "loss": 0.5271,
+      "step": 6492
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.174511060647648e-06,
+      "loss": 0.5013,
+      "step": 6493
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.173440013698673e-06,
+      "loss": 0.7922,
+      "step": 6494
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1723689097640815e-06,
+      "loss": 0.5872,
+      "step": 6495
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.171297748895889e-06,
+      "loss": 0.868,
+      "step": 6496
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.170226531146111e-06,
+      "loss": 0.6441,
+      "step": 6497
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.169155256566772e-06,
+      "loss": 0.6549,
+      "step": 6498
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.168083925209896e-06,
+      "loss": 0.5832,
+      "step": 6499
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.167012537127507e-06,
+      "loss": 0.8318,
+      "step": 6500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.165941092371637e-06,
+      "loss": 0.7915,
+      "step": 6501
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.164869590994318e-06,
+      "loss": 0.7076,
+      "step": 6502
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.163798033047586e-06,
+      "loss": 0.7839,
+      "step": 6503
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.162726418583477e-06,
+      "loss": 0.9583,
+      "step": 6504
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.161654747654033e-06,
+      "loss": 0.679,
+      "step": 6505
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.160583020311297e-06,
+      "loss": 0.849,
+      "step": 6506
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1595112366073164e-06,
+      "loss": 0.3912,
+      "step": 6507
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.158439396594138e-06,
+      "loss": 0.8061,
+      "step": 6508
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.157367500323816e-06,
+      "loss": 0.8793,
+      "step": 6509
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.156295547848403e-06,
+      "loss": 0.9616,
+      "step": 6510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.155223539219954e-06,
+      "loss": 0.9599,
+      "step": 6511
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.154151474490534e-06,
+      "loss": 0.868,
+      "step": 6512
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.153079353712201e-06,
+      "loss": 0.5217,
+      "step": 6513
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.152007176937024e-06,
+      "loss": 0.4802,
+      "step": 6514
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.150934944217067e-06,
+      "loss": 1.0783,
+      "step": 6515
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.149862655604404e-06,
+      "loss": 0.7079,
+      "step": 6516
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.148790311151107e-06,
+      "loss": 0.7676,
+      "step": 6517
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14771791090925e-06,
+      "loss": 0.7164,
+      "step": 6518
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.146645454930916e-06,
+      "loss": 0.7821,
+      "step": 6519
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.145572943268186e-06,
+      "loss": 0.7518,
+      "step": 6520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.14450037597314e-06,
+      "loss": 0.5218,
+      "step": 6521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.143427753097868e-06,
+      "loss": 1.0171,
+      "step": 6522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.142355074694461e-06,
+      "loss": 0.8348,
+      "step": 6523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.141282340815009e-06,
+      "loss": 0.8759,
+      "step": 6524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.140209551511609e-06,
+      "loss": 0.7661,
+      "step": 6525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.139136706836356e-06,
+      "loss": 0.8384,
+      "step": 6526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.138063806841353e-06,
+      "loss": 0.9669,
+      "step": 6527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.136990851578701e-06,
+      "loss": 0.7196,
+      "step": 6528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.135917841100508e-06,
+      "loss": 0.5947,
+      "step": 6529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1348447754588805e-06,
+      "loss": 0.6676,
+      "step": 6530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.133771654705932e-06,
+      "loss": 0.6735,
+      "step": 6531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.132698478893773e-06,
+      "loss": 0.7134,
+      "step": 6532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.131625248074522e-06,
+      "loss": 0.768,
+      "step": 6533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.1305519623003e-06,
+      "loss": 0.4865,
+      "step": 6534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.129478621623224e-06,
+      "loss": 0.8604,
+      "step": 6535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.128405226095424e-06,
+      "loss": 0.8683,
+      "step": 6536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.127331775769023e-06,
+      "loss": 0.8418,
+      "step": 6537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.126258270696152e-06,
+      "loss": 1.1099,
+      "step": 6538
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.125184710928945e-06,
+      "loss": 0.8731,
+      "step": 6539
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.124111096519536e-06,
+      "loss": 0.5684,
+      "step": 6540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.123037427520062e-06,
+      "loss": 0.7985,
+      "step": 6541
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.121963703982664e-06,
+      "loss": 0.5942,
+      "step": 6542
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.120889925959486e-06,
+      "loss": 0.7888,
+      "step": 6543
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.119816093502674e-06,
+      "loss": 0.6764,
+      "step": 6544
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.118742206664373e-06,
+      "loss": 0.7317,
+      "step": 6545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.117668265496738e-06,
+      "loss": 0.9297,
+      "step": 6546
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.116594270051922e-06,
+      "loss": 0.7045,
+      "step": 6547
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.11552022038208e-06,
+      "loss": 0.6808,
+      "step": 6548
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.114446116539369e-06,
+      "loss": 0.7282,
+      "step": 6549
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.113371958575955e-06,
+      "loss": 0.8669,
+      "step": 6550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.112297746544e-06,
+      "loss": 0.8887,
+      "step": 6551
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.1112234804956715e-06,
+      "loss": 0.9118,
+      "step": 6552
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.110149160483139e-06,
+      "loss": 0.9439,
+      "step": 6553
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.109074786558572e-06,
+      "loss": 0.6094,
+      "step": 6554
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.10800035877415e-06,
+      "loss": 0.9245,
+      "step": 6555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.106925877182044e-06,
+      "loss": 0.8637,
+      "step": 6556
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.105851341834439e-06,
+      "loss": 0.6603,
+      "step": 6557
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.104776752783518e-06,
+      "loss": 0.7839,
+      "step": 6558
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.10370211008146e-06,
+      "loss": 0.7521,
+      "step": 6559
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.10262741378046e-06,
+      "loss": 0.8255,
+      "step": 6560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.101552663932704e-06,
+      "loss": 0.7852,
+      "step": 6561
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.100477860590386e-06,
+      "loss": 0.9679,
+      "step": 6562
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0994030038057015e-06,
+      "loss": 0.602,
+      "step": 6563
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.09832809363085e-06,
+      "loss": 0.8781,
+      "step": 6564
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.097253130118029e-06,
+      "loss": 1.1827,
+      "step": 6565
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.096178113319446e-06,
+      "loss": 0.6747,
+      "step": 6566
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.095103043287303e-06,
+      "loss": 0.9304,
+      "step": 6567
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0940279200738105e-06,
+      "loss": 0.7801,
+      "step": 6568
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.092952743731179e-06,
+      "loss": 0.6085,
+      "step": 6569
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.091877514311623e-06,
+      "loss": 0.7435,
+      "step": 6570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.090802231867358e-06,
+      "loss": 0.6527,
+      "step": 6571
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.089726896450604e-06,
+      "loss": 0.5868,
+      "step": 6572
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.088651508113579e-06,
+      "loss": 0.5246,
+      "step": 6573
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.087576066908512e-06,
+      "loss": 0.7022,
+      "step": 6574
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0865005728876245e-06,
+      "loss": 0.8465,
+      "step": 6575
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.085425026103149e-06,
+      "loss": 0.9057,
+      "step": 6576
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.084349426607314e-06,
+      "loss": 0.7398,
+      "step": 6577
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.083273774452359e-06,
+      "loss": 0.6375,
+      "step": 6578
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0821980696905145e-06,
+      "loss": 0.7611,
+      "step": 6579
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.081122312374025e-06,
+      "loss": 0.4623,
+      "step": 6580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.080046502555129e-06,
+      "loss": 0.5389,
+      "step": 6581
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.078970640286073e-06,
+      "loss": 0.8338,
+      "step": 6582
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0778947256191016e-06,
+      "loss": 0.5008,
+      "step": 6583
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.076818758606467e-06,
+      "loss": 0.9076,
+      "step": 6584
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.07574273930042e-06,
+      "loss": 0.8693,
+      "step": 6585
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.074666667753214e-06,
+      "loss": 0.8002,
+      "step": 6586
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.073590544017107e-06,
+      "loss": 0.6423,
+      "step": 6587
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.072514368144361e-06,
+      "loss": 0.6828,
+      "step": 6588
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.071438140187235e-06,
+      "loss": 0.9357,
+      "step": 6589
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0703618601979944e-06,
+      "loss": 0.814,
+      "step": 6590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.069285528228908e-06,
+      "loss": 0.8468,
+      "step": 6591
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0682091443322435e-06,
+      "loss": 0.8143,
+      "step": 6592
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0671327085602746e-06,
+      "loss": 0.8107,
+      "step": 6593
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.066056220965275e-06,
+      "loss": 0.6712,
+      "step": 6594
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.064979681599523e-06,
+      "loss": 0.6133,
+      "step": 6595
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0639030905153e-06,
+      "loss": 0.9102,
+      "step": 6596
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.062826447764883e-06,
+      "loss": 0.6619,
+      "step": 6597
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.061749753400563e-06,
+      "loss": 0.6591,
+      "step": 6598
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.060673007474624e-06,
+      "loss": 0.676,
+      "step": 6599
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.059596210039358e-06,
+      "loss": 0.5719,
+      "step": 6600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.058519361147055e-06,
+      "loss": 1.0816,
+      "step": 6601
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.05744246085001e-06,
+      "loss": 0.6556,
+      "step": 6602
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.056365509200523e-06,
+      "loss": 0.745,
+      "step": 6603
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.05528850625089e-06,
+      "loss": 0.8168,
+      "step": 6604
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0542114520534165e-06,
+      "loss": 0.7067,
+      "step": 6605
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.053134346660407e-06,
+      "loss": 0.6524,
+      "step": 6606
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.052057190124169e-06,
+      "loss": 1.145,
+      "step": 6607
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.050979982497008e-06,
+      "loss": 0.9643,
+      "step": 6608
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.049902723831243e-06,
+      "loss": 0.7383,
+      "step": 6609
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.048825414179184e-06,
+      "loss": 0.8257,
+      "step": 6610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0477480535931495e-06,
+      "loss": 0.9355,
+      "step": 6611
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.046670642125461e-06,
+      "loss": 0.6973,
+      "step": 6612
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.045593179828438e-06,
+      "loss": 0.4176,
+      "step": 6613
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.044515666754407e-06,
+      "loss": 0.8168,
+      "step": 6614
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0434381029556945e-06,
+      "loss": 0.8592,
+      "step": 6615
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.04236048848463e-06,
+      "loss": 0.6276,
+      "step": 6616
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.041282823393546e-06,
+      "loss": 0.6642,
+      "step": 6617
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.040205107734776e-06,
+      "loss": 0.7908,
+      "step": 6618
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.039127341560657e-06,
+      "loss": 0.7934,
+      "step": 6619
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.038049524923532e-06,
+      "loss": 0.92,
+      "step": 6620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.036971657875736e-06,
+      "loss": 0.7817,
+      "step": 6621
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0358937404696195e-06,
+      "loss": 0.8117,
+      "step": 6622
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.034815772757528e-06,
+      "loss": 0.831,
+      "step": 6623
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.033737754791809e-06,
+      "loss": 0.5419,
+      "step": 6624
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.032659686624813e-06,
+      "loss": 0.7339,
+      "step": 6625
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0315815683088986e-06,
+      "loss": 0.6698,
+      "step": 6626
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.030503399896418e-06,
+      "loss": 0.5635,
+      "step": 6627
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.029425181439732e-06,
+      "loss": 0.7971,
+      "step": 6628
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.028346912991202e-06,
+      "loss": 0.635,
+      "step": 6629
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.027268594603192e-06,
+      "loss": 0.8867,
+      "step": 6630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.026190226328069e-06,
+      "loss": 0.9836,
+      "step": 6631
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.025111808218198e-06,
+      "loss": 0.6998,
+      "step": 6632
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.024033340325954e-06,
+      "loss": 0.7636,
+      "step": 6633
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0229548227037095e-06,
+      "loss": 0.5199,
+      "step": 6634
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.021876255403839e-06,
+      "loss": 0.4088,
+      "step": 6635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.020797638478721e-06,
+      "loss": 0.778,
+      "step": 6636
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01971897198074e-06,
+      "loss": 0.9223,
+      "step": 6637
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.0186402559622755e-06,
+      "loss": 0.7902,
+      "step": 6638
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.017561490475713e-06,
+      "loss": 0.9594,
+      "step": 6639
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01648267557344e-06,
+      "loss": 0.6455,
+      "step": 6640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.01540381130785e-06,
+      "loss": 0.7135,
+      "step": 6641
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.014324897731334e-06,
+      "loss": 0.6696,
+      "step": 6642
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.013245934896286e-06,
+      "loss": 0.6452,
+      "step": 6643
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.012166922855105e-06,
+      "loss": 0.8498,
+      "step": 6644
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.011087861660191e-06,
+      "loss": 0.709,
+      "step": 6645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.010008751363944e-06,
+      "loss": 0.7102,
+      "step": 6646
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.008929592018773e-06,
+      "loss": 0.8943,
+      "step": 6647
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.007850383677081e-06,
+      "loss": 0.8315,
+      "step": 6648
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.006771126391278e-06,
+      "loss": 0.7713,
+      "step": 6649
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.005691820213778e-06,
+      "loss": 0.7804,
+      "step": 6650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.004612465196994e-06,
+      "loss": 0.7323,
+      "step": 6651
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.003533061393343e-06,
+      "loss": 0.6594,
+      "step": 6652
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.002453608855243e-06,
+      "loss": 0.8389,
+      "step": 6653
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.001374107635116e-06,
+      "loss": 0.7502,
+      "step": 6654
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.000294557785386e-06,
+      "loss": 0.5126,
+      "step": 6655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.999214959358477e-06,
+      "loss": 0.5759,
+      "step": 6656
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.998135312406821e-06,
+      "loss": 0.7504,
+      "step": 6657
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.997055616982847e-06,
+      "loss": 0.7797,
+      "step": 6658
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.995975873138985e-06,
+      "loss": 0.6392,
+      "step": 6659
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.994896080927675e-06,
+      "loss": 1.0513,
+      "step": 6660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9938162404013544e-06,
+      "loss": 0.9759,
+      "step": 6661
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.992736351612461e-06,
+      "loss": 0.7016,
+      "step": 6662
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.991656414613437e-06,
+      "loss": 0.6156,
+      "step": 6663
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.990576429456732e-06,
+      "loss": 0.9373,
+      "step": 6664
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.989496396194787e-06,
+      "loss": 0.6326,
+      "step": 6665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.988416314880055e-06,
+      "loss": 0.8338,
+      "step": 6666
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9873361855649885e-06,
+      "loss": 0.8529,
+      "step": 6667
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9862560083020395e-06,
+      "loss": 0.6425,
+      "step": 6668
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.985175783143666e-06,
+      "loss": 0.9062,
+      "step": 6669
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9840955101423244e-06,
+      "loss": 0.6348,
+      "step": 6670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.983015189350479e-06,
+      "loss": 0.7663,
+      "step": 6671
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.981934820820593e-06,
+      "loss": 0.7145,
+      "step": 6672
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.980854404605129e-06,
+      "loss": 0.5487,
+      "step": 6673
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9797739407565595e-06,
+      "loss": 0.846,
+      "step": 6674
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.978693429327351e-06,
+      "loss": 0.7987,
+      "step": 6675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9776128703699785e-06,
+      "loss": 0.5517,
+      "step": 6676
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.976532263936916e-06,
+      "loss": 0.8008,
+      "step": 6677
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.975451610080643e-06,
+      "loss": 0.742,
+      "step": 6678
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.974370908853635e-06,
+      "loss": 0.8841,
+      "step": 6679
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.973290160308378e-06,
+      "loss": 0.9855,
+      "step": 6680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.972209364497355e-06,
+      "loss": 0.886,
+      "step": 6681
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.971128521473052e-06,
+      "loss": 0.7858,
+      "step": 6682
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.970047631287958e-06,
+      "loss": 0.894,
+      "step": 6683
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.968966693994563e-06,
+      "loss": 0.7997,
+      "step": 6684
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.967885709645364e-06,
+      "loss": 0.6586,
+      "step": 6685
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.9668046782928545e-06,
+      "loss": 0.5054,
+      "step": 6686
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.965723599989529e-06,
+      "loss": 0.4971,
+      "step": 6687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.964642474787895e-06,
+      "loss": 0.7212,
+      "step": 6688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9635613027404495e-06,
+      "loss": 0.8261,
+      "step": 6689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.962480083899699e-06,
+      "loss": 0.6358,
+      "step": 6690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.961398818318151e-06,
+      "loss": 0.6861,
+      "step": 6691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.960317506048313e-06,
+      "loss": 0.6279,
+      "step": 6692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.959236147142701e-06,
+      "loss": 0.8862,
+      "step": 6693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.958154741653822e-06,
+      "loss": 0.728,
+      "step": 6694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.957073289634199e-06,
+      "loss": 0.863,
+      "step": 6695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9559917911363475e-06,
+      "loss": 0.7493,
+      "step": 6696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.954910246212787e-06,
+      "loss": 0.7447,
+      "step": 6697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.953828654916042e-06,
+      "loss": 0.9119,
+      "step": 6698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.952747017298638e-06,
+      "loss": 0.5601,
+      "step": 6699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.951665333413101e-06,
+      "loss": 0.9012,
+      "step": 6700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9505836033119616e-06,
+      "loss": 0.8576,
+      "step": 6701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.949501827047753e-06,
+      "loss": 1.0142,
+      "step": 6702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.948420004673006e-06,
+      "loss": 0.7582,
+      "step": 6703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.947338136240259e-06,
+      "loss": 0.6983,
+      "step": 6704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.946256221802052e-06,
+      "loss": 0.8063,
+      "step": 6705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.945174261410923e-06,
+      "loss": 0.7292,
+      "step": 6706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.944092255119418e-06,
+      "loss": 0.6794,
+      "step": 6707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.94301020298008e-06,
+      "loss": 0.6388,
+      "step": 6708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.941928105045457e-06,
+      "loss": 0.7132,
+      "step": 6709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9408459613681e-06,
+      "loss": 0.8541,
+      "step": 6710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93976377200056e-06,
+      "loss": 0.8556,
+      "step": 6711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.938681536995392e-06,
+      "loss": 0.9052,
+      "step": 6712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.937599256405151e-06,
+      "loss": 0.7707,
+      "step": 6713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.936516930282397e-06,
+      "loss": 0.7687,
+      "step": 6714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.93543455867969e-06,
+      "loss": 0.7832,
+      "step": 6715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.934352141649594e-06,
+      "loss": 0.9727,
+      "step": 6716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.933269679244673e-06,
+      "loss": 1.0204,
+      "step": 6717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.932187171517496e-06,
+      "loss": 0.4043,
+      "step": 6718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9311046185206325e-06,
+      "loss": 0.8758,
+      "step": 6719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.930022020306654e-06,
+      "loss": 0.6999,
+      "step": 6720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.928939376928134e-06,
+      "loss": 0.5786,
+      "step": 6721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9278566884376476e-06,
+      "loss": 0.97,
+      "step": 6722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.926773954887777e-06,
+      "loss": 0.7263,
+      "step": 6723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9256911763310984e-06,
+      "loss": 0.9606,
+      "step": 6724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.924608352820198e-06,
+      "loss": 0.6775,
+      "step": 6725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.923525484407661e-06,
+      "loss": 0.8381,
+      "step": 6726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.922442571146072e-06,
+      "loss": 0.7025,
+      "step": 6727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9213596130880225e-06,
+      "loss": 0.7848,
+      "step": 6728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.920276610286102e-06,
+      "loss": 0.7033,
+      "step": 6729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.919193562792907e-06,
+      "loss": 0.6791,
+      "step": 6730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9181104706610315e-06,
+      "loss": 0.832,
+      "step": 6731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.917027333943073e-06,
+      "loss": 0.7878,
+      "step": 6732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.915944152691634e-06,
+      "loss": 0.7251,
+      "step": 6733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.914860926959316e-06,
+      "loss": 0.8037,
+      "step": 6734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.913777656798722e-06,
+      "loss": 0.6575,
+      "step": 6735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.912694342262458e-06,
+      "loss": 1.0135,
+      "step": 6736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9116109834031386e-06,
+      "loss": 0.7212,
+      "step": 6737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.910527580273369e-06,
+      "loss": 0.873,
+      "step": 6738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.909444132925765e-06,
+      "loss": 0.8877,
+      "step": 6739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90836064141294e-06,
+      "loss": 0.656,
+      "step": 6740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.907277105787513e-06,
+      "loss": 0.8331,
+      "step": 6741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.906193526102104e-06,
+      "loss": 0.81,
+      "step": 6742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.905109902409334e-06,
+      "loss": 1.0735,
+      "step": 6743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.904026234761827e-06,
+      "loss": 0.7943,
+      "step": 6744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.90294252321221e-06,
+      "loss": 0.6847,
+      "step": 6745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.901858767813107e-06,
+      "loss": 0.6119,
+      "step": 6746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.9007749686171534e-06,
+      "loss": 0.5537,
+      "step": 6747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8996911256769796e-06,
+      "loss": 0.5776,
+      "step": 6748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.898607239045219e-06,
+      "loss": 0.8224,
+      "step": 6749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.897523308774511e-06,
+      "loss": 0.6053,
+      "step": 6750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.896439334917492e-06,
+      "loss": 0.8974,
+      "step": 6751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.895355317526803e-06,
+      "loss": 0.7198,
+      "step": 6752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.894271256655089e-06,
+      "loss": 0.5684,
+      "step": 6753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.893187152354992e-06,
+      "loss": 0.831,
+      "step": 6754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.892103004679161e-06,
+      "loss": 0.8255,
+      "step": 6755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.891018813680246e-06,
+      "loss": 0.6288,
+      "step": 6756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.889934579410898e-06,
+      "loss": 0.8196,
+      "step": 6757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8888503019237696e-06,
+      "loss": 0.525,
+      "step": 6758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.887765981271518e-06,
+      "loss": 0.9982,
+      "step": 6759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.886681617506797e-06,
+      "loss": 0.5473,
+      "step": 6760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.885597210682273e-06,
+      "loss": 0.9597,
+      "step": 6761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.884512760850602e-06,
+      "loss": 0.6707,
+      "step": 6762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.883428268064451e-06,
+      "loss": 0.6377,
+      "step": 6763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.882343732376487e-06,
+      "loss": 0.8042,
+      "step": 6764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8812591538393735e-06,
+      "loss": 0.4434,
+      "step": 6765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.880174532505786e-06,
+      "loss": 0.6822,
+      "step": 6766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.879089868428394e-06,
+      "loss": 0.8299,
+      "step": 6767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.878005161659872e-06,
+      "loss": 0.495,
+      "step": 6768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.876920412252899e-06,
+      "loss": 0.5873,
+      "step": 6769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.875835620260148e-06,
+      "loss": 0.8226,
+      "step": 6770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.874750785734305e-06,
+      "loss": 0.8734,
+      "step": 6771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8736659087280525e-06,
+      "loss": 0.6607,
+      "step": 6772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.872580989294072e-06,
+      "loss": 0.6541,
+      "step": 6773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8714960274850506e-06,
+      "loss": 0.6883,
+      "step": 6774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.870411023353681e-06,
+      "loss": 0.6504,
+      "step": 6775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.86932597695265e-06,
+      "loss": 0.9898,
+      "step": 6776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8682408883346535e-06,
+      "loss": 0.7284,
+      "step": 6777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.867155757552383e-06,
+      "loss": 0.995,
+      "step": 6778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.866070584658539e-06,
+      "loss": 0.6269,
+      "step": 6779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.864985369705819e-06,
+      "loss": 0.7387,
+      "step": 6780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.863900112746925e-06,
+      "loss": 1.0291,
+      "step": 6781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.86281481383456e-06,
+      "loss": 0.7204,
+      "step": 6782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.861729473021429e-06,
+      "loss": 0.7514,
+      "step": 6783
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.860644090360237e-06,
+      "loss": 0.9066,
+      "step": 6784
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.859558665903699e-06,
+      "loss": 0.7449,
+      "step": 6785
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.858473199704519e-06,
+      "loss": 0.6398,
+      "step": 6786
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.857387691815416e-06,
+      "loss": 0.6067,
+      "step": 6787
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8563021422891056e-06,
+      "loss": 0.8532,
+      "step": 6788
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8552165511783e-06,
+      "loss": 0.9672,
+      "step": 6789
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.854130918535723e-06,
+      "loss": 0.5906,
+      "step": 6790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.853045244414095e-06,
+      "loss": 0.5346,
+      "step": 6791
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.851959528866139e-06,
+      "loss": 0.7794,
+      "step": 6792
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.850873771944581e-06,
+      "loss": 0.4493,
+      "step": 6793
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.849787973702148e-06,
+      "loss": 0.6248,
+      "step": 6794
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.848702134191571e-06,
+      "loss": 0.4961,
+      "step": 6795
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8476162534655786e-06,
+      "loss": 0.6233,
+      "step": 6796
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.846530331576907e-06,
+      "loss": 0.8204,
+      "step": 6797
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.845444368578288e-06,
+      "loss": 0.6184,
+      "step": 6798
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.844358364522466e-06,
+      "loss": 0.6344,
+      "step": 6799
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.843272319462173e-06,
+      "loss": 0.6237,
+      "step": 6800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.842186233450153e-06,
+      "loss": 0.7465,
+      "step": 6801
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.841100106539153e-06,
+      "loss": 0.8341,
+      "step": 6802
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.840013938781913e-06,
+      "loss": 0.7686,
+      "step": 6803
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.838927730231183e-06,
+      "loss": 0.9448,
+      "step": 6804
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8378414809397135e-06,
+      "loss": 0.8117,
+      "step": 6805
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.836755190960255e-06,
+      "loss": 0.5811,
+      "step": 6806
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.835668860345561e-06,
+      "loss": 0.6352,
+      "step": 6807
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8345824891483836e-06,
+      "loss": 0.8717,
+      "step": 6808
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.833496077421485e-06,
+      "loss": 0.6743,
+      "step": 6809
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.832409625217623e-06,
+      "loss": 0.7254,
+      "step": 6810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.831323132589557e-06,
+      "loss": 0.8706,
+      "step": 6811
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.83023659959005e-06,
+      "loss": 0.8761,
+      "step": 6812
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.829150026271871e-06,
+      "loss": 0.814,
+      "step": 6813
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.828063412687785e-06,
+      "loss": 0.8293,
+      "step": 6814
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.826976758890559e-06,
+      "loss": 0.9737,
+      "step": 6815
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.825890064932968e-06,
+      "loss": 0.6585,
+      "step": 6816
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.824803330867781e-06,
+      "loss": 0.6708,
+      "step": 6817
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.823716556747775e-06,
+      "loss": 0.9782,
+      "step": 6818
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.822629742625726e-06,
+      "loss": 0.7171,
+      "step": 6819
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.821542888554414e-06,
+      "loss": 0.7838,
+      "step": 6820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.820455994586621e-06,
+      "loss": 0.7224,
+      "step": 6821
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8193690607751255e-06,
+      "loss": 0.7637,
+      "step": 6822
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8182820871727166e-06,
+      "loss": 0.6584,
+      "step": 6823
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.817195073832177e-06,
+      "loss": 0.7586,
+      "step": 6824
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.816108020806297e-06,
+      "loss": 0.8776,
+      "step": 6825
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.815020928147867e-06,
+      "loss": 0.7165,
+      "step": 6826
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.813933795909681e-06,
+      "loss": 0.9735,
+      "step": 6827
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.81284662414453e-06,
+      "loss": 0.7934,
+      "step": 6828
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.811759412905213e-06,
+      "loss": 0.8922,
+      "step": 6829
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.8106721622445254e-06,
+      "loss": 0.9087,
+      "step": 6830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.80958487221527e-06,
+      "loss": 0.7724,
+      "step": 6831
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.808497542870246e-06,
+      "loss": 0.9932,
+      "step": 6832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.807410174262259e-06,
+      "loss": 0.8401,
+      "step": 6833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.806322766444116e-06,
+      "loss": 0.8624,
+      "step": 6834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.805235319468622e-06,
+      "loss": 0.8928,
+      "step": 6835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.804147833388586e-06,
+      "loss": 0.629,
+      "step": 6836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.803060308256824e-06,
+      "loss": 1.0357,
+      "step": 6837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.801972744126144e-06,
+      "loss": 0.982,
+      "step": 6838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.800885141049364e-06,
+      "loss": 0.7383,
+      "step": 6839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.799797499079301e-06,
+      "loss": 0.7459,
+      "step": 6840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.798709818268775e-06,
+      "loss": 0.7111,
+      "step": 6841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.797622098670604e-06,
+      "loss": 0.6607,
+      "step": 6842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.796534340337614e-06,
+      "loss": 0.7406,
+      "step": 6843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.795446543322627e-06,
+      "loss": 0.8277,
+      "step": 6844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.794358707678471e-06,
+      "loss": 0.7122,
+      "step": 6845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.793270833457973e-06,
+      "loss": 0.9294,
+      "step": 6846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.792182920713966e-06,
+      "loss": 0.7857,
+      "step": 6847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.791094969499281e-06,
+      "loss": 0.655,
+      "step": 6848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.79000697986675e-06,
+      "loss": 0.8651,
+      "step": 6849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.788918951869211e-06,
+      "loss": 0.8665,
+      "step": 6850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7878308855595025e-06,
+      "loss": 0.8144,
+      "step": 6851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.786742780990462e-06,
+      "loss": 0.6801,
+      "step": 6852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7856546382149316e-06,
+      "loss": 0.9429,
+      "step": 6853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7845664572857544e-06,
+      "loss": 0.7751,
+      "step": 6854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.783478238255777e-06,
+      "loss": 0.5906,
+      "step": 6855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.782389981177845e-06,
+      "loss": 0.5401,
+      "step": 6856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.781301686104808e-06,
+      "loss": 0.7847,
+      "step": 6857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.780213353089517e-06,
+      "loss": 0.8742,
+      "step": 6858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.779124982184825e-06,
+      "loss": 0.6399,
+      "step": 6859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.778036573443583e-06,
+      "loss": 0.6921,
+      "step": 6860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.776948126918652e-06,
+      "loss": 0.641,
+      "step": 6861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.775859642662888e-06,
+      "loss": 0.7742,
+      "step": 6862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.774771120729149e-06,
+      "loss": 0.7241,
+      "step": 6863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.773682561170301e-06,
+      "loss": 0.9614,
+      "step": 6864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.772593964039203e-06,
+      "loss": 1.0333,
+      "step": 6865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.771505329388724e-06,
+      "loss": 0.7488,
+      "step": 6866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7704166572717295e-06,
+      "loss": 0.8874,
+      "step": 6867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7693279477410876e-06,
+      "loss": 0.9097,
+      "step": 6868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.768239200849672e-06,
+      "loss": 0.8245,
+      "step": 6869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.767150416650354e-06,
+      "loss": 0.7702,
+      "step": 6870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.766061595196006e-06,
+      "loss": 0.7987,
+      "step": 6871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.764972736539508e-06,
+      "loss": 0.7819,
+      "step": 6872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.763883840733736e-06,
+      "loss": 0.8096,
+      "step": 6873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.762794907831568e-06,
+      "loss": 0.6523,
+      "step": 6874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.761705937885891e-06,
+      "loss": 0.701,
+      "step": 6875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.760616930949584e-06,
+      "loss": 0.7708,
+      "step": 6876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.759527887075533e-06,
+      "loss": 0.8081,
+      "step": 6877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.758438806316628e-06,
+      "loss": 0.6256,
+      "step": 6878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.757349688725754e-06,
+      "loss": 0.7537,
+      "step": 6879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.756260534355804e-06,
+      "loss": 0.625,
+      "step": 6880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.755171343259669e-06,
+      "loss": 0.5608,
+      "step": 6881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.754082115490245e-06,
+      "loss": 0.6137,
+      "step": 6882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.752992851100427e-06,
+      "loss": 0.6171,
+      "step": 6883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.751903550143113e-06,
+      "loss": 0.7921,
+      "step": 6884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.750814212671202e-06,
+      "loss": 0.8106,
+      "step": 6885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.749724838737597e-06,
+      "loss": 0.8104,
+      "step": 6886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7486354283952e-06,
+      "loss": 0.7103,
+      "step": 6887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7475459816969136e-06,
+      "loss": 0.7631,
+      "step": 6888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.746456498695648e-06,
+      "loss": 0.5347,
+      "step": 6889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.745366979444311e-06,
+      "loss": 0.7745,
+      "step": 6890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.744277423995812e-06,
+      "loss": 0.7631,
+      "step": 6891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.743187832403063e-06,
+      "loss": 1.0933,
+      "step": 6892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.742098204718978e-06,
+      "loss": 0.9006,
+      "step": 6893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.741008540996472e-06,
+      "loss": 0.5044,
+      "step": 6894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.739918841288462e-06,
+      "loss": 0.5775,
+      "step": 6895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.738829105647868e-06,
+      "loss": 0.6715,
+      "step": 6896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.737739334127611e-06,
+      "loss": 0.5301,
+      "step": 6897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.736649526780611e-06,
+      "loss": 0.5935,
+      "step": 6898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.735559683659795e-06,
+      "loss": 0.5968,
+      "step": 6899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7344698048180865e-06,
+      "loss": 0.8358,
+      "step": 6900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7333798903084146e-06,
+      "loss": 0.587,
+      "step": 6901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.732289940183708e-06,
+      "loss": 0.6963,
+      "step": 6902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.731199954496899e-06,
+      "loss": 0.6089,
+      "step": 6903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.730109933300919e-06,
+      "loss": 0.6802,
+      "step": 6904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.729019876648704e-06,
+      "loss": 0.767,
+      "step": 6905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.727929784593188e-06,
+      "loss": 0.827,
+      "step": 6906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.726839657187313e-06,
+      "loss": 0.4184,
+      "step": 6907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.725749494484013e-06,
+      "loss": 0.9593,
+      "step": 6908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.724659296536234e-06,
+      "loss": 0.4991,
+      "step": 6909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.723569063396919e-06,
+      "loss": 0.5292,
+      "step": 6910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.72247879511901e-06,
+      "loss": 0.6724,
+      "step": 6911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.721388491755455e-06,
+      "loss": 0.765,
+      "step": 6912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.720298153359204e-06,
+      "loss": 0.6744,
+      "step": 6913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.719207779983203e-06,
+      "loss": 0.7606,
+      "step": 6914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.718117371680408e-06,
+      "loss": 0.8245,
+      "step": 6915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.71702692850377e-06,
+      "loss": 0.5704,
+      "step": 6916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.715936450506244e-06,
+      "loss": 0.704,
+      "step": 6917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.714845937740787e-06,
+      "loss": 1.0226,
+      "step": 6918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.713755390260358e-06,
+      "loss": 0.6239,
+      "step": 6919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.712664808117918e-06,
+      "loss": 0.6316,
+      "step": 6920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.711574191366427e-06,
+      "loss": 0.5512,
+      "step": 6921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.710483540058847e-06,
+      "loss": 0.7469,
+      "step": 6922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.709392854248147e-06,
+      "loss": 0.785,
+      "step": 6923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.708302133987293e-06,
+      "loss": 0.8893,
+      "step": 6924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.707211379329252e-06,
+      "loss": 0.7887,
+      "step": 6925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7061205903269935e-06,
+      "loss": 0.6261,
+      "step": 6926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.7050297670334905e-06,
+      "loss": 0.8318,
+      "step": 6927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.703938909501717e-06,
+      "loss": 0.777,
+      "step": 6928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.702848017784648e-06,
+      "loss": 0.5144,
+      "step": 6929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.70175709193526e-06,
+      "loss": 0.5462,
+      "step": 6930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.700666132006531e-06,
+      "loss": 0.5842,
+      "step": 6931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.699575138051443e-06,
+      "loss": 0.8528,
+      "step": 6932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.698484110122976e-06,
+      "loss": 0.5926,
+      "step": 6933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.697393048274114e-06,
+      "loss": 0.5696,
+      "step": 6934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6963019525578435e-06,
+      "loss": 0.844,
+      "step": 6935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.695210823027147e-06,
+      "loss": 0.6371,
+      "step": 6936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.694119659735018e-06,
+      "loss": 0.601,
+      "step": 6937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.693028462734443e-06,
+      "loss": 0.7043,
+      "step": 6938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.691937232078415e-06,
+      "loss": 0.7817,
+      "step": 6939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.690845967819928e-06,
+      "loss": 0.4241,
+      "step": 6940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.689754670011976e-06,
+      "loss": 0.801,
+      "step": 6941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.688663338707554e-06,
+      "loss": 0.9896,
+      "step": 6942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.687571973959663e-06,
+      "loss": 0.7054,
+      "step": 6943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.686480575821303e-06,
+      "loss": 0.7589,
+      "step": 6944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.685389144345474e-06,
+      "loss": 0.6694,
+      "step": 6945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.684297679585176e-06,
+      "loss": 0.6527,
+      "step": 6946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.683206181593419e-06,
+      "loss": 0.7112,
+      "step": 6947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.682114650423208e-06,
+      "loss": 0.7388,
+      "step": 6948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.681023086127548e-06,
+      "loss": 0.5149,
+      "step": 6949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.679931488759452e-06,
+      "loss": 0.4608,
+      "step": 6950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.67883985837193e-06,
+      "loss": 0.8257,
+      "step": 6951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.677748195017992e-06,
+      "loss": 1.0014,
+      "step": 6952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6766564987506564e-06,
+      "loss": 0.9455,
+      "step": 6953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.675564769622937e-06,
+      "loss": 1.0015,
+      "step": 6954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6744730076878505e-06,
+      "loss": 0.735,
+      "step": 6955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.673381212998419e-06,
+      "loss": 0.6009,
+      "step": 6956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6722893856076596e-06,
+      "loss": 0.8783,
+      "step": 6957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6711975255685965e-06,
+      "loss": 0.4944,
+      "step": 6958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6701056329342556e-06,
+      "loss": 0.6632,
+      "step": 6959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.669013707757657e-06,
+      "loss": 0.7525,
+      "step": 6960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.667921750091833e-06,
+      "loss": 0.7091,
+      "step": 6961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.666829759989809e-06,
+      "loss": 0.5408,
+      "step": 6962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.665737737504616e-06,
+      "loss": 0.7082,
+      "step": 6963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.664645682689287e-06,
+      "loss": 0.5134,
+      "step": 6964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.663553595596853e-06,
+      "loss": 0.9246,
+      "step": 6965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.662461476280351e-06,
+      "loss": 0.7173,
+      "step": 6966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.661369324792816e-06,
+      "loss": 0.7316,
+      "step": 6967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.660277141187287e-06,
+      "loss": 0.6046,
+      "step": 6968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.659184925516802e-06,
+      "loss": 0.797,
+      "step": 6969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.658092677834405e-06,
+      "loss": 0.7313,
+      "step": 6970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.6570003981931355e-06,
+      "loss": 0.7103,
+      "step": 6971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.655908086646042e-06,
+      "loss": 0.5769,
+      "step": 6972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.654815743246164e-06,
+      "loss": 0.7592,
+      "step": 6973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.653723368046553e-06,
+      "loss": 0.7238,
+      "step": 6974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.65263096110026e-06,
+      "loss": 0.9017,
+      "step": 6975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.651538522460329e-06,
+      "loss": 0.9237,
+      "step": 6976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.650446052179817e-06,
+      "loss": 0.8351,
+      "step": 6977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.649353550311777e-06,
+      "loss": 0.6844,
+      "step": 6978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.648261016909263e-06,
+      "loss": 0.8063,
+      "step": 6979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.64716845202533e-06,
+      "loss": 0.7933,
+      "step": 6980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.646075855713039e-06,
+      "loss": 0.9433,
+      "step": 6981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.644983228025449e-06,
+      "loss": 0.6269,
+      "step": 6982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6438905690156196e-06,
+      "loss": 0.6196,
+      "step": 6983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.642797878736614e-06,
+      "loss": 0.7126,
+      "step": 6984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.641705157241497e-06,
+      "loss": 0.6875,
+      "step": 6985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.640612404583337e-06,
+      "loss": 0.9724,
+      "step": 6986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.639519620815195e-06,
+      "loss": 0.6423,
+      "step": 6987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.638426805990142e-06,
+      "loss": 0.8925,
+      "step": 6988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.637333960161252e-06,
+      "loss": 0.8543,
+      "step": 6989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.636241083381593e-06,
+      "loss": 0.8088,
+      "step": 6990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.635148175704239e-06,
+      "loss": 0.6942,
+      "step": 6991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6340552371822655e-06,
+      "loss": 0.4835,
+      "step": 6992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.632962267868747e-06,
+      "loss": 0.9645,
+      "step": 6993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.631869267816763e-06,
+      "loss": 0.7165,
+      "step": 6994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.630776237079392e-06,
+      "loss": 0.8404,
+      "step": 6995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.629683175709715e-06,
+      "loss": 0.5412,
+      "step": 6996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.628590083760815e-06,
+      "loss": 0.9064,
+      "step": 6997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.627496961285772e-06,
+      "loss": 0.6511,
+      "step": 6998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.626403808337677e-06,
+      "loss": 0.4439,
+      "step": 6999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.625310624969612e-06,
+      "loss": 0.8065,
+      "step": 7000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.624217411234667e-06,
+      "loss": 0.7651,
+      "step": 7001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.62312416718593e-06,
+      "loss": 0.6359,
+      "step": 7002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.622030892876494e-06,
+      "loss": 0.7672,
+      "step": 7003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.620937588359452e-06,
+      "loss": 0.7516,
+      "step": 7004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.619844253687896e-06,
+      "loss": 0.9442,
+      "step": 7005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.618750888914923e-06,
+      "loss": 0.5875,
+      "step": 7006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.617657494093628e-06,
+      "loss": 0.8435,
+      "step": 7007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.616564069277111e-06,
+      "loss": 0.7685,
+      "step": 7008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.615470614518472e-06,
+      "loss": 0.7461,
+      "step": 7009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.614377129870813e-06,
+      "loss": 0.6718,
+      "step": 7010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6132836153872335e-06,
+      "loss": 0.8406,
+      "step": 7011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.61219007112084e-06,
+      "loss": 0.537,
+      "step": 7012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.61109649712474e-06,
+      "loss": 0.7124,
+      "step": 7013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.610002893452036e-06,
+      "loss": 0.7072,
+      "step": 7014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6089092601558396e-06,
+      "loss": 0.7845,
+      "step": 7015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.607815597289261e-06,
+      "loss": 0.874,
+      "step": 7016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.60672190490541e-06,
+      "loss": 0.8696,
+      "step": 7017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.605628183057401e-06,
+      "loss": 0.7365,
+      "step": 7018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6045344317983476e-06,
+      "loss": 0.6807,
+      "step": 7019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.603440651181366e-06,
+      "loss": 0.8569,
+      "step": 7020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.602346841259571e-06,
+      "loss": 0.8143,
+      "step": 7021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.601253002086084e-06,
+      "loss": 0.61,
+      "step": 7022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.6001591337140234e-06,
+      "loss": 0.9149,
+      "step": 7023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.599065236196513e-06,
+      "loss": 0.7726,
+      "step": 7024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.597971309586672e-06,
+      "loss": 0.808,
+      "step": 7025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5968773539376266e-06,
+      "loss": 0.8383,
+      "step": 7026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.595783369302503e-06,
+      "loss": 0.855,
+      "step": 7027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.594689355734426e-06,
+      "loss": 0.6595,
+      "step": 7028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.593595313286526e-06,
+      "loss": 0.9332,
+      "step": 7029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.592501242011933e-06,
+      "loss": 0.6927,
+      "step": 7030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.591407141963777e-06,
+      "loss": 0.5899,
+      "step": 7031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.590313013195191e-06,
+      "loss": 0.7695,
+      "step": 7032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.58921885575931e-06,
+      "loss": 0.7678,
+      "step": 7033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.588124669709268e-06,
+      "loss": 0.6307,
+      "step": 7034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.587030455098203e-06,
+      "loss": 0.8225,
+      "step": 7035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.585936211979252e-06,
+      "loss": 0.8453,
+      "step": 7036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5848419404055555e-06,
+      "loss": 0.6746,
+      "step": 7037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.583747640430255e-06,
+      "loss": 0.7301,
+      "step": 7038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.582653312106491e-06,
+      "loss": 0.909,
+      "step": 7039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5815589554874085e-06,
+      "loss": 1.05,
+      "step": 7040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5804645706261515e-06,
+      "loss": 0.6198,
+      "step": 7041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5793701575758684e-06,
+      "loss": 0.6823,
+      "step": 7042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.578275716389706e-06,
+      "loss": 0.7461,
+      "step": 7043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.577181247120814e-06,
+      "loss": 0.6303,
+      "step": 7044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.576086749822343e-06,
+      "loss": 0.6769,
+      "step": 7045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.574992224547442e-06,
+      "loss": 0.6406,
+      "step": 7046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.573897671349269e-06,
+      "loss": 0.6642,
+      "step": 7047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.572803090280976e-06,
+      "loss": 0.5763,
+      "step": 7048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.571708481395719e-06,
+      "loss": 0.8555,
+      "step": 7049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.570613844746654e-06,
+      "loss": 0.9465,
+      "step": 7050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5695191803869445e-06,
+      "loss": 0.7595,
+      "step": 7051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.568424488369746e-06,
+      "loss": 0.6868,
+      "step": 7052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.567329768748221e-06,
+      "loss": 0.7313,
+      "step": 7053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5662350215755335e-06,
+      "loss": 0.4358,
+      "step": 7054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5651402469048455e-06,
+      "loss": 0.7912,
+      "step": 7055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.564045444789324e-06,
+      "loss": 0.8426,
+      "step": 7056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.562950615282136e-06,
+      "loss": 0.7905,
+      "step": 7057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.561855758436449e-06,
+      "loss": 0.9215,
+      "step": 7058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.560760874305432e-06,
+      "loss": 0.8742,
+      "step": 7059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.559665962942254e-06,
+      "loss": 0.6145,
+      "step": 7060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.558571024400091e-06,
+      "loss": 0.7995,
+      "step": 7061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.557476058732115e-06,
+      "loss": 0.5438,
+      "step": 7062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5563810659915e-06,
+      "loss": 0.7513,
+      "step": 7063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.555286046231421e-06,
+      "loss": 0.5401,
+      "step": 7064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5541909995050554e-06,
+      "loss": 0.9673,
+      "step": 7065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.553095925865584e-06,
+      "loss": 0.7721,
+      "step": 7066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.552000825366186e-06,
+      "loss": 0.6092,
+      "step": 7067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.550905698060041e-06,
+      "loss": 0.8777,
+      "step": 7068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.549810544000334e-06,
+      "loss": 0.9848,
+      "step": 7069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5487153632402455e-06,
+      "loss": 0.8896,
+      "step": 7070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.547620155832964e-06,
+      "loss": 0.7078,
+      "step": 7071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.546524921831674e-06,
+      "loss": 0.7907,
+      "step": 7072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.545429661289565e-06,
+      "loss": 1.0871,
+      "step": 7073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.544334374259823e-06,
+      "loss": 0.8225,
+      "step": 7074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.543239060795642e-06,
+      "loss": 0.894,
+      "step": 7075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.54214372095021e-06,
+      "loss": 0.6557,
+      "step": 7076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.541048354776721e-06,
+      "loss": 0.7479,
+      "step": 7077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.539952962328371e-06,
+      "loss": 0.8064,
+      "step": 7078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.538857543658353e-06,
+      "loss": 0.7153,
+      "step": 7079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.537762098819865e-06,
+      "loss": 0.7334,
+      "step": 7080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.536666627866104e-06,
+      "loss": 0.5751,
+      "step": 7081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.535571130850271e-06,
+      "loss": 0.9908,
+      "step": 7082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.534475607825566e-06,
+      "loss": 0.7918,
+      "step": 7083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.533380058845187e-06,
+      "loss": 0.8469,
+      "step": 7084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.532284483962341e-06,
+      "loss": 0.8783,
+      "step": 7085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.531188883230234e-06,
+      "loss": 0.6788,
+      "step": 7086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.530093256702066e-06,
+      "loss": 0.8738,
+      "step": 7087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.528997604431047e-06,
+      "loss": 0.9031,
+      "step": 7088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.527901926470386e-06,
+      "loss": 0.8492,
+      "step": 7089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5268062228732885e-06,
+      "loss": 0.7996,
+      "step": 7090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.52571049369297e-06,
+      "loss": 0.8877,
+      "step": 7091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.524614738982637e-06,
+      "loss": 0.8963,
+      "step": 7092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.523518958795507e-06,
+      "loss": 0.8012,
+      "step": 7093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.522423153184792e-06,
+      "loss": 0.8044,
+      "step": 7094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.521327322203708e-06,
+      "loss": 0.7827,
+      "step": 7095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.520231465905471e-06,
+      "loss": 0.861,
+      "step": 7096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.519135584343301e-06,
+      "loss": 0.6209,
+      "step": 7097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.518039677570413e-06,
+      "loss": 0.4939,
+      "step": 7098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.516943745640032e-06,
+      "loss": 0.7094,
+      "step": 7099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.515847788605379e-06,
+      "loss": 0.7859,
+      "step": 7100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.514751806519673e-06,
+      "loss": 0.7072,
+      "step": 7101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.513655799436142e-06,
+      "loss": 0.7547,
+      "step": 7102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.512559767408008e-06,
+      "loss": 0.8804,
+      "step": 7103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5114637104885e-06,
+      "loss": 0.8271,
+      "step": 7104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.510367628730845e-06,
+      "loss": 0.6822,
+      "step": 7105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.509271522188271e-06,
+      "loss": 0.7538,
+      "step": 7106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50817539091401e-06,
+      "loss": 0.7843,
+      "step": 7107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.507079234961291e-06,
+      "loss": 0.7119,
+      "step": 7108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5059830543833485e-06,
+      "loss": 0.8353,
+      "step": 7109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5048868492334165e-06,
+      "loss": 0.6479,
+      "step": 7110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5037906195647276e-06,
+      "loss": 0.5906,
+      "step": 7111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.502694365430517e-06,
+      "loss": 0.5908,
+      "step": 7112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.5015980868840254e-06,
+      "loss": 0.6345,
+      "step": 7113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.50050178397849e-06,
+      "loss": 0.7846,
+      "step": 7114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.49940545676715e-06,
+      "loss": 0.6244,
+      "step": 7115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.498309105303245e-06,
+      "loss": 0.6001,
+      "step": 7116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.49721272964002e-06,
+      "loss": 0.6855,
+      "step": 7117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.496116329830717e-06,
+      "loss": 0.7029,
+      "step": 7118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.495019905928578e-06,
+      "loss": 0.776,
+      "step": 7119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.493923457986851e-06,
+      "loss": 0.7911,
+      "step": 7120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.492826986058782e-06,
+      "loss": 0.5042,
+      "step": 7121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.491730490197618e-06,
+      "loss": 0.7427,
+      "step": 7122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.490633970456609e-06,
+      "loss": 0.6442,
+      "step": 7123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4895374268890065e-06,
+      "loss": 1.1152,
+      "step": 7124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.488440859548059e-06,
+      "loss": 0.6343,
+      "step": 7125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4873442684870186e-06,
+      "loss": 0.7737,
+      "step": 7126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.486247653759142e-06,
+      "loss": 0.8076,
+      "step": 7127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.485151015417683e-06,
+      "loss": 0.4904,
+      "step": 7128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.484054353515896e-06,
+      "loss": 0.8037,
+      "step": 7129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.48295766810704e-06,
+      "loss": 0.6169,
+      "step": 7130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.48186095924437e-06,
+      "loss": 0.8119,
+      "step": 7131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.48076422698115e-06,
+      "loss": 0.5632,
+      "step": 7132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.479667471370637e-06,
+      "loss": 0.9102,
+      "step": 7133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.478570692466093e-06,
+      "loss": 0.6148,
+      "step": 7134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.477473890320784e-06,
+      "loss": 0.7189,
+      "step": 7135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.476377064987968e-06,
+      "loss": 0.6993,
+      "step": 7136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.475280216520913e-06,
+      "loss": 0.7021,
+      "step": 7137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.474183344972887e-06,
+      "loss": 0.6399,
+      "step": 7138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.473086450397156e-06,
+      "loss": 0.5604,
+      "step": 7139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.471989532846987e-06,
+      "loss": 0.8414,
+      "step": 7140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4708925923756505e-06,
+      "loss": 0.8443,
+      "step": 7141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.469795629036416e-06,
+      "loss": 0.628,
+      "step": 7142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.468698642882557e-06,
+      "loss": 0.733,
+      "step": 7143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.467601633967346e-06,
+      "loss": 0.7753,
+      "step": 7144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.466504602344055e-06,
+      "loss": 0.7402,
+      "step": 7145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.465407548065961e-06,
+      "loss": 0.8597,
+      "step": 7146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4643104711863395e-06,
+      "loss": 0.8382,
+      "step": 7147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.463213371758468e-06,
+      "loss": 0.8566,
+      "step": 7148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.462116249835624e-06,
+      "loss": 0.7496,
+      "step": 7149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.461019105471086e-06,
+      "loss": 0.7949,
+      "step": 7150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.459921938718138e-06,
+      "loss": 0.7646,
+      "step": 7151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.458824749630059e-06,
+      "loss": 0.7351,
+      "step": 7152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4577275382601315e-06,
+      "loss": 0.7094,
+      "step": 7153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.45663030466164e-06,
+      "loss": 0.7059,
+      "step": 7154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.455533048887868e-06,
+      "loss": 0.6865,
+      "step": 7155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.454435770992104e-06,
+      "loss": 0.8081,
+      "step": 7156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.453338471027633e-06,
+      "loss": 0.607,
+      "step": 7157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.452241149047742e-06,
+      "loss": 0.6819,
+      "step": 7158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.451143805105724e-06,
+      "loss": 0.9042,
+      "step": 7159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4500464392548645e-06,
+      "loss": 0.8338,
+      "step": 7160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.448949051548459e-06,
+      "loss": 0.75,
+      "step": 7161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4478516420397975e-06,
+      "loss": 1.0632,
+      "step": 7162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.446754210782173e-06,
+      "loss": 0.6613,
+      "step": 7163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.44565675782888e-06,
+      "loss": 0.8744,
+      "step": 7164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.444559283233216e-06,
+      "loss": 0.4653,
+      "step": 7165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.443461787048475e-06,
+      "loss": 0.6771,
+      "step": 7166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4423642693279556e-06,
+      "loss": 0.4915,
+      "step": 7167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.441266730124957e-06,
+      "loss": 1.0827,
+      "step": 7168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.440169169492779e-06,
+      "loss": 1.0016,
+      "step": 7169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.43907158748472e-06,
+      "loss": 0.7939,
+      "step": 7170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.437973984154084e-06,
+      "loss": 0.9624,
+      "step": 7171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.436876359554173e-06,
+      "loss": 0.9469,
+      "step": 7172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.435778713738292e-06,
+      "loss": 0.8461,
+      "step": 7173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4346810467597435e-06,
+      "loss": 0.9806,
+      "step": 7174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.433583358671836e-06,
+      "loss": 0.7238,
+      "step": 7175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4324856495278735e-06,
+      "loss": 0.743,
+      "step": 7176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.431387919381166e-06,
+      "loss": 0.7743,
+      "step": 7177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.430290168285022e-06,
+      "loss": 0.806,
+      "step": 7178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.42919239629275e-06,
+      "loss": 0.6227,
+      "step": 7179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4280946034576635e-06,
+      "loss": 0.834,
+      "step": 7180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.426996789833072e-06,
+      "loss": 0.5312,
+      "step": 7181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.425898955472291e-06,
+      "loss": 0.7407,
+      "step": 7182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.424801100428633e-06,
+      "loss": 0.5689,
+      "step": 7183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.423703224755414e-06,
+      "loss": 0.8461,
+      "step": 7184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4226053285059466e-06,
+      "loss": 0.6077,
+      "step": 7185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.421507411733553e-06,
+      "loss": 0.7392,
+      "step": 7186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.420409474491548e-06,
+      "loss": 0.8056,
+      "step": 7187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.41931151683325e-06,
+      "loss": 0.6661,
+      "step": 7188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.418213538811982e-06,
+      "loss": 0.7301,
+      "step": 7189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.417115540481062e-06,
+      "loss": 0.6692,
+      "step": 7190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.416017521893813e-06,
+      "loss": 0.6561,
+      "step": 7191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.414919483103557e-06,
+      "loss": 0.7357,
+      "step": 7192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.41382142416362e-06,
+      "loss": 0.7294,
+      "step": 7193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.412723345127325e-06,
+      "loss": 0.758,
+      "step": 7194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.411625246048e-06,
+      "loss": 0.6169,
+      "step": 7195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.41052712697897e-06,
+      "loss": 0.5509,
+      "step": 7196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.409428987973564e-06,
+      "loss": 0.7271,
+      "step": 7197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4083308290851085e-06,
+      "loss": 0.8595,
+      "step": 7198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.407232650366936e-06,
+      "loss": 0.7933,
+      "step": 7199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4061344518723765e-06,
+      "loss": 0.8138,
+      "step": 7200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.40503623365476e-06,
+      "loss": 0.6985,
+      "step": 7201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.403937995767421e-06,
+      "loss": 0.7763,
+      "step": 7202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.402839738263693e-06,
+      "loss": 0.5654,
+      "step": 7203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.40174146119691e-06,
+      "loss": 0.4192,
+      "step": 7204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.4006431646204075e-06,
+      "loss": 0.7882,
+      "step": 7205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3995448485875205e-06,
+      "loss": 0.5784,
+      "step": 7206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.398446513151589e-06,
+      "loss": 0.7678,
+      "step": 7207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.397348158365949e-06,
+      "loss": 0.5742,
+      "step": 7208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.396249784283943e-06,
+      "loss": 1.1703,
+      "step": 7209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.395151390958908e-06,
+      "loss": 0.6696,
+      "step": 7210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.394052978444187e-06,
+      "loss": 0.5561,
+      "step": 7211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.39295454679312e-06,
+      "loss": 0.7087,
+      "step": 7212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3918560960590526e-06,
+      "loss": 0.7363,
+      "step": 7213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.390757626295327e-06,
+      "loss": 0.9712,
+      "step": 7214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.389659137555287e-06,
+      "loss": 0.7078,
+      "step": 7215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.388560629892282e-06,
+      "loss": 0.7826,
+      "step": 7216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.387462103359655e-06,
+      "loss": 0.734,
+      "step": 7217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.386363558010756e-06,
+      "loss": 0.6357,
+      "step": 7218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3852649938989335e-06,
+      "loss": 0.7846,
+      "step": 7219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3841664110775334e-06,
+      "loss": 0.6584,
+      "step": 7220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3830678095999125e-06,
+      "loss": 0.8496,
+      "step": 7221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.381969189519415e-06,
+      "loss": 0.581,
+      "step": 7222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.380870550889398e-06,
+      "loss": 0.6771,
+      "step": 7223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.379771893763212e-06,
+      "loss": 0.6611,
+      "step": 7224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3786732181942135e-06,
+      "loss": 0.7257,
+      "step": 7225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3775745242357525e-06,
+      "loss": 1.007,
+      "step": 7226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3764758119411905e-06,
+      "loss": 0.7649,
+      "step": 7227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3753770813638815e-06,
+      "loss": 0.794,
+      "step": 7228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.374278332557183e-06,
+      "loss": 0.724,
+      "step": 7229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.373179565574453e-06,
+      "loss": 0.6055,
+      "step": 7230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.37208078046905e-06,
+      "loss": 0.7275,
+      "step": 7231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.370981977294336e-06,
+      "loss": 0.6894,
+      "step": 7232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.369883156103672e-06,
+      "loss": 0.7184,
+      "step": 7233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.36878431695042e-06,
+      "loss": 0.719,
+      "step": 7234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.367685459887941e-06,
+      "loss": 0.7555,
+      "step": 7235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3665865849695995e-06,
+      "loss": 0.6083,
+      "step": 7236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.365487692248761e-06,
+      "loss": 0.6888,
+      "step": 7237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.364388781778791e-06,
+      "loss": 0.8837,
+      "step": 7238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.363289853613054e-06,
+      "loss": 1.0,
+      "step": 7239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.362190907804919e-06,
+      "loss": 0.9777,
+      "step": 7240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.361091944407751e-06,
+      "loss": 0.8037,
+      "step": 7241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3599929634749215e-06,
+      "loss": 0.6306,
+      "step": 7242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3588939650598e-06,
+      "loss": 0.797,
+      "step": 7243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.357794949215756e-06,
+      "loss": 0.6103,
+      "step": 7244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.356695915996162e-06,
+      "loss": 0.6575,
+      "step": 7245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.35559686545439e-06,
+      "loss": 0.7246,
+      "step": 7246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.354497797643812e-06,
+      "loss": 1.0417,
+      "step": 7247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.353398712617802e-06,
+      "loss": 0.5668,
+      "step": 7248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.352299610429737e-06,
+      "loss": 0.7654,
+      "step": 7249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.351200491132988e-06,
+      "loss": 0.7077,
+      "step": 7250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3501013547809376e-06,
+      "loss": 0.8143,
+      "step": 7251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.349002201426957e-06,
+      "loss": 0.6909,
+      "step": 7252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.347903031124428e-06,
+      "loss": 0.8384,
+      "step": 7253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.346803843926728e-06,
+      "loss": 0.8078,
+      "step": 7254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.345704639887237e-06,
+      "loss": 0.5397,
+      "step": 7255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.344605419059336e-06,
+      "loss": 0.7115,
+      "step": 7256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.343506181496405e-06,
+      "loss": 0.6996,
+      "step": 7257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.342406927251827e-06,
+      "loss": 0.7846,
+      "step": 7258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.341307656378988e-06,
+      "loss": 0.7392,
+      "step": 7259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3402083689312644e-06,
+      "loss": 0.7437,
+      "step": 7260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.339109064962047e-06,
+      "loss": 0.7358,
+      "step": 7261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.33800974452472e-06,
+      "loss": 0.6628,
+      "step": 7262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.336910407672668e-06,
+      "loss": 0.6928,
+      "step": 7263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.335811054459277e-06,
+      "loss": 0.9761,
+      "step": 7264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.334711684937939e-06,
+      "loss": 0.8011,
+      "step": 7265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.333612299162039e-06,
+      "loss": 0.81,
+      "step": 7266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.332512897184967e-06,
+      "loss": 0.8454,
+      "step": 7267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3314134790601145e-06,
+      "loss": 0.8352,
+      "step": 7268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.330314044840871e-06,
+      "loss": 0.6424,
+      "step": 7269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3292145945806275e-06,
+      "loss": 0.552,
+      "step": 7270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.328115128332778e-06,
+      "loss": 0.911,
+      "step": 7271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.327015646150716e-06,
+      "loss": 0.6512,
+      "step": 7272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3259161480878354e-06,
+      "loss": 0.5867,
+      "step": 7273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.3248166341975275e-06,
+      "loss": 0.7302,
+      "step": 7274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.323717104533193e-06,
+      "loss": 0.926,
+      "step": 7275
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.322617559148225e-06,
+      "loss": 0.7736,
+      "step": 7276
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.321517998096023e-06,
+      "loss": 0.7877,
+      "step": 7277
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.320418421429982e-06,
+      "loss": 0.633,
+      "step": 7278
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.319318829203501e-06,
+      "loss": 0.7623,
+      "step": 7279
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.3182192214699825e-06,
+      "loss": 0.7011,
+      "step": 7280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.317119598282823e-06,
+      "loss": 0.5809,
+      "step": 7281
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.316019959695426e-06,
+      "loss": 0.7597,
+      "step": 7282
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.314920305761191e-06,
+      "loss": 0.797,
+      "step": 7283
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.313820636533521e-06,
+      "loss": 0.7725,
+      "step": 7284
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.312720952065819e-06,
+      "loss": 0.7621,
+      "step": 7285
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.311621252411491e-06,
+      "loss": 0.5408,
+      "step": 7286
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.310521537623938e-06,
+      "loss": 0.9781,
+      "step": 7287
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.309421807756568e-06,
+      "loss": 0.7862,
+      "step": 7288
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.308322062862786e-06,
+      "loss": 0.751,
+      "step": 7289
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.307222302995998e-06,
+      "loss": 0.8159,
+      "step": 7290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.306122528209613e-06,
+      "loss": 0.6645,
+      "step": 7291
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.305022738557039e-06,
+      "loss": 0.7991,
+      "step": 7292
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.303922934091684e-06,
+      "loss": 0.9329,
+      "step": 7293
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.302823114866959e-06,
+      "loss": 0.8983,
+      "step": 7294
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.301723280936271e-06,
+      "loss": 0.7725,
+      "step": 7295
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.300623432353036e-06,
+      "loss": 0.7046,
+      "step": 7296
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2995235691706645e-06,
+      "loss": 0.7809,
+      "step": 7297
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.298423691442566e-06,
+      "loss": 0.674,
+      "step": 7298
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.297323799222156e-06,
+      "loss": 0.6787,
+      "step": 7299
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.296223892562849e-06,
+      "loss": 0.8469,
+      "step": 7300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.295123971518058e-06,
+      "loss": 1.0231,
+      "step": 7301
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.294024036141199e-06,
+      "loss": 0.8936,
+      "step": 7302
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.292924086485689e-06,
+      "loss": 0.8668,
+      "step": 7303
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.291824122604944e-06,
+      "loss": 0.6587,
+      "step": 7304
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.290724144552379e-06,
+      "loss": 0.8349,
+      "step": 7305
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.289624152381416e-06,
+      "loss": 0.6516,
+      "step": 7306
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.288524146145473e-06,
+      "loss": 0.7395,
+      "step": 7307
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.287424125897967e-06,
+      "loss": 0.8186,
+      "step": 7308
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.286324091692321e-06,
+      "loss": 0.5829,
+      "step": 7309
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.285224043581954e-06,
+      "loss": 0.9535,
+      "step": 7310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.284123981620289e-06,
+      "loss": 0.6646,
+      "step": 7311
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2830239058607446e-06,
+      "loss": 0.437,
+      "step": 7312
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.281923816356748e-06,
+      "loss": 0.7079,
+      "step": 7313
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.280823713161721e-06,
+      "loss": 0.5479,
+      "step": 7314
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2797235963290885e-06,
+      "loss": 0.6564,
+      "step": 7315
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.278623465912273e-06,
+      "loss": 0.8085,
+      "step": 7316
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.277523321964701e-06,
+      "loss": 0.5198,
+      "step": 7317
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.276423164539801e-06,
+      "loss": 0.8617,
+      "step": 7318
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.275322993690998e-06,
+      "loss": 0.9833,
+      "step": 7319
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.274222809471719e-06,
+      "loss": 0.6882,
+      "step": 7320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2731226119353915e-06,
+      "loss": 0.9727,
+      "step": 7321
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.272022401135446e-06,
+      "loss": 1.0454,
+      "step": 7322
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.270922177125312e-06,
+      "loss": 0.4973,
+      "step": 7323
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2698219399584195e-06,
+      "loss": 0.7438,
+      "step": 7324
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.268721689688197e-06,
+      "loss": 0.6303,
+      "step": 7325
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.267621426368076e-06,
+      "loss": 0.7029,
+      "step": 7326
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.266521150051492e-06,
+      "loss": 0.7849,
+      "step": 7327
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2654208607918745e-06,
+      "loss": 0.7829,
+      "step": 7328
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.264320558642657e-06,
+      "loss": 0.5021,
+      "step": 7329
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.263220243657274e-06,
+      "loss": 0.7508,
+      "step": 7330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.262119915889161e-06,
+      "loss": 0.9317,
+      "step": 7331
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.261019575391751e-06,
+      "loss": 0.7189,
+      "step": 7332
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.25991922221848e-06,
+      "loss": 0.5384,
+      "step": 7333
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.258818856422785e-06,
+      "loss": 0.5088,
+      "step": 7334
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.257718478058105e-06,
+      "loss": 0.4803,
+      "step": 7335
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.256618087177872e-06,
+      "loss": 0.6791,
+      "step": 7336
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.255517683835528e-06,
+      "loss": 0.9218,
+      "step": 7337
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.254417268084514e-06,
+      "loss": 1.1387,
+      "step": 7338
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2533168399782646e-06,
+      "loss": 0.6337,
+      "step": 7339
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.25221639957022e-06,
+      "loss": 0.9125,
+      "step": 7340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.251115946913826e-06,
+      "loss": 0.9676,
+      "step": 7341
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.250015482062518e-06,
+      "loss": 0.8494,
+      "step": 7342
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.248915005069741e-06,
+      "loss": 0.8329,
+      "step": 7343
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.247814515988936e-06,
+      "loss": 0.8074,
+      "step": 7344
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.246714014873545e-06,
+      "loss": 0.6812,
+      "step": 7345
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.245613501777014e-06,
+      "loss": 1.0409,
+      "step": 7346
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2445129767527844e-06,
+      "loss": 0.6106,
+      "step": 7347
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.243412439854304e-06,
+      "loss": 0.8921,
+      "step": 7348
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.242311891135016e-06,
+      "loss": 0.5003,
+      "step": 7349
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.241211330648365e-06,
+      "loss": 0.5798,
+      "step": 7350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2401107584478e-06,
+      "loss": 0.5697,
+      "step": 7351
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.239010174586767e-06,
+      "loss": 0.9384,
+      "step": 7352
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.237909579118713e-06,
+      "loss": 0.6234,
+      "step": 7353
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.236808972097085e-06,
+      "loss": 0.7158,
+      "step": 7354
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.235708353575335e-06,
+      "loss": 0.6284,
+      "step": 7355
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.234607723606908e-06,
+      "loss": 0.6482,
+      "step": 7356
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.233507082245258e-06,
+      "loss": 0.6823,
+      "step": 7357
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.232406429543833e-06,
+      "loss": 0.7616,
+      "step": 7358
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.231305765556083e-06,
+      "loss": 0.6914,
+      "step": 7359
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.230205090335461e-06,
+      "loss": 0.6345,
+      "step": 7360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.229104403935419e-06,
+      "loss": 0.8332,
+      "step": 7361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.228003706409409e-06,
+      "loss": 0.7014,
+      "step": 7362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.226902997810883e-06,
+      "loss": 0.7428,
+      "step": 7363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.225802278193296e-06,
+      "loss": 0.7539,
+      "step": 7364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2247015476101025e-06,
+      "loss": 0.6642,
+      "step": 7365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.223600806114756e-06,
+      "loss": 0.6749,
+      "step": 7366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2225000537607115e-06,
+      "loss": 0.9208,
+      "step": 7367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.221399290601426e-06,
+      "loss": 0.8276,
+      "step": 7368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.220298516690353e-06,
+      "loss": 1.0633,
+      "step": 7369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.219197732080954e-06,
+      "loss": 0.7429,
+      "step": 7370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.218096936826681e-06,
+      "loss": 0.6633,
+      "step": 7371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.216996130980995e-06,
+      "loss": 0.6241,
+      "step": 7372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.215895314597355e-06,
+      "loss": 0.7674,
+      "step": 7373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2147944877292145e-06,
+      "loss": 0.895,
+      "step": 7374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.21369365043004e-06,
+      "loss": 1.0185,
+      "step": 7375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2125928027532865e-06,
+      "loss": 0.5324,
+      "step": 7376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2114919447524155e-06,
+      "loss": 0.7959,
+      "step": 7377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.210391076480888e-06,
+      "loss": 0.6819,
+      "step": 7378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.209290197992167e-06,
+      "loss": 0.8603,
+      "step": 7379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.208189309339713e-06,
+      "loss": 0.7058,
+      "step": 7380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.207088410576986e-06,
+      "loss": 0.6903,
+      "step": 7381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.2059875017574525e-06,
+      "loss": 0.8166,
+      "step": 7382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.204886582934573e-06,
+      "loss": 0.8884,
+      "step": 7383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.203785654161814e-06,
+      "loss": 0.784,
+      "step": 7384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.202684715492635e-06,
+      "loss": 0.9714,
+      "step": 7385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.201583766980506e-06,
+      "loss": 0.7744,
+      "step": 7386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.200482808678892e-06,
+      "loss": 0.5048,
+      "step": 7387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.199381840641253e-06,
+      "loss": 0.6929,
+      "step": 7388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.198280862921062e-06,
+      "loss": 0.7961,
+      "step": 7389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.197179875571782e-06,
+      "loss": 0.831,
+      "step": 7390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.196078878646881e-06,
+      "loss": 0.5596,
+      "step": 7391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.194977872199826e-06,
+      "loss": 0.7222,
+      "step": 7392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.193876856284085e-06,
+      "loss": 0.5798,
+      "step": 7393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.192775830953126e-06,
+      "loss": 0.7135,
+      "step": 7394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.19167479626042e-06,
+      "loss": 0.8446,
+      "step": 7395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.190573752259436e-06,
+      "loss": 1.0469,
+      "step": 7396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.189472699003642e-06,
+      "loss": 0.4403,
+      "step": 7397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.188371636546508e-06,
+      "loss": 0.7095,
+      "step": 7398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.187270564941509e-06,
+      "loss": 0.6395,
+      "step": 7399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.186169484242113e-06,
+      "loss": 0.6431,
+      "step": 7400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.185068394501791e-06,
+      "loss": 0.8404,
+      "step": 7401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1839672957740165e-06,
+      "loss": 0.6236,
+      "step": 7402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.182866188112262e-06,
+      "loss": 0.4962,
+      "step": 7403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.18176507157e-06,
+      "loss": 0.7305,
+      "step": 7404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1806639462007024e-06,
+      "loss": 0.8418,
+      "step": 7405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.179562812057845e-06,
+      "loss": 0.7313,
+      "step": 7406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.178461669194903e-06,
+      "loss": 0.8523,
+      "step": 7407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.17736051766535e-06,
+      "loss": 0.6789,
+      "step": 7408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.176259357522658e-06,
+      "loss": 0.9304,
+      "step": 7409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.175158188820307e-06,
+      "loss": 0.8562,
+      "step": 7410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.174057011611771e-06,
+      "loss": 0.7349,
+      "step": 7411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.172955825950526e-06,
+      "loss": 0.5917,
+      "step": 7412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.171854631890049e-06,
+      "loss": 0.7869,
+      "step": 7413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.17075342948382e-06,
+      "loss": 0.7335,
+      "step": 7414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.169652218785313e-06,
+      "loss": 0.8441,
+      "step": 7415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.168550999848005e-06,
+      "loss": 0.5818,
+      "step": 7416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1674497727253766e-06,
+      "loss": 0.8897,
+      "step": 7417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.1663485374709065e-06,
+      "loss": 0.6675,
+      "step": 7418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.165247294138074e-06,
+      "loss": 0.6454,
+      "step": 7419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.164146042780358e-06,
+      "loss": 0.7878,
+      "step": 7420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.163044783451239e-06,
+      "loss": 0.5803,
+      "step": 7421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.161943516204197e-06,
+      "loss": 0.7005,
+      "step": 7422
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.16084224109271e-06,
+      "loss": 0.54,
+      "step": 7423
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.159740958170265e-06,
+      "loss": 0.5869,
+      "step": 7424
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.15863966749034e-06,
+      "loss": 0.5978,
+      "step": 7425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.157538369106414e-06,
+      "loss": 0.7242,
+      "step": 7426
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1564370630719755e-06,
+      "loss": 0.8411,
+      "step": 7427
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1553357494405024e-06,
+      "loss": 0.6917,
+      "step": 7428
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.154234428265478e-06,
+      "loss": 0.8958,
+      "step": 7429
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.153133099600388e-06,
+      "loss": 0.7947,
+      "step": 7430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.152031763498715e-06,
+      "loss": 0.7043,
+      "step": 7431
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.150930420013942e-06,
+      "loss": 0.8923,
+      "step": 7432
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.149829069199555e-06,
+      "loss": 0.8099,
+      "step": 7433
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1487277111090375e-06,
+      "loss": 0.6185,
+      "step": 7434
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.147626345795876e-06,
+      "loss": 0.5358,
+      "step": 7435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.146524973313553e-06,
+      "loss": 0.7594,
+      "step": 7436
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.145423593715558e-06,
+      "loss": 0.6148,
+      "step": 7437
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.144322207055375e-06,
+      "loss": 0.9189,
+      "step": 7438
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.143220813386491e-06,
+      "loss": 0.8064,
+      "step": 7439
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.142119412762392e-06,
+      "loss": 0.6074,
+      "step": 7440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.141018005236567e-06,
+      "loss": 0.8501,
+      "step": 7441
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.139916590862504e-06,
+      "loss": 0.9951,
+      "step": 7442
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.138815169693687e-06,
+      "loss": 0.8359,
+      "step": 7443
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.137713741783608e-06,
+      "loss": 0.9816,
+      "step": 7444
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.136612307185753e-06,
+      "loss": 0.712,
+      "step": 7445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1355108659536135e-06,
+      "loss": 0.7941,
+      "step": 7446
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.134409418140675e-06,
+      "loss": 0.895,
+      "step": 7447
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.133307963800429e-06,
+      "loss": 0.7744,
+      "step": 7448
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.132206502986368e-06,
+      "loss": 0.9718,
+      "step": 7449
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.131105035751977e-06,
+      "loss": 0.8622,
+      "step": 7450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.130003562150748e-06,
+      "loss": 0.759,
+      "step": 7451
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1289020822361755e-06,
+      "loss": 0.7876,
+      "step": 7452
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1278005960617465e-06,
+      "loss": 0.5714,
+      "step": 7453
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.126699103680952e-06,
+      "loss": 0.6231,
+      "step": 7454
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1255976051472865e-06,
+      "loss": 0.767,
+      "step": 7455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.12449610051424e-06,
+      "loss": 0.8496,
+      "step": 7456
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1233945898353055e-06,
+      "loss": 1.0513,
+      "step": 7457
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.122293073163974e-06,
+      "loss": 0.8668,
+      "step": 7458
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.121191550553741e-06,
+      "loss": 0.8056,
+      "step": 7459
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.120090022058099e-06,
+      "loss": 0.8593,
+      "step": 7460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.118988487730537e-06,
+      "loss": 0.5268,
+      "step": 7461
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1178869476245554e-06,
+      "loss": 0.7409,
+      "step": 7462
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.116785401793645e-06,
+      "loss": 0.8445,
+      "step": 7463
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.115683850291298e-06,
+      "loss": 0.5498,
+      "step": 7464
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.114582293171012e-06,
+      "loss": 0.6884,
+      "step": 7465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.113480730486281e-06,
+      "loss": 0.7448,
+      "step": 7466
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.112379162290599e-06,
+      "loss": 0.7274,
+      "step": 7467
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.111277588637462e-06,
+      "loss": 0.9593,
+      "step": 7468
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.110176009580365e-06,
+      "loss": 0.7727,
+      "step": 7469
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.109074425172806e-06,
+      "loss": 0.7305,
+      "step": 7470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.107972835468279e-06,
+      "loss": 0.6494,
+      "step": 7471
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.10687124052028e-06,
+      "loss": 0.8601,
+      "step": 7472
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.105769640382309e-06,
+      "loss": 0.6649,
+      "step": 7473
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.1046680351078584e-06,
+      "loss": 0.6075,
+      "step": 7474
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.103566424750427e-06,
+      "loss": 0.7803,
+      "step": 7475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.102464809363514e-06,
+      "loss": 0.4694,
+      "step": 7476
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.101363189000614e-06,
+      "loss": 0.8587,
+      "step": 7477
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.100261563715227e-06,
+      "loss": 0.9147,
+      "step": 7478
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.099159933560851e-06,
+      "loss": 0.8047,
+      "step": 7479
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.098058298590983e-06,
+      "loss": 0.7654,
+      "step": 7480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.096956658859122e-06,
+      "loss": 0.7084,
+      "step": 7481
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.095855014418768e-06,
+      "loss": 0.7345,
+      "step": 7482
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.094753365323419e-06,
+      "loss": 0.6485,
+      "step": 7483
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.093651711626575e-06,
+      "loss": 0.5931,
+      "step": 7484
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.092550053381733e-06,
+      "loss": 0.8541,
+      "step": 7485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.091448390642395e-06,
+      "loss": 0.7227,
+      "step": 7486
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.090346723462062e-06,
+      "loss": 0.8029,
+      "step": 7487
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089245051894231e-06,
+      "loss": 0.8445,
+      "step": 7488
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.088143375992404e-06,
+      "loss": 0.7003,
+      "step": 7489
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.087041695810082e-06,
+      "loss": 0.6397,
+      "step": 7490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0859400114007654e-06,
+      "loss": 0.7398,
+      "step": 7491
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.084838322817954e-06,
+      "loss": 1.02,
+      "step": 7492
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.083736630115151e-06,
+      "loss": 0.7619,
+      "step": 7493
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.082634933345856e-06,
+      "loss": 0.6195,
+      "step": 7494
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0815332325635715e-06,
+      "loss": 0.6495,
+      "step": 7495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.080431527821798e-06,
+      "loss": 0.8017,
+      "step": 7496
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.07932981917404e-06,
+      "loss": 0.8151,
+      "step": 7497
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.078228106673798e-06,
+      "loss": 0.7887,
+      "step": 7498
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.077126390374573e-06,
+      "loss": 0.8572,
+      "step": 7499
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.076024670329871e-06,
+      "loss": 0.746,
+      "step": 7500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.074922946593192e-06,
+      "loss": 0.7937,
+      "step": 7501
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.073821219218037e-06,
+      "loss": 1.0072,
+      "step": 7502
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.072719488257915e-06,
+      "loss": 0.9544,
+      "step": 7503
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.071617753766324e-06,
+      "loss": 0.8775,
+      "step": 7504
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0705160157967685e-06,
+      "loss": 0.7003,
+      "step": 7505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.069414274402753e-06,
+      "loss": 0.573,
+      "step": 7506
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.068312529637781e-06,
+      "loss": 0.7422,
+      "step": 7507
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0672107815553576e-06,
+      "loss": 0.8875,
+      "step": 7508
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.066109030208985e-06,
+      "loss": 0.4633,
+      "step": 7509
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.065007275652166e-06,
+      "loss": 0.8879,
+      "step": 7510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.06390551793841e-06,
+      "loss": 0.6673,
+      "step": 7511
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.062803757121216e-06,
+      "loss": 0.5129,
+      "step": 7512
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.061701993254092e-06,
+      "loss": 0.6631,
+      "step": 7513
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.060600226390543e-06,
+      "loss": 0.6394,
+      "step": 7514
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.059498456584072e-06,
+      "loss": 0.5213,
+      "step": 7515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.058396683888184e-06,
+      "loss": 0.5229,
+      "step": 7516
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.057294908356387e-06,
+      "loss": 0.6982,
+      "step": 7517
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.056193130042184e-06,
+      "loss": 0.7812,
+      "step": 7518
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.05509134899908e-06,
+      "loss": 0.7626,
+      "step": 7519
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0539895652805834e-06,
+      "loss": 0.5447,
+      "step": 7520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.052887778940199e-06,
+      "loss": 0.7061,
+      "step": 7521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0517859900314316e-06,
+      "loss": 0.685,
+      "step": 7522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.050684198607786e-06,
+      "loss": 0.6089,
+      "step": 7523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.049582404722771e-06,
+      "loss": 0.9142,
+      "step": 7524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.048480608429893e-06,
+      "loss": 1.003,
+      "step": 7525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.047378809782656e-06,
+      "loss": 0.5332,
+      "step": 7526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.046277008834568e-06,
+      "loss": 0.5546,
+      "step": 7527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0451752056391365e-06,
+      "loss": 0.7482,
+      "step": 7528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.044073400249867e-06,
+      "loss": 0.8212,
+      "step": 7529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.042971592720267e-06,
+      "loss": 0.5354,
+      "step": 7530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.041869783103842e-06,
+      "loss": 0.6427,
+      "step": 7531
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0407679714541e-06,
+      "loss": 0.5968,
+      "step": 7532
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.039666157824549e-06,
+      "loss": 0.6412,
+      "step": 7533
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.038564342268695e-06,
+      "loss": 0.7758,
+      "step": 7534
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.037462524840045e-06,
+      "loss": 0.7287,
+      "step": 7535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.03636070559211e-06,
+      "loss": 0.9336,
+      "step": 7536
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.035258884578392e-06,
+      "loss": 0.7978,
+      "step": 7537
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0341570618524014e-06,
+      "loss": 0.607,
+      "step": 7538
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.033055237467646e-06,
+      "loss": 0.8106,
+      "step": 7539
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.031953411477633e-06,
+      "loss": 0.6572,
+      "step": 7540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.030851583935872e-06,
+      "loss": 0.8816,
+      "step": 7541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.029749754895869e-06,
+      "loss": 0.8131,
+      "step": 7542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.028647924411132e-06,
+      "loss": 0.705,
+      "step": 7543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0275460925351694e-06,
+      "loss": 0.602,
+      "step": 7544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.026444259321489e-06,
+      "loss": 0.6442,
+      "step": 7545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0253424248236e-06,
+      "loss": 0.7547,
+      "step": 7546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.02424058909501e-06,
+      "loss": 0.8364,
+      "step": 7547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.023138752189227e-06,
+      "loss": 0.7501,
+      "step": 7548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.022036914159762e-06,
+      "loss": 0.8655,
+      "step": 7549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.020935075060117e-06,
+      "loss": 0.8306,
+      "step": 7550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.019833234943806e-06,
+      "loss": 0.6687,
+      "step": 7551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.018731393864339e-06,
+      "loss": 0.5789,
+      "step": 7552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.017629551875219e-06,
+      "loss": 0.9726,
+      "step": 7553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.016527709029957e-06,
+      "loss": 0.6215,
+      "step": 7554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.015425865382064e-06,
+      "loss": 0.8175,
+      "step": 7555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.014324020985045e-06,
+      "loss": 0.7006,
+      "step": 7556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.013222175892411e-06,
+      "loss": 0.4354,
+      "step": 7557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.012120330157671e-06,
+      "loss": 0.5866,
+      "step": 7558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.011018483834331e-06,
+      "loss": 0.8917,
+      "step": 7559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.009916636975903e-06,
+      "loss": 0.6304,
+      "step": 7560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.008814789635894e-06,
+      "loss": 0.7379,
+      "step": 7561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.007712941867814e-06,
+      "loss": 0.8366,
+      "step": 7562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.006611093725172e-06,
+      "loss": 0.7476,
+      "step": 7563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.005509245261475e-06,
+      "loss": 0.7416,
+      "step": 7564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.004407396530234e-06,
+      "loss": 0.7897,
+      "step": 7565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.0033055475849575e-06,
+      "loss": 0.5161,
+      "step": 7566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.002203698479154e-06,
+      "loss": 0.3551,
+      "step": 7567
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.001101849266332e-06,
+      "loss": 0.7371,
+      "step": 7568
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5e-06,
+      "loss": 0.7516,
+      "step": 7569
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.998898150733671e-06,
+      "loss": 0.7818,
+      "step": 7570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.997796301520848e-06,
+      "loss": 0.6086,
+      "step": 7571
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.996694452415044e-06,
+      "loss": 0.5691,
+      "step": 7572
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9955926034697675e-06,
+      "loss": 0.6279,
+      "step": 7573
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.994490754738526e-06,
+      "loss": 0.533,
+      "step": 7574
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.993388906274829e-06,
+      "loss": 0.694,
+      "step": 7575
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.992287058132188e-06,
+      "loss": 0.622,
+      "step": 7576
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9911852103641065e-06,
+      "loss": 0.7682,
+      "step": 7577
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.990083363024098e-06,
+      "loss": 0.8093,
+      "step": 7578
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98898151616567e-06,
+      "loss": 0.669,
+      "step": 7579
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.987879669842333e-06,
+      "loss": 0.6275,
+      "step": 7580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.98677782410759e-06,
+      "loss": 0.771,
+      "step": 7581
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.985675979014956e-06,
+      "loss": 0.6716,
+      "step": 7582
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9845741346179386e-06,
+      "loss": 0.6631,
+      "step": 7583
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.983472290970043e-06,
+      "loss": 0.6412,
+      "step": 7584
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.982370448124783e-06,
+      "loss": 0.7205,
+      "step": 7585
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.981268606135664e-06,
+      "loss": 0.6577,
+      "step": 7586
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.980166765056194e-06,
+      "loss": 0.7821,
+      "step": 7587
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.979064924939884e-06,
+      "loss": 0.796,
+      "step": 7588
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.977963085840242e-06,
+      "loss": 0.7587,
+      "step": 7589
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9768612478107735e-06,
+      "loss": 0.3871,
+      "step": 7590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.975759410904992e-06,
+      "loss": 0.6674,
+      "step": 7591
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.974657575176401e-06,
+      "loss": 0.5517,
+      "step": 7592
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.973555740678512e-06,
+      "loss": 0.687,
+      "step": 7593
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.972453907464832e-06,
+      "loss": 0.8512,
+      "step": 7594
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.97135207558887e-06,
+      "loss": 0.636,
+      "step": 7595
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.970250245104132e-06,
+      "loss": 0.8849,
+      "step": 7596
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.969148416064129e-06,
+      "loss": 0.6213,
+      "step": 7597
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.968046588522368e-06,
+      "loss": 0.5466,
+      "step": 7598
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.966944762532356e-06,
+      "loss": 0.8454,
+      "step": 7599
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.965842938147599e-06,
+      "loss": 0.8147,
+      "step": 7600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.96474111542161e-06,
+      "loss": 0.7458,
+      "step": 7601
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.963639294407893e-06,
+      "loss": 0.6843,
+      "step": 7602
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9625374751599546e-06,
+      "loss": 0.6568,
+      "step": 7603
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.961435657731307e-06,
+      "loss": 0.7931,
+      "step": 7604
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.960333842175453e-06,
+      "loss": 0.7776,
+      "step": 7605
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9592320285459e-06,
+      "loss": 0.5535,
+      "step": 7606
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9581302168961595e-06,
+      "loss": 1.0089,
+      "step": 7607
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9570284072797356e-06,
+      "loss": 0.6762,
+      "step": 7608
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.955926599750134e-06,
+      "loss": 0.7578,
+      "step": 7609
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.954824794360864e-06,
+      "loss": 0.7576,
+      "step": 7610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.953722991165433e-06,
+      "loss": 0.3719,
+      "step": 7611
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.952621190217344e-06,
+      "loss": 0.5504,
+      "step": 7612
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.951519391570108e-06,
+      "loss": 0.6405,
+      "step": 7613
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.95041759527723e-06,
+      "loss": 0.6549,
+      "step": 7614
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.949315801392215e-06,
+      "loss": 0.8572,
+      "step": 7615
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.94821400996857e-06,
+      "loss": 0.9144,
+      "step": 7616
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.947112221059803e-06,
+      "loss": 0.9693,
+      "step": 7617
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.946010434719419e-06,
+      "loss": 0.7393,
+      "step": 7618
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.944908651000921e-06,
+      "loss": 1.0144,
+      "step": 7619
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9438068699578175e-06,
+      "loss": 0.8301,
+      "step": 7620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9427050916436155e-06,
+      "loss": 0.7658,
+      "step": 7621
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.941603316111817e-06,
+      "loss": 0.9539,
+      "step": 7622
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.940501543415929e-06,
+      "loss": 0.584,
+      "step": 7623
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.939399773609459e-06,
+      "loss": 0.7017,
+      "step": 7624
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.938298006745909e-06,
+      "loss": 1.0206,
+      "step": 7625
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.937196242878785e-06,
+      "loss": 0.8725,
+      "step": 7626
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.936094482061592e-06,
+      "loss": 0.8798,
+      "step": 7627
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.934992724347834e-06,
+      "loss": 0.8182,
+      "step": 7628
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.933890969791018e-06,
+      "loss": 1.0827,
+      "step": 7629
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.932789218444644e-06,
+      "loss": 0.8015,
+      "step": 7630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.931687470362219e-06,
+      "loss": 0.6256,
+      "step": 7631
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9305857255972485e-06,
+      "loss": 0.5784,
+      "step": 7632
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.929483984203232e-06,
+      "loss": 0.8957,
+      "step": 7633
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.92838224623368e-06,
+      "loss": 0.7228,
+      "step": 7634
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.927280511742087e-06,
+      "loss": 0.698,
+      "step": 7635
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.926178780781965e-06,
+      "loss": 0.8248,
+      "step": 7636
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9250770534068115e-06,
+      "loss": 0.4748,
+      "step": 7637
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9239753296701305e-06,
+      "loss": 0.7998,
+      "step": 7638
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.922873609625428e-06,
+      "loss": 0.7532,
+      "step": 7639
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.921771893326204e-06,
+      "loss": 0.7278,
+      "step": 7640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9206701808259605e-06,
+      "loss": 0.7638,
+      "step": 7641
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.9195684721782025e-06,
+      "loss": 0.6202,
+      "step": 7642
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91846676743643e-06,
+      "loss": 0.6916,
+      "step": 7643
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.917365066654144e-06,
+      "loss": 0.5515,
+      "step": 7644
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.916263369884851e-06,
+      "loss": 0.733,
+      "step": 7645
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.915161677182048e-06,
+      "loss": 0.8241,
+      "step": 7646
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.914059988599235e-06,
+      "loss": 0.8602,
+      "step": 7647
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.912958304189919e-06,
+      "loss": 0.8883,
+      "step": 7648
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.911856624007598e-06,
+      "loss": 0.7047,
+      "step": 7649
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.91075494810577e-06,
+      "loss": 0.9486,
+      "step": 7650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.90965327653794e-06,
+      "loss": 0.5987,
+      "step": 7651
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.908551609357606e-06,
+      "loss": 0.5963,
+      "step": 7652
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.907449946618268e-06,
+      "loss": 0.72,
+      "step": 7653
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.906348288373426e-06,
+      "loss": 0.5958,
+      "step": 7654
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.905246634676582e-06,
+      "loss": 0.8658,
+      "step": 7655
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.904144985581234e-06,
+      "loss": 0.7458,
+      "step": 7656
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.903043341140879e-06,
+      "loss": 0.5441,
+      "step": 7657
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.901941701409018e-06,
+      "loss": 0.7212,
+      "step": 7658
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.900840066439151e-06,
+      "loss": 0.8536,
+      "step": 7659
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.899738436284774e-06,
+      "loss": 0.8426,
+      "step": 7660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8986368109993865e-06,
+      "loss": 0.8184,
+      "step": 7661
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.897535190636489e-06,
+      "loss": 0.8215,
+      "step": 7662
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.896433575249574e-06,
+      "loss": 0.7698,
+      "step": 7663
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.895331964892144e-06,
+      "loss": 0.7413,
+      "step": 7664
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.894230359617693e-06,
+      "loss": 0.7976,
+      "step": 7665
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.893128759479719e-06,
+      "loss": 0.5359,
+      "step": 7666
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.892027164531722e-06,
+      "loss": 0.5772,
+      "step": 7667
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.890925574827195e-06,
+      "loss": 0.6775,
+      "step": 7668
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.889823990419635e-06,
+      "loss": 0.6714,
+      "step": 7669
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.888722411362539e-06,
+      "loss": 0.8836,
+      "step": 7670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8876208377094025e-06,
+      "loss": 0.8585,
+      "step": 7671
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.886519269513721e-06,
+      "loss": 0.6886,
+      "step": 7672
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.885417706828989e-06,
+      "loss": 0.8066,
+      "step": 7673
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.884316149708704e-06,
+      "loss": 0.6975,
+      "step": 7674
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8832145982063575e-06,
+      "loss": 0.9089,
+      "step": 7675
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.882113052375445e-06,
+      "loss": 0.729,
+      "step": 7676
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881011512269464e-06,
+      "loss": 0.7594,
+      "step": 7677
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.879909977941904e-06,
+      "loss": 0.4678,
+      "step": 7678
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.878808449446259e-06,
+      "loss": 0.798,
+      "step": 7679
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8777069268360264e-06,
+      "loss": 0.7582,
+      "step": 7680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.876605410164696e-06,
+      "loss": 0.8045,
+      "step": 7681
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.87550389948576e-06,
+      "loss": 0.8745,
+      "step": 7682
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.874402394852715e-06,
+      "loss": 0.9464,
+      "step": 7683
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.87330089631905e-06,
+      "loss": 0.7823,
+      "step": 7684
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.872199403938255e-06,
+      "loss": 0.9005,
+      "step": 7685
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.871097917763825e-06,
+      "loss": 0.7033,
+      "step": 7686
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8699964378492535e-06,
+      "loss": 0.9758,
+      "step": 7687
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.868894964248024e-06,
+      "loss": 0.6426,
+      "step": 7688
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.867793497013634e-06,
+      "loss": 0.5582,
+      "step": 7689
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.866692036199573e-06,
+      "loss": 0.6714,
+      "step": 7690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.865590581859327e-06,
+      "loss": 0.6945,
+      "step": 7691
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.864489134046388e-06,
+      "loss": 0.8144,
+      "step": 7692
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.863387692814249e-06,
+      "loss": 0.7455,
+      "step": 7693
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.862286258216396e-06,
+      "loss": 0.7138,
+      "step": 7694
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.861184830306314e-06,
+      "loss": 0.4454,
+      "step": 7695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8600834091374986e-06,
+      "loss": 0.8906,
+      "step": 7696
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.858981994763435e-06,
+      "loss": 0.7212,
+      "step": 7697
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8578805872376085e-06,
+      "loss": 0.855,
+      "step": 7698
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.85677918661351e-06,
+      "loss": 0.7632,
+      "step": 7699
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.855677792944627e-06,
+      "loss": 0.5261,
+      "step": 7700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.854576406284443e-06,
+      "loss": 0.7779,
+      "step": 7701
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.853475026686449e-06,
+      "loss": 0.7887,
+      "step": 7702
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.852373654204126e-06,
+      "loss": 0.9662,
+      "step": 7703
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.851272288890963e-06,
+      "loss": 0.9399,
+      "step": 7704
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.850170930800447e-06,
+      "loss": 0.7983,
+      "step": 7705
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.84906957998606e-06,
+      "loss": 0.7999,
+      "step": 7706
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8479682365012855e-06,
+      "loss": 1.0233,
+      "step": 7707
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.846866900399613e-06,
+      "loss": 0.6734,
+      "step": 7708
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.845765571734522e-06,
+      "loss": 0.753,
+      "step": 7709
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8446642505595e-06,
+      "loss": 0.8964,
+      "step": 7710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.843562936928026e-06,
+      "loss": 0.7544,
+      "step": 7711
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.842461630893587e-06,
+      "loss": 0.4849,
+      "step": 7712
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.841360332509663e-06,
+      "loss": 0.7999,
+      "step": 7713
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.840259041829736e-06,
+      "loss": 0.7958,
+      "step": 7714
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.8391577589072905e-06,
+      "loss": 0.8141,
+      "step": 7715
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.838056483795805e-06,
+      "loss": 0.7885,
+      "step": 7716
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.836955216548762e-06,
+      "loss": 0.7776,
+      "step": 7717
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.835853957219644e-06,
+      "loss": 0.5748,
+      "step": 7718
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.834752705861927e-06,
+      "loss": 0.6821,
+      "step": 7719
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8336514625290935e-06,
+      "loss": 0.6226,
+      "step": 7720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.832550227274624e-06,
+      "loss": 0.7148,
+      "step": 7721
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.831449000151997e-06,
+      "loss": 0.6475,
+      "step": 7722
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8303477812146895e-06,
+      "loss": 0.7775,
+      "step": 7723
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8292465705161815e-06,
+      "loss": 0.69,
+      "step": 7724
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8281453681099515e-06,
+      "loss": 0.3286,
+      "step": 7725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8270441740494736e-06,
+      "loss": 0.5209,
+      "step": 7726
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.825942988388229e-06,
+      "loss": 0.7865,
+      "step": 7727
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.824841811179695e-06,
+      "loss": 0.899,
+      "step": 7728
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.823740642477343e-06,
+      "loss": 0.5884,
+      "step": 7729
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.822639482334652e-06,
+      "loss": 0.6359,
+      "step": 7730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.821538330805098e-06,
+      "loss": 0.9035,
+      "step": 7731
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.820437187942157e-06,
+      "loss": 0.6845,
+      "step": 7732
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.819336053799299e-06,
+      "loss": 0.8523,
+      "step": 7733
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.818234928430003e-06,
+      "loss": 0.8557,
+      "step": 7734
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.817133811887741e-06,
+      "loss": 0.9341,
+      "step": 7735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.816032704225984e-06,
+      "loss": 1.0528,
+      "step": 7736
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.81493160549821e-06,
+      "loss": 0.6518,
+      "step": 7737
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.813830515757889e-06,
+      "loss": 0.6201,
+      "step": 7738
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.812729435058492e-06,
+      "loss": 0.4829,
+      "step": 7739
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8116283634534924e-06,
+      "loss": 0.6799,
+      "step": 7740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.81052730099636e-06,
+      "loss": 0.5653,
+      "step": 7741
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.809426247740565e-06,
+      "loss": 0.7233,
+      "step": 7742
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.808325203739581e-06,
+      "loss": 0.634,
+      "step": 7743
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.807224169046875e-06,
+      "loss": 0.6262,
+      "step": 7744
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.806123143715916e-06,
+      "loss": 0.5937,
+      "step": 7745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.805022127800176e-06,
+      "loss": 0.8894,
+      "step": 7746
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.803921121353121e-06,
+      "loss": 0.7185,
+      "step": 7747
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.802820124428221e-06,
+      "loss": 0.6822,
+      "step": 7748
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.801719137078939e-06,
+      "loss": 0.8599,
+      "step": 7749
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.800618159358748e-06,
+      "loss": 0.8039,
+      "step": 7750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.799517191321111e-06,
+      "loss": 0.6046,
+      "step": 7751
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.798416233019494e-06,
+      "loss": 0.7576,
+      "step": 7752
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7973152845073666e-06,
+      "loss": 0.8146,
+      "step": 7753
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.796214345838189e-06,
+      "loss": 0.8354,
+      "step": 7754
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.795113417065427e-06,
+      "loss": 0.8287,
+      "step": 7755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.794012498242549e-06,
+      "loss": 0.89,
+      "step": 7756
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.792911589423015e-06,
+      "loss": 0.596,
+      "step": 7757
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.791810690660288e-06,
+      "loss": 0.7831,
+      "step": 7758
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.790709802007834e-06,
+      "loss": 0.5875,
+      "step": 7759
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.789608923519113e-06,
+      "loss": 0.9819,
+      "step": 7760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.788508055247585e-06,
+      "loss": 0.8122,
+      "step": 7761
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.787407197246714e-06,
+      "loss": 0.5153,
+      "step": 7762
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.786306349569963e-06,
+      "loss": 0.6039,
+      "step": 7763
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.785205512270787e-06,
+      "loss": 0.6183,
+      "step": 7764
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.784104685402648e-06,
+      "loss": 0.8682,
+      "step": 7765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7830038690190076e-06,
+      "loss": 0.509,
+      "step": 7766
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.781903063173321e-06,
+      "loss": 0.8771,
+      "step": 7767
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.780802267919048e-06,
+      "loss": 0.7833,
+      "step": 7768
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.779701483309648e-06,
+      "loss": 0.7117,
+      "step": 7769
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7786007093985775e-06,
+      "loss": 0.7181,
+      "step": 7770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.77749994623929e-06,
+      "loss": 0.9564,
+      "step": 7771
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.776399193885246e-06,
+      "loss": 0.8117,
+      "step": 7772
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7752984523899e-06,
+      "loss": 0.4514,
+      "step": 7773
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7741977218067045e-06,
+      "loss": 0.7362,
+      "step": 7774
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.773097002189118e-06,
+      "loss": 0.6699,
+      "step": 7775
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.771996293590593e-06,
+      "loss": 0.6717,
+      "step": 7776
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.770895596064582e-06,
+      "loss": 0.7742,
+      "step": 7777
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.76979490966454e-06,
+      "loss": 0.7274,
+      "step": 7778
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.768694234443919e-06,
+      "loss": 0.8235,
+      "step": 7779
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.767593570456168e-06,
+      "loss": 0.6479,
+      "step": 7780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.766492917754743e-06,
+      "loss": 0.7506,
+      "step": 7781
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.765392276393092e-06,
+      "loss": 0.6898,
+      "step": 7782
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.764291646424668e-06,
+      "loss": 0.7756,
+      "step": 7783
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.763191027902916e-06,
+      "loss": 0.5746,
+      "step": 7784
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.762090420881289e-06,
+      "loss": 0.7717,
+      "step": 7785
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.760989825413236e-06,
+      "loss": 0.478,
+      "step": 7786
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7598892415522005e-06,
+      "loss": 0.8205,
+      "step": 7787
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.758788669351637e-06,
+      "loss": 0.9408,
+      "step": 7788
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7576881088649865e-06,
+      "loss": 0.7151,
+      "step": 7789
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7565875601456965e-06,
+      "loss": 0.7629,
+      "step": 7790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.755487023247216e-06,
+      "loss": 0.5757,
+      "step": 7791
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.754386498222988e-06,
+      "loss": 0.5177,
+      "step": 7792
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.753285985126455e-06,
+      "loss": 0.7233,
+      "step": 7793
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.752185484011066e-06,
+      "loss": 0.4733,
+      "step": 7794
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.75108499493026e-06,
+      "loss": 0.9361,
+      "step": 7795
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.749984517937482e-06,
+      "loss": 0.704,
+      "step": 7796
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7488840530861755e-06,
+      "loss": 0.5821,
+      "step": 7797
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.747783600429781e-06,
+      "loss": 0.4881,
+      "step": 7798
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.746683160021736e-06,
+      "loss": 0.8211,
+      "step": 7799
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.745582731915488e-06,
+      "loss": 0.7639,
+      "step": 7800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7444823161644725e-06,
+      "loss": 0.8299,
+      "step": 7801
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.74338191282213e-06,
+      "loss": 0.7841,
+      "step": 7802
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.742281521941897e-06,
+      "loss": 0.9903,
+      "step": 7803
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.741181143577216e-06,
+      "loss": 0.5305,
+      "step": 7804
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7400807777815215e-06,
+      "loss": 0.7105,
+      "step": 7805
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.73898042460825e-06,
+      "loss": 0.8379,
+      "step": 7806
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.73788008411084e-06,
+      "loss": 0.7704,
+      "step": 7807
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.736779756342727e-06,
+      "loss": 0.6338,
+      "step": 7808
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.735679441357343e-06,
+      "loss": 0.7282,
+      "step": 7809
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.734579139208127e-06,
+      "loss": 0.4132,
+      "step": 7810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.73347884994851e-06,
+      "loss": 0.3063,
+      "step": 7811
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7323785736319246e-06,
+      "loss": 0.6017,
+      "step": 7812
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.731278310311805e-06,
+      "loss": 0.8189,
+      "step": 7813
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.730178060041584e-06,
+      "loss": 0.8542,
+      "step": 7814
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.729077822874689e-06,
+      "loss": 0.6816,
+      "step": 7815
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.727977598864555e-06,
+      "loss": 0.8135,
+      "step": 7816
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.726877388064609e-06,
+      "loss": 0.8293,
+      "step": 7817
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.725777190528282e-06,
+      "loss": 0.7628,
+      "step": 7818
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.724677006309004e-06,
+      "loss": 0.7035,
+      "step": 7819
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7235768354602005e-06,
+      "loss": 0.6498,
+      "step": 7820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7224766780353005e-06,
+      "loss": 0.6888,
+      "step": 7821
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.721376534087729e-06,
+      "loss": 0.4624,
+      "step": 7822
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.720276403670914e-06,
+      "loss": 0.8428,
+      "step": 7823
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.719176286838282e-06,
+      "loss": 1.0109,
+      "step": 7824
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.718076183643252e-06,
+      "loss": 0.6857,
+      "step": 7825
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.716976094139257e-06,
+      "loss": 1.0252,
+      "step": 7826
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.715876018379714e-06,
+      "loss": 0.701,
+      "step": 7827
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.714775956418047e-06,
+      "loss": 0.6614,
+      "step": 7828
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.713675908307681e-06,
+      "loss": 0.7593,
+      "step": 7829
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.712575874102034e-06,
+      "loss": 0.7028,
+      "step": 7830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.711475853854528e-06,
+      "loss": 0.7838,
+      "step": 7831
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.710375847618585e-06,
+      "loss": 0.5806,
+      "step": 7832
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7092758554476215e-06,
+      "loss": 0.5566,
+      "step": 7833
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.708175877395057e-06,
+      "loss": 0.8787,
+      "step": 7834
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.707075913514312e-06,
+      "loss": 0.5823,
+      "step": 7835
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.705975963858803e-06,
+      "loss": 0.5254,
+      "step": 7836
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.704876028481943e-06,
+      "loss": 0.5688,
+      "step": 7837
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7037761074371524e-06,
+      "loss": 1.0122,
+      "step": 7838
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.7026762007778455e-06,
+      "loss": 0.7931,
+      "step": 7839
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.701576308557436e-06,
+      "loss": 0.705,
+      "step": 7840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.700476430829337e-06,
+      "loss": 0.5641,
+      "step": 7841
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.699376567646965e-06,
+      "loss": 0.6599,
+      "step": 7842
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.698276719063729e-06,
+      "loss": 0.8221,
+      "step": 7843
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.697176885133042e-06,
+      "loss": 0.7411,
+      "step": 7844
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.696077065908317e-06,
+      "loss": 0.8467,
+      "step": 7845
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.694977261442964e-06,
+      "loss": 0.5925,
+      "step": 7846
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.693877471790389e-06,
+      "loss": 0.422,
+      "step": 7847
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.692777697004003e-06,
+      "loss": 0.8123,
+      "step": 7848
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.691677937137217e-06,
+      "loss": 0.7163,
+      "step": 7849
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.690578192243433e-06,
+      "loss": 1.0144,
+      "step": 7850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.689478462376063e-06,
+      "loss": 0.6772,
+      "step": 7851
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.688378747588512e-06,
+      "loss": 0.6871,
+      "step": 7852
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6872790479341815e-06,
+      "loss": 0.7355,
+      "step": 7853
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.686179363466481e-06,
+      "loss": 0.7503,
+      "step": 7854
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.685079694238811e-06,
+      "loss": 0.9404,
+      "step": 7855
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.683980040304575e-06,
+      "loss": 0.8798,
+      "step": 7856
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.682880401717178e-06,
+      "loss": 0.5335,
+      "step": 7857
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.681780778530019e-06,
+      "loss": 0.6693,
+      "step": 7858
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.680681170796501e-06,
+      "loss": 0.8086,
+      "step": 7859
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.679581578570019e-06,
+      "loss": 0.7169,
+      "step": 7860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6784820019039786e-06,
+      "loss": 0.6231,
+      "step": 7861
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6773824408517765e-06,
+      "loss": 0.8853,
+      "step": 7862
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.676282895466807e-06,
+      "loss": 0.8542,
+      "step": 7863
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.675183365802473e-06,
+      "loss": 0.6948,
+      "step": 7864
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.674083851912167e-06,
+      "loss": 0.7664,
+      "step": 7865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.672984353849285e-06,
+      "loss": 0.7151,
+      "step": 7866
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.671884871667223e-06,
+      "loss": 0.7142,
+      "step": 7867
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.670785405419374e-06,
+      "loss": 0.7892,
+      "step": 7868
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.66968595515913e-06,
+      "loss": 0.6646,
+      "step": 7869
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.668586520939887e-06,
+      "loss": 0.7647,
+      "step": 7870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6674871028150335e-06,
+      "loss": 0.584,
+      "step": 7871
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.666387700837961e-06,
+      "loss": 0.624,
+      "step": 7872
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.665288315062062e-06,
+      "loss": 0.7999,
+      "step": 7873
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.664188945540724e-06,
+      "loss": 0.7139,
+      "step": 7874
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.663089592327333e-06,
+      "loss": 0.6591,
+      "step": 7875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6619902554752815e-06,
+      "loss": 0.8795,
+      "step": 7876
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.660890935037954e-06,
+      "loss": 0.672,
+      "step": 7877
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.659791631068737e-06,
+      "loss": 0.9586,
+      "step": 7878
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.658692343621014e-06,
+      "loss": 0.4618,
+      "step": 7879
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6575930727481736e-06,
+      "loss": 0.5382,
+      "step": 7880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6564938185035954e-06,
+      "loss": 0.6679,
+      "step": 7881
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6553945809406646e-06,
+      "loss": 0.8692,
+      "step": 7882
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6542953601127635e-06,
+      "loss": 0.9071,
+      "step": 7883
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.653196156073274e-06,
+      "loss": 0.8911,
+      "step": 7884
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.652096968875574e-06,
+      "loss": 0.5366,
+      "step": 7885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.650997798573044e-06,
+      "loss": 0.6189,
+      "step": 7886
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.649898645219065e-06,
+      "loss": 0.6881,
+      "step": 7887
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.648799508867012e-06,
+      "loss": 0.6708,
+      "step": 7888
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.647700389570264e-06,
+      "loss": 0.9333,
+      "step": 7889
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6466012873822e-06,
+      "loss": 0.7601,
+      "step": 7890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.645502202356189e-06,
+      "loss": 1.1316,
+      "step": 7891
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.644403134545612e-06,
+      "loss": 0.7113,
+      "step": 7892
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.643304084003839e-06,
+      "loss": 0.7588,
+      "step": 7893
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6422050507842455e-06,
+      "loss": 0.7141,
+      "step": 7894
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6411060349402015e-06,
+      "loss": 0.6742,
+      "step": 7895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.640007036525079e-06,
+      "loss": 0.9858,
+      "step": 7896
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.638908055592252e-06,
+      "loss": 0.5708,
+      "step": 7897
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.637809092195084e-06,
+      "loss": 0.7866,
+      "step": 7898
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.636710146386948e-06,
+      "loss": 0.592,
+      "step": 7899
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.635611218221212e-06,
+      "loss": 0.6188,
+      "step": 7900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6345123077512396e-06,
+      "loss": 0.6728,
+      "step": 7901
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.633413415030401e-06,
+      "loss": 0.8947,
+      "step": 7902
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.63231454011206e-06,
+      "loss": 0.8019,
+      "step": 7903
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.631215683049581e-06,
+      "loss": 0.6272,
+      "step": 7904
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.630116843896329e-06,
+      "loss": 0.5817,
+      "step": 7905
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6290180227056644e-06,
+      "loss": 0.7122,
+      "step": 7906
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6279192195309505e-06,
+      "loss": 0.9629,
+      "step": 7907
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.626820434425549e-06,
+      "loss": 0.7071,
+      "step": 7908
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.625721667442819e-06,
+      "loss": 0.7787,
+      "step": 7909
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6246229186361184e-06,
+      "loss": 0.6923,
+      "step": 7910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.62352418805881e-06,
+      "loss": 0.5382,
+      "step": 7911
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.622425475764248e-06,
+      "loss": 1.0761,
+      "step": 7912
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.62132678180579e-06,
+      "loss": 0.9715,
+      "step": 7913
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.620228106236788e-06,
+      "loss": 0.6233,
+      "step": 7914
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.619129449110604e-06,
+      "loss": 0.7964,
+      "step": 7915
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.618030810480586e-06,
+      "loss": 0.8006,
+      "step": 7916
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.616932190400089e-06,
+      "loss": 0.8822,
+      "step": 7917
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.615833588922467e-06,
+      "loss": 1.1029,
+      "step": 7918
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.614735006101069e-06,
+      "loss": 0.9714,
+      "step": 7919
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.613636441989244e-06,
+      "loss": 0.5644,
+      "step": 7920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.6125378966403465e-06,
+      "loss": 0.6385,
+      "step": 7921
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.611439370107721e-06,
+      "loss": 0.8357,
+      "step": 7922
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.610340862444713e-06,
+      "loss": 0.7261,
+      "step": 7923
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.609242373704675e-06,
+      "loss": 0.7066,
+      "step": 7924
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.60814390394095e-06,
+      "loss": 0.7612,
+      "step": 7925
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.607045453206881e-06,
+      "loss": 0.727,
+      "step": 7926
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.605947021555814e-06,
+      "loss": 0.7918,
+      "step": 7927
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.604848609041094e-06,
+      "loss": 0.7332,
+      "step": 7928
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.603750215716057e-06,
+      "loss": 0.976,
+      "step": 7929
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.602651841634052e-06,
+      "loss": 0.6976,
+      "step": 7930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.601553486848412e-06,
+      "loss": 0.5148,
+      "step": 7931
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.600455151412482e-06,
+      "loss": 0.5754,
+      "step": 7932
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.599356835379595e-06,
+      "loss": 0.6284,
+      "step": 7933
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.598258538803092e-06,
+      "loss": 0.8076,
+      "step": 7934
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.597160261736309e-06,
+      "loss": 0.6897,
+      "step": 7935
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.59606200423258e-06,
+      "loss": 0.7413,
+      "step": 7936
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.594963766345241e-06,
+      "loss": 0.8256,
+      "step": 7937
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.593865548127626e-06,
+      "loss": 0.768,
+      "step": 7938
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.592767349633065e-06,
+      "loss": 0.7969,
+      "step": 7939
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.591669170914893e-06,
+      "loss": 0.7666,
+      "step": 7940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5905710120264385e-06,
+      "loss": 0.5082,
+      "step": 7941
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.58947287302103e-06,
+      "loss": 0.6419,
+      "step": 7942
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.588374753952001e-06,
+      "loss": 0.7599,
+      "step": 7943
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5872766548726755e-06,
+      "loss": 0.717,
+      "step": 7944
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.58617857583638e-06,
+      "loss": 0.581,
+      "step": 7945
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.585080516896444e-06,
+      "loss": 0.7841,
+      "step": 7946
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.583982478106189e-06,
+      "loss": 0.7612,
+      "step": 7947
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.582884459518939e-06,
+      "loss": 0.8232,
+      "step": 7948
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.58178646118802e-06,
+      "loss": 0.7834,
+      "step": 7949
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.580688483166752e-06,
+      "loss": 0.7742,
+      "step": 7950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.579590525508455e-06,
+      "loss": 0.6675,
+      "step": 7951
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.578492588266448e-06,
+      "loss": 0.6358,
+      "step": 7952
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.577394671494054e-06,
+      "loss": 0.5541,
+      "step": 7953
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.576296775244589e-06,
+      "loss": 0.8073,
+      "step": 7954
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.575198899571368e-06,
+      "loss": 0.8818,
+      "step": 7955
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.57410104452771e-06,
+      "loss": 0.7834,
+      "step": 7956
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.573003210166929e-06,
+      "loss": 0.7309,
+      "step": 7957
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.571905396542337e-06,
+      "loss": 0.9873,
+      "step": 7958
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.570807603707251e-06,
+      "loss": 0.7218,
+      "step": 7959
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.569709831714982e-06,
+      "loss": 0.604,
+      "step": 7960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.568612080618836e-06,
+      "loss": 0.5829,
+      "step": 7961
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.567514350472128e-06,
+      "loss": 0.7411,
+      "step": 7962
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.566416641328167e-06,
+      "loss": 0.7161,
+      "step": 7963
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.565318953240257e-06,
+      "loss": 0.5709,
+      "step": 7964
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.564221286261709e-06,
+      "loss": 0.7956,
+      "step": 7965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.563123640445829e-06,
+      "loss": 0.9137,
+      "step": 7966
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.562026015845916e-06,
+      "loss": 0.8675,
+      "step": 7967
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5609284125152815e-06,
+      "loss": 0.5331,
+      "step": 7968
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.559830830507223e-06,
+      "loss": 0.733,
+      "step": 7969
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5587332698750455e-06,
+      "loss": 0.8431,
+      "step": 7970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.557635730672045e-06,
+      "loss": 0.6833,
+      "step": 7971
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.556538212951526e-06,
+      "loss": 0.8594,
+      "step": 7972
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.555440716766787e-06,
+      "loss": 0.7938,
+      "step": 7973
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.554343242171121e-06,
+      "loss": 0.3685,
+      "step": 7974
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.553245789217828e-06,
+      "loss": 0.811,
+      "step": 7975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.552148357960205e-06,
+      "loss": 0.6782,
+      "step": 7976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.551050948451542e-06,
+      "loss": 0.6204,
+      "step": 7977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.549953560745136e-06,
+      "loss": 0.6935,
+      "step": 7978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.548856194894279e-06,
+      "loss": 0.5547,
+      "step": 7979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.547758850952258e-06,
+      "loss": 0.544,
+      "step": 7980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5466615289723695e-06,
+      "loss": 0.8726,
+      "step": 7981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5455642290078985e-06,
+      "loss": 0.7575,
+      "step": 7982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.544466951112132e-06,
+      "loss": 1.1031,
+      "step": 7983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.543369695338362e-06,
+      "loss": 0.6875,
+      "step": 7984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.54227246173987e-06,
+      "loss": 0.8895,
+      "step": 7985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.541175250369941e-06,
+      "loss": 0.7485,
+      "step": 7986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.540078061281863e-06,
+      "loss": 0.6112,
+      "step": 7987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.538980894528916e-06,
+      "loss": 0.857,
+      "step": 7988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5378837501643785e-06,
+      "loss": 0.6576,
+      "step": 7989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.536786628241534e-06,
+      "loss": 0.7313,
+      "step": 7990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.535689528813662e-06,
+      "loss": 0.7947,
+      "step": 7991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.534592451934041e-06,
+      "loss": 0.8377,
+      "step": 7992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.533495397655946e-06,
+      "loss": 0.543,
+      "step": 7993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.532398366032656e-06,
+      "loss": 0.6612,
+      "step": 7994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.531301357117444e-06,
+      "loss": 0.872,
+      "step": 7995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.530204370963585e-06,
+      "loss": 0.7517,
+      "step": 7996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.529107407624351e-06,
+      "loss": 0.5681,
+      "step": 7997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.528010467153015e-06,
+      "loss": 0.8742,
+      "step": 7998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.526913549602846e-06,
+      "loss": 0.6216,
+      "step": 7999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.525816655027113e-06,
+      "loss": 0.7259,
+      "step": 8000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.524719783479088e-06,
+      "loss": 1.001,
+      "step": 8001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.523622935012033e-06,
+      "loss": 0.8014,
+      "step": 8002
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.522526109679218e-06,
+      "loss": 0.5951,
+      "step": 8003
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.521429307533908e-06,
+      "loss": 0.55,
+      "step": 8004
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.520332528629363e-06,
+      "loss": 0.8858,
+      "step": 8005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.5192357730188515e-06,
+      "loss": 0.9011,
+      "step": 8006
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.518139040755631e-06,
+      "loss": 0.8756,
+      "step": 8007
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.517042331892963e-06,
+      "loss": 0.8417,
+      "step": 8008
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.515945646484105e-06,
+      "loss": 0.6066,
+      "step": 8009
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.514848984582318e-06,
+      "loss": 0.8118,
+      "step": 8010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.51375234624086e-06,
+      "loss": 0.5704,
+      "step": 8011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.512655731512982e-06,
+      "loss": 0.7605,
+      "step": 8012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5115591404519435e-06,
+      "loss": 0.5287,
+      "step": 8013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.510462573110996e-06,
+      "loss": 0.8086,
+      "step": 8014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.509366029543392e-06,
+      "loss": 0.7556,
+      "step": 8015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.508269509802384e-06,
+      "loss": 0.8013,
+      "step": 8016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50717301394122e-06,
+      "loss": 0.4404,
+      "step": 8017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50607654201315e-06,
+      "loss": 0.7247,
+      "step": 8018
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.504980094071424e-06,
+      "loss": 0.6066,
+      "step": 8019
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.503883670169286e-06,
+      "loss": 0.5811,
+      "step": 8020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.50278727035998e-06,
+      "loss": 0.5996,
+      "step": 8021
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.5016908946967556e-06,
+      "loss": 0.8029,
+      "step": 8022
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.500594543232851e-06,
+      "loss": 0.5897,
+      "step": 8023
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.499498216021512e-06,
+      "loss": 0.6917,
+      "step": 8024
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.498401913115975e-06,
+      "loss": 0.8775,
+      "step": 8025
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.497305634569485e-06,
+      "loss": 0.6647,
+      "step": 8026
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.496209380435276e-06,
+      "loss": 0.7706,
+      "step": 8027
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.495113150766585e-06,
+      "loss": 0.7331,
+      "step": 8028
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.494016945616652e-06,
+      "loss": 0.9436,
+      "step": 8029
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.49292076503871e-06,
+      "loss": 0.5794,
+      "step": 8030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4918246090859905e-06,
+      "loss": 0.6768,
+      "step": 8031
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.49072847781173e-06,
+      "loss": 0.9358,
+      "step": 8032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4896323712691566e-06,
+      "loss": 0.6847,
+      "step": 8033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.488536289511501e-06,
+      "loss": 0.6207,
+      "step": 8034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.487440232591994e-06,
+      "loss": 0.8042,
+      "step": 8035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.486344200563861e-06,
+      "loss": 0.7169,
+      "step": 8036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.485248193480328e-06,
+      "loss": 0.8593,
+      "step": 8037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.484152211394623e-06,
+      "loss": 0.6303,
+      "step": 8038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.483056254359969e-06,
+      "loss": 0.947,
+      "step": 8039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.481960322429587e-06,
+      "loss": 0.6888,
+      "step": 8040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4808644156567e-06,
+      "loss": 0.8275,
+      "step": 8041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.47976853409453e-06,
+      "loss": 0.7074,
+      "step": 8042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.478672677796294e-06,
+      "loss": 0.6537,
+      "step": 8043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.477576846815209e-06,
+      "loss": 0.7692,
+      "step": 8044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.476481041204494e-06,
+      "loss": 0.8445,
+      "step": 8045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.475385261017364e-06,
+      "loss": 0.8501,
+      "step": 8046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.474289506307032e-06,
+      "loss": 0.7227,
+      "step": 8047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.473193777126712e-06,
+      "loss": 0.7839,
+      "step": 8048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.472098073529617e-06,
+      "loss": 0.7107,
+      "step": 8049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.471002395568954e-06,
+      "loss": 0.9943,
+      "step": 8050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.469906743297935e-06,
+      "loss": 0.9012,
+      "step": 8051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.468811116769769e-06,
+      "loss": 0.8731,
+      "step": 8052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.467715516037659e-06,
+      "loss": 0.8235,
+      "step": 8053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.466619941154815e-06,
+      "loss": 0.6546,
+      "step": 8054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.465524392174437e-06,
+      "loss": 0.7657,
+      "step": 8055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.46442886914973e-06,
+      "loss": 0.6899,
+      "step": 8056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.463333372133897e-06,
+      "loss": 0.737,
+      "step": 8057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.462237901180136e-06,
+      "loss": 0.9518,
+      "step": 8058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4611424563416476e-06,
+      "loss": 0.7084,
+      "step": 8059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.46004703767163e-06,
+      "loss": 0.7838,
+      "step": 8060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.458951645223279e-06,
+      "loss": 0.6436,
+      "step": 8061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4578562790497924e-06,
+      "loss": 0.8845,
+      "step": 8062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4567609392043595e-06,
+      "loss": 0.5937,
+      "step": 8063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4556656257401786e-06,
+      "loss": 0.8281,
+      "step": 8064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.454570338710437e-06,
+      "loss": 0.5864,
+      "step": 8065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.453475078168326e-06,
+      "loss": 0.6505,
+      "step": 8066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4523798441670375e-06,
+      "loss": 0.6344,
+      "step": 8067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.451284636759755e-06,
+      "loss": 0.7262,
+      "step": 8068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.450189455999667e-06,
+      "loss": 0.7294,
+      "step": 8069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.44909430193996e-06,
+      "loss": 0.8534,
+      "step": 8070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.447999174633817e-06,
+      "loss": 0.8191,
+      "step": 8071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.446904074134417e-06,
+      "loss": 0.9817,
+      "step": 8072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.445809000494945e-06,
+      "loss": 0.9231,
+      "step": 8073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.444713953768582e-06,
+      "loss": 0.8422,
+      "step": 8074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.443618934008502e-06,
+      "loss": 0.8099,
+      "step": 8075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.442523941267886e-06,
+      "loss": 0.6182,
+      "step": 8076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.44142897559991e-06,
+      "loss": 0.9251,
+      "step": 8077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.440334037057746e-06,
+      "loss": 0.5688,
+      "step": 8078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.43923912569457e-06,
+      "loss": 0.5342,
+      "step": 8079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.438144241563554e-06,
+      "loss": 0.479,
+      "step": 8080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.437049384717867e-06,
+      "loss": 0.6317,
+      "step": 8081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.435954555210677e-06,
+      "loss": 0.556,
+      "step": 8082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.434859753095155e-06,
+      "loss": 0.6764,
+      "step": 8083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.43376497842447e-06,
+      "loss": 0.5896,
+      "step": 8084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.43267023125178e-06,
+      "loss": 0.6973,
+      "step": 8085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.431575511630256e-06,
+      "loss": 0.5875,
+      "step": 8086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.430480819613058e-06,
+      "loss": 0.5821,
+      "step": 8087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.429386155253346e-06,
+      "loss": 0.8577,
+      "step": 8088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.428291518604283e-06,
+      "loss": 0.8629,
+      "step": 8089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.427196909719026e-06,
+      "loss": 0.6518,
+      "step": 8090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.426102328650732e-06,
+      "loss": 0.7049,
+      "step": 8091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4250077754525595e-06,
+      "loss": 0.6809,
+      "step": 8092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.42391325017766e-06,
+      "loss": 0.5934,
+      "step": 8093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.422818752879187e-06,
+      "loss": 0.8024,
+      "step": 8094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.421724283610295e-06,
+      "loss": 0.7429,
+      "step": 8095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.420629842424133e-06,
+      "loss": 0.5429,
+      "step": 8096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4195354293738484e-06,
+      "loss": 0.7704,
+      "step": 8097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.418441044512594e-06,
+      "loss": 0.8414,
+      "step": 8098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.417346687893511e-06,
+      "loss": 0.7209,
+      "step": 8099
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.416252359569748e-06,
+      "loss": 0.7426,
+      "step": 8100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.415158059594445e-06,
+      "loss": 0.5935,
+      "step": 8101
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.414063788020751e-06,
+      "loss": 0.4793,
+      "step": 8102
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.412969544901799e-06,
+      "loss": 0.6535,
+      "step": 8103
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.411875330290732e-06,
+      "loss": 0.7909,
+      "step": 8104
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.410781144240692e-06,
+      "loss": 0.9743,
+      "step": 8105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.40968698680481e-06,
+      "loss": 1.0962,
+      "step": 8106
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.408592858036223e-06,
+      "loss": 0.6132,
+      "step": 8107
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.407498757988068e-06,
+      "loss": 0.7343,
+      "step": 8108
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4064046867134755e-06,
+      "loss": 0.7398,
+      "step": 8109
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.4053106442655745e-06,
+      "loss": 0.7235,
+      "step": 8110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.404216630697499e-06,
+      "loss": 0.7641,
+      "step": 8111
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.403122646062375e-06,
+      "loss": 0.8935,
+      "step": 8112
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.402028690413329e-06,
+      "loss": 0.492,
+      "step": 8113
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.400934763803488e-06,
+      "loss": 0.8273,
+      "step": 8114
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.399840866285977e-06,
+      "loss": 0.6973,
+      "step": 8115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.398746997913916e-06,
+      "loss": 0.6816,
+      "step": 8116
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.39765315874043e-06,
+      "loss": 0.7536,
+      "step": 8117
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.396559348818636e-06,
+      "loss": 0.4386,
+      "step": 8118
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.395465568201655e-06,
+      "loss": 0.7375,
+      "step": 8119
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3943718169426e-06,
+      "loss": 0.7481,
+      "step": 8120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.393278095094591e-06,
+      "loss": 0.39,
+      "step": 8121
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.392184402710742e-06,
+      "loss": 0.738,
+      "step": 8122
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.391090739844161e-06,
+      "loss": 0.7727,
+      "step": 8123
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.389997106547966e-06,
+      "loss": 0.5137,
+      "step": 8124
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3889035028752635e-06,
+      "loss": 0.8415,
+      "step": 8125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.387809928879161e-06,
+      "loss": 0.5146,
+      "step": 8126
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.386716384612768e-06,
+      "loss": 0.5205,
+      "step": 8127
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3856228701291905e-06,
+      "loss": 0.7646,
+      "step": 8128
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.384529385481528e-06,
+      "loss": 0.7378,
+      "step": 8129
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.38343593072289e-06,
+      "loss": 0.9407,
+      "step": 8130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.382342505906374e-06,
+      "loss": 0.808,
+      "step": 8131
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.381249111085078e-06,
+      "loss": 0.6088,
+      "step": 8132
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.380155746312106e-06,
+      "loss": 1.0463,
+      "step": 8133
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3790624116405494e-06,
+      "loss": 0.7317,
+      "step": 8134
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.377969107123505e-06,
+      "loss": 0.8812,
+      "step": 8135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.376875832814071e-06,
+      "loss": 0.9381,
+      "step": 8136
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3757825887653345e-06,
+      "loss": 1.018,
+      "step": 8137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.374689375030391e-06,
+      "loss": 0.5483,
+      "step": 8138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.373596191662324e-06,
+      "loss": 0.7509,
+      "step": 8139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3725030387142295e-06,
+      "loss": 0.6475,
+      "step": 8140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.371409916239188e-06,
+      "loss": 0.6979,
+      "step": 8141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.370316824290285e-06,
+      "loss": 0.7221,
+      "step": 8142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.369223762920609e-06,
+      "loss": 0.7301,
+      "step": 8143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3681307321832385e-06,
+      "loss": 0.71,
+      "step": 8144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.367037732131254e-06,
+      "loss": 0.7102,
+      "step": 8145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.365944762817736e-06,
+      "loss": 0.8291,
+      "step": 8146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.364851824295764e-06,
+      "loss": 0.8486,
+      "step": 8147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.363758916618409e-06,
+      "loss": 0.6699,
+      "step": 8148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.36266603983875e-06,
+      "loss": 0.9399,
+      "step": 8149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.361573194009859e-06,
+      "loss": 0.8851,
+      "step": 8150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.360480379184807e-06,
+      "loss": 0.7672,
+      "step": 8151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.359387595416665e-06,
+      "loss": 0.9476,
+      "step": 8152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.358294842758504e-06,
+      "loss": 0.5784,
+      "step": 8153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3572021212633865e-06,
+      "loss": 0.696,
+      "step": 8154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.356109430984381e-06,
+      "loss": 0.4672,
+      "step": 8155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.355016771974553e-06,
+      "loss": 0.6684,
+      "step": 8156
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.353924144286963e-06,
+      "loss": 0.7524,
+      "step": 8157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.352831547974671e-06,
+      "loss": 1.0172,
+      "step": 8158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.351738983090739e-06,
+      "loss": 0.594,
+      "step": 8159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.350646449688226e-06,
+      "loss": 0.6577,
+      "step": 8160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3495539478201834e-06,
+      "loss": 0.7814,
+      "step": 8161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.348461477539672e-06,
+      "loss": 0.8141,
+      "step": 8162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.347369038899744e-06,
+      "loss": 0.7463,
+      "step": 8163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.346276631953447e-06,
+      "loss": 0.8488,
+      "step": 8164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.345184256753837e-06,
+      "loss": 0.6454,
+      "step": 8165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.344091913353962e-06,
+      "loss": 0.5826,
+      "step": 8166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.342999601806865e-06,
+      "loss": 0.6088,
+      "step": 8167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3419073221655975e-06,
+      "loss": 0.6827,
+      "step": 8168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.340815074483199e-06,
+      "loss": 0.742,
+      "step": 8169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3397228588127136e-06,
+      "loss": 0.904,
+      "step": 8170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.338630675207186e-06,
+      "loss": 0.6907,
+      "step": 8171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.337538523719651e-06,
+      "loss": 0.6858,
+      "step": 8172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3364464044031494e-06,
+      "loss": 0.5337,
+      "step": 8173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.335354317310715e-06,
+      "loss": 0.9552,
+      "step": 8174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.334262262495385e-06,
+      "loss": 0.7115,
+      "step": 8175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.333170240010193e-06,
+      "loss": 0.8428,
+      "step": 8176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.332078249908168e-06,
+      "loss": 0.5807,
+      "step": 8177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.330986292242344e-06,
+      "loss": 0.8161,
+      "step": 8178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.329894367065747e-06,
+      "loss": 0.81,
+      "step": 8179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3288024744314034e-06,
+      "loss": 0.5318,
+      "step": 8180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.327710614392341e-06,
+      "loss": 0.6997,
+      "step": 8181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.326618787001583e-06,
+      "loss": 0.7506,
+      "step": 8182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3255269923121494e-06,
+      "loss": 0.7743,
+      "step": 8183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.324435230377065e-06,
+      "loss": 0.8733,
+      "step": 8184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.323343501249346e-06,
+      "loss": 0.6495,
+      "step": 8185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.322251804982009e-06,
+      "loss": 0.6702,
+      "step": 8186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.321160141628073e-06,
+      "loss": 0.6713,
+      "step": 8187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.320068511240551e-06,
+      "loss": 0.5384,
+      "step": 8188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3189769138724525e-06,
+      "loss": 0.9877,
+      "step": 8189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.317885349576794e-06,
+      "loss": 0.5514,
+      "step": 8190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.316793818406583e-06,
+      "loss": 0.7665,
+      "step": 8191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.315702320414825e-06,
+      "loss": 0.7473,
+      "step": 8192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.314610855654528e-06,
+      "loss": 0.7527,
+      "step": 8193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.313519424178699e-06,
+      "loss": 0.6067,
+      "step": 8194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.312428026040339e-06,
+      "loss": 0.7271,
+      "step": 8195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.311336661292447e-06,
+      "loss": 0.7075,
+      "step": 8196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.310245329988026e-06,
+      "loss": 0.5588,
+      "step": 8197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.309154032180074e-06,
+      "loss": 1.0875,
+      "step": 8198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.308062767921586e-06,
+      "loss": 0.9302,
+      "step": 8199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.306971537265558e-06,
+      "loss": 0.7907,
+      "step": 8200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.305880340264985e-06,
+      "loss": 0.7306,
+      "step": 8201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.304789176972853e-06,
+      "loss": 0.4973,
+      "step": 8202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.303698047442158e-06,
+      "loss": 0.6516,
+      "step": 8203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.302606951725887e-06,
+      "loss": 0.5569,
+      "step": 8204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3015158898770244e-06,
+      "loss": 0.7039,
+      "step": 8205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.300424861948558e-06,
+      "loss": 0.605,
+      "step": 8206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2993338679934695e-06,
+      "loss": 0.7503,
+      "step": 8207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.29824290806474e-06,
+      "loss": 0.9439,
+      "step": 8208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.297151982215353e-06,
+      "loss": 0.9811,
+      "step": 8209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.296061090498284e-06,
+      "loss": 0.6514,
+      "step": 8210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.294970232966512e-06,
+      "loss": 0.8832,
+      "step": 8211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.293879409673008e-06,
+      "loss": 0.7055,
+      "step": 8212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.292788620670751e-06,
+      "loss": 0.5191,
+      "step": 8213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.291697866012709e-06,
+      "loss": 0.5995,
+      "step": 8214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.290607145751854e-06,
+      "loss": 0.8125,
+      "step": 8215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.289516459941153e-06,
+      "loss": 0.7214,
+      "step": 8216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2884258086335755e-06,
+      "loss": 0.7221,
+      "step": 8217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.287335191882083e-06,
+      "loss": 0.6101,
+      "step": 8218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286244609739643e-06,
+      "loss": 0.761,
+      "step": 8219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.285154062259214e-06,
+      "loss": 0.6818,
+      "step": 8220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.284063549493756e-06,
+      "loss": 0.6853,
+      "step": 8221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2829730714962314e-06,
+      "loss": 0.7978,
+      "step": 8222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.281882628319594e-06,
+      "loss": 0.6564,
+      "step": 8223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.280792220016797e-06,
+      "loss": 0.7803,
+      "step": 8224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.279701846640798e-06,
+      "loss": 0.7645,
+      "step": 8225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.278611508244547e-06,
+      "loss": 0.5311,
+      "step": 8226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.277521204880991e-06,
+      "loss": 0.6442,
+      "step": 8227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.276430936603082e-06,
+      "loss": 0.6524,
+      "step": 8228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.275340703463767e-06,
+      "loss": 0.4217,
+      "step": 8229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.274250505515988e-06,
+      "loss": 0.6696,
+      "step": 8230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.273160342812689e-06,
+      "loss": 0.9419,
+      "step": 8231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.272070215406813e-06,
+      "loss": 0.7866,
+      "step": 8232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.270980123351299e-06,
+      "loss": 0.7573,
+      "step": 8233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.269890066699082e-06,
+      "loss": 0.6289,
+      "step": 8234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.268800045503102e-06,
+      "loss": 0.6473,
+      "step": 8235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.267710059816294e-06,
+      "loss": 0.7674,
+      "step": 8236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.266620109691586e-06,
+      "loss": 0.9146,
+      "step": 8237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.265530195181915e-06,
+      "loss": 0.7311,
+      "step": 8238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.264440316340208e-06,
+      "loss": 0.8836,
+      "step": 8239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.26335047321939e-06,
+      "loss": 0.824,
+      "step": 8240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.26226066587239e-06,
+      "loss": 0.6187,
+      "step": 8241
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.261170894352133e-06,
+      "loss": 0.7418,
+      "step": 8242
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.260081158711538e-06,
+      "loss": 0.86,
+      "step": 8243
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.258991459003529e-06,
+      "loss": 0.7286,
+      "step": 8244
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.257901795281024e-06,
+      "loss": 0.7065,
+      "step": 8245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.256812167596937e-06,
+      "loss": 0.6147,
+      "step": 8246
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.255722576004189e-06,
+      "loss": 0.7234,
+      "step": 8247
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.25463302055569e-06,
+      "loss": 0.7438,
+      "step": 8248
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2535435013043535e-06,
+      "loss": 0.9295,
+      "step": 8249
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.252454018303087e-06,
+      "loss": 0.8376,
+      "step": 8250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251364571604803e-06,
+      "loss": 0.5434,
+      "step": 8251
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2502751612624055e-06,
+      "loss": 0.769,
+      "step": 8252
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.249185787328798e-06,
+      "loss": 0.6656,
+      "step": 8253
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.248096449856889e-06,
+      "loss": 0.9062,
+      "step": 8254
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.247007148899575e-06,
+      "loss": 0.5854,
+      "step": 8255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.245917884509756e-06,
+      "loss": 0.556,
+      "step": 8256
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2448286567403315e-06,
+      "loss": 0.6182,
+      "step": 8257
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.243739465644197e-06,
+      "loss": 0.7888,
+      "step": 8258
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.242650311274246e-06,
+      "loss": 0.9456,
+      "step": 8259
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.241561193683373e-06,
+      "loss": 0.9127,
+      "step": 8260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.240472112924469e-06,
+      "loss": 0.5798,
+      "step": 8261
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239383069050417e-06,
+      "loss": 0.8681,
+      "step": 8262
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.23829406211411e-06,
+      "loss": 0.7226,
+      "step": 8263
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.237205092168433e-06,
+      "loss": 0.7229,
+      "step": 8264
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2361161592662655e-06,
+      "loss": 0.8957,
+      "step": 8265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2350272634604924e-06,
+      "loss": 0.752,
+      "step": 8266
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.233938404803995e-06,
+      "loss": 0.4675,
+      "step": 8267
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.232849583349649e-06,
+      "loss": 0.869,
+      "step": 8268
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2317607991503285e-06,
+      "loss": 0.557,
+      "step": 8269
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.230672052258913e-06,
+      "loss": 0.98,
+      "step": 8270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.229583342728273e-06,
+      "loss": 0.597,
+      "step": 8271
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.228494670611278e-06,
+      "loss": 0.9621,
+      "step": 8272
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.227406035960798e-06,
+      "loss": 0.5899,
+      "step": 8273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.226317438829702e-06,
+      "loss": 0.5995,
+      "step": 8274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2252288792708515e-06,
+      "loss": 0.7105,
+      "step": 8275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.224140357337114e-06,
+      "loss": 0.8118,
+      "step": 8276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2230518730813495e-06,
+      "loss": 0.679,
+      "step": 8277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.221963426556417e-06,
+      "loss": 0.7836,
+      "step": 8278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.220875017815177e-06,
+      "loss": 0.7547,
+      "step": 8279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.219786646910484e-06,
+      "loss": 0.8886,
+      "step": 8280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.218698313895192e-06,
+      "loss": 0.9381,
+      "step": 8281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.217610018822156e-06,
+      "loss": 0.7217,
+      "step": 8282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.216521761744224e-06,
+      "loss": 0.6825,
+      "step": 8283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.215433542714248e-06,
+      "loss": 0.893,
+      "step": 8284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21434536178507e-06,
+      "loss": 0.733,
+      "step": 8285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21325721900954e-06,
+      "loss": 0.75,
+      "step": 8286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2121691144405e-06,
+      "loss": 0.5601,
+      "step": 8287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21108104813079e-06,
+      "loss": 0.7784,
+      "step": 8288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.209993020133251e-06,
+      "loss": 0.9692,
+      "step": 8289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.208905030500722e-06,
+      "loss": 0.7906,
+      "step": 8290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.207817079286034e-06,
+      "loss": 0.7457,
+      "step": 8291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.206729166542028e-06,
+      "loss": 0.8662,
+      "step": 8292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.205641292321531e-06,
+      "loss": 0.8072,
+      "step": 8293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.204553456677374e-06,
+      "loss": 0.5857,
+      "step": 8294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.203465659662388e-06,
+      "loss": 0.6412,
+      "step": 8295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.202377901329397e-06,
+      "loss": 0.6039,
+      "step": 8296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2012901817312255e-06,
+      "loss": 0.5853,
+      "step": 8297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2002025009207e-06,
+      "loss": 0.7423,
+      "step": 8298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.199114858950638e-06,
+      "loss": 0.6427,
+      "step": 8299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.198027255873857e-06,
+      "loss": 0.9514,
+      "step": 8300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.196939691743178e-06,
+      "loss": 0.5842,
+      "step": 8301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.195852166611416e-06,
+      "loss": 0.7297,
+      "step": 8302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.194764680531381e-06,
+      "loss": 0.7088,
+      "step": 8303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.193677233555886e-06,
+      "loss": 0.983,
+      "step": 8304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.192589825737742e-06,
+      "loss": 0.6839,
+      "step": 8305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.191502457129756e-06,
+      "loss": 0.7454,
+      "step": 8306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.190415127784731e-06,
+      "loss": 1.1431,
+      "step": 8307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.189327837755475e-06,
+      "loss": 0.6876,
+      "step": 8308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.18824058709479e-06,
+      "loss": 0.7879,
+      "step": 8309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.187153375855471e-06,
+      "loss": 0.7745,
+      "step": 8310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.18606620409032e-06,
+      "loss": 0.4125,
+      "step": 8311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.184979071852134e-06,
+      "loss": 0.6152,
+      "step": 8312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.183891979193703e-06,
+      "loss": 0.656,
+      "step": 8313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.182804926167824e-06,
+      "loss": 0.8333,
+      "step": 8314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.181717912827286e-06,
+      "loss": 0.7626,
+      "step": 8315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1806309392248745e-06,
+      "loss": 0.5058,
+      "step": 8316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.17954400541338e-06,
+      "loss": 0.791,
+      "step": 8317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.178457111445587e-06,
+      "loss": 0.7966,
+      "step": 8318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.177370257374273e-06,
+      "loss": 0.926,
+      "step": 8319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.176283443252227e-06,
+      "loss": 0.6531,
+      "step": 8320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.175196669132221e-06,
+      "loss": 0.627,
+      "step": 8321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.174109935067036e-06,
+      "loss": 0.7219,
+      "step": 8322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.173023241109442e-06,
+      "loss": 0.5824,
+      "step": 8323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.171936587312216e-06,
+      "loss": 0.5861,
+      "step": 8324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1708499737281305e-06,
+      "loss": 0.6725,
+      "step": 8325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.16976340040995e-06,
+      "loss": 0.7743,
+      "step": 8326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.168676867410445e-06,
+      "loss": 0.5196,
+      "step": 8327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1675903747823795e-06,
+      "loss": 0.738,
+      "step": 8328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.166503922578516e-06,
+      "loss": 0.5058,
+      "step": 8329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.165417510851618e-06,
+      "loss": 0.8041,
+      "step": 8330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.164331139654442e-06,
+      "loss": 0.8881,
+      "step": 8331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.163244809039746e-06,
+      "loss": 0.666,
+      "step": 8332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.162158519060287e-06,
+      "loss": 0.8172,
+      "step": 8333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.161072269768818e-06,
+      "loss": 0.6095,
+      "step": 8334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.159986061218087e-06,
+      "loss": 0.7157,
+      "step": 8335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.158899893460849e-06,
+      "loss": 0.6439,
+      "step": 8336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1578137665498485e-06,
+      "loss": 1.042,
+      "step": 8337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.156727680537829e-06,
+      "loss": 0.6432,
+      "step": 8338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.155641635477536e-06,
+      "loss": 0.8061,
+      "step": 8339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.154555631421713e-06,
+      "loss": 0.5896,
+      "step": 8340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.153469668423096e-06,
+      "loss": 0.5437,
+      "step": 8341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.152383746534422e-06,
+      "loss": 0.6433,
+      "step": 8342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.151297865808432e-06,
+      "loss": 0.9019,
+      "step": 8343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.150212026297853e-06,
+      "loss": 0.6817,
+      "step": 8344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.149126228055419e-06,
+      "loss": 0.6659,
+      "step": 8345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.148040471133862e-06,
+      "loss": 0.6256,
+      "step": 8346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.146954755585908e-06,
+      "loss": 0.778,
+      "step": 8347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.145869081464278e-06,
+      "loss": 0.8176,
+      "step": 8348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1447834488217005e-06,
+      "loss": 0.5439,
+      "step": 8349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.143697857710898e-06,
+      "loss": 0.7561,
+      "step": 8350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.142612308184584e-06,
+      "loss": 0.7437,
+      "step": 8351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.141526800295481e-06,
+      "loss": 0.449,
+      "step": 8352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.140441334096304e-06,
+      "loss": 0.6051,
+      "step": 8353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.139355909639763e-06,
+      "loss": 0.9096,
+      "step": 8354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.138270526978573e-06,
+      "loss": 0.5267,
+      "step": 8355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.137185186165442e-06,
+      "loss": 0.9226,
+      "step": 8356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.136099887253076e-06,
+      "loss": 0.5913,
+      "step": 8357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1350146302941815e-06,
+      "loss": 0.766,
+      "step": 8358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.133929415341462e-06,
+      "loss": 0.6397,
+      "step": 8359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.132844242447619e-06,
+      "loss": 1.0979,
+      "step": 8360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.131759111665349e-06,
+      "loss": 0.6906,
+      "step": 8361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.130674023047351e-06,
+      "loss": 0.7748,
+      "step": 8362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.129588976646322e-06,
+      "loss": 0.9206,
+      "step": 8363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.12850397251495e-06,
+      "loss": 0.4183,
+      "step": 8364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.12741901070593e-06,
+      "loss": 0.7125,
+      "step": 8365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.12633409127195e-06,
+      "loss": 0.7079,
+      "step": 8366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1252492142656955e-06,
+      "loss": 0.7537,
+      "step": 8367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1241643797398535e-06,
+      "loss": 0.608,
+      "step": 8368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.123079587747104e-06,
+      "loss": 0.7801,
+      "step": 8369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.121994838340128e-06,
+      "loss": 0.6882,
+      "step": 8370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.120910131571608e-06,
+      "loss": 0.8727,
+      "step": 8371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.119825467494216e-06,
+      "loss": 0.67,
+      "step": 8372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.118740846160626e-06,
+      "loss": 0.4483,
+      "step": 8373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.117656267623516e-06,
+      "loss": 0.6899,
+      "step": 8374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.116571731935551e-06,
+      "loss": 0.8614,
+      "step": 8375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.115487239149399e-06,
+      "loss": 0.672,
+      "step": 8376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.114402789317729e-06,
+      "loss": 0.6895,
+      "step": 8377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.113318382493204e-06,
+      "loss": 0.5437,
+      "step": 8378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1122340187284845e-06,
+      "loss": 0.8579,
+      "step": 8379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.111149698076231e-06,
+      "loss": 0.5665,
+      "step": 8380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.110065420589104e-06,
+      "loss": 0.6185,
+      "step": 8381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.108981186319755e-06,
+      "loss": 1.0078,
+      "step": 8382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1078969953208394e-06,
+      "loss": 0.8203,
+      "step": 8383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.10681284764501e-06,
+      "loss": 0.8796,
+      "step": 8384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1057287433449145e-06,
+      "loss": 0.6972,
+      "step": 8385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.104644682473198e-06,
+      "loss": 0.577,
+      "step": 8386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.103560665082509e-06,
+      "loss": 0.4092,
+      "step": 8387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.1024766912254915e-06,
+      "loss": 0.9293,
+      "step": 8388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.101392760954781e-06,
+      "loss": 0.7286,
+      "step": 8389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.100308874323021e-06,
+      "loss": 0.7944,
+      "step": 8390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.099225031382848e-06,
+      "loss": 0.6427,
+      "step": 8391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.098141232186893e-06,
+      "loss": 0.6235,
+      "step": 8392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.097057476787792e-06,
+      "loss": 0.9586,
+      "step": 8393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0959737652381745e-06,
+      "loss": 0.7546,
+      "step": 8394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.094890097590666e-06,
+      "loss": 0.869,
+      "step": 8395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.093806473897897e-06,
+      "loss": 0.6044,
+      "step": 8396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.092722894212488e-06,
+      "loss": 0.8233,
+      "step": 8397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.091639358587062e-06,
+      "loss": 0.7768,
+      "step": 8398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.090555867074237e-06,
+      "loss": 0.7523,
+      "step": 8399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.089472419726632e-06,
+      "loss": 0.7203,
+      "step": 8400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.088389016596864e-06,
+      "loss": 0.6209,
+      "step": 8401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0873056577375416e-06,
+      "loss": 0.6859,
+      "step": 8402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.086222343201279e-06,
+      "loss": 0.6771,
+      "step": 8403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.085139073040687e-06,
+      "loss": 0.6263,
+      "step": 8404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.084055847308367e-06,
+      "loss": 0.5364,
+      "step": 8405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0829726660569285e-06,
+      "loss": 0.7679,
+      "step": 8406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.08188952933897e-06,
+      "loss": 0.8259,
+      "step": 8407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.080806437207094e-06,
+      "loss": 0.6146,
+      "step": 8408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.079723389713899e-06,
+      "loss": 0.9545,
+      "step": 8409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.078640386911979e-06,
+      "loss": 0.7901,
+      "step": 8410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0775574288539285e-06,
+      "loss": 0.8205,
+      "step": 8411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.076474515592341e-06,
+      "loss": 0.7108,
+      "step": 8412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.075391647179804e-06,
+      "loss": 0.7923,
+      "step": 8413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.074308823668903e-06,
+      "loss": 0.784,
+      "step": 8414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.073226045112225e-06,
+      "loss": 0.6248,
+      "step": 8415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.072143311562354e-06,
+      "loss": 0.6707,
+      "step": 8416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.071060623071869e-06,
+      "loss": 1.0021,
+      "step": 8417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.069977979693347e-06,
+      "loss": 0.8485,
+      "step": 8418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.068895381479369e-06,
+      "loss": 0.6045,
+      "step": 8419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0678128284825045e-06,
+      "loss": 0.7557,
+      "step": 8420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.066730320755327e-06,
+      "loss": 0.6753,
+      "step": 8421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.065647858350408e-06,
+      "loss": 0.6796,
+      "step": 8422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.064565441320313e-06,
+      "loss": 0.7302,
+      "step": 8423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.063483069717604e-06,
+      "loss": 0.6377,
+      "step": 8424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.06240074359485e-06,
+      "loss": 1.01,
+      "step": 8425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0613184630046105e-06,
+      "loss": 0.915,
+      "step": 8426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.060236227999441e-06,
+      "loss": 0.5714,
+      "step": 8427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0591540386319005e-06,
+      "loss": 0.7384,
+      "step": 8428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.058071894954544e-06,
+      "loss": 0.7162,
+      "step": 8429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.056989797019921e-06,
+      "loss": 0.9418,
+      "step": 8430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.055907744880583e-06,
+      "loss": 0.7312,
+      "step": 8431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.054825738589078e-06,
+      "loss": 0.8779,
+      "step": 8432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.053743778197951e-06,
+      "loss": 0.7316,
+      "step": 8433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.052661863759742e-06,
+      "loss": 0.8379,
+      "step": 8434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.051579995326995e-06,
+      "loss": 0.8508,
+      "step": 8435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.05049817295225e-06,
+      "loss": 0.9134,
+      "step": 8436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.049416396688039e-06,
+      "loss": 0.7104,
+      "step": 8437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0483346665869e-06,
+      "loss": 0.8824,
+      "step": 8438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.047252982701364e-06,
+      "loss": 0.9201,
+      "step": 8439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.046171345083959e-06,
+      "loss": 0.57,
+      "step": 8440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.045089753787214e-06,
+      "loss": 0.6513,
+      "step": 8441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.044008208863655e-06,
+      "loss": 0.6319,
+      "step": 8442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.042926710365802e-06,
+      "loss": 0.7298,
+      "step": 8443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.041845258346179e-06,
+      "loss": 0.7081,
+      "step": 8444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0407638528573025e-06,
+      "loss": 0.8585,
+      "step": 8445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.039682493951687e-06,
+      "loss": 0.8534,
+      "step": 8446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.038601181681851e-06,
+      "loss": 0.7708,
+      "step": 8447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.037519916100303e-06,
+      "loss": 0.9103,
+      "step": 8448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.036438697259551e-06,
+      "loss": 0.6293,
+      "step": 8449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.035357525212107e-06,
+      "loss": 0.65,
+      "step": 8450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0342764000104715e-06,
+      "loss": 0.7027,
+      "step": 8451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.033195321707148e-06,
+      "loss": 0.7636,
+      "step": 8452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0321142903546365e-06,
+      "loss": 0.4413,
+      "step": 8453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0310333060054375e-06,
+      "loss": 0.6129,
+      "step": 8454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.029952368712044e-06,
+      "loss": 0.7877,
+      "step": 8455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.028871478526949e-06,
+      "loss": 0.5538,
+      "step": 8456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.027790635502646e-06,
+      "loss": 0.6372,
+      "step": 8457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.026709839691623e-06,
+      "loss": 0.8935,
+      "step": 8458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0256290911463646e-06,
+      "loss": 0.8493,
+      "step": 8459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.02454838991936e-06,
+      "loss": 0.7359,
+      "step": 8460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0234677360630865e-06,
+      "loss": 0.6786,
+      "step": 8461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.022387129630023e-06,
+      "loss": 0.5765,
+      "step": 8462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.021306570672651e-06,
+      "loss": 0.8637,
+      "step": 8463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.020226059243444e-06,
+      "loss": 0.8453,
+      "step": 8464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.019145595394871e-06,
+      "loss": 0.7669,
+      "step": 8465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.018065179179409e-06,
+      "loss": 0.858,
+      "step": 8466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.016984810649522e-06,
+      "loss": 0.8982,
+      "step": 8467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.015904489857676e-06,
+      "loss": 0.7505,
+      "step": 8468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.014824216856336e-06,
+      "loss": 0.6673,
+      "step": 8469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.013743991697963e-06,
+      "loss": 0.747,
+      "step": 8470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.012663814435014e-06,
+      "loss": 0.7483,
+      "step": 8471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.011583685119946e-06,
+      "loss": 0.945,
+      "step": 8472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.010503603805214e-06,
+      "loss": 0.6445,
+      "step": 8473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.009423570543271e-06,
+      "loss": 0.7608,
+      "step": 8474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.008343585386562e-06,
+      "loss": 0.7915,
+      "step": 8475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.0072636483875404e-06,
+      "loss": 0.8805,
+      "step": 8476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.006183759598648e-06,
+      "loss": 0.7305,
+      "step": 8477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.005103919072325e-06,
+      "loss": 0.6885,
+      "step": 8478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.004024126861016e-06,
+      "loss": 0.5725,
+      "step": 8479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.002944383017156e-06,
+      "loss": 0.5597,
+      "step": 8480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.00186468759318e-06,
+      "loss": 0.9685,
+      "step": 8481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.000785040641524e-06,
+      "loss": 0.6216,
+      "step": 8482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.999705442214616e-06,
+      "loss": 0.8913,
+      "step": 8483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9986258923648845e-06,
+      "loss": 0.7071,
+      "step": 8484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9975463911447585e-06,
+      "loss": 0.6147,
+      "step": 8485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.996466938606659e-06,
+      "loss": 0.5955,
+      "step": 8486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.995387534803006e-06,
+      "loss": 0.5698,
+      "step": 8487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.994308179786223e-06,
+      "loss": 0.5489,
+      "step": 8488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.993228873608724e-06,
+      "loss": 0.527,
+      "step": 8489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.992149616322923e-06,
+      "loss": 0.7528,
+      "step": 8490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.991070407981229e-06,
+      "loss": 0.7108,
+      "step": 8491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.989991248636058e-06,
+      "loss": 0.5951,
+      "step": 8492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.988912138339812e-06,
+      "loss": 0.5965,
+      "step": 8493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.987833077144896e-06,
+      "loss": 0.7409,
+      "step": 8494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.986754065103715e-06,
+      "loss": 0.7298,
+      "step": 8495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.985675102268668e-06,
+      "loss": 0.688,
+      "step": 8496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.98459618869215e-06,
+      "loss": 0.9679,
+      "step": 8497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9835173244265605e-06,
+      "loss": 0.8959,
+      "step": 8498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.98243850952429e-06,
+      "loss": 0.536,
+      "step": 8499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.981359744037726e-06,
+      "loss": 0.64,
+      "step": 8500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.980281028019261e-06,
+      "loss": 0.9259,
+      "step": 8501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.97920236152128e-06,
+      "loss": 0.6417,
+      "step": 8502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.978123744596162e-06,
+      "loss": 1.0418,
+      "step": 8503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.977045177296291e-06,
+      "loss": 0.5818,
+      "step": 8504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.975966659674048e-06,
+      "loss": 0.5336,
+      "step": 8505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.974888191781803e-06,
+      "loss": 0.6824,
+      "step": 8506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.973809773671933e-06,
+      "loss": 0.763,
+      "step": 8507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.972731405396809e-06,
+      "loss": 0.5436,
+      "step": 8508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9716530870088e-06,
+      "loss": 0.6721,
+      "step": 8509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.970574818560269e-06,
+      "loss": 0.6858,
+      "step": 8510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.969496600103583e-06,
+      "loss": 0.7016,
+      "step": 8511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.968418431691105e-06,
+      "loss": 0.3888,
+      "step": 8512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9673403133751874e-06,
+      "loss": 0.7977,
+      "step": 8513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.966262245208193e-06,
+      "loss": 0.6491,
+      "step": 8514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9651842272424744e-06,
+      "loss": 0.8695,
+      "step": 8515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9641062595303805e-06,
+      "loss": 0.776,
+      "step": 8516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9630283421242645e-06,
+      "loss": 0.9751,
+      "step": 8517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9619504750764705e-06,
+      "loss": 0.5366,
+      "step": 8518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.960872658439343e-06,
+      "loss": 0.7613,
+      "step": 8519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.959794892265226e-06,
+      "loss": 0.5432,
+      "step": 8520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.958717176606456e-06,
+      "loss": 0.7493,
+      "step": 8521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.95763951151537e-06,
+      "loss": 0.7826,
+      "step": 8522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.956561897044306e-06,
+      "loss": 0.9185,
+      "step": 8523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.955484333245594e-06,
+      "loss": 0.9491,
+      "step": 8524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.954406820171564e-06,
+      "loss": 0.7149,
+      "step": 8525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.95332935787454e-06,
+      "loss": 0.7044,
+      "step": 8526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.952251946406852e-06,
+      "loss": 0.9172,
+      "step": 8527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.951174585820818e-06,
+      "loss": 0.7875,
+      "step": 8528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.950097276168758e-06,
+      "loss": 0.7473,
+      "step": 8529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.949020017502993e-06,
+      "loss": 0.7401,
+      "step": 8530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.947942809875834e-06,
+      "loss": 0.7924,
+      "step": 8531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.946865653339594e-06,
+      "loss": 0.6121,
+      "step": 8532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.945788547946584e-06,
+      "loss": 0.5746,
+      "step": 8533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.944711493749111e-06,
+      "loss": 0.7751,
+      "step": 8534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.943634490799478e-06,
+      "loss": 0.7226,
+      "step": 8535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.942557539149991e-06,
+      "loss": 0.8263,
+      "step": 8536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.941480638852948e-06,
+      "loss": 0.9107,
+      "step": 8537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9404037899606434e-06,
+      "loss": 0.7451,
+      "step": 8538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.939326992525377e-06,
+      "loss": 0.801,
+      "step": 8539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.938250246599439e-06,
+      "loss": 0.5153,
+      "step": 8540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.937173552235117e-06,
+      "loss": 0.9094,
+      "step": 8541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.936096909484702e-06,
+      "loss": 0.5589,
+      "step": 8542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.935020318400478e-06,
+      "loss": 0.5186,
+      "step": 8543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.933943779034727e-06,
+      "loss": 0.9441,
+      "step": 8544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.932867291439726e-06,
+      "loss": 0.8461,
+      "step": 8545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.931790855667759e-06,
+      "loss": 0.8176,
+      "step": 8546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.930714471771096e-06,
+      "loss": 0.8648,
+      "step": 8547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.929638139802007e-06,
+      "loss": 0.6601,
+      "step": 8548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.928561859812767e-06,
+      "loss": 0.6215,
+      "step": 8549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.927485631855642e-06,
+      "loss": 0.8354,
+      "step": 8550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.926409455982894e-06,
+      "loss": 0.4471,
+      "step": 8551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.925333332246787e-06,
+      "loss": 0.7358,
+      "step": 8552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.924257260699583e-06,
+      "loss": 0.7609,
+      "step": 8553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.923181241393534e-06,
+      "loss": 0.5732,
+      "step": 8554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9221052743809e-06,
+      "loss": 0.6654,
+      "step": 8555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.921029359713929e-06,
+      "loss": 0.742,
+      "step": 8556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.919953497444872e-06,
+      "loss": 0.7433,
+      "step": 8557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.918877687625977e-06,
+      "loss": 0.7201,
+      "step": 8558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.917801930309486e-06,
+      "loss": 0.691,
+      "step": 8559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.916726225547642e-06,
+      "loss": 0.691,
+      "step": 8560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.915650573392687e-06,
+      "loss": 0.7118,
+      "step": 8561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.914574973896853e-06,
+      "loss": 0.9259,
+      "step": 8562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.913499427112378e-06,
+      "loss": 1.054,
+      "step": 8563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.91242393309149e-06,
+      "loss": 0.7728,
+      "step": 8564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.911348491886423e-06,
+      "loss": 0.7832,
+      "step": 8565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.910273103549399e-06,
+      "loss": 0.7963,
+      "step": 8566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.909197768132643e-06,
+      "loss": 0.9768,
+      "step": 8567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.908122485688379e-06,
+      "loss": 0.5606,
+      "step": 8568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.907047256268822e-06,
+      "loss": 0.5181,
+      "step": 8569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.90597207992619e-06,
+      "loss": 0.7295,
+      "step": 8570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.904896956712699e-06,
+      "loss": 0.9962,
+      "step": 8571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.903821886680557e-06,
+      "loss": 0.4712,
+      "step": 8572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.902746869881971e-06,
+      "loss": 0.6235,
+      "step": 8573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.901671906369152e-06,
+      "loss": 0.4235,
+      "step": 8574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9005969961943e-06,
+      "loss": 0.7452,
+      "step": 8575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.899522139409615e-06,
+      "loss": 0.5352,
+      "step": 8576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.898447336067297e-06,
+      "loss": 0.8152,
+      "step": 8577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.897372586219542e-06,
+      "loss": 0.8911,
+      "step": 8578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.89629788991854e-06,
+      "loss": 0.7404,
+      "step": 8579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.895223247216484e-06,
+      "loss": 0.6225,
+      "step": 8580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.894148658165562e-06,
+      "loss": 0.9959,
+      "step": 8581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.893074122817957e-06,
+      "loss": 0.8587,
+      "step": 8582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.891999641225852e-06,
+      "loss": 0.7669,
+      "step": 8583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.890925213441429e-06,
+      "loss": 0.7188,
+      "step": 8584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8898508395168645e-06,
+      "loss": 0.8486,
+      "step": 8585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.888776519504329e-06,
+      "loss": 0.7317,
+      "step": 8586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.887702253456001e-06,
+      "loss": 0.9584,
+      "step": 8587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8866280414240465e-06,
+      "loss": 0.4556,
+      "step": 8588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.885553883460631e-06,
+      "loss": 0.8572,
+      "step": 8589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.884479779617923e-06,
+      "loss": 0.9045,
+      "step": 8590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.883405729948081e-06,
+      "loss": 0.8224,
+      "step": 8591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.882331734503263e-06,
+      "loss": 0.7873,
+      "step": 8592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8812577933356284e-06,
+      "loss": 0.8426,
+      "step": 8593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.880183906497329e-06,
+      "loss": 0.8359,
+      "step": 8594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.879110074040515e-06,
+      "loss": 0.7598,
+      "step": 8595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.878036296017337e-06,
+      "loss": 0.6506,
+      "step": 8596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8769625724799395e-06,
+      "loss": 0.7047,
+      "step": 8597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.875888903480465e-06,
+      "loss": 0.8372,
+      "step": 8598
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.874815289071056e-06,
+      "loss": 0.7817,
+      "step": 8599
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8737417293038484e-06,
+      "loss": 0.8654,
+      "step": 8600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.872668224230979e-06,
+      "loss": 0.6413,
+      "step": 8601
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.871594773904578e-06,
+      "loss": 0.6978,
+      "step": 8602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.870521378376777e-06,
+      "loss": 0.8797,
+      "step": 8603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.869448037699702e-06,
+      "loss": 0.9398,
+      "step": 8604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.868374751925478e-06,
+      "loss": 0.8936,
+      "step": 8605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.867301521106229e-06,
+      "loss": 0.9044,
+      "step": 8606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8662283452940705e-06,
+      "loss": 0.6016,
+      "step": 8607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8651552245411194e-06,
+      "loss": 0.5512,
+      "step": 8608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.864082158899494e-06,
+      "loss": 0.857,
+      "step": 8609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8630091484213e-06,
+      "loss": 0.6138,
+      "step": 8610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.861936193158647e-06,
+      "loss": 0.5966,
+      "step": 8611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.860863293163645e-06,
+      "loss": 0.5373,
+      "step": 8612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.859790448488394e-06,
+      "loss": 0.7881,
+      "step": 8613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.858717659184992e-06,
+      "loss": 0.8131,
+      "step": 8614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8576449253055395e-06,
+      "loss": 0.8404,
+      "step": 8615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.856572246902133e-06,
+      "loss": 0.8389,
+      "step": 8616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.855499624026861e-06,
+      "loss": 0.6652,
+      "step": 8617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.854427056731817e-06,
+      "loss": 0.7337,
+      "step": 8618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.853354545069086e-06,
+      "loss": 0.7269,
+      "step": 8619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.852282089090751e-06,
+      "loss": 0.7945,
+      "step": 8620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.851209688848895e-06,
+      "loss": 0.7746,
+      "step": 8621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.850137344395598e-06,
+      "loss": 0.6233,
+      "step": 8622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8490650557829355e-06,
+      "loss": 0.6513,
+      "step": 8623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.847992823062978e-06,
+      "loss": 0.618,
+      "step": 8624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8469206462878e-06,
+      "loss": 0.7734,
+      "step": 8625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.845848525509468e-06,
+      "loss": 0.8461,
+      "step": 8626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.844776460780047e-06,
+      "loss": 0.812,
+      "step": 8627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.843704452151598e-06,
+      "loss": 0.7175,
+      "step": 8628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842632499676186e-06,
+      "loss": 0.7648,
+      "step": 8629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.841560603405862e-06,
+      "loss": 0.6943,
+      "step": 8630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.840488763392685e-06,
+      "loss": 0.8086,
+      "step": 8631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.839416979688705e-06,
+      "loss": 0.6685,
+      "step": 8632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.838345252345968e-06,
+      "loss": 0.5896,
+      "step": 8633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8372735814165245e-06,
+      "loss": 0.8583,
+      "step": 8634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.836201966952416e-06,
+      "loss": 0.7322,
+      "step": 8635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.835130409005682e-06,
+      "loss": 0.797,
+      "step": 8636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.834058907628365e-06,
+      "loss": 0.5387,
+      "step": 8637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.832987462872495e-06,
+      "loss": 0.7348,
+      "step": 8638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8319160747901074e-06,
+      "loss": 0.6043,
+      "step": 8639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8308447434332286e-06,
+      "loss": 0.6986,
+      "step": 8640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.82977346885389e-06,
+      "loss": 0.9305,
+      "step": 8641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.828702251104113e-06,
+      "loss": 0.8359,
+      "step": 8642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8276310902359185e-06,
+      "loss": 0.6088,
+      "step": 8643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.826559986301328e-06,
+      "loss": 0.7263,
+      "step": 8644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.825488939352355e-06,
+      "loss": 0.7875,
+      "step": 8645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.824417949441012e-06,
+      "loss": 1.0138,
+      "step": 8646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.823347016619313e-06,
+      "loss": 0.7905,
+      "step": 8647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.822276140939261e-06,
+      "loss": 0.99,
+      "step": 8648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.821205322452863e-06,
+      "loss": 0.5903,
+      "step": 8649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.820134561212123e-06,
+      "loss": 0.5124,
+      "step": 8650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.819063857269039e-06,
+      "loss": 0.6807,
+      "step": 8651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.817993210675603e-06,
+      "loss": 0.6962,
+      "step": 8652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.816922621483815e-06,
+      "loss": 0.7099,
+      "step": 8653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.815852089745663e-06,
+      "loss": 0.6777,
+      "step": 8654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8147816155131345e-06,
+      "loss": 0.8661,
+      "step": 8655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.813711198838215e-06,
+      "loss": 0.6166,
+      "step": 8656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8126408397728898e-06,
+      "loss": 0.4536,
+      "step": 8657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8115705383691354e-06,
+      "loss": 0.811,
+      "step": 8658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.810500294678929e-06,
+      "loss": 0.5955,
+      "step": 8659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8094301087542475e-06,
+      "loss": 0.7299,
+      "step": 8660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8083599806470594e-06,
+      "loss": 0.6415,
+      "step": 8661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8072899104093325e-06,
+      "loss": 0.5672,
+      "step": 8662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.806219898093035e-06,
+      "loss": 0.6032,
+      "step": 8663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.80514994375013e-06,
+      "loss": 0.8527,
+      "step": 8664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.804080047432574e-06,
+      "loss": 0.5052,
+      "step": 8665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8030102091923276e-06,
+      "loss": 0.659,
+      "step": 8666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8019404290813456e-06,
+      "loss": 0.6486,
+      "step": 8667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.800870707151576e-06,
+      "loss": 0.763,
+      "step": 8668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7998010434549716e-06,
+      "loss": 0.7186,
+      "step": 8669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7987314380434753e-06,
+      "loss": 0.6753,
+      "step": 8670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7976618909690304e-06,
+      "loss": 0.5784,
+      "step": 8671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7965924022835802e-06,
+      "loss": 0.7155,
+      "step": 8672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7955229720390595e-06,
+      "loss": 0.4767,
+      "step": 8673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7944536002874043e-06,
+      "loss": 0.8008,
+      "step": 8674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7933842870805438e-06,
+      "loss": 0.8918,
+      "step": 8675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7923150324704095e-06,
+      "loss": 0.7505,
+      "step": 8676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.791245836508928e-06,
+      "loss": 0.8191,
+      "step": 8677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.790176699248018e-06,
+      "loss": 0.5782,
+      "step": 8678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.789107620739606e-06,
+      "loss": 0.8511,
+      "step": 8679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7880386010356056e-06,
+      "loss": 0.8275,
+      "step": 8680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.786969640187932e-06,
+      "loss": 0.7495,
+      "step": 8681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7859007382484987e-06,
+      "loss": 1.089,
+      "step": 8682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7848318952692124e-06,
+      "loss": 0.6303,
+      "step": 8683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7837631113019797e-06,
+      "loss": 0.8734,
+      "step": 8684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.782694386398706e-06,
+      "loss": 0.9922,
+      "step": 8685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7816257206112893e-06,
+      "loss": 1.037,
+      "step": 8686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7805571139916266e-06,
+      "loss": 0.8223,
+      "step": 8687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7794885665916155e-06,
+      "loss": 0.8315,
+      "step": 8688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7784200784631465e-06,
+      "loss": 0.9134,
+      "step": 8689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7773516496581055e-06,
+      "loss": 0.6632,
+      "step": 8690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.776283280228381e-06,
+      "loss": 0.7422,
+      "step": 8691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7752149702258577e-06,
+      "loss": 0.8679,
+      "step": 8692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7741467197024133e-06,
+      "loss": 0.7029,
+      "step": 8693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.773078528709925e-06,
+      "loss": 0.6665,
+      "step": 8694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7720103973002698e-06,
+      "loss": 0.6402,
+      "step": 8695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7709423255253157e-06,
+      "loss": 0.6361,
+      "step": 8696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.769874313436933e-06,
+      "loss": 0.8601,
+      "step": 8697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7688063610869898e-06,
+      "loss": 0.7864,
+      "step": 8698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.767738468527346e-06,
+      "loss": 0.889,
+      "step": 8699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7666706358098602e-06,
+      "loss": 0.8444,
+      "step": 8700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7656028629863927e-06,
+      "loss": 0.6262,
+      "step": 8701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.764535150108798e-06,
+      "loss": 0.67,
+      "step": 8702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7634674972289227e-06,
+      "loss": 0.8446,
+      "step": 8703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.762399904398619e-06,
+      "loss": 0.4516,
+      "step": 8704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.761332371669734e-06,
+      "loss": 0.5331,
+      "step": 8705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7602648990941037e-06,
+      "loss": 0.6874,
+      "step": 8706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.759197486723574e-06,
+      "loss": 0.5797,
+      "step": 8707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.758130134609978e-06,
+      "loss": 0.863,
+      "step": 8708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7570628428051494e-06,
+      "loss": 0.6378,
+      "step": 8709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.755995611360922e-06,
+      "loss": 0.8059,
+      "step": 8710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.75492844032912e-06,
+      "loss": 0.7076,
+      "step": 8711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.753861329761571e-06,
+      "loss": 0.8226,
+      "step": 8712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.752794279710094e-06,
+      "loss": 0.8981,
+      "step": 8713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7517272902265107e-06,
+      "loss": 0.6848,
+      "step": 8714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.750660361362637e-06,
+      "loss": 0.7636,
+      "step": 8715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7495934931702838e-06,
+      "loss": 0.6835,
+      "step": 8716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7485266857012644e-06,
+      "loss": 0.7421,
+      "step": 8717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7474599390073833e-06,
+      "loss": 0.6539,
+      "step": 8718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.746393253140445e-06,
+      "loss": 0.7467,
+      "step": 8719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7453266281522537e-06,
+      "loss": 0.5732,
+      "step": 8720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7442600640946045e-06,
+      "loss": 0.8048,
+      "step": 8721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7431935610192936e-06,
+      "loss": 0.8993,
+      "step": 8722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.742127118978116e-06,
+      "loss": 0.7758,
+      "step": 8723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.741060738022858e-06,
+      "loss": 0.8029,
+      "step": 8724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.739994418205306e-06,
+      "loss": 0.6499,
+      "step": 8725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.738928159577247e-06,
+      "loss": 0.7816,
+      "step": 8726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7378619621904593e-06,
+      "loss": 0.6647,
+      "step": 8727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7367958260967187e-06,
+      "loss": 0.4777,
+      "step": 8728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.735729751347803e-06,
+      "loss": 0.6784,
+      "step": 8729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.734663737995483e-06,
+      "loss": 0.7253,
+      "step": 8730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.733597786091526e-06,
+      "loss": 0.7964,
+      "step": 8731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7325318956876985e-06,
+      "loss": 0.5692,
+      "step": 8732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.731466066835765e-06,
+      "loss": 0.7265,
+      "step": 8733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7304002995874824e-06,
+      "loss": 0.6533,
+      "step": 8734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.729334593994607e-06,
+      "loss": 0.8244,
+      "step": 8735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7282689501088964e-06,
+      "loss": 0.878,
+      "step": 8736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7272033679820997e-06,
+      "loss": 0.4956,
+      "step": 8737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7261378476659614e-06,
+      "loss": 0.8039,
+      "step": 8738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.72507238921223e-06,
+      "loss": 0.6879,
+      "step": 8739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7240069926726473e-06,
+      "loss": 0.6575,
+      "step": 8740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.722941658098948e-06,
+      "loss": 0.9394,
+      "step": 8741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.721876385542872e-06,
+      "loss": 0.7169,
+      "step": 8742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7208111750561515e-06,
+      "loss": 0.619,
+      "step": 8743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7197460266905127e-06,
+      "loss": 0.9502,
+      "step": 8744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7186809404976877e-06,
+      "loss": 0.6309,
+      "step": 8745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7176159165293955e-06,
+      "loss": 0.8764,
+      "step": 8746
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.716550954837358e-06,
+      "loss": 0.7117,
+      "step": 8747
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.715486055473295e-06,
+      "loss": 0.4702,
+      "step": 8748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7144212184889185e-06,
+      "loss": 0.6614,
+      "step": 8749
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7133564439359417e-06,
+      "loss": 0.8013,
+      "step": 8750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.71229173186607e-06,
+      "loss": 0.6457,
+      "step": 8751
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7112270823310128e-06,
+      "loss": 0.9815,
+      "step": 8752
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.710162495382471e-06,
+      "loss": 0.7644,
+      "step": 8753
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7090979710721424e-06,
+      "loss": 0.5916,
+      "step": 8754
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.708033509451727e-06,
+      "loss": 0.7901,
+      "step": 8755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7069691105729142e-06,
+      "loss": 0.5339,
+      "step": 8756
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.705904774487396e-06,
+      "loss": 1.0607,
+      "step": 8757
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7048405012468613e-06,
+      "loss": 0.8964,
+      "step": 8758
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7037762909029915e-06,
+      "loss": 0.9036,
+      "step": 8759
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7027121435074683e-06,
+      "loss": 0.6158,
+      "step": 8760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.701648059111972e-06,
+      "loss": 0.8782,
+      "step": 8761
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7005840377681744e-06,
+      "loss": 0.8937,
+      "step": 8762
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6995200795277485e-06,
+      "loss": 0.8775,
+      "step": 8763
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6984561844423648e-06,
+      "loss": 0.7734,
+      "step": 8764
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.697392352563689e-06,
+      "loss": 0.9013,
+      "step": 8765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6963285839433805e-06,
+      "loss": 0.4663,
+      "step": 8766
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6952648786331015e-06,
+      "loss": 0.6908,
+      "step": 8767
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6942012366845097e-06,
+      "loss": 0.7925,
+      "step": 8768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6931376581492564e-06,
+      "loss": 0.528,
+      "step": 8769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.692074143078991e-06,
+      "loss": 0.8276,
+      "step": 8770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6910106915253652e-06,
+      "loss": 0.6579,
+      "step": 8771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6899473035400197e-06,
+      "loss": 0.8098,
+      "step": 8772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6888839791745955e-06,
+      "loss": 0.8039,
+      "step": 8773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.687820718480734e-06,
+      "loss": 0.5788,
+      "step": 8774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.686757521510068e-06,
+      "loss": 0.7037,
+      "step": 8775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6856943883142277e-06,
+      "loss": 0.8517,
+      "step": 8776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6846313189448447e-06,
+      "loss": 0.894,
+      "step": 8777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6835683134535437e-06,
+      "loss": 0.6067,
+      "step": 8778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.682505371891947e-06,
+      "loss": 0.6427,
+      "step": 8779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.681442494311674e-06,
+      "loss": 0.9692,
+      "step": 8780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6803796807643426e-06,
+      "loss": 0.7719,
+      "step": 8781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.679316931301563e-06,
+      "loss": 0.6518,
+      "step": 8782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6782542459749493e-06,
+      "loss": 0.912,
+      "step": 8783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.677191624836106e-06,
+      "loss": 0.8297,
+      "step": 8784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6761290679366384e-06,
+      "loss": 0.779,
+      "step": 8785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6750665753281445e-06,
+      "loss": 0.5707,
+      "step": 8786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6740041470622245e-06,
+      "loss": 0.595,
+      "step": 8787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.672941783190474e-06,
+      "loss": 0.6786,
+      "step": 8788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6718794837644813e-06,
+      "loss": 0.6966,
+      "step": 8789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6708172488358364e-06,
+      "loss": 0.6419,
+      "step": 8790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6697550784561254e-06,
+      "loss": 0.7651,
+      "step": 8791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6686929726769284e-06,
+      "loss": 0.6711,
+      "step": 8792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.667630931549826e-06,
+      "loss": 1.1181,
+      "step": 8793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.666568955126393e-06,
+      "loss": 0.7072,
+      "step": 8794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.665507043458201e-06,
+      "loss": 0.745,
+      "step": 8795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.664445196596823e-06,
+      "loss": 0.7373,
+      "step": 8796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.663383414593822e-06,
+      "loss": 0.7193,
+      "step": 8797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6623216975007603e-06,
+      "loss": 0.8184,
+      "step": 8798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.661260045369203e-06,
+      "loss": 0.5635,
+      "step": 8799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.660198458250702e-06,
+      "loss": 0.5336,
+      "step": 8800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6591369361968127e-06,
+      "loss": 0.9108,
+      "step": 8801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6580754792590877e-06,
+      "loss": 0.9651,
+      "step": 8802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6570140874890724e-06,
+      "loss": 1.0459,
+      "step": 8803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6559527609383096e-06,
+      "loss": 0.769,
+      "step": 8804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6548914996583417e-06,
+      "loss": 0.6133,
+      "step": 8805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6538303037007084e-06,
+      "loss": 0.6286,
+      "step": 8806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6527691731169425e-06,
+      "loss": 0.7261,
+      "step": 8807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6517081079585736e-06,
+      "loss": 0.6005,
+      "step": 8808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6506471082771357e-06,
+      "loss": 0.5681,
+      "step": 8809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.649586174124149e-06,
+      "loss": 0.8173,
+      "step": 8810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.648525305551136e-06,
+      "loss": 0.9109,
+      "step": 8811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.647464502609619e-06,
+      "loss": 0.718,
+      "step": 8812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.646403765351112e-06,
+      "loss": 0.8009,
+      "step": 8813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.645343093827124e-06,
+      "loss": 0.8936,
+      "step": 8814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.644282488089168e-06,
+      "loss": 0.6376,
+      "step": 8815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6432219481887504e-06,
+      "loss": 0.5336,
+      "step": 8816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6421614741773702e-06,
+      "loss": 0.7124,
+      "step": 8817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6411010661065318e-06,
+      "loss": 0.4648,
+      "step": 8818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6400407240277295e-06,
+      "loss": 0.9989,
+      "step": 8819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.638980447992455e-06,
+      "loss": 0.7036,
+      "step": 8820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6379202380522017e-06,
+      "loss": 0.4211,
+      "step": 8821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6368600942584536e-06,
+      "loss": 0.5033,
+      "step": 8822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6358000166626966e-06,
+      "loss": 0.8241,
+      "step": 8823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6347400053164077e-06,
+      "loss": 0.6553,
+      "step": 8824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6336800602710676e-06,
+      "loss": 0.6284,
+      "step": 8825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.63262018157815e-06,
+      "loss": 1.0109,
+      "step": 8826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6315603692891222e-06,
+      "loss": 0.9042,
+      "step": 8827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.630500623455455e-06,
+      "loss": 0.7889,
+      "step": 8828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6294409441286133e-06,
+      "loss": 0.6835,
+      "step": 8829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6283813313600546e-06,
+      "loss": 0.8246,
+      "step": 8830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.627321785201241e-06,
+      "loss": 0.7526,
+      "step": 8831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6262623057036238e-06,
+      "loss": 0.7315,
+      "step": 8832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6252028929186546e-06,
+      "loss": 1.0781,
+      "step": 8833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.624143546897784e-06,
+      "loss": 0.7819,
+      "step": 8834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.623084267692455e-06,
+      "loss": 0.6855,
+      "step": 8835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.622025055354108e-06,
+      "loss": 0.9684,
+      "step": 8836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6209659099341855e-06,
+      "loss": 0.9895,
+      "step": 8837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.619906831484119e-06,
+      "loss": 0.8078,
+      "step": 8838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6188478200553403e-06,
+      "loss": 1.0344,
+      "step": 8839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.617788875699282e-06,
+      "loss": 0.6735,
+      "step": 8840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6167299984673655e-06,
+      "loss": 0.7768,
+      "step": 8841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.615671188411014e-06,
+      "loss": 0.6964,
+      "step": 8842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6146124455816455e-06,
+      "loss": 0.624,
+      "step": 8843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6135537700306785e-06,
+      "loss": 0.7233,
+      "step": 8844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6124951618095224e-06,
+      "loss": 0.8765,
+      "step": 8845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.611436620969586e-06,
+      "loss": 0.7813,
+      "step": 8846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.610378147562279e-06,
+      "loss": 0.8816,
+      "step": 8847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.609319741639e-06,
+      "loss": 0.6933,
+      "step": 8848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6082614032511475e-06,
+      "loss": 0.768,
+      "step": 8849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6072031324501234e-06,
+      "loss": 0.9833,
+      "step": 8850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.606144929287315e-06,
+      "loss": 0.8714,
+      "step": 8851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6050867938141116e-06,
+      "loss": 0.6178,
+      "step": 8852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6040287260819017e-06,
+      "loss": 0.7459,
+      "step": 8853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.602970726142069e-06,
+      "loss": 0.9577,
+      "step": 8854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.601912794045988e-06,
+      "loss": 0.3697,
+      "step": 8855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6008549298450403e-06,
+      "loss": 0.3788,
+      "step": 8856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5997971335905966e-06,
+      "loss": 0.5758,
+      "step": 8857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5987394053340254e-06,
+      "loss": 0.784,
+      "step": 8858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.597681745126696e-06,
+      "loss": 0.9815,
+      "step": 8859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5966241530199686e-06,
+      "loss": 0.6195,
+      "step": 8860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.595566629065205e-06,
+      "loss": 0.8532,
+      "step": 8861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.594509173313758e-06,
+      "loss": 0.8883,
+      "step": 8862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.593451785816985e-06,
+      "loss": 0.6341,
+      "step": 8863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5923944666262358e-06,
+      "loss": 1.0086,
+      "step": 8864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5913372157928515e-06,
+      "loss": 0.6431,
+      "step": 8865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5902800333681807e-06,
+      "loss": 0.7496,
+      "step": 8866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.589222919403562e-06,
+      "loss": 0.9254,
+      "step": 8867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.58816587395033e-06,
+      "loss": 0.92,
+      "step": 8868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5871088970598208e-06,
+      "loss": 0.7901,
+      "step": 8869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5860519887833612e-06,
+      "loss": 0.684,
+      "step": 8870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5849951491722777e-06,
+      "loss": 0.7562,
+      "step": 8871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5839383782778973e-06,
+      "loss": 0.7718,
+      "step": 8872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.582881676151536e-06,
+      "loss": 0.6645,
+      "step": 8873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.581825042844511e-06,
+      "loss": 0.5784,
+      "step": 8874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5807684784081385e-06,
+      "loss": 0.6509,
+      "step": 8875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5797119828937242e-06,
+      "loss": 0.7097,
+      "step": 8876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5786555563525748e-06,
+      "loss": 0.7685,
+      "step": 8877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5775991988359983e-06,
+      "loss": 0.9681,
+      "step": 8878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.57654291039529e-06,
+      "loss": 0.8525,
+      "step": 8879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5754866910817463e-06,
+      "loss": 0.8015,
+      "step": 8880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.574430540946661e-06,
+      "loss": 0.6708,
+      "step": 8881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5733744600413247e-06,
+      "loss": 0.4824,
+      "step": 8882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.572318448417023e-06,
+      "loss": 0.5714,
+      "step": 8883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.571262506125038e-06,
+      "loss": 0.8883,
+      "step": 8884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.570206633216652e-06,
+      "loss": 0.7724,
+      "step": 8885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5691508297431375e-06,
+      "loss": 0.9808,
+      "step": 8886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.56809509575577e-06,
+      "loss": 0.944,
+      "step": 8887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5670394313058186e-06,
+      "loss": 0.8924,
+      "step": 8888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5659838364445505e-06,
+      "loss": 0.4424,
+      "step": 8889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5649283112232235e-06,
+      "loss": 0.5428,
+      "step": 8890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5638728556931023e-06,
+      "loss": 0.8418,
+      "step": 8891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5628174699054424e-06,
+      "loss": 0.7473,
+      "step": 8892
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5617621539114923e-06,
+      "loss": 0.6666,
+      "step": 8893
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5607069077625044e-06,
+      "loss": 0.7168,
+      "step": 8894
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5596517315097257e-06,
+      "loss": 0.8446,
+      "step": 8895
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5585966252043946e-06,
+      "loss": 0.6081,
+      "step": 8896
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5575415888977536e-06,
+      "loss": 0.6344,
+      "step": 8897
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5564866226410368e-06,
+      "loss": 0.7017,
+      "step": 8898
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5554317264854778e-06,
+      "loss": 0.8243,
+      "step": 8899
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5543769004823013e-06,
+      "loss": 0.7122,
+      "step": 8900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.553322144682737e-06,
+      "loss": 0.5671,
+      "step": 8901
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.552267459138006e-06,
+      "loss": 0.6553,
+      "step": 8902
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.551212843899324e-06,
+      "loss": 0.7981,
+      "step": 8903
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5501582990179105e-06,
+      "loss": 0.8019,
+      "step": 8904
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.549103824544975e-06,
+      "loss": 0.7238,
+      "step": 8905
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5480494205317252e-06,
+      "loss": 0.8101,
+      "step": 8906
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.546995087029368e-06,
+      "loss": 0.5831,
+      "step": 8907
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5459408240891024e-06,
+      "loss": 0.6825,
+      "step": 8908
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.544886631762127e-06,
+      "loss": 0.6686,
+      "step": 8909
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5438325100996384e-06,
+      "loss": 0.7543,
+      "step": 8910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.542778459152826e-06,
+      "loss": 0.6587,
+      "step": 8911
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5417244789728766e-06,
+      "loss": 0.8867,
+      "step": 8912
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5406705696109777e-06,
+      "loss": 0.5356,
+      "step": 8913
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5396167311183082e-06,
+      "loss": 0.9224,
+      "step": 8914
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.538562963546046e-06,
+      "loss": 0.7953,
+      "step": 8915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5375092669453624e-06,
+      "loss": 0.9506,
+      "step": 8916
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5364556413674334e-06,
+      "loss": 0.8069,
+      "step": 8917
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5354020868634207e-06,
+      "loss": 0.6901,
+      "step": 8918
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5343486034844897e-06,
+      "loss": 0.7991,
+      "step": 8919
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5332951912818038e-06,
+      "loss": 0.551,
+      "step": 8920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5322418503065148e-06,
+      "loss": 0.5693,
+      "step": 8921
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.531188580609778e-06,
+      "loss": 0.9758,
+      "step": 8922
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5301353822427454e-06,
+      "loss": 0.7948,
+      "step": 8923
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5290822552565594e-06,
+      "loss": 0.6953,
+      "step": 8924
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5280291997023653e-06,
+      "loss": 0.6111,
+      "step": 8925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5269762156313018e-06,
+      "loss": 0.8187,
+      "step": 8926
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5259233030945064e-06,
+      "loss": 0.7653,
+      "step": 8927
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.524870462143108e-06,
+      "loss": 0.8772,
+      "step": 8928
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5238176928282384e-06,
+      "loss": 0.7157,
+      "step": 8929
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.522764995201024e-06,
+      "loss": 0.8193,
+      "step": 8930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5217123693125827e-06,
+      "loss": 0.7949,
+      "step": 8931
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.520659815214037e-06,
+      "loss": 0.5965,
+      "step": 8932
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.519607332956502e-06,
+      "loss": 0.8304,
+      "step": 8933
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5185549225910865e-06,
+      "loss": 0.6558,
+      "step": 8934
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5175025841688986e-06,
+      "loss": 0.8975,
+      "step": 8935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.516450317741045e-06,
+      "loss": 0.7389,
+      "step": 8936
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5153981233586277e-06,
+      "loss": 0.7897,
+      "step": 8937
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5143460010727392e-06,
+      "loss": 0.8176,
+      "step": 8938
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5132939509344787e-06,
+      "loss": 0.8143,
+      "step": 8939
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5122419729949366e-06,
+      "loss": 0.7967,
+      "step": 8940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5111900673051953e-06,
+      "loss": 0.8597,
+      "step": 8941
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5101382339163424e-06,
+      "loss": 0.7919,
+      "step": 8942
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5090864728794583e-06,
+      "loss": 0.6333,
+      "step": 8943
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5080347842456164e-06,
+      "loss": 0.7966,
+      "step": 8944
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5069831680658932e-06,
+      "loss": 0.9173,
+      "step": 8945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5059316243913556e-06,
+      "loss": 0.7506,
+      "step": 8946
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.50488015327307e-06,
+      "loss": 0.8913,
+      "step": 8947
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5038287547621007e-06,
+      "loss": 0.8888,
+      "step": 8948
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5027774289095046e-06,
+      "loss": 0.7324,
+      "step": 8949
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5017261757663366e-06,
+      "loss": 0.4513,
+      "step": 8950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5006749953836515e-06,
+      "loss": 0.6307,
+      "step": 8951
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4996238878124955e-06,
+      "loss": 0.4572,
+      "step": 8952
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.498572853103915e-06,
+      "loss": 0.721,
+      "step": 8953
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4975218913089475e-06,
+      "loss": 0.8327,
+      "step": 8954
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4964710024786354e-06,
+      "loss": 0.5504,
+      "step": 8955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.49542018666401e-06,
+      "loss": 0.6011,
+      "step": 8956
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4943694439161015e-06,
+      "loss": 0.6823,
+      "step": 8957
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4933187742859407e-06,
+      "loss": 0.7912,
+      "step": 8958
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4922681778245473e-06,
+      "loss": 0.6587,
+      "step": 8959
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4912176545829412e-06,
+      "loss": 0.7469,
+      "step": 8960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4901672046121428e-06,
+      "loss": 0.5236,
+      "step": 8961
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4891168279631628e-06,
+      "loss": 0.772,
+      "step": 8962
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.488066524687007e-06,
+      "loss": 0.8093,
+      "step": 8963
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.487016294834686e-06,
+      "loss": 0.4754,
+      "step": 8964
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.485966138457201e-06,
+      "loss": 0.9957,
+      "step": 8965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4849160556055483e-06,
+      "loss": 0.5635,
+      "step": 8966
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.483866046330725e-06,
+      "loss": 0.9477,
+      "step": 8967
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4828161106837234e-06,
+      "loss": 0.59,
+      "step": 8968
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.481766248715528e-06,
+      "loss": 0.7687,
+      "step": 8969
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.480716460477126e-06,
+      "loss": 0.6479,
+      "step": 8970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4796667460194976e-06,
+      "loss": 0.5814,
+      "step": 8971
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4786171053936214e-06,
+      "loss": 0.8337,
+      "step": 8972
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.477567538650466e-06,
+      "loss": 0.9045,
+      "step": 8973
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4765180458410053e-06,
+      "loss": 0.8732,
+      "step": 8974
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.475468627016207e-06,
+      "loss": 0.7552,
+      "step": 8975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4744192822270294e-06,
+      "loss": 0.9767,
+      "step": 8976
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.473370011524435e-06,
+      "loss": 0.8403,
+      "step": 8977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4723208149593797e-06,
+      "loss": 0.5735,
+      "step": 8978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.471271692582812e-06,
+      "loss": 0.737,
+      "step": 8979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4702226444456844e-06,
+      "loss": 0.5742,
+      "step": 8980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.469173670598941e-06,
+      "loss": 0.8125,
+      "step": 8981
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4681247710935197e-06,
+      "loss": 0.7075,
+      "step": 8982
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4670759459803615e-06,
+      "loss": 0.6345,
+      "step": 8983
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.466027195310399e-06,
+      "loss": 0.8583,
+      "step": 8984
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4649785191345613e-06,
+      "loss": 0.6728,
+      "step": 8985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4639299175037788e-06,
+      "loss": 1.0568,
+      "step": 8986
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4628813904689716e-06,
+      "loss": 0.6372,
+      "step": 8987
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.461832938081059e-06,
+      "loss": 0.5541,
+      "step": 8988
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4607845603909594e-06,
+      "loss": 0.8588,
+      "step": 8989
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4597362574495824e-06,
+      "loss": 0.9593,
+      "step": 8990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.458688029307839e-06,
+      "loss": 0.4542,
+      "step": 8991
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4576398760166306e-06,
+      "loss": 0.8504,
+      "step": 8992
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4565917976268627e-06,
+      "loss": 0.7295,
+      "step": 8993
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.45554379418943e-06,
+      "loss": 0.7827,
+      "step": 8994
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4544958657552263e-06,
+      "loss": 0.7076,
+      "step": 8995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4534480123751457e-06,
+      "loss": 0.8934,
+      "step": 8996
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.452400234100072e-06,
+      "loss": 0.7454,
+      "step": 8997
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.451352530980887e-06,
+      "loss": 0.6233,
+      "step": 8998
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.450304903068475e-06,
+      "loss": 0.5162,
+      "step": 8999
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4492573504137085e-06,
+      "loss": 0.7477,
+      "step": 9000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4482098730674577e-06,
+      "loss": 0.8051,
+      "step": 9001
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.447162471080595e-06,
+      "loss": 0.7728,
+      "step": 9002
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4461151445039844e-06,
+      "loss": 0.9003,
+      "step": 9003
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4450678933884846e-06,
+      "loss": 0.5398,
+      "step": 9004
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.444020717784956e-06,
+      "loss": 0.8493,
+      "step": 9005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.442973617744252e-06,
+      "loss": 0.5905,
+      "step": 9006
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.441926593317221e-06,
+      "loss": 0.8217,
+      "step": 9007
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.440879644554711e-06,
+      "loss": 0.5665,
+      "step": 9008
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.439832771507565e-06,
+      "loss": 0.6276,
+      "step": 9009
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4387859742266226e-06,
+      "loss": 0.8185,
+      "step": 9010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.437739252762716e-06,
+      "loss": 0.7662,
+      "step": 9011
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4366926071666804e-06,
+      "loss": 0.6229,
+      "step": 9012
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4356460374893443e-06,
+      "loss": 0.52,
+      "step": 9013
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4345995437815283e-06,
+      "loss": 0.6427,
+      "step": 9014
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4335531260940567e-06,
+      "loss": 0.7296,
+      "step": 9015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4325067844777467e-06,
+      "loss": 0.595,
+      "step": 9016
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4314605189834076e-06,
+      "loss": 0.8615,
+      "step": 9017
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.430414329661853e-06,
+      "loss": 0.7618,
+      "step": 9018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4293682165638898e-06,
+      "loss": 0.764,
+      "step": 9019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.428322179740315e-06,
+      "loss": 0.8034,
+      "step": 9020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.427276219241933e-06,
+      "loss": 0.9231,
+      "step": 9021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.426230335119534e-06,
+      "loss": 0.6459,
+      "step": 9022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4251845274239105e-06,
+      "loss": 0.8336,
+      "step": 9023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4241387962058524e-06,
+      "loss": 0.6793,
+      "step": 9024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.423093141516141e-06,
+      "loss": 0.7028,
+      "step": 9025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4220475634055563e-06,
+      "loss": 0.7145,
+      "step": 9026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4210020619248762e-06,
+      "loss": 0.6243,
+      "step": 9027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.419956637124871e-06,
+      "loss": 0.7057,
+      "step": 9028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4189112890563118e-06,
+      "loss": 0.7473,
+      "step": 9029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4178660177699608e-06,
+      "loss": 0.8899,
+      "step": 9030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4168208233165835e-06,
+      "loss": 0.7476,
+      "step": 9031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.415775705746934e-06,
+      "loss": 0.7376,
+      "step": 9032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4147306651117663e-06,
+      "loss": 0.7769,
+      "step": 9033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.413685701461833e-06,
+      "loss": 0.7758,
+      "step": 9034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.412640814847879e-06,
+      "loss": 0.842,
+      "step": 9035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4115960053206463e-06,
+      "loss": 0.6158,
+      "step": 9036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4105512729308764e-06,
+      "loss": 0.6894,
+      "step": 9037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4095066177293033e-06,
+      "loss": 0.5613,
+      "step": 9038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4084620397666556e-06,
+      "loss": 0.8816,
+      "step": 9039
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.407417539093665e-06,
+      "loss": 0.7139,
+      "step": 9040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4063731157610545e-06,
+      "loss": 0.5576,
+      "step": 9041
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4053287698195415e-06,
+      "loss": 0.9025,
+      "step": 9042
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4042845013198457e-06,
+      "loss": 0.9048,
+      "step": 9043
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.40324031031268e-06,
+      "loss": 0.6272,
+      "step": 9044
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.402196196848751e-06,
+      "loss": 0.9026,
+      "step": 9045
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.4011521609787633e-06,
+      "loss": 0.672,
+      "step": 9046
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.400108202753423e-06,
+      "loss": 0.7154,
+      "step": 9047
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.399064322223424e-06,
+      "loss": 0.7074,
+      "step": 9048
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.398020519439459e-06,
+      "loss": 0.789,
+      "step": 9049
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3969767944522204e-06,
+      "loss": 0.9107,
+      "step": 9050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3959331473123947e-06,
+      "loss": 0.5351,
+      "step": 9051
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3948895780706616e-06,
+      "loss": 0.6373,
+      "step": 9052
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.393846086777703e-06,
+      "loss": 0.7044,
+      "step": 9053
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3928026734841935e-06,
+      "loss": 0.6866,
+      "step": 9054
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.391759338240801e-06,
+      "loss": 0.7833,
+      "step": 9055
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3907160810981963e-06,
+      "loss": 0.9672,
+      "step": 9056
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.389672902107044e-06,
+      "loss": 0.9112,
+      "step": 9057
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3886298013179987e-06,
+      "loss": 0.9716,
+      "step": 9058
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3875867787817217e-06,
+      "loss": 0.5651,
+      "step": 9059
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.386543834548862e-06,
+      "loss": 0.6964,
+      "step": 9060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.385500968670068e-06,
+      "loss": 0.8118,
+      "step": 9061
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3844581811959863e-06,
+      "loss": 0.5968,
+      "step": 9062
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.383415472177256e-06,
+      "loss": 0.7015,
+      "step": 9063
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3823728416645156e-06,
+      "loss": 0.7708,
+      "step": 9064
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3813302897083955e-06,
+      "loss": 0.5715,
+      "step": 9065
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3802878163595273e-06,
+      "loss": 0.8694,
+      "step": 9066
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3792454216685365e-06,
+      "loss": 0.9411,
+      "step": 9067
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3782031056860427e-06,
+      "loss": 0.7157,
+      "step": 9068
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3771608684626667e-06,
+      "loss": 0.8962,
+      "step": 9069
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3761187100490204e-06,
+      "loss": 0.6954,
+      "step": 9070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.375076630495714e-06,
+      "loss": 0.689,
+      "step": 9071
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.374034629853356e-06,
+      "loss": 0.7845,
+      "step": 9072
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3729927081725465e-06,
+      "loss": 0.5848,
+      "step": 9073
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3719508655038847e-06,
+      "loss": 0.5722,
+      "step": 9074
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3709091018979677e-06,
+      "loss": 0.8022,
+      "step": 9075
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.369867417405385e-06,
+      "loss": 0.8348,
+      "step": 9076
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3688258120767207e-06,
+      "loss": 0.8461,
+      "step": 9077
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3677842859625626e-06,
+      "loss": 0.7522,
+      "step": 9078
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3667428391134897e-06,
+      "loss": 0.6286,
+      "step": 9079
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3657014715800746e-06,
+      "loss": 0.7164,
+      "step": 9080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3646601834128924e-06,
+      "loss": 0.6701,
+      "step": 9081
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3636189746625105e-06,
+      "loss": 0.568,
+      "step": 9082
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3625778453794904e-06,
+      "loss": 0.798,
+      "step": 9083
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.361536795614394e-06,
+      "loss": 0.5833,
+      "step": 9084
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3604958254177794e-06,
+      "loss": 0.6806,
+      "step": 9085
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.359454934840198e-06,
+      "loss": 0.5993,
+      "step": 9086
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3584141239321953e-06,
+      "loss": 0.8614,
+      "step": 9087
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3573733927443203e-06,
+      "loss": 0.6797,
+      "step": 9088
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.356332741327113e-06,
+      "loss": 0.9402,
+      "step": 9089
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.355292169731108e-06,
+      "loss": 0.8123,
+      "step": 9090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.354251678006841e-06,
+      "loss": 0.9012,
+      "step": 9091
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.353211266204841e-06,
+      "loss": 0.653,
+      "step": 9092
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3521709343756317e-06,
+      "loss": 0.7321,
+      "step": 9093
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3511306825697365e-06,
+      "loss": 0.5412,
+      "step": 9094
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.350090510837673e-06,
+      "loss": 0.7685,
+      "step": 9095
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.349050419229952e-06,
+      "loss": 0.8002,
+      "step": 9096
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.348010407797088e-06,
+      "loss": 0.9461,
+      "step": 9097
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.346970476589583e-06,
+      "loss": 0.6891,
+      "step": 9098
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3459306256579405e-06,
+      "loss": 0.6319,
+      "step": 9099
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3448908550526595e-06,
+      "loss": 0.7427,
+      "step": 9100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3438511648242335e-06,
+      "loss": 0.6674,
+      "step": 9101
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.342811555023153e-06,
+      "loss": 0.6614,
+      "step": 9102
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.341772025699902e-06,
+      "loss": 0.699,
+      "step": 9103
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3407325769049662e-06,
+      "loss": 0.7473,
+      "step": 9104
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3396932086888245e-06,
+      "loss": 0.8119,
+      "step": 9105
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.338653921101949e-06,
+      "loss": 0.7819,
+      "step": 9106
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3376147141948127e-06,
+      "loss": 0.7369,
+      "step": 9107
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.336575588017881e-06,
+      "loss": 0.6984,
+      "step": 9108
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.335536542621617e-06,
+      "loss": 0.7689,
+      "step": 9109
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3344975780564813e-06,
+      "loss": 0.4666,
+      "step": 9110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.333458694372928e-06,
+      "loss": 0.833,
+      "step": 9111
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.332419891621407e-06,
+      "loss": 0.8799,
+      "step": 9112
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3313811698523677e-06,
+      "loss": 0.8222,
+      "step": 9113
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.330342529116253e-06,
+      "loss": 0.8351,
+      "step": 9114
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3293039694634997e-06,
+      "loss": 0.8106,
+      "step": 9115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3282654909445467e-06,
+      "loss": 0.6135,
+      "step": 9116
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3272270936098243e-06,
+      "loss": 1.0251,
+      "step": 9117
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.326188777509758e-06,
+      "loss": 0.6196,
+      "step": 9118
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3251505426947744e-06,
+      "loss": 0.8894,
+      "step": 9119
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3241123892152925e-06,
+      "loss": 1.0355,
+      "step": 9120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3230743171217274e-06,
+      "loss": 0.6946,
+      "step": 9121
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3220363264644895e-06,
+      "loss": 0.6805,
+      "step": 9122
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.32099841729399e-06,
+      "loss": 1.0783,
+      "step": 9123
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.319960589660631e-06,
+      "loss": 0.7646,
+      "step": 9124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.31892284361481e-06,
+      "loss": 0.9545,
+      "step": 9125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3178851792069265e-06,
+      "loss": 0.8184,
+      "step": 9126
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3168475964873713e-06,
+      "loss": 0.6788,
+      "step": 9127
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3158100955065307e-06,
+      "loss": 0.5489,
+      "step": 9128
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3147726763147913e-06,
+      "loss": 0.737,
+      "step": 9129
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3137353389625337e-06,
+      "loss": 0.9272,
+      "step": 9130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3126980835001297e-06,
+      "loss": 0.6864,
+      "step": 9131
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.311660909977955e-06,
+      "loss": 0.8006,
+      "step": 9132
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.310623818446378e-06,
+      "loss": 0.6477,
+      "step": 9133
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3095868089557594e-06,
+      "loss": 0.8983,
+      "step": 9134
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3085498815564644e-06,
+      "loss": 0.6319,
+      "step": 9135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3075130362988438e-06,
+      "loss": 0.7937,
+      "step": 9136
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3064762732332528e-06,
+      "loss": 0.7529,
+      "step": 9137
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.305439592410041e-06,
+      "loss": 0.5907,
+      "step": 9138
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.304402993879549e-06,
+      "loss": 0.6086,
+      "step": 9139
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.30336647769212e-06,
+      "loss": 0.9368,
+      "step": 9140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.302330043898088e-06,
+      "loss": 0.8315,
+      "step": 9141
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3012936925477856e-06,
+      "loss": 0.8094,
+      "step": 9142
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.3002574236915438e-06,
+      "loss": 0.7433,
+      "step": 9143
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2992212373796816e-06,
+      "loss": 0.5072,
+      "step": 9144
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.298185133662525e-06,
+      "loss": 0.6847,
+      "step": 9145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2971491125903854e-06,
+      "loss": 0.4514,
+      "step": 9146
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2961131742135766e-06,
+      "loss": 0.6849,
+      "step": 9147
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2950773185824094e-06,
+      "loss": 0.7704,
+      "step": 9148
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.294041545747183e-06,
+      "loss": 0.6128,
+      "step": 9149
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2930058557582e-06,
+      "loss": 0.7923,
+      "step": 9150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.291970248665759e-06,
+      "loss": 0.4775,
+      "step": 9151
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.290934724520149e-06,
+      "loss": 0.9034,
+      "step": 9152
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.289899283371657e-06,
+      "loss": 0.7124,
+      "step": 9153
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.288863925270569e-06,
+      "loss": 0.3587,
+      "step": 9154
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2878286502671665e-06,
+      "loss": 0.8903,
+      "step": 9155
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.28679345841172e-06,
+      "loss": 0.7102,
+      "step": 9156
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.285758349754507e-06,
+      "loss": 0.7512,
+      "step": 9157
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2847233243457944e-06,
+      "loss": 0.9525,
+      "step": 9158
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2836883822358436e-06,
+      "loss": 0.8338,
+      "step": 9159
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.282653523474914e-06,
+      "loss": 0.7742,
+      "step": 9160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2816187481132655e-06,
+      "loss": 0.8009,
+      "step": 9161
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.280584056201147e-06,
+      "loss": 0.6149,
+      "step": 9162
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2795494477888045e-06,
+      "loss": 0.4455,
+      "step": 9163
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2785149229264844e-06,
+      "loss": 0.7739,
+      "step": 9164
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2774804816644255e-06,
+      "loss": 0.7046,
+      "step": 9165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2764461240528617e-06,
+      "loss": 0.6657,
+      "step": 9166
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2754118501420263e-06,
+      "loss": 0.4943,
+      "step": 9167
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.274377659982147e-06,
+      "loss": 1.0327,
+      "step": 9168
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2733435536234438e-06,
+      "loss": 0.7739,
+      "step": 9169
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.272309531116139e-06,
+      "loss": 0.8205,
+      "step": 9170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2712755925104478e-06,
+      "loss": 0.8282,
+      "step": 9171
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2702417378565775e-06,
+      "loss": 0.6968,
+      "step": 9172
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2692079672047407e-06,
+      "loss": 0.562,
+      "step": 9173
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2681742806051354e-06,
+      "loss": 0.6381,
+      "step": 9174
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.267140678107964e-06,
+      "loss": 0.9785,
+      "step": 9175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.266107159763417e-06,
+      "loss": 0.8113,
+      "step": 9176
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2650737256216885e-06,
+      "loss": 0.6054,
+      "step": 9177
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.264040375732965e-06,
+      "loss": 0.5603,
+      "step": 9178
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2630071101474258e-06,
+      "loss": 0.5806,
+      "step": 9179
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.261973928915253e-06,
+      "loss": 0.4266,
+      "step": 9180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.26094083208662e-06,
+      "loss": 0.4756,
+      "step": 9181
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.259907819711694e-06,
+      "loss": 0.735,
+      "step": 9182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2588748918406454e-06,
+      "loss": 0.8181,
+      "step": 9183
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.257842048523633e-06,
+      "loss": 0.5724,
+      "step": 9184
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2568092898108143e-06,
+      "loss": 0.8727,
+      "step": 9185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2557766157523467e-06,
+      "loss": 0.7671,
+      "step": 9186
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2547440263983763e-06,
+      "loss": 0.6583,
+      "step": 9187
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.253711521799049e-06,
+      "loss": 0.4806,
+      "step": 9188
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.252679102004509e-06,
+      "loss": 0.8544,
+      "step": 9189
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.251646767064891e-06,
+      "loss": 0.6886,
+      "step": 9190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.250614517030327e-06,
+      "loss": 0.8449,
+      "step": 9191
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.249582351950949e-06,
+      "loss": 0.4715,
+      "step": 9192
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2485502718768814e-06,
+      "loss": 0.6206,
+      "step": 9193
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2475182768582435e-06,
+      "loss": 0.7767,
+      "step": 9194
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2464863669451508e-06,
+      "loss": 0.8036,
+      "step": 9195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2454545421877204e-06,
+      "loss": 0.8743,
+      "step": 9196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.244422802636057e-06,
+      "loss": 0.8837,
+      "step": 9197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.243391148340264e-06,
+      "loss": 0.5256,
+      "step": 9198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.242359579350446e-06,
+      "loss": 0.6436,
+      "step": 9199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.241328095716696e-06,
+      "loss": 0.5375,
+      "step": 9200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.240296697489104e-06,
+      "loss": 0.7427,
+      "step": 9201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2392653847177603e-06,
+      "loss": 0.5617,
+      "step": 9202
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.238234157452749e-06,
+      "loss": 0.8861,
+      "step": 9203
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.237203015744147e-06,
+      "loss": 0.748,
+      "step": 9204
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2361719596420303e-06,
+      "loss": 0.7859,
+      "step": 9205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2351409891964723e-06,
+      "loss": 0.7446,
+      "step": 9206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.234110104457536e-06,
+      "loss": 0.8257,
+      "step": 9207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2330793054752873e-06,
+      "loss": 0.6901,
+      "step": 9208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2320485922997842e-06,
+      "loss": 0.5551,
+      "step": 9209
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2310179649810785e-06,
+      "loss": 0.6045,
+      "step": 9210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.229987423569224e-06,
+      "loss": 0.6436,
+      "step": 9211
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2289569681142653e-06,
+      "loss": 0.6244,
+      "step": 9212
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.227926598666244e-06,
+      "loss": 0.7763,
+      "step": 9213
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2268963152751966e-06,
+      "loss": 0.5941,
+      "step": 9214
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2258661179911583e-06,
+      "loss": 0.8827,
+      "step": 9215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2248360068641595e-06,
+      "loss": 0.7159,
+      "step": 9216
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2238059819442214e-06,
+      "loss": 0.6151,
+      "step": 9217
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2227760432813694e-06,
+      "loss": 0.8397,
+      "step": 9218
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2217461909256186e-06,
+      "loss": 0.5635,
+      "step": 9219
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2207164249269795e-06,
+      "loss": 0.9133,
+      "step": 9220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.219686745335465e-06,
+      "loss": 0.5075,
+      "step": 9221
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2186571522010744e-06,
+      "loss": 0.9596,
+      "step": 9222
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2176276455738103e-06,
+      "loss": 0.6752,
+      "step": 9223
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.21659822550367e-06,
+      "loss": 0.7033,
+      "step": 9224
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2155688920406415e-06,
+      "loss": 0.7876,
+      "step": 9225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2145396452347133e-06,
+      "loss": 0.6728,
+      "step": 9226
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2135104851358713e-06,
+      "loss": 0.7962,
+      "step": 9227
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2124814117940923e-06,
+      "loss": 0.4191,
+      "step": 9228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2114524252593494e-06,
+      "loss": 0.7776,
+      "step": 9229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.210423525581615e-06,
+      "loss": 0.8062,
+      "step": 9230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.209394712810857e-06,
+      "loss": 0.7365,
+      "step": 9231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.208365986997035e-06,
+      "loss": 0.5652,
+      "step": 9232
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.207337348190107e-06,
+      "loss": 0.6838,
+      "step": 9233
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2063087964400284e-06,
+      "loss": 1.0438,
+      "step": 9234
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2052803317967475e-06,
+      "loss": 0.8776,
+      "step": 9235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.204251954310208e-06,
+      "loss": 0.7168,
+      "step": 9236
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.203223664030355e-06,
+      "loss": 0.511,
+      "step": 9237
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2021954610071227e-06,
+      "loss": 0.7881,
+      "step": 9238
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2011673452904412e-06,
+      "loss": 0.6083,
+      "step": 9239
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.200139316930242e-06,
+      "loss": 0.7693,
+      "step": 9240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1991113759764493e-06,
+      "loss": 0.8414,
+      "step": 9241
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.19808352247898e-06,
+      "loss": 0.6689,
+      "step": 9242
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1970557564877524e-06,
+      "loss": 0.7066,
+      "step": 9243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1960280780526774e-06,
+      "loss": 0.5132,
+      "step": 9244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.19500048722366e-06,
+      "loss": 0.7482,
+      "step": 9245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1939729840506052e-06,
+      "loss": 0.8556,
+      "step": 9246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1929455685834114e-06,
+      "loss": 0.8481,
+      "step": 9247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.191918240871971e-06,
+      "loss": 0.8003,
+      "step": 9248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.190891000966176e-06,
+      "loss": 0.496,
+      "step": 9249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1898638489159106e-06,
+      "loss": 0.4722,
+      "step": 9250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.188836784771058e-06,
+      "loss": 0.8198,
+      "step": 9251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1878098085814926e-06,
+      "loss": 0.5867,
+      "step": 9252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.186782920397089e-06,
+      "loss": 0.7963,
+      "step": 9253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.185756120267718e-06,
+      "loss": 0.7715,
+      "step": 9254
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1847294082432388e-06,
+      "loss": 0.8083,
+      "step": 9255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.183702784373516e-06,
+      "loss": 0.6352,
+      "step": 9256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1826762487084053e-06,
+      "loss": 0.8465,
+      "step": 9257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1816498012977536e-06,
+      "loss": 0.8368,
+      "step": 9258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1806234421914144e-06,
+      "loss": 0.6427,
+      "step": 9259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1795971714392264e-06,
+      "loss": 0.8795,
+      "step": 9260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.178570989091028e-06,
+      "loss": 0.769,
+      "step": 9261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1775448951966577e-06,
+      "loss": 0.8653,
+      "step": 9262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.176518889805941e-06,
+      "loss": 0.7524,
+      "step": 9263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1754929729687057e-06,
+      "loss": 0.856,
+      "step": 9264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.174467144734774e-06,
+      "loss": 0.8664,
+      "step": 9265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.173441405153963e-06,
+      "loss": 0.9132,
+      "step": 9266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.172415754276082e-06,
+      "loss": 0.7931,
+      "step": 9267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1713901921509436e-06,
+      "loss": 0.8942,
+      "step": 9268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1703647188283515e-06,
+      "loss": 0.6248,
+      "step": 9269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.169339334358105e-06,
+      "loss": 0.5407,
+      "step": 9270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.168314038789997e-06,
+      "loss": 0.4908,
+      "step": 9271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1672888321738237e-06,
+      "loss": 0.7365,
+      "step": 9272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.16626371455937e-06,
+      "loss": 0.5731,
+      "step": 9273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1652386859964155e-06,
+      "loss": 0.7544,
+      "step": 9274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1642137465347444e-06,
+      "loss": 0.6921,
+      "step": 9275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.163188896224128e-06,
+      "loss": 0.7748,
+      "step": 9276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.162164135114333e-06,
+      "loss": 0.7858,
+      "step": 9277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1611394632551282e-06,
+      "loss": 0.8187,
+      "step": 9278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1601148806962757e-06,
+      "loss": 0.5894,
+      "step": 9279
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.159090387487528e-06,
+      "loss": 0.6962,
+      "step": 9280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1580659836786397e-06,
+      "loss": 0.7064,
+      "step": 9281
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1570416693193605e-06,
+      "loss": 0.7195,
+      "step": 9282
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1560174444594304e-06,
+      "loss": 0.8344,
+      "step": 9283
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1549933091485912e-06,
+      "loss": 0.5816,
+      "step": 9284
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1539692634365788e-06,
+      "loss": 0.4628,
+      "step": 9285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1529453073731198e-06,
+      "loss": 0.6046,
+      "step": 9286
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.151921441007945e-06,
+      "loss": 0.7118,
+      "step": 9287
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1508976643907733e-06,
+      "loss": 0.4703,
+      "step": 9288
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.149873977571324e-06,
+      "loss": 0.7872,
+      "step": 9289
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.148850380599308e-06,
+      "loss": 0.6858,
+      "step": 9290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.147826873524436e-06,
+      "loss": 0.8701,
+      "step": 9291
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1468034563964135e-06,
+      "loss": 0.9225,
+      "step": 9292
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1457801292649365e-06,
+      "loss": 0.6657,
+      "step": 9293
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1447568921797047e-06,
+      "loss": 0.6471,
+      "step": 9294
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1437337451904083e-06,
+      "loss": 0.735,
+      "step": 9295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1427106883467327e-06,
+      "loss": 0.8365,
+      "step": 9296
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.141687721698363e-06,
+      "loss": 0.535,
+      "step": 9297
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1406648452949764e-06,
+      "loss": 0.5587,
+      "step": 9298
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.139642059186244e-06,
+      "loss": 0.6298,
+      "step": 9299
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1386193634218406e-06,
+      "loss": 0.6609,
+      "step": 9300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1375967580514265e-06,
+      "loss": 0.4904,
+      "step": 9301
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.136574243124664e-06,
+      "loss": 0.8546,
+      "step": 9302
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.135551818691211e-06,
+      "loss": 0.4547,
+      "step": 9303
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.134529484800718e-06,
+      "loss": 0.7463,
+      "step": 9304
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.133507241502832e-06,
+      "loss": 0.8148,
+      "step": 9305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1324850888471945e-06,
+      "loss": 0.846,
+      "step": 9306
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.131463026883449e-06,
+      "loss": 0.7698,
+      "step": 9307
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1304410556612253e-06,
+      "loss": 0.67,
+      "step": 9308
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.129419175230154e-06,
+      "loss": 0.6035,
+      "step": 9309
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1283973856398632e-06,
+      "loss": 1.083,
+      "step": 9310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.127375686939971e-06,
+      "loss": 0.658,
+      "step": 9311
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1263540791800947e-06,
+      "loss": 0.6722,
+      "step": 9312
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1253325624098485e-06,
+      "loss": 0.8523,
+      "step": 9313
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1243111366788394e-06,
+      "loss": 0.8792,
+      "step": 9314
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.123289802036667e-06,
+      "loss": 0.7862,
+      "step": 9315
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1222685585329347e-06,
+      "loss": 0.726,
+      "step": 9316
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1212474062172365e-06,
+      "loss": 0.6652,
+      "step": 9317
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1202263451391603e-06,
+      "loss": 0.5915,
+      "step": 9318
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.119205375348293e-06,
+      "loss": 0.6345,
+      "step": 9319
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.118184496894217e-06,
+      "loss": 0.896,
+      "step": 9320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1171637098265063e-06,
+      "loss": 0.4336,
+      "step": 9321
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.116143014194736e-06,
+      "loss": 0.6942,
+      "step": 9322
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1151224100484734e-06,
+      "loss": 0.8075,
+      "step": 9323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1141018974372816e-06,
+      "loss": 0.637,
+      "step": 9324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1130814764107174e-06,
+      "loss": 0.7749,
+      "step": 9325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.112061147018339e-06,
+      "loss": 0.9033,
+      "step": 9326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.111040909309696e-06,
+      "loss": 0.7367,
+      "step": 9327
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.110020763334331e-06,
+      "loss": 0.6351,
+      "step": 9328
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1090007091417884e-06,
+      "loss": 0.6594,
+      "step": 9329
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1079807467816043e-06,
+      "loss": 0.6364,
+      "step": 9330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.106960876303309e-06,
+      "loss": 0.8337,
+      "step": 9331
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1059410977564323e-06,
+      "loss": 0.6025,
+      "step": 9332
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.104921411190499e-06,
+      "loss": 0.7069,
+      "step": 9333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.103901816655024e-06,
+      "loss": 0.7993,
+      "step": 9334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.102882314199526e-06,
+      "loss": 0.9179,
+      "step": 9335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1018629038735117e-06,
+      "loss": 0.5337,
+      "step": 9336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.1008435857264862e-06,
+      "loss": 0.8866,
+      "step": 9337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0998243598079546e-06,
+      "loss": 0.7464,
+      "step": 9338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.09880522616741e-06,
+      "loss": 0.8099,
+      "step": 9339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0977861848543446e-06,
+      "loss": 0.664,
+      "step": 9340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.096767235918249e-06,
+      "loss": 0.8829,
+      "step": 9341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.095748379408603e-06,
+      "loss": 0.8241,
+      "step": 9342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0947296153748864e-06,
+      "loss": 1.0216,
+      "step": 9343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.093710943866571e-06,
+      "loss": 0.839,
+      "step": 9344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.092692364933132e-06,
+      "loss": 0.6963,
+      "step": 9345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.091673878624029e-06,
+      "loss": 0.8319,
+      "step": 9346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.090655484988724e-06,
+      "loss": 0.9818,
+      "step": 9347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0896371840766758e-06,
+      "loss": 0.744,
+      "step": 9348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0886189759373337e-06,
+      "loss": 0.9705,
+      "step": 9349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0876008606201433e-06,
+      "loss": 0.7518,
+      "step": 9350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0865828381745515e-06,
+      "loss": 0.8422,
+      "step": 9351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.085564908649994e-06,
+      "loss": 0.7021,
+      "step": 9352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0845470720959027e-06,
+      "loss": 0.719,
+      "step": 9353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0835293285617085e-06,
+      "loss": 0.9409,
+      "step": 9354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.082511678096838e-06,
+      "loss": 0.6649,
+      "step": 9355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0814941207507053e-06,
+      "loss": 0.8613,
+      "step": 9356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.080476656572732e-06,
+      "loss": 0.7697,
+      "step": 9357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0794592856123272e-06,
+      "loss": 1.0087,
+      "step": 9358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.078442007918895e-06,
+      "loss": 0.7143,
+      "step": 9359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.07742482354184e-06,
+      "loss": 0.5353,
+      "step": 9360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.07640773253056e-06,
+      "loss": 0.5866,
+      "step": 9361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.075390734934446e-06,
+      "loss": 0.6556,
+      "step": 9362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.074373830802886e-06,
+      "loss": 0.7284,
+      "step": 9363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.073357020185265e-06,
+      "loss": 0.7705,
+      "step": 9364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0723403031309636e-06,
+      "loss": 0.8619,
+      "step": 9365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.071323679689352e-06,
+      "loss": 0.8196,
+      "step": 9366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.070307149909805e-06,
+      "loss": 0.6272,
+      "step": 9367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0692907138416874e-06,
+      "loss": 0.6199,
+      "step": 9368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0682743715343565e-06,
+      "loss": 0.8097,
+      "step": 9369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0672581230371747e-06,
+      "loss": 0.8447,
+      "step": 9370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0662419683994893e-06,
+      "loss": 0.6374,
+      "step": 9371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.065225907670648e-06,
+      "loss": 0.6762,
+      "step": 9372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0642099408999982e-06,
+      "loss": 0.5683,
+      "step": 9373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0631940681368723e-06,
+      "loss": 0.8184,
+      "step": 9374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.062178289430606e-06,
+      "loss": 0.7486,
+      "step": 9375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.061162604830531e-06,
+      "loss": 0.6726,
+      "step": 9376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.060147014385969e-06,
+      "loss": 0.5084,
+      "step": 9377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.05913151814624e-06,
+      "loss": 0.6977,
+      "step": 9378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0581161161606623e-06,
+      "loss": 0.996,
+      "step": 9379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0571008084785443e-06,
+      "loss": 0.6238,
+      "step": 9380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.056085595149192e-06,
+      "loss": 0.7006,
+      "step": 9381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.055070476221906e-06,
+      "loss": 0.9397,
+      "step": 9382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0540554517459874e-06,
+      "loss": 0.5825,
+      "step": 9383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.053040521770726e-06,
+      "loss": 0.7112,
+      "step": 9384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0520256863454077e-06,
+      "loss": 0.6768,
+      "step": 9385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0510109455193217e-06,
+      "loss": 0.9407,
+      "step": 9386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.049996299341742e-06,
+      "loss": 0.8095,
+      "step": 9387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0489817478619428e-06,
+      "loss": 0.7415,
+      "step": 9388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.047967291129197e-06,
+      "loss": 0.6095,
+      "step": 9389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.046952929192767e-06,
+      "loss": 0.6274,
+      "step": 9390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0459386621019114e-06,
+      "loss": 0.7646,
+      "step": 9391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0449244899058895e-06,
+      "loss": 0.7602,
+      "step": 9392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.043910412653952e-06,
+      "loss": 0.8095,
+      "step": 9393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.042896430395342e-06,
+      "loss": 0.5893,
+      "step": 9394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0418825431793054e-06,
+      "loss": 0.5893,
+      "step": 9395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0408687510550777e-06,
+      "loss": 0.7058,
+      "step": 9396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0398550540718908e-06,
+      "loss": 0.6863,
+      "step": 9397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0388414522789733e-06,
+      "loss": 0.8466,
+      "step": 9398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.037827945725551e-06,
+      "loss": 0.6901,
+      "step": 9399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.03681453446084e-06,
+      "loss": 0.5322,
+      "step": 9400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.035801218534054e-06,
+      "loss": 0.7043,
+      "step": 9401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.034787997994404e-06,
+      "loss": 0.7929,
+      "step": 9402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0337748728910956e-06,
+      "loss": 0.8676,
+      "step": 9403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.032761843273326e-06,
+      "loss": 0.698,
+      "step": 9404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0317489091902936e-06,
+      "loss": 0.7947,
+      "step": 9405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0307360706911897e-06,
+      "loss": 0.5546,
+      "step": 9406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.029723327825198e-06,
+      "loss": 0.5751,
+      "step": 9407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0287106806415035e-06,
+      "loss": 0.6838,
+      "step": 9408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.02769812918928e-06,
+      "loss": 0.8624,
+      "step": 9409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.026685673517701e-06,
+      "loss": 0.6033,
+      "step": 9410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0256733136759363e-06,
+      "loss": 0.7393,
+      "step": 9411
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0246610497131463e-06,
+      "loss": 0.5112,
+      "step": 9412
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0236488816784894e-06,
+      "loss": 0.4878,
+      "step": 9413
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0226368096211234e-06,
+      "loss": 0.6677,
+      "step": 9414
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.021624833590193e-06,
+      "loss": 0.7485,
+      "step": 9415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.020612953634845e-06,
+      "loss": 0.529,
+      "step": 9416
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.019601169804216e-06,
+      "loss": 0.3841,
+      "step": 9417
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.018589482147446e-06,
+      "loss": 0.7674,
+      "step": 9418
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0175778907136623e-06,
+      "loss": 0.7165,
+      "step": 9419
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0165663955519896e-06,
+      "loss": 0.7202,
+      "step": 9420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0155549967115536e-06,
+      "loss": 0.9044,
+      "step": 9421
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0145436942414662e-06,
+      "loss": 0.6381,
+      "step": 9422
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0135324881908405e-06,
+      "loss": 0.6838,
+      "step": 9423
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.012521378608786e-06,
+      "loss": 0.7581,
+      "step": 9424
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0115103655444017e-06,
+      "loss": 0.7557,
+      "step": 9425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0104994490467853e-06,
+      "loss": 0.8288,
+      "step": 9426
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0094886291650337e-06,
+      "loss": 0.6646,
+      "step": 9427
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.008477905948232e-06,
+      "loss": 0.8355,
+      "step": 9428
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.007467279445463e-06,
+      "loss": 1.0899,
+      "step": 9429
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0064567497058085e-06,
+      "loss": 0.6069,
+      "step": 9430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.005446316778341e-06,
+      "loss": 0.6017,
+      "step": 9431
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0044359807121295e-06,
+      "loss": 0.6432,
+      "step": 9432
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.00342574155624e-06,
+      "loss": 0.5805,
+      "step": 9433
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.002415599359734e-06,
+      "loss": 0.9452,
+      "step": 9434
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.0014055541716637e-06,
+      "loss": 0.7352,
+      "step": 9435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.00039560604108e-06,
+      "loss": 0.5404,
+      "step": 9436
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9993857550170323e-06,
+      "loss": 0.5885,
+      "step": 9437
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9983760011485607e-06,
+      "loss": 0.7649,
+      "step": 9438
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.997366344484698e-06,
+      "loss": 0.7305,
+      "step": 9439
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9963567850744797e-06,
+      "loss": 0.7529,
+      "step": 9440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.995347322966933e-06,
+      "loss": 0.6085,
+      "step": 9441
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9943379582110778e-06,
+      "loss": 0.7809,
+      "step": 9442
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9933286908559335e-06,
+      "loss": 0.7181,
+      "step": 9443
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.992319520950514e-06,
+      "loss": 0.7192,
+      "step": 9444
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.991310448543825e-06,
+      "loss": 0.8031,
+      "step": 9445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9903014736848724e-06,
+      "loss": 0.8306,
+      "step": 9446
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9892925964226527e-06,
+      "loss": 0.6583,
+      "step": 9447
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9882838168061606e-06,
+      "loss": 0.642,
+      "step": 9448
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9872751348843875e-06,
+      "loss": 0.9416,
+      "step": 9449
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.986266550706315e-06,
+      "loss": 0.6578,
+      "step": 9450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.985258064320922e-06,
+      "loss": 0.8694,
+      "step": 9451
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9842496757771883e-06,
+      "loss": 0.8275,
+      "step": 9452
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9832413851240803e-06,
+      "loss": 0.9497,
+      "step": 9453
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.982233192410565e-06,
+      "loss": 0.9408,
+      "step": 9454
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9812250976856006e-06,
+      "loss": 0.8483,
+      "step": 9455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9802171009981475e-06,
+      "loss": 0.915,
+      "step": 9456
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.979209202397153e-06,
+      "loss": 0.7649,
+      "step": 9457
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9782014019315635e-06,
+      "loss": 0.7925,
+      "step": 9458
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.977193699650325e-06,
+      "loss": 0.6853,
+      "step": 9459
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.97618609560237e-06,
+      "loss": 0.5868,
+      "step": 9460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.975178589836632e-06,
+      "loss": 0.6482,
+      "step": 9461
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9741711824020405e-06,
+      "loss": 0.3434,
+      "step": 9462
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9731638733475142e-06,
+      "loss": 0.6372,
+      "step": 9463
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9721566627219734e-06,
+      "loss": 0.6815,
+      "step": 9464
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9711495505743317e-06,
+      "loss": 0.8231,
+      "step": 9465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9701425369534964e-06,
+      "loss": 0.8274,
+      "step": 9466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9691356219083687e-06,
+      "loss": 0.6602,
+      "step": 9467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.968128805487851e-06,
+      "loss": 0.4782,
+      "step": 9468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9671220877408367e-06,
+      "loss": 0.9309,
+      "step": 9469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9661154687162117e-06,
+      "loss": 0.846,
+      "step": 9470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.965108948462864e-06,
+      "loss": 0.7438,
+      "step": 9471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9641025270296725e-06,
+      "loss": 0.7184,
+      "step": 9472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9630962044655097e-06,
+      "loss": 0.6935,
+      "step": 9473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.962089980819246e-06,
+      "loss": 0.7412,
+      "step": 9474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9610838561397492e-06,
+      "loss": 0.7166,
+      "step": 9475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.960077830475878e-06,
+      "loss": 0.6424,
+      "step": 9476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.959071903876486e-06,
+      "loss": 0.9014,
+      "step": 9477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9580660763904257e-06,
+      "loss": 0.3998,
+      "step": 9478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.957060348066545e-06,
+      "loss": 0.9055,
+      "step": 9479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9560547189536797e-06,
+      "loss": 0.9623,
+      "step": 9480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9550491891006704e-06,
+      "loss": 0.6866,
+      "step": 9481
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9540437585563487e-06,
+      "loss": 0.8144,
+      "step": 9482
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9530384273695373e-06,
+      "loss": 0.6343,
+      "step": 9483
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9520331955890624e-06,
+      "loss": 0.785,
+      "step": 9484
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9510280632637386e-06,
+      "loss": 0.5062,
+      "step": 9485
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9500230304423765e-06,
+      "loss": 1.0155,
+      "step": 9486
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9490180971737882e-06,
+      "loss": 0.8475,
+      "step": 9487
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9480132635067717e-06,
+      "loss": 0.7869,
+      "step": 9488
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9470085294901244e-06,
+      "loss": 0.6848,
+      "step": 9489
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.946003895172643e-06,
+      "loss": 0.6592,
+      "step": 9490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9449993606031137e-06,
+      "loss": 0.7123,
+      "step": 9491
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9439949258303187e-06,
+      "loss": 0.7218,
+      "step": 9492
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.942990590903035e-06,
+      "loss": 0.6889,
+      "step": 9493
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9419863558700397e-06,
+      "loss": 0.6959,
+      "step": 9494
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9409822207800976e-06,
+      "loss": 0.6031,
+      "step": 9495
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9399781856819744e-06,
+      "loss": 0.5895,
+      "step": 9496
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.938974250624429e-06,
+      "loss": 0.76,
+      "step": 9497
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.937970415656215e-06,
+      "loss": 0.4435,
+      "step": 9498
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.93696668082608e-06,
+      "loss": 0.8779,
+      "step": 9499
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9359630461827714e-06,
+      "loss": 0.5445,
+      "step": 9500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9349595117750267e-06,
+      "loss": 0.8067,
+      "step": 9501
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9339560776515797e-06,
+      "loss": 0.9286,
+      "step": 9502
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.932952743861162e-06,
+      "loss": 0.7481,
+      "step": 9503
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9319495104524974e-06,
+      "loss": 0.6224,
+      "step": 9504
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9309463774743047e-06,
+      "loss": 0.7734,
+      "step": 9505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9299433449753e-06,
+      "loss": 1.0021,
+      "step": 9506
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.928940413004195e-06,
+      "loss": 0.7366,
+      "step": 9507
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.927937581609691e-06,
+      "loss": 0.7434,
+      "step": 9508
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9269348508404916e-06,
+      "loss": 0.8849,
+      "step": 9509
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9259322207452922e-06,
+      "loss": 0.5642,
+      "step": 9510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.924929691372782e-06,
+      "loss": 0.7079,
+      "step": 9511
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9239272627716462e-06,
+      "loss": 0.9265,
+      "step": 9512
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9229249349905686e-06,
+      "loss": 0.9257,
+      "step": 9513
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.921922708078223e-06,
+      "loss": 0.5522,
+      "step": 9514
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9209205820832788e-06,
+      "loss": 0.6045,
+      "step": 9515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9199185570544054e-06,
+      "loss": 0.6408,
+      "step": 9516
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.918916633040263e-06,
+      "loss": 0.6069,
+      "step": 9517
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.917914810089505e-06,
+      "loss": 0.729,
+      "step": 9518
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.916913088250789e-06,
+      "loss": 0.8608,
+      "step": 9519
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9159114675727553e-06,
+      "loss": 0.7706,
+      "step": 9520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9149099481040465e-06,
+      "loss": 0.6854,
+      "step": 9521
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9139085298933043e-06,
+      "loss": 0.7619,
+      "step": 9522
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9129072129891558e-06,
+      "loss": 0.6943,
+      "step": 9523
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9119059974402277e-06,
+      "loss": 0.8904,
+      "step": 9524
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9109048832951435e-06,
+      "loss": 0.6164,
+      "step": 9525
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9099038706025197e-06,
+      "loss": 0.545,
+      "step": 9526
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9089029594109686e-06,
+      "loss": 0.517,
+      "step": 9527
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9079021497690972e-06,
+      "loss": 0.7027,
+      "step": 9528
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.906901441725507e-06,
+      "loss": 0.9375,
+      "step": 9529
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9059008353287963e-06,
+      "loss": 0.8596,
+      "step": 9530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9049003306275575e-06,
+      "loss": 0.6611,
+      "step": 9531
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.903899927670377e-06,
+      "loss": 0.6971,
+      "step": 9532
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9028996265058395e-06,
+      "loss": 0.8445,
+      "step": 9533
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9018994271825172e-06,
+      "loss": 0.7455,
+      "step": 9534
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9008993297489885e-06,
+      "loss": 0.6858,
+      "step": 9535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8998993342538197e-06,
+      "loss": 0.7969,
+      "step": 9536
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.898899440745569e-06,
+      "loss": 0.7523,
+      "step": 9537
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8978996492727994e-06,
+      "loss": 0.7442,
+      "step": 9538
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.896899959884064e-06,
+      "loss": 0.7632,
+      "step": 9539
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8959003726279044e-06,
+      "loss": 0.6863,
+      "step": 9540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8949008875528706e-06,
+      "loss": 0.6856,
+      "step": 9541
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.893901504707496e-06,
+      "loss": 0.7337,
+      "step": 9542
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8929022241403126e-06,
+      "loss": 0.4509,
+      "step": 9543
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.891903045899854e-06,
+      "loss": 0.6692,
+      "step": 9544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8909039700346385e-06,
+      "loss": 0.6382,
+      "step": 9545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.889904996593185e-06,
+      "loss": 0.7815,
+      "step": 9546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.888906125624007e-06,
+      "loss": 0.8899,
+      "step": 9547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.887907357175612e-06,
+      "loss": 0.5965,
+      "step": 9548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.886908691296504e-06,
+      "loss": 0.7671,
+      "step": 9549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.88591012803518e-06,
+      "loss": 0.6598,
+      "step": 9550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8849116674401333e-06,
+      "loss": 0.8138,
+      "step": 9551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8839133095598527e-06,
+      "loss": 0.7117,
+      "step": 9552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.88291505444282e-06,
+      "loss": 0.7955,
+      "step": 9553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8819169021375154e-06,
+      "loss": 0.517,
+      "step": 9554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8809188526924105e-06,
+      "loss": 0.7575,
+      "step": 9555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.879920906155974e-06,
+      "loss": 0.9346,
+      "step": 9556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8789230625766686e-06,
+      "loss": 0.5271,
+      "step": 9557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.877925322002953e-06,
+      "loss": 0.7185,
+      "step": 9558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.87692768448328e-06,
+      "loss": 0.7325,
+      "step": 9559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8759301500660975e-06,
+      "loss": 0.5982,
+      "step": 9560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8749327187998516e-06,
+      "loss": 0.5276,
+      "step": 9561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8739353907329736e-06,
+      "loss": 0.6884,
+      "step": 9562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8729381659139023e-06,
+      "loss": 0.7957,
+      "step": 9563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.871941044391067e-06,
+      "loss": 0.7016,
+      "step": 9564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8709440262128862e-06,
+      "loss": 0.5746,
+      "step": 9565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.869947111427777e-06,
+      "loss": 0.6422,
+      "step": 9566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.86895030008416e-06,
+      "loss": 0.6683,
+      "step": 9567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8679535922304367e-06,
+      "loss": 0.9435,
+      "step": 9568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8669569879150096e-06,
+      "loss": 0.6594,
+      "step": 9569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8659604871862834e-06,
+      "loss": 0.7651,
+      "step": 9570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.864964090092644e-06,
+      "loss": 1.0326,
+      "step": 9571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.863967796682482e-06,
+      "loss": 0.5365,
+      "step": 9572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.862971607004181e-06,
+      "loss": 0.9501,
+      "step": 9573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8619755211061175e-06,
+      "loss": 0.7389,
+      "step": 9574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8609795390366645e-06,
+      "loss": 0.5181,
+      "step": 9575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.859983660844191e-06,
+      "loss": 0.7802,
+      "step": 9576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.858987886577058e-06,
+      "loss": 0.6893,
+      "step": 9577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8579922162836245e-06,
+      "loss": 0.5418,
+      "step": 9578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8569966500122424e-06,
+      "loss": 0.9092,
+      "step": 9579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8560011878112594e-06,
+      "loss": 1.0023,
+      "step": 9580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855005829729018e-06,
+      "loss": 0.7112,
+      "step": 9581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.854010575813856e-06,
+      "loss": 0.5588,
+      "step": 9582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8530154261141054e-06,
+      "loss": 0.5567,
+      "step": 9583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.852020380678095e-06,
+      "loss": 0.6105,
+      "step": 9584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.851025439554142e-06,
+      "loss": 0.9037,
+      "step": 9585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.850030602790569e-06,
+      "loss": 0.6649,
+      "step": 9586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8490358704356892e-06,
+      "loss": 0.9631,
+      "step": 9587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8480412425378024e-06,
+      "loss": 0.8637,
+      "step": 9588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.847046719145219e-06,
+      "loss": 0.7027,
+      "step": 9589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8460523003062303e-06,
+      "loss": 0.7003,
+      "step": 9590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8450579860691276e-06,
+      "loss": 0.7202,
+      "step": 9591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.844063776482203e-06,
+      "loss": 0.7116,
+      "step": 9592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.843069671593734e-06,
+      "loss": 0.5499,
+      "step": 9593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.842075671451996e-06,
+      "loss": 0.6986,
+      "step": 9594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.841081776105267e-06,
+      "loss": 0.9924,
+      "step": 9595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8400879856018064e-06,
+      "loss": 0.8121,
+      "step": 9596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.839094299989877e-06,
+      "loss": 0.9104,
+      "step": 9597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.83810071931774e-06,
+      "loss": 0.6548,
+      "step": 9598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8371072436336412e-06,
+      "loss": 0.6494,
+      "step": 9599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.836113872985827e-06,
+      "loss": 0.7159,
+      "step": 9600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8351206074225408e-06,
+      "loss": 0.5768,
+      "step": 9601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8341274469920165e-06,
+      "loss": 0.5857,
+      "step": 9602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8331343917424857e-06,
+      "loss": 0.7149,
+      "step": 9603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.832141441722175e-06,
+      "loss": 0.8547,
+      "step": 9604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.831148596979302e-06,
+      "loss": 0.8962,
+      "step": 9605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8301558575620846e-06,
+      "loss": 0.6024,
+      "step": 9606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8291632235187328e-06,
+      "loss": 0.9356,
+      "step": 9607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.828170694897451e-06,
+      "loss": 0.5815,
+      "step": 9608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8271782717464413e-06,
+      "loss": 0.7653,
+      "step": 9609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8261859541138927e-06,
+      "loss": 0.7455,
+      "step": 9610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8251937420480014e-06,
+      "loss": 0.7397,
+      "step": 9611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8242016355969517e-06,
+      "loss": 0.7349,
+      "step": 9612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8232096348089172e-06,
+      "loss": 0.6773,
+      "step": 9613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8222177397320782e-06,
+      "loss": 0.7839,
+      "step": 9614
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8212259504146045e-06,
+      "loss": 0.9624,
+      "step": 9615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8202342669046538e-06,
+      "loss": 0.4796,
+      "step": 9616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8192426892503933e-06,
+      "loss": 0.7189,
+      "step": 9617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.818251217499971e-06,
+      "loss": 0.5659,
+      "step": 9618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.817259851701535e-06,
+      "loss": 0.7657,
+      "step": 9619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8162685919032346e-06,
+      "loss": 0.7514,
+      "step": 9620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8152774381532033e-06,
+      "loss": 0.7058,
+      "step": 9621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.814286390499576e-06,
+      "loss": 0.7612,
+      "step": 9622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8132954489904808e-06,
+      "loss": 0.5447,
+      "step": 9623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8123046136740407e-06,
+      "loss": 0.7428,
+      "step": 9624
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.811313884598373e-06,
+      "loss": 0.9026,
+      "step": 9625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8103232618115906e-06,
+      "loss": 0.7235,
+      "step": 9626
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.809332745361802e-06,
+      "loss": 0.6762,
+      "step": 9627
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8083423352971085e-06,
+      "loss": 0.7702,
+      "step": 9628
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8073520316656067e-06,
+      "loss": 0.8311,
+      "step": 9629
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8063618345153897e-06,
+      "loss": 0.7848,
+      "step": 9630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805371743894544e-06,
+      "loss": 0.9522,
+      "step": 9631
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8043817598511514e-06,
+      "loss": 0.6126,
+      "step": 9632
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803391882433288e-06,
+      "loss": 0.6768,
+      "step": 9633
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.802402111689025e-06,
+      "loss": 0.9119,
+      "step": 9634
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.801412447666429e-06,
+      "loss": 0.9222,
+      "step": 9635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8004228904135606e-06,
+      "loss": 0.9093,
+      "step": 9636
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7994334399784773e-06,
+      "loss": 0.4846,
+      "step": 9637
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7984440964092245e-06,
+      "loss": 0.8022,
+      "step": 9638
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.797454859753853e-06,
+      "loss": 0.7066,
+      "step": 9639
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7964657300604026e-06,
+      "loss": 0.9855,
+      "step": 9640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.795476707376905e-06,
+      "loss": 0.8,
+      "step": 9641
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.79448779175139e-06,
+      "loss": 0.4999,
+      "step": 9642
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.793498983231888e-06,
+      "loss": 0.8588,
+      "step": 9643
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7925102818664112e-06,
+      "loss": 0.8219,
+      "step": 9644
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7915216877029754e-06,
+      "loss": 1.1701,
+      "step": 9645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.790533200789595e-06,
+      "loss": 0.6473,
+      "step": 9646
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.789544821174268e-06,
+      "loss": 0.8177,
+      "step": 9647
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7885565489049948e-06,
+      "loss": 0.5164,
+      "step": 9648
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7875683840297684e-06,
+      "loss": 0.7437,
+      "step": 9649
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.786580326596577e-06,
+      "loss": 0.6753,
+      "step": 9650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7855923766534037e-06,
+      "loss": 0.6175,
+      "step": 9651
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7846045342482265e-06,
+      "loss": 0.8044,
+      "step": 9652
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7836167994290175e-06,
+      "loss": 0.7154,
+      "step": 9653
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7826291722437437e-06,
+      "loss": 0.6132,
+      "step": 9654
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.781641652740368e-06,
+      "loss": 0.7804,
+      "step": 9655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7806542409668457e-06,
+      "loss": 0.6479,
+      "step": 9656
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7796669369711294e-06,
+      "loss": 0.5647,
+      "step": 9657
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7786797408011656e-06,
+      "loss": 0.7457,
+      "step": 9658
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7776926525048953e-06,
+      "loss": 0.7059,
+      "step": 9659
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7767056721302553e-06,
+      "loss": 0.7677,
+      "step": 9660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7757187997251717e-06,
+      "loss": 0.7411,
+      "step": 9661
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.774732035337575e-06,
+      "loss": 0.6327,
+      "step": 9662
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7737453790153853e-06,
+      "loss": 0.6157,
+      "step": 9663
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.772758830806511e-06,
+      "loss": 0.8869,
+      "step": 9664
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.771772390758871e-06,
+      "loss": 0.808,
+      "step": 9665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.770786058920364e-06,
+      "loss": 0.8482,
+      "step": 9666
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7697998353388865e-06,
+      "loss": 0.6198,
+      "step": 9667
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7688137200623406e-06,
+      "loss": 0.9281,
+      "step": 9668
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7678277131386085e-06,
+      "loss": 0.9135,
+      "step": 9669
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.766841814615573e-06,
+      "loss": 0.6197,
+      "step": 9670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7658560245411183e-06,
+      "loss": 0.6334,
+      "step": 9671
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7648703429631114e-06,
+      "loss": 0.7269,
+      "step": 9672
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7638847699294196e-06,
+      "loss": 0.5372,
+      "step": 9673
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7628993054879115e-06,
+      "loss": 0.8335,
+      "step": 9674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.761913949686438e-06,
+      "loss": 0.5786,
+      "step": 9675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7609287025728524e-06,
+      "loss": 0.7226,
+      "step": 9676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7599435641950023e-06,
+      "loss": 0.751,
+      "step": 9677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7589585346007275e-06,
+      "loss": 0.7299,
+      "step": 9678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.757973613837864e-06,
+      "loss": 0.4871,
+      "step": 9679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7569888019542427e-06,
+      "loss": 0.8624,
+      "step": 9680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7560040989976894e-06,
+      "loss": 0.559,
+      "step": 9681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.755019505016023e-06,
+      "loss": 0.7387,
+      "step": 9682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7540350200570597e-06,
+      "loss": 0.7144,
+      "step": 9683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.753050644168608e-06,
+      "loss": 0.8283,
+      "step": 9684
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7520663773984733e-06,
+      "loss": 0.8204,
+      "step": 9685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7510822197944498e-06,
+      "loss": 0.6029,
+      "step": 9686
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7500981714043367e-06,
+      "loss": 0.5625,
+      "step": 9687
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.749114232275922e-06,
+      "loss": 0.5236,
+      "step": 9688
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7481304024569823e-06,
+      "loss": 0.4644,
+      "step": 9689
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7471466819953007e-06,
+      "loss": 0.5633,
+      "step": 9690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.746163070938651e-06,
+      "loss": 0.6095,
+      "step": 9691
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.745179569334794e-06,
+      "loss": 0.7718,
+      "step": 9692
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.744196177231498e-06,
+      "loss": 0.7673,
+      "step": 9693
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.743212894676514e-06,
+      "loss": 0.5485,
+      "step": 9694
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.742229721717595e-06,
+      "loss": 0.8678,
+      "step": 9695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7412466584024878e-06,
+      "loss": 0.6619,
+      "step": 9696
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7402637047789314e-06,
+      "loss": 0.666,
+      "step": 9697
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7392808608946614e-06,
+      "loss": 0.4359,
+      "step": 9698
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7382981267974073e-06,
+      "loss": 0.7435,
+      "step": 9699
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7373155025348943e-06,
+      "loss": 0.6634,
+      "step": 9700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7363329881548395e-06,
+      "loss": 0.5966,
+      "step": 9701
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.735350583704959e-06,
+      "loss": 0.6377,
+      "step": 9702
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.734368289232959e-06,
+      "loss": 0.8341,
+      "step": 9703
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.733386104786545e-06,
+      "loss": 0.7557,
+      "step": 9704
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7324040304134125e-06,
+      "loss": 0.7289,
+      "step": 9705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7314220661612543e-06,
+      "loss": 0.7658,
+      "step": 9706
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7304402120777584e-06,
+      "loss": 0.9975,
+      "step": 9707
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.729458468210606e-06,
+      "loss": 0.7735,
+      "step": 9708
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7284768346074732e-06,
+      "loss": 0.6968,
+      "step": 9709
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7274953113160303e-06,
+      "loss": 0.595,
+      "step": 9710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.726513898383944e-06,
+      "loss": 0.8706,
+      "step": 9711
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7255325958588742e-06,
+      "loss": 0.6197,
+      "step": 9712
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7245514037884773e-06,
+      "loss": 0.8945,
+      "step": 9713
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.723570322220399e-06,
+      "loss": 0.7125,
+      "step": 9714
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7225893512022846e-06,
+      "loss": 0.6563,
+      "step": 9715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7216084907817775e-06,
+      "loss": 0.8087,
+      "step": 9716
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.720627741006505e-06,
+      "loss": 0.7096,
+      "step": 9717
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7196471019240963e-06,
+      "loss": 0.6968,
+      "step": 9718
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.718666573582179e-06,
+      "loss": 0.6934,
+      "step": 9719
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7176861560283646e-06,
+      "loss": 0.6012,
+      "step": 9720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.716705849310265e-06,
+      "loss": 0.6925,
+      "step": 9721
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7157256534754933e-06,
+      "loss": 0.7145,
+      "step": 9722
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.714745568571644e-06,
+      "loss": 0.7451,
+      "step": 9723
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.713765594646315e-06,
+      "loss": 0.4959,
+      "step": 9724
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7127857317470967e-06,
+      "loss": 0.9344,
+      "step": 9725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7118059799215746e-06,
+      "loss": 0.7303,
+      "step": 9726
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.710826339217327e-06,
+      "loss": 0.4692,
+      "step": 9727
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7098468096819286e-06,
+      "loss": 0.5986,
+      "step": 9728
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.708867391362948e-06,
+      "loss": 0.7977,
+      "step": 9729
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.70788808430795e-06,
+      "loss": 0.8526,
+      "step": 9730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.706908888564491e-06,
+      "loss": 1.0205,
+      "step": 9731
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.705929804180124e-06,
+      "loss": 0.7795,
+      "step": 9732
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7049508312023977e-06,
+      "loss": 0.715,
+      "step": 9733
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7039719696788488e-06,
+      "loss": 0.8081,
+      "step": 9734
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.7029932196570188e-06,
+      "loss": 0.7991,
+      "step": 9735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.702014581184439e-06,
+      "loss": 0.6261,
+      "step": 9736
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.701036054308629e-06,
+      "loss": 0.8675,
+      "step": 9737
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.700057639077115e-06,
+      "loss": 0.7328,
+      "step": 9738
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6990793355374105e-06,
+      "loss": 0.8241,
+      "step": 9739
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6981011437370204e-06,
+      "loss": 0.6857,
+      "step": 9740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6971230637234554e-06,
+      "loss": 0.7886,
+      "step": 9741
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.696145095544208e-06,
+      "loss": 0.7476,
+      "step": 9742
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6951672392467715e-06,
+      "loss": 0.6935,
+      "step": 9743
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6941894948786397e-06,
+      "loss": 0.7653,
+      "step": 9744
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.693211862487287e-06,
+      "loss": 0.6552,
+      "step": 9745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6922343421201924e-06,
+      "loss": 0.5579,
+      "step": 9746
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6912569338248317e-06,
+      "loss": 0.5713,
+      "step": 9747
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.690279637648664e-06,
+      "loss": 0.8185,
+      "step": 9748
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6893024536391517e-06,
+      "loss": 0.9248,
+      "step": 9749
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.688325381843754e-06,
+      "loss": 0.769,
+      "step": 9750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6873484223099143e-06,
+      "loss": 0.6681,
+      "step": 9751
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6863715750850795e-06,
+      "loss": 0.7237,
+      "step": 9752
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.685394840216688e-06,
+      "loss": 0.8942,
+      "step": 9753
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6844182177521716e-06,
+      "loss": 0.6468,
+      "step": 9754
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.68344170773896e-06,
+      "loss": 0.7474,
+      "step": 9755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.682465310224473e-06,
+      "loss": 0.8839,
+      "step": 9756
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6814890252561288e-06,
+      "loss": 1.1139,
+      "step": 9757
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6805128528813382e-06,
+      "loss": 0.7369,
+      "step": 9758
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.679536793147507e-06,
+      "loss": 0.7461,
+      "step": 9759
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.678560846102035e-06,
+      "loss": 0.9162,
+      "step": 9760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.67758501179232e-06,
+      "loss": 0.8707,
+      "step": 9761
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6766092902657447e-06,
+      "loss": 0.8261,
+      "step": 9762
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.675633681569698e-06,
+      "loss": 0.6842,
+      "step": 9763
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6746581857515607e-06,
+      "loss": 0.6507,
+      "step": 9764
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.673682802858697e-06,
+      "loss": 0.5884,
+      "step": 9765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.672707532938482e-06,
+      "loss": 0.6509,
+      "step": 9766
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.671732376038277e-06,
+      "loss": 0.8696,
+      "step": 9767
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6707573322054327e-06,
+      "loss": 0.7297,
+      "step": 9768
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6697824014873076e-06,
+      "loss": 0.8662,
+      "step": 9769
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.668807583931241e-06,
+      "loss": 0.921,
+      "step": 9770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.667832879584575e-06,
+      "loss": 0.7,
+      "step": 9771
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6668582884946454e-06,
+      "loss": 0.6889,
+      "step": 9772
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.665883810708779e-06,
+      "loss": 0.6517,
+      "step": 9773
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.664909446274301e-06,
+      "loss": 0.6813,
+      "step": 9774
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.663935195238528e-06,
+      "loss": 0.892,
+      "step": 9775
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6629610576487737e-06,
+      "loss": 0.5928,
+      "step": 9776
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6619870335523434e-06,
+      "loss": 0.7151,
+      "step": 9777
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.66101312299654e-06,
+      "loss": 0.5055,
+      "step": 9778
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.660039326028659e-06,
+      "loss": 0.6491,
+      "step": 9779
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6590656426959906e-06,
+      "loss": 0.6065,
+      "step": 9780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6580920730458194e-06,
+      "loss": 0.6421,
+      "step": 9781
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.657118617125425e-06,
+      "loss": 0.6029,
+      "step": 9782
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.656145274982081e-06,
+      "loss": 0.613,
+      "step": 9783
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.655172046663056e-06,
+      "loss": 0.6126,
+      "step": 9784
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.654198932215613e-06,
+      "loss": 0.8687,
+      "step": 9785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.653225931687009e-06,
+      "loss": 0.6071,
+      "step": 9786
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6522530451244955e-06,
+      "loss": 0.8194,
+      "step": 9787
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6512802725753183e-06,
+      "loss": 0.8123,
+      "step": 9788
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.650307614086721e-06,
+      "loss": 0.787,
+      "step": 9789
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.649335069705934e-06,
+      "loss": 0.7361,
+      "step": 9790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6483626394801877e-06,
+      "loss": 0.7262,
+      "step": 9791
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.64739032345671e-06,
+      "loss": 0.5079,
+      "step": 9792
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6464181216827164e-06,
+      "loss": 0.5051,
+      "step": 9793
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6454460342054174e-06,
+      "loss": 0.8928,
+      "step": 9794
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.644474061072027e-06,
+      "loss": 0.5986,
+      "step": 9795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.643502202329742e-06,
+      "loss": 0.7277,
+      "step": 9796
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.642530458025759e-06,
+      "loss": 0.4758,
+      "step": 9797
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6415588282072702e-06,
+      "loss": 1.1365,
+      "step": 9798
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6405873129214595e-06,
+      "loss": 0.6318,
+      "step": 9799
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6396159122155074e-06,
+      "loss": 0.7603,
+      "step": 9800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6386446261365874e-06,
+      "loss": 1.0521,
+      "step": 9801
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6376734547318685e-06,
+      "loss": 0.594,
+      "step": 9802
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6367023980485122e-06,
+      "loss": 0.7925,
+      "step": 9803
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6357314561336777e-06,
+      "loss": 0.7442,
+      "step": 9804
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6347606290345153e-06,
+      "loss": 0.6764,
+      "step": 9805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.633789916798174e-06,
+      "loss": 0.7813,
+      "step": 9806
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6328193194717875e-06,
+      "loss": 0.6854,
+      "step": 9807
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6318488371024973e-06,
+      "loss": 0.7252,
+      "step": 9808
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6308784697374335e-06,
+      "loss": 0.7911,
+      "step": 9809
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6299082174237135e-06,
+      "loss": 0.791,
+      "step": 9810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6289380802084607e-06,
+      "loss": 0.6742,
+      "step": 9811
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6279680581387885e-06,
+      "loss": 0.7676,
+      "step": 9812
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.626998151261798e-06,
+      "loss": 0.7811,
+      "step": 9813
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.626028359624596e-06,
+      "loss": 0.6052,
+      "step": 9814
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6250586832742792e-06,
+      "loss": 0.3344,
+      "step": 9815
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6240891222579314e-06,
+      "loss": 0.6708,
+      "step": 9816
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.623119676622645e-06,
+      "loss": 0.7767,
+      "step": 9817
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.622150346415494e-06,
+      "loss": 0.7212,
+      "step": 9818
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.621181131683551e-06,
+      "loss": 0.6818,
+      "step": 9819
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.620212032473891e-06,
+      "loss": 0.8714,
+      "step": 9820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6192430488335687e-06,
+      "loss": 0.7727,
+      "step": 9821
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6182741808096414e-06,
+      "loss": 0.8723,
+      "step": 9822
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.617305428449166e-06,
+      "loss": 0.8476,
+      "step": 9823
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6163367917991824e-06,
+      "loss": 0.6896,
+      "step": 9824
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6153682709067317e-06,
+      "loss": 0.7386,
+      "step": 9825
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6143998658188485e-06,
+      "loss": 0.6609,
+      "step": 9826
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6134315765825613e-06,
+      "loss": 0.5816,
+      "step": 9827
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6124634032448924e-06,
+      "loss": 0.7747,
+      "step": 9828
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6114953458528604e-06,
+      "loss": 0.7808,
+      "step": 9829
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6105274044534757e-06,
+      "loss": 0.5768,
+      "step": 9830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.609559579093745e-06,
+      "loss": 0.5892,
+      "step": 9831
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6085918698206686e-06,
+      "loss": 0.8099,
+      "step": 9832
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.607624276681241e-06,
+      "loss": 0.7995,
+      "step": 9833
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.606656799722453e-06,
+      "loss": 0.9611,
+      "step": 9834
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6056894389912826e-06,
+      "loss": 0.7202,
+      "step": 9835
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6047221945347133e-06,
+      "loss": 0.7268,
+      "step": 9836
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.603755066399718e-06,
+      "loss": 0.7424,
+      "step": 9837
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.602788054633256e-06,
+      "loss": 0.6544,
+      "step": 9838
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.601821159282295e-06,
+      "loss": 0.8038,
+      "step": 9839
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.6008543803937903e-06,
+      "loss": 0.7775,
+      "step": 9840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5998877180146854e-06,
+      "loss": 0.8822,
+      "step": 9841
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5989211721919296e-06,
+      "loss": 0.724,
+      "step": 9842
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.597954742972461e-06,
+      "loss": 0.7062,
+      "step": 9843
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5969884304032093e-06,
+      "loss": 0.6224,
+      "step": 9844
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5960222345311026e-06,
+      "loss": 0.6115,
+      "step": 9845
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.595056155403063e-06,
+      "loss": 0.8515,
+      "step": 9846
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5940901930660052e-06,
+      "loss": 0.6794,
+      "step": 9847
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.593124347566839e-06,
+      "loss": 0.6802,
+      "step": 9848
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5921586189524694e-06,
+      "loss": 0.8701,
+      "step": 9849
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5911930072697944e-06,
+      "loss": 0.7957,
+      "step": 9850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5902275125657067e-06,
+      "loss": 0.6397,
+      "step": 9851
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.589262134887094e-06,
+      "loss": 0.7753,
+      "step": 9852
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5882968742808377e-06,
+      "loss": 0.5474,
+      "step": 9853
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5873317307938138e-06,
+      "loss": 0.5973,
+      "step": 9854
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5863667044728924e-06,
+      "loss": 0.7474,
+      "step": 9855
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5854017953649375e-06,
+      "loss": 0.6537,
+      "step": 9856
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5844370035168077e-06,
+      "loss": 0.8503,
+      "step": 9857
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.583472328975357e-06,
+      "loss": 0.5903,
+      "step": 9858
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.582507771787432e-06,
+      "loss": 0.6802,
+      "step": 9859
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.581543331999875e-06,
+      "loss": 0.8872,
+      "step": 9860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5805790096595205e-06,
+      "loss": 0.8214,
+      "step": 9861
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.579614804813201e-06,
+      "loss": 0.9272,
+      "step": 9862
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5786507175077414e-06,
+      "loss": 0.7057,
+      "step": 9863
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5776867477899554e-06,
+      "loss": 0.8512,
+      "step": 9864
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5767228957066635e-06,
+      "loss": 0.6272,
+      "step": 9865
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.575759161304667e-06,
+      "loss": 0.7856,
+      "step": 9866
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.574795544630769e-06,
+      "loss": 0.8818,
+      "step": 9867
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.57383204573177e-06,
+      "loss": 0.7789,
+      "step": 9868
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5728686646544556e-06,
+      "loss": 0.7318,
+      "step": 9869
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.571905401445608e-06,
+      "loss": 0.6894,
+      "step": 9870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5709422561520147e-06,
+      "loss": 0.9052,
+      "step": 9871
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.569979228820441e-06,
+      "loss": 0.7772,
+      "step": 9872
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5690163194976576e-06,
+      "loss": 0.8844,
+      "step": 9873
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.568053528230425e-06,
+      "loss": 0.6962,
+      "step": 9874
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5670908550655e-06,
+      "loss": 0.7641,
+      "step": 9875
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5661283000496324e-06,
+      "loss": 0.7004,
+      "step": 9876
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5651658632295662e-06,
+      "loss": 0.8983,
+      "step": 9877
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.564203544652041e-06,
+      "loss": 0.6511,
+      "step": 9878
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5632413443637887e-06,
+      "loss": 0.6337,
+      "step": 9879
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5622792624115377e-06,
+      "loss": 0.7236,
+      "step": 9880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.561317298842008e-06,
+      "loss": 0.944,
+      "step": 9881
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5603554537019194e-06,
+      "loss": 0.83,
+      "step": 9882
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.559393727037973e-06,
+      "loss": 0.9048,
+      "step": 9883
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.558432118896882e-06,
+      "loss": 0.5389,
+      "step": 9884
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.557470629325343e-06,
+      "loss": 0.8577,
+      "step": 9885
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5565092583700436e-06,
+      "loss": 0.9533,
+      "step": 9886
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.555548006077676e-06,
+      "loss": 0.6877,
+      "step": 9887
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5545868724949212e-06,
+      "loss": 0.6659,
+      "step": 9888
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5536258576684497e-06,
+      "loss": 0.7716,
+      "step": 9889
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.552664961644936e-06,
+      "loss": 0.4642,
+      "step": 9890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5517041844710456e-06,
+      "loss": 0.7376,
+      "step": 9891
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.550743526193429e-06,
+      "loss": 0.964,
+      "step": 9892
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5497829868587466e-06,
+      "loss": 0.6074,
+      "step": 9893
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.548822566513639e-06,
+      "loss": 0.7722,
+      "step": 9894
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.547862265204748e-06,
+      "loss": 0.7159,
+      "step": 9895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.546902082978714e-06,
+      "loss": 0.5546,
+      "step": 9896
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5459420198821604e-06,
+      "loss": 0.627,
+      "step": 9897
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.54498207596171e-06,
+      "loss": 0.7256,
+      "step": 9898
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5440222512639868e-06,
+      "loss": 0.8428,
+      "step": 9899
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5430625458355965e-06,
+      "loss": 0.7859,
+      "step": 9900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5421029597231476e-06,
+      "loss": 1.0024,
+      "step": 9901
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5411434929732403e-06,
+      "loss": 0.6634,
+      "step": 9902
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5401841456324683e-06,
+      "loss": 0.5953,
+      "step": 9903
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5392249177474217e-06,
+      "loss": 0.5354,
+      "step": 9904
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.538265809364681e-06,
+      "loss": 0.5827,
+      "step": 9905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.537306820530826e-06,
+      "loss": 0.6854,
+      "step": 9906
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.536347951292426e-06,
+      "loss": 0.7052,
+      "step": 9907
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.535389201696047e-06,
+      "loss": 0.9273,
+      "step": 9908
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5344305717882487e-06,
+      "loss": 0.7852,
+      "step": 9909
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5334720616155877e-06,
+      "loss": 0.8378,
+      "step": 9910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5325136712246044e-06,
+      "loss": 0.6985,
+      "step": 9911
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5315554006618487e-06,
+      "loss": 0.5974,
+      "step": 9912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.530597249973856e-06,
+      "loss": 0.9204,
+      "step": 9913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5296392192071508e-06,
+      "loss": 0.6899,
+      "step": 9914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.528681308408264e-06,
+      "loss": 0.5804,
+      "step": 9915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.527723517623714e-06,
+      "loss": 0.8267,
+      "step": 9916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.526765846900009e-06,
+      "loss": 0.831,
+      "step": 9917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5258082962836616e-06,
+      "loss": 0.4958,
+      "step": 9918
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5248508658211727e-06,
+      "loss": 0.6085,
+      "step": 9919
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.523893555559035e-06,
+      "loss": 0.4037,
+      "step": 9920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5229363655437394e-06,
+      "loss": 0.6059,
+      "step": 9921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5219792958217703e-06,
+      "loss": 0.8358,
+      "step": 9922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5210223464396055e-06,
+      "loss": 0.6846,
+      "step": 9923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5200655174437165e-06,
+      "loss": 0.5516,
+      "step": 9924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5191088088805716e-06,
+      "loss": 0.457,
+      "step": 9925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.518152220796629e-06,
+      "loss": 0.7959,
+      "step": 9926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5171957532383453e-06,
+      "loss": 0.6961,
+      "step": 9927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.516239406252168e-06,
+      "loss": 0.3944,
+      "step": 9928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.51528317988454e-06,
+      "loss": 0.5917,
+      "step": 9929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5143270741818994e-06,
+      "loss": 0.6293,
+      "step": 9930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5133710891906773e-06,
+      "loss": 0.8562,
+      "step": 9931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.512415224957298e-06,
+      "loss": 0.8727,
+      "step": 9932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.511459481528182e-06,
+      "loss": 0.7233,
+      "step": 9933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.510503858949743e-06,
+      "loss": 0.7914,
+      "step": 9934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5095483572683886e-06,
+      "loss": 0.6882,
+      "step": 9935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.508592976530522e-06,
+      "loss": 0.5818,
+      "step": 9936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.507637716782535e-06,
+      "loss": 0.7115,
+      "step": 9937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5066825780708215e-06,
+      "loss": 1.0233,
+      "step": 9938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.505727560441767e-06,
+      "loss": 0.6193,
+      "step": 9939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5047726639417437e-06,
+      "loss": 0.8012,
+      "step": 9940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5038178886171327e-06,
+      "loss": 1.2005,
+      "step": 9941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5028632345142935e-06,
+      "loss": 1.0005,
+      "step": 9942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.501908701679588e-06,
+      "loss": 0.353,
+      "step": 9943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5009542901593763e-06,
+      "loss": 0.5868,
+      "step": 9944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.5000000000000015e-06,
+      "loss": 0.7583,
+      "step": 9945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4990458312478067e-06,
+      "loss": 0.6976,
+      "step": 9946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.498091783949135e-06,
+      "loss": 0.5875,
+      "step": 9947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.497137858150312e-06,
+      "loss": 0.7057,
+      "step": 9948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4961840538976643e-06,
+      "loss": 0.7697,
+      "step": 9949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.495230371237513e-06,
+      "loss": 0.7345,
+      "step": 9950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4942768102161704e-06,
+      "loss": 0.53,
+      "step": 9951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4933233708799438e-06,
+      "loss": 0.6767,
+      "step": 9952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4923700532751355e-06,
+      "loss": 0.4953,
+      "step": 9953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.491416857448042e-06,
+      "loss": 1.006,
+      "step": 9954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4904637834449525e-06,
+      "loss": 0.7244,
+      "step": 9955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4895108313121508e-06,
+      "loss": 0.8158,
+      "step": 9956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4885580010959153e-06,
+      "loss": 0.8643,
+      "step": 9957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4876052928425202e-06,
+      "loss": 0.473,
+      "step": 9958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.486652706598226e-06,
+      "loss": 0.5715,
+      "step": 9959
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4857002424092986e-06,
+      "loss": 1.0103,
+      "step": 9960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4847479003219926e-06,
+      "loss": 0.5274,
+      "step": 9961
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4837956803825504e-06,
+      "loss": 0.631,
+      "step": 9962
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4828435826372204e-06,
+      "loss": 0.6118,
+      "step": 9963
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4818916071322397e-06,
+      "loss": 0.751,
+      "step": 9964
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4809397539138324e-06,
+      "loss": 0.6497,
+      "step": 9965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4799880230282293e-06,
+      "loss": 0.7049,
+      "step": 9966
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4790364145216495e-06,
+      "loss": 0.6868,
+      "step": 9967
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4780849284402996e-06,
+      "loss": 0.676,
+      "step": 9968
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4771335648303944e-06,
+      "loss": 0.8275,
+      "step": 9969
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.47618232373813e-06,
+      "loss": 0.9373,
+      "step": 9970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4752312052096996e-06,
+      "loss": 0.6186,
+      "step": 9971
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.474280209291299e-06,
+      "loss": 0.9825,
+      "step": 9972
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4733293360291055e-06,
+      "loss": 0.6961,
+      "step": 9973
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.472378585469298e-06,
+      "loss": 0.8383,
+      "step": 9974
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4714279576580487e-06,
+      "loss": 0.5592,
+      "step": 9975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.470477452641521e-06,
+      "loss": 0.5018,
+      "step": 9976
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4695270704658753e-06,
+      "loss": 0.616,
+      "step": 9977
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4685768111772647e-06,
+      "loss": 0.9564,
+      "step": 9978
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4676266748218357e-06,
+      "loss": 0.6753,
+      "step": 9979
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4666766614457316e-06,
+      "loss": 0.6701,
+      "step": 9980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.465726771095086e-06,
+      "loss": 0.4366,
+      "step": 9981
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4647770038160292e-06,
+      "loss": 0.5721,
+      "step": 9982
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4638273596546835e-06,
+      "loss": 0.7986,
+      "step": 9983
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4628778386571676e-06,
+      "loss": 0.715,
+      "step": 9984
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.461928440869593e-06,
+      "loss": 0.5816,
+      "step": 9985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4609791663380668e-06,
+      "loss": 0.4666,
+      "step": 9986
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.460030015108682e-06,
+      "loss": 0.7434,
+      "step": 9987
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4590809872275385e-06,
+      "loss": 0.6836,
+      "step": 9988
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.458132082740724e-06,
+      "loss": 0.6264,
+      "step": 9989
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4571833016943136e-06,
+      "loss": 0.6182,
+      "step": 9990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.456234644134389e-06,
+      "loss": 0.6491,
+      "step": 9991
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4552861101070207e-06,
+      "loss": 0.5793,
+      "step": 9992
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.454337699658267e-06,
+      "loss": 0.5713,
+      "step": 9993
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.453389412834186e-06,
+      "loss": 0.7927,
+      "step": 9994
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4524412496808345e-06,
+      "loss": 0.8358,
+      "step": 9995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.451493210244253e-06,
+      "loss": 0.7748,
+      "step": 9996
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.450545294570482e-06,
+      "loss": 0.6471,
+      "step": 9997
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4495975027055567e-06,
+      "loss": 0.8563,
+      "step": 9998
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.448649834695503e-06,
+      "loss": 0.5284,
+      "step": 9999
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.447702290586343e-06,
+      "loss": 1.0682,
+      "step": 10000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4467548704240916e-06,
+      "loss": 0.7754,
+      "step": 10001
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.445807574254759e-06,
+      "loss": 0.9199,
+      "step": 10002
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4448604021243485e-06,
+      "loss": 0.7184,
+      "step": 10003
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4439133540788566e-06,
+      "loss": 0.7709,
+      "step": 10004
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4429664301642763e-06,
+      "loss": 0.6028,
+      "step": 10005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4420196304265916e-06,
+      "loss": 0.77,
+      "step": 10006
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4410729549117824e-06,
+      "loss": 0.6651,
+      "step": 10007
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4401264036658216e-06,
+      "loss": 0.5642,
+      "step": 10008
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.439179976734677e-06,
+      "loss": 0.8333,
+      "step": 10009
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4382336741643093e-06,
+      "loss": 1.1526,
+      "step": 10010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.437287496000674e-06,
+      "loss": 0.7273,
+      "step": 10011
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4363414422897218e-06,
+      "loss": 0.7927,
+      "step": 10012
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4353955130773906e-06,
+      "loss": 0.7807,
+      "step": 10013
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4344497084096224e-06,
+      "loss": 0.9098,
+      "step": 10014
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.433504028332349e-06,
+      "loss": 0.6575,
+      "step": 10015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.432558472891489e-06,
+      "loss": 0.8339,
+      "step": 10016
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4316130421329696e-06,
+      "loss": 0.677,
+      "step": 10017
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.430667736102697e-06,
+      "loss": 0.7239,
+      "step": 10018
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4297225548465786e-06,
+      "loss": 0.6906,
+      "step": 10019
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.428777498410521e-06,
+      "loss": 0.5862,
+      "step": 10020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.427832566840412e-06,
+      "loss": 0.5669,
+      "step": 10021
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4268877601821422e-06,
+      "loss": 0.4955,
+      "step": 10022
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.425943078481598e-06,
+      "loss": 0.8091,
+      "step": 10023
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4249985217846504e-06,
+      "loss": 0.737,
+      "step": 10024
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4240540901371727e-06,
+      "loss": 1.0341,
+      "step": 10025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.423109783585029e-06,
+      "loss": 0.7628,
+      "step": 10026
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4221656021740765e-06,
+      "loss": 0.9103,
+      "step": 10027
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4212215459501676e-06,
+      "loss": 0.607,
+      "step": 10028
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4202776149591494e-06,
+      "loss": 0.5194,
+      "step": 10029
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.419333809246861e-06,
+      "loss": 0.6895,
+      "step": 10030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4183901288591364e-06,
+      "loss": 0.8242,
+      "step": 10031
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4174465738418044e-06,
+      "loss": 1.0327,
+      "step": 10032
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4165031442406857e-06,
+      "loss": 0.6122,
+      "step": 10033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.415559840101598e-06,
+      "loss": 0.6894,
+      "step": 10034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4146166614703454e-06,
+      "loss": 0.7945,
+      "step": 10035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4136736083927364e-06,
+      "loss": 0.6893,
+      "step": 10036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4127306809145695e-06,
+      "loss": 0.8013,
+      "step": 10037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4117878790816295e-06,
+      "loss": 0.5806,
+      "step": 10038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.410845202939708e-06,
+      "loss": 0.6411,
+      "step": 10039
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.409902652534583e-06,
+      "loss": 0.6278,
+      "step": 10040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4089602279120224e-06,
+      "loss": 0.7849,
+      "step": 10041
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4080179291177983e-06,
+      "loss": 0.9617,
+      "step": 10042
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.407075756197672e-06,
+      "loss": 0.7843,
+      "step": 10043
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.406133709197392e-06,
+      "loss": 0.6535,
+      "step": 10044
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.405191788162715e-06,
+      "loss": 0.8391,
+      "step": 10045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.404249993139377e-06,
+      "loss": 0.6492,
+      "step": 10046
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.403308324173115e-06,
+      "loss": 0.6957,
+      "step": 10047
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4023667813096636e-06,
+      "loss": 0.6168,
+      "step": 10048
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.401425364594742e-06,
+      "loss": 0.4996,
+      "step": 10049
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.4004840740740702e-06,
+      "loss": 0.8149,
+      "step": 10050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3995429097933596e-06,
+      "loss": 0.5321,
+      "step": 10051
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3986018717983166e-06,
+      "loss": 0.8419,
+      "step": 10052
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3976609601346395e-06,
+      "loss": 0.7825,
+      "step": 10053
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3967201748480223e-06,
+      "loss": 0.6461,
+      "step": 10054
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.395779515984152e-06,
+      "loss": 0.8344,
+      "step": 10055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39483898358871e-06,
+      "loss": 0.9558,
+      "step": 10056
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.393898577707371e-06,
+      "loss": 1.0005,
+      "step": 10057
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.392958298385804e-06,
+      "loss": 0.8305,
+      "step": 10058
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3920181456696717e-06,
+      "loss": 0.9059,
+      "step": 10059
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39107811960463e-06,
+      "loss": 0.7099,
+      "step": 10060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3901382202363306e-06,
+      "loss": 0.7378,
+      "step": 10061
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.389198447610418e-06,
+      "loss": 0.5264,
+      "step": 10062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3882588017725263e-06,
+      "loss": 0.5352,
+      "step": 10063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3873192827682913e-06,
+      "loss": 0.7971,
+      "step": 10064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3863798906433396e-06,
+      "loss": 0.7388,
+      "step": 10065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.385440625443287e-06,
+      "loss": 0.825,
+      "step": 10066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3845014872137473e-06,
+      "loss": 1.089,
+      "step": 10067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.383562476000333e-06,
+      "loss": 0.9315,
+      "step": 10068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.382623591848639e-06,
+      "loss": 0.4911,
+      "step": 10069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.381684834804261e-06,
+      "loss": 0.9595,
+      "step": 10070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3807462049127934e-06,
+      "loss": 0.6709,
+      "step": 10071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3798077022198125e-06,
+      "loss": 0.632,
+      "step": 10072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3788693267708975e-06,
+      "loss": 0.7537,
+      "step": 10073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.377931078611617e-06,
+      "loss": 0.7806,
+      "step": 10074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3769929577875365e-06,
+      "loss": 0.7408,
+      "step": 10075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.376054964344214e-06,
+      "loss": 0.426,
+      "step": 10076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3751170983272e-06,
+      "loss": 0.8513,
+      "step": 10077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.374179359782041e-06,
+      "loss": 0.6854,
+      "step": 10078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3732417487542746e-06,
+      "loss": 0.433,
+      "step": 10079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.372304265289436e-06,
+      "loss": 0.8947,
+      "step": 10080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.371366909433051e-06,
+      "loss": 0.6247,
+      "step": 10081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3704296812306405e-06,
+      "loss": 0.5238,
+      "step": 10082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.369492580727719e-06,
+      "loss": 0.65,
+      "step": 10083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3685556079697948e-06,
+      "loss": 0.7604,
+      "step": 10084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3676187630023717e-06,
+      "loss": 0.7783,
+      "step": 10085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.36668204587094e-06,
+      "loss": 0.7687,
+      "step": 10086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3657454566209952e-06,
+      "loss": 0.5171,
+      "step": 10087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3648089952980204e-06,
+      "loss": 0.4979,
+      "step": 10088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.363872661947488e-06,
+      "loss": 0.8408,
+      "step": 10089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3629364566148735e-06,
+      "loss": 0.9195,
+      "step": 10090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3620003793456426e-06,
+      "loss": 0.7863,
+      "step": 10091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3610644301852483e-06,
+      "loss": 0.4361,
+      "step": 10092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3601286091791496e-06,
+      "loss": 0.6907,
+      "step": 10093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3591929163727876e-06,
+      "loss": 0.4981,
+      "step": 10094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.358257351811603e-06,
+      "loss": 0.5981,
+      "step": 10095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.357321915541034e-06,
+      "loss": 0.6287,
+      "step": 10096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.356386607606503e-06,
+      "loss": 0.7058,
+      "step": 10097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3554514280534306e-06,
+      "loss": 0.9854,
+      "step": 10098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3545163769272383e-06,
+      "loss": 0.8718,
+      "step": 10099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.353581454273329e-06,
+      "loss": 0.5776,
+      "step": 10100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3526466601371067e-06,
+      "loss": 0.7808,
+      "step": 10101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.351711994563968e-06,
+      "loss": 0.5399,
+      "step": 10102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.350777457599303e-06,
+      "loss": 0.612,
+      "step": 10103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3498430492884953e-06,
+      "loss": 0.8764,
+      "step": 10104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3489087696769225e-06,
+      "loss": 0.7918,
+      "step": 10105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.347974618809956e-06,
+      "loss": 0.7532,
+      "step": 10106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3470405967329605e-06,
+      "loss": 0.4974,
+      "step": 10107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.346106703491296e-06,
+      "loss": 0.5785,
+      "step": 10108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3451729391303137e-06,
+      "loss": 0.6807,
+      "step": 10109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3442393036953614e-06,
+      "loss": 0.5066,
+      "step": 10110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.343305797231775e-06,
+      "loss": 0.6462,
+      "step": 10111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.342372419784893e-06,
+      "loss": 0.8738,
+      "step": 10112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3414391714000435e-06,
+      "loss": 0.7449,
+      "step": 10113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.340506052122541e-06,
+      "loss": 0.4941,
+      "step": 10114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.339573061997707e-06,
+      "loss": 0.9178,
+      "step": 10115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3386402010708496e-06,
+      "loss": 0.7362,
+      "step": 10116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3377074693872664e-06,
+      "loss": 1.0007,
+      "step": 10117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3367748669922583e-06,
+      "loss": 0.6155,
+      "step": 10118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.335842393931116e-06,
+      "loss": 0.6819,
+      "step": 10119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3349100502491163e-06,
+      "loss": 0.9648,
+      "step": 10120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.333977835991545e-06,
+      "loss": 0.6044,
+      "step": 10121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.333045751203667e-06,
+      "loss": 0.9521,
+      "step": 10122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.332113795930749e-06,
+      "loss": 0.3229,
+      "step": 10123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3311819702180494e-06,
+      "loss": 0.7577,
+      "step": 10124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.33025027411082e-06,
+      "loss": 0.5125,
+      "step": 10125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3293187076543076e-06,
+      "loss": 0.5509,
+      "step": 10126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.328387270893751e-06,
+      "loss": 0.7645,
+      "step": 10127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3274559638743834e-06,
+      "loss": 0.7789,
+      "step": 10128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.326524786641432e-06,
+      "loss": 0.5738,
+      "step": 10129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3255937392401174e-06,
+      "loss": 0.6227,
+      "step": 10130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3246628217156536e-06,
+      "loss": 0.5061,
+      "step": 10131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.32373203411325e-06,
+      "loss": 1.072,
+      "step": 10132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.322801376478106e-06,
+      "loss": 0.8194,
+      "step": 10133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.321870848855419e-06,
+      "loss": 0.7769,
+      "step": 10134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3209404512903764e-06,
+      "loss": 0.3704,
+      "step": 10135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.320010183828163e-06,
+      "loss": 0.6532,
+      "step": 10136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.319080046513954e-06,
+      "loss": 0.8495,
+      "step": 10137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3181500393929215e-06,
+      "loss": 0.6614,
+      "step": 10138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3172201625102235e-06,
+      "loss": 0.8691,
+      "step": 10139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.316290415911023e-06,
+      "loss": 0.6157,
+      "step": 10140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3153607996404726e-06,
+      "loss": 0.8185,
+      "step": 10141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3144313137437125e-06,
+      "loss": 0.6667,
+      "step": 10142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3135019582658803e-06,
+      "loss": 0.6553,
+      "step": 10143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.312572733252115e-06,
+      "loss": 0.4164,
+      "step": 10144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.311643638747536e-06,
+      "loss": 0.7596,
+      "step": 10145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3107146747972637e-06,
+      "loss": 0.6851,
+      "step": 10146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3097858414464157e-06,
+      "loss": 0.4551,
+      "step": 10147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.308857138740095e-06,
+      "loss": 0.7815,
+      "step": 10148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3079285667234017e-06,
+      "loss": 0.6788,
+      "step": 10149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3070001254414317e-06,
+      "loss": 0.8022,
+      "step": 10150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3060718149392713e-06,
+      "loss": 0.7729,
+      "step": 10151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.305143635262004e-06,
+      "loss": 0.7104,
+      "step": 10152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3042155864547024e-06,
+      "loss": 1.0209,
+      "step": 10153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.303287668562437e-06,
+      "loss": 0.6479,
+      "step": 10154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3023598816302694e-06,
+      "loss": 0.8996,
+      "step": 10155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3014322257032555e-06,
+      "loss": 0.7722,
+      "step": 10156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.3005047008264457e-06,
+      "loss": 0.8066,
+      "step": 10157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.299577307044883e-06,
+      "loss": 0.5984,
+      "step": 10158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2986500444036037e-06,
+      "loss": 0.6815,
+      "step": 10159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2977229129476386e-06,
+      "loss": 0.5535,
+      "step": 10160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.296795912722014e-06,
+      "loss": 0.5808,
+      "step": 10161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.295869043771742e-06,
+      "loss": 0.9922,
+      "step": 10162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2949423061418397e-06,
+      "loss": 0.6315,
+      "step": 10163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2940156998773117e-06,
+      "loss": 0.6253,
+      "step": 10164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.293089225023152e-06,
+      "loss": 0.681,
+      "step": 10165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2921628816243566e-06,
+      "loss": 0.6176,
+      "step": 10166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2912366697259137e-06,
+      "loss": 0.5267,
+      "step": 10167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2903105893727962e-06,
+      "loss": 0.706,
+      "step": 10168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2893846406099847e-06,
+      "loss": 0.623,
+      "step": 10169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2884588234824395e-06,
+      "loss": 0.4648,
+      "step": 10170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2875331380351225e-06,
+      "loss": 0.8786,
+      "step": 10171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.286607584312992e-06,
+      "loss": 0.6268,
+      "step": 10172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.285682162360991e-06,
+      "loss": 0.6489,
+      "step": 10173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2847568722240592e-06,
+      "loss": 0.543,
+      "step": 10174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2838317139471385e-06,
+      "loss": 0.7583,
+      "step": 10175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.282906687575151e-06,
+      "loss": 0.7914,
+      "step": 10176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.28198179315302e-06,
+      "loss": 0.6419,
+      "step": 10177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.281057030725662e-06,
+      "loss": 0.6136,
+      "step": 10178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2801324003379854e-06,
+      "loss": 0.7432,
+      "step": 10179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.279207902034893e-06,
+      "loss": 0.8418,
+      "step": 10180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.278283535861281e-06,
+      "loss": 0.5974,
+      "step": 10181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.27735930186204e-06,
+      "loss": 0.6093,
+      "step": 10182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2764352000820526e-06,
+      "loss": 0.7571,
+      "step": 10183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2755112305661966e-06,
+      "loss": 0.6214,
+      "step": 10184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.274587393359342e-06,
+      "loss": 0.494,
+      "step": 10185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2736636885063545e-06,
+      "loss": 0.8009,
+      "step": 10186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.272740116052088e-06,
+      "loss": 0.5566,
+      "step": 10187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2718166760413974e-06,
+      "loss": 0.7195,
+      "step": 10188
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2708933685191285e-06,
+      "loss": 0.6407,
+      "step": 10189
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2699701935301144e-06,
+      "loss": 0.8626,
+      "step": 10190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2690471511191918e-06,
+      "loss": 0.5303,
+      "step": 10191
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.268124241331188e-06,
+      "loss": 0.7183,
+      "step": 10192
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.267201464210915e-06,
+      "loss": 0.7423,
+      "step": 10193
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.266278819803191e-06,
+      "loss": 0.5012,
+      "step": 10194
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2653563081528236e-06,
+      "loss": 0.599,
+      "step": 10195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.264433929304608e-06,
+      "loss": 0.8363,
+      "step": 10196
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.263511683303339e-06,
+      "loss": 1.06,
+      "step": 10197
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.262589570193805e-06,
+      "loss": 0.6803,
+      "step": 10198
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.261667590020785e-06,
+      "loss": 0.7427,
+      "step": 10199
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.260745742829054e-06,
+      "loss": 0.8549,
+      "step": 10200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2598240286633787e-06,
+      "loss": 1.0277,
+      "step": 10201
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2589024475685206e-06,
+      "loss": 0.7536,
+      "step": 10202
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2579809995892343e-06,
+      "loss": 0.8573,
+      "step": 10203
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2570596847702687e-06,
+      "loss": 0.7442,
+      "step": 10204
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.256138503156364e-06,
+      "loss": 0.8743,
+      "step": 10205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.255217454792256e-06,
+      "loss": 0.6939,
+      "step": 10206
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.254296539722674e-06,
+      "loss": 0.8654,
+      "step": 10207
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2533757579923405e-06,
+      "loss": 0.9377,
+      "step": 10208
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2524551096459703e-06,
+      "loss": 0.9328,
+      "step": 10209
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.251534594728273e-06,
+      "loss": 0.8112,
+      "step": 10210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2506142132839516e-06,
+      "loss": 0.8442,
+      "step": 10211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2496939653577027e-06,
+      "loss": 0.2898,
+      "step": 10212
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2487738509942157e-06,
+      "loss": 0.633,
+      "step": 10213
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.247853870238176e-06,
+      "loss": 0.6386,
+      "step": 10214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.246934023134257e-06,
+      "loss": 0.7704,
+      "step": 10215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.24601430972713e-06,
+      "loss": 0.9708,
+      "step": 10216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.245094730061463e-06,
+      "loss": 0.725,
+      "step": 10217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.244175284181908e-06,
+      "loss": 1.0033,
+      "step": 10218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2432559721331165e-06,
+      "loss": 0.8437,
+      "step": 10219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2423367939597384e-06,
+      "loss": 0.5898,
+      "step": 10220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.241417749706406e-06,
+      "loss": 0.8336,
+      "step": 10221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.240498839417751e-06,
+      "loss": 0.9311,
+      "step": 10222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.239580063138403e-06,
+      "loss": 0.9785,
+      "step": 10223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2386614209129754e-06,
+      "loss": 0.7461,
+      "step": 10224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.237742912786082e-06,
+      "loss": 0.5989,
+      "step": 10225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2368245388023293e-06,
+      "loss": 0.504,
+      "step": 10226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.235906299006314e-06,
+      "loss": 0.914,
+      "step": 10227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.23498819344263e-06,
+      "loss": 0.7589,
+      "step": 10228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2340702221558626e-06,
+      "loss": 0.7687,
+      "step": 10229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2331523851905913e-06,
+      "loss": 0.6383,
+      "step": 10230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.232234682591389e-06,
+      "loss": 0.4938,
+      "step": 10231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.231317114402822e-06,
+      "loss": 0.9202,
+      "step": 10232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.230399680669449e-06,
+      "loss": 0.7008,
+      "step": 10233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2294823814358263e-06,
+      "loss": 0.6874,
+      "step": 10234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.228565216746495e-06,
+      "loss": 0.5032,
+      "step": 10235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.227648186646e-06,
+      "loss": 0.8099,
+      "step": 10236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.226731291178875e-06,
+      "loss": 1.0565,
+      "step": 10237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2258145303896427e-06,
+      "loss": 0.6419,
+      "step": 10238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.224897904322828e-06,
+      "loss": 0.6754,
+      "step": 10239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2239814130229453e-06,
+      "loss": 0.6381,
+      "step": 10240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.223065056534497e-06,
+      "loss": 0.5597,
+      "step": 10241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2221488349019903e-06,
+      "loss": 0.745,
+      "step": 10242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2212327481699146e-06,
+      "loss": 0.6547,
+      "step": 10243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.220316796382758e-06,
+      "loss": 0.9064,
+      "step": 10244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.219400979585007e-06,
+      "loss": 0.5128,
+      "step": 10245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.21848529782113e-06,
+      "loss": 0.5893,
+      "step": 10246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.217569751135597e-06,
+      "loss": 0.8209,
+      "step": 10247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2166543395728733e-06,
+      "loss": 0.6578,
+      "step": 10248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.215739063177409e-06,
+      "loss": 0.8459,
+      "step": 10249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.214823921993653e-06,
+      "loss": 0.6329,
+      "step": 10250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.213908916066052e-06,
+      "loss": 0.5418,
+      "step": 10251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2129940454390364e-06,
+      "loss": 0.9739,
+      "step": 10252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2120793101570366e-06,
+      "loss": 0.6174,
+      "step": 10253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2111647102644744e-06,
+      "loss": 0.9471,
+      "step": 10254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2102502458057656e-06,
+      "loss": 0.8812,
+      "step": 10255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2093359168253197e-06,
+      "loss": 0.7448,
+      "step": 10256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2084217233675386e-06,
+      "loss": 0.6141,
+      "step": 10257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.207507665476818e-06,
+      "loss": 0.6665,
+      "step": 10258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.206593743197548e-06,
+      "loss": 0.6206,
+      "step": 10259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2056799565741105e-06,
+      "loss": 0.6427,
+      "step": 10260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.204766305650882e-06,
+      "loss": 0.7379,
+      "step": 10261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2038527904722334e-06,
+      "loss": 0.7258,
+      "step": 10262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2029394110825226e-06,
+      "loss": 0.965,
+      "step": 10263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2020261675261112e-06,
+      "loss": 0.895,
+      "step": 10264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2011130598473498e-06,
+      "loss": 0.5961,
+      "step": 10265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.2002000880905745e-06,
+      "loss": 0.8004,
+      "step": 10266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.199287252300128e-06,
+      "loss": 0.4771,
+      "step": 10267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1983745525203404e-06,
+      "loss": 0.9348,
+      "step": 10268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1974619887955294e-06,
+      "loss": 0.5877,
+      "step": 10269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.196549561170017e-06,
+      "loss": 0.7969,
+      "step": 10270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1956372696881145e-06,
+      "loss": 0.7203,
+      "step": 10271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.19472511439412e-06,
+      "loss": 0.9231,
+      "step": 10272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.193813095332333e-06,
+      "loss": 0.6768,
+      "step": 10273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.192901212547044e-06,
+      "loss": 0.7568,
+      "step": 10274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1919894660825362e-06,
+      "loss": 0.6075,
+      "step": 10275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.191077855983087e-06,
+      "loss": 0.5501,
+      "step": 10276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1901663822929663e-06,
+      "loss": 0.6738,
+      "step": 10277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.189255045056439e-06,
+      "loss": 0.7393,
+      "step": 10278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1883438443177607e-06,
+      "loss": 0.66,
+      "step": 10279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1874327801211832e-06,
+      "loss": 0.4455,
+      "step": 10280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1865218525109496e-06,
+      "loss": 0.4304,
+      "step": 10281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1856110615312977e-06,
+      "loss": 1.081,
+      "step": 10282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.184700407226457e-06,
+      "loss": 0.5799,
+      "step": 10283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1837898896406536e-06,
+      "loss": 0.7935,
+      "step": 10284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1828795088181026e-06,
+      "loss": 0.5623,
+      "step": 10285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.181969264803017e-06,
+      "loss": 0.6564,
+      "step": 10286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1810591576395985e-06,
+      "loss": 0.6841,
+      "step": 10287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1801491873720456e-06,
+      "loss": 0.9261,
+      "step": 10288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1792393540445493e-06,
+      "loss": 0.7475,
+      "step": 10289
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1783296577012957e-06,
+      "loss": 0.6621,
+      "step": 10290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.177420098386458e-06,
+      "loss": 0.645,
+      "step": 10291
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.176510676144207e-06,
+      "loss": 0.8385,
+      "step": 10292
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.175601391018713e-06,
+      "loss": 0.8363,
+      "step": 10293
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.174692243054127e-06,
+      "loss": 0.7055,
+      "step": 10294
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.173783232294601e-06,
+      "loss": 0.8586,
+      "step": 10295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.172874358784284e-06,
+      "loss": 0.7656,
+      "step": 10296
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.171965622567308e-06,
+      "loss": 0.6846,
+      "step": 10297
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1710570236878042e-06,
+      "loss": 0.9716,
+      "step": 10298
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1701485621899014e-06,
+      "loss": 0.6793,
+      "step": 10299
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.169240238117713e-06,
+      "loss": 0.9688,
+      "step": 10300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1683320515153505e-06,
+      "loss": 0.5199,
+      "step": 10301
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.167424002426919e-06,
+      "loss": 0.661,
+      "step": 10302
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.166516090896515e-06,
+      "loss": 0.6549,
+      "step": 10303
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1656083169682296e-06,
+      "loss": 0.9817,
+      "step": 10304
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1647006806861472e-06,
+      "loss": 0.4107,
+      "step": 10305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1637931820943446e-06,
+      "loss": 0.7316,
+      "step": 10306
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.162885821236893e-06,
+      "loss": 0.671,
+      "step": 10307
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.161978598157857e-06,
+      "loss": 0.6341,
+      "step": 10308
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1610715129012934e-06,
+      "loss": 0.652,
+      "step": 10309
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.160164565511254e-06,
+      "loss": 0.5289,
+      "step": 10310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1592577560317784e-06,
+      "loss": 0.6798,
+      "step": 10311
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1583510845069094e-06,
+      "loss": 0.9538,
+      "step": 10312
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1574445509806764e-06,
+      "loss": 0.9052,
+      "step": 10313
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1565381554970987e-06,
+      "loss": 0.7117,
+      "step": 10314
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.155631898100199e-06,
+      "loss": 0.2946,
+      "step": 10315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.154725778833987e-06,
+      "loss": 0.5729,
+      "step": 10316
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.153819797742462e-06,
+      "loss": 0.8527,
+      "step": 10317
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1529139548696277e-06,
+      "loss": 0.8952,
+      "step": 10318
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1520082502594686e-06,
+      "loss": 0.6132,
+      "step": 10319
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1511026839559696e-06,
+      "loss": 0.5684,
+      "step": 10320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1501972560031115e-06,
+      "loss": 0.6753,
+      "step": 10321
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.14929196644486e-06,
+      "loss": 0.6049,
+      "step": 10322
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.148386815325179e-06,
+      "loss": 0.7921,
+      "step": 10323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1474818026880297e-06,
+      "loss": 0.9051,
+      "step": 10324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.146576928577357e-06,
+      "loss": 0.7477,
+      "step": 10325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1456721930371054e-06,
+      "loss": 0.7362,
+      "step": 10326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.144767596111213e-06,
+      "loss": 0.5395,
+      "step": 10327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1438631378436083e-06,
+      "loss": 0.6539,
+      "step": 10328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1429588182782147e-06,
+      "loss": 0.637,
+      "step": 10329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1420546374589486e-06,
+      "loss": 0.929,
+      "step": 10330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1411505954297197e-06,
+      "loss": 0.6996,
+      "step": 10331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1402466922344306e-06,
+      "loss": 0.7445,
+      "step": 10332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.139342927916978e-06,
+      "loss": 0.4711,
+      "step": 10333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.138439302521251e-06,
+      "loss": 0.4928,
+      "step": 10334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1375358160911325e-06,
+      "loss": 0.5397,
+      "step": 10335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1366324686704974e-06,
+      "loss": 0.5488,
+      "step": 10336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1357292603032158e-06,
+      "loss": 0.6424,
+      "step": 10337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1348261910331517e-06,
+      "loss": 0.9495,
+      "step": 10338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1339232609041553e-06,
+      "loss": 0.597,
+      "step": 10339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1330204699600815e-06,
+      "loss": 0.4666,
+      "step": 10340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.132117818244771e-06,
+      "loss": 0.468,
+      "step": 10341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1312153058020546e-06,
+      "loss": 0.7446,
+      "step": 10342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1303129326757663e-06,
+      "loss": 0.6424,
+      "step": 10343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1294106989097286e-06,
+      "loss": 0.7434,
+      "step": 10344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1285086045477515e-06,
+      "loss": 0.651,
+      "step": 10345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.127606649633644e-06,
+      "loss": 0.5728,
+      "step": 10346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1267048342112135e-06,
+      "loss": 0.7465,
+      "step": 10347
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1258031583242483e-06,
+      "loss": 0.5877,
+      "step": 10348
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.124901622016539e-06,
+      "loss": 0.6722,
+      "step": 10349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.124000225331867e-06,
+      "loss": 0.5196,
+      "step": 10350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.123098968314006e-06,
+      "loss": 0.7681,
+      "step": 10351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.122197851006724e-06,
+      "loss": 0.5538,
+      "step": 10352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1212968734537813e-06,
+      "loss": 0.785,
+      "step": 10353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.120396035698933e-06,
+      "loss": 0.872,
+      "step": 10354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.119495337785925e-06,
+      "loss": 0.5201,
+      "step": 10355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1185947797584995e-06,
+      "loss": 0.5683,
+      "step": 10356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1176943616603887e-06,
+      "loss": 0.857,
+      "step": 10357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.11679408353532e-06,
+      "loss": 0.7611,
+      "step": 10358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1158939454270138e-06,
+      "loss": 0.9483,
+      "step": 10359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.114993947379183e-06,
+      "loss": 0.832,
+      "step": 10360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1140940894355345e-06,
+      "loss": 0.7876,
+      "step": 10361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1131943716397673e-06,
+      "loss": 0.6499,
+      "step": 10362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.112294794035575e-06,
+      "loss": 0.6482,
+      "step": 10363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.111395356666645e-06,
+      "loss": 0.5796,
+      "step": 10364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.110496059576651e-06,
+      "loss": 0.6768,
+      "step": 10365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.109596902809273e-06,
+      "loss": 0.6794,
+      "step": 10366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.108697886408172e-06,
+      "loss": 0.7356,
+      "step": 10367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1077990104170053e-06,
+      "loss": 0.7497,
+      "step": 10368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1069002748794316e-06,
+      "loss": 0.7938,
+      "step": 10369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.106001679839089e-06,
+      "loss": 0.7436,
+      "step": 10370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.105103225339618e-06,
+      "loss": 0.797,
+      "step": 10371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1042049114246543e-06,
+      "loss": 0.819,
+      "step": 10372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.103306738137817e-06,
+      "loss": 0.7136,
+      "step": 10373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.102408705522725e-06,
+      "loss": 0.6089,
+      "step": 10374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.1015108136229934e-06,
+      "loss": 0.6867,
+      "step": 10375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.100613062482222e-06,
+      "loss": 0.8092,
+      "step": 10376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.09971545214401e-06,
+      "loss": 1.1937,
+      "step": 10377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0988179826519472e-06,
+      "loss": 0.8807,
+      "step": 10378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.097920654049618e-06,
+      "loss": 0.842,
+      "step": 10379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.097023466380599e-06,
+      "loss": 0.8073,
+      "step": 10380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0961264196884603e-06,
+      "loss": 0.5771,
+      "step": 10381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0952295140167645e-06,
+      "loss": 0.7912,
+      "step": 10382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0943327494090687e-06,
+      "loss": 0.894,
+      "step": 10383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.093436125908922e-06,
+      "loss": 0.7427,
+      "step": 10384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0925396435598665e-06,
+      "loss": 0.841,
+      "step": 10385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.091643302405441e-06,
+      "loss": 0.888,
+      "step": 10386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0907471024891684e-06,
+      "loss": 0.6223,
+      "step": 10387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0898510438545754e-06,
+      "loss": 0.7912,
+      "step": 10388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0889551265451787e-06,
+      "loss": 0.8519,
+      "step": 10389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.08805935060448e-06,
+      "loss": 0.9317,
+      "step": 10390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0871637160759863e-06,
+      "loss": 0.5866,
+      "step": 10391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0862682230031923e-06,
+      "loss": 0.682,
+      "step": 10392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0853728714295807e-06,
+      "loss": 0.5534,
+      "step": 10393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0844776613986394e-06,
+      "loss": 0.5744,
+      "step": 10394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.083582592953837e-06,
+      "loss": 0.6351,
+      "step": 10395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0826876661386397e-06,
+      "loss": 0.7635,
+      "step": 10396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0817928809965136e-06,
+      "loss": 0.8616,
+      "step": 10397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.080898237570908e-06,
+      "loss": 0.6317,
+      "step": 10398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.080003735905268e-06,
+      "loss": 0.7098,
+      "step": 10399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.079109376043038e-06,
+      "loss": 0.8374,
+      "step": 10400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0782151580276467e-06,
+      "loss": 0.689,
+      "step": 10401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.077321081902521e-06,
+      "loss": 0.7381,
+      "step": 10402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.07642714771108e-06,
+      "loss": 0.6304,
+      "step": 10403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.075533355496735e-06,
+      "loss": 0.953,
+      "step": 10404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.074639705302892e-06,
+      "loss": 0.7702,
+      "step": 10405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0737461971729496e-06,
+      "loss": 0.8037,
+      "step": 10406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0728528311502977e-06,
+      "loss": 0.5978,
+      "step": 10407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0719596072783223e-06,
+      "loss": 0.7735,
+      "step": 10408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0710665256003994e-06,
+      "loss": 0.7345,
+      "step": 10409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0701735861599005e-06,
+      "loss": 0.6643,
+      "step": 10410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.069280789000189e-06,
+      "loss": 0.7303,
+      "step": 10411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0683881341646216e-06,
+      "loss": 0.8305,
+      "step": 10412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0674956216965484e-06,
+      "loss": 0.5694,
+      "step": 10413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.066603251639315e-06,
+      "loss": 0.6803,
+      "step": 10414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0657110240362506e-06,
+      "loss": 0.8202,
+      "step": 10415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.06481893893069e-06,
+      "loss": 0.6774,
+      "step": 10416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0639269963659563e-06,
+      "loss": 0.6661,
+      "step": 10417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.063035196385359e-06,
+      "loss": 1.0236,
+      "step": 10418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0621435390322116e-06,
+      "loss": 0.568,
+      "step": 10419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0612520243498154e-06,
+      "loss": 0.7265,
+      "step": 10420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.060360652381461e-06,
+      "loss": 0.5899,
+      "step": 10421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0594694231704375e-06,
+      "loss": 0.8162,
+      "step": 10422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.05857833676003e-06,
+      "loss": 0.7652,
+      "step": 10423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0576873931935065e-06,
+      "loss": 0.804,
+      "step": 10424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0567965925141366e-06,
+      "loss": 0.9514,
+      "step": 10425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.055905934765179e-06,
+      "loss": 0.8227,
+      "step": 10426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0550154199898863e-06,
+      "loss": 0.6226,
+      "step": 10427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.054125048231506e-06,
+      "loss": 0.8454,
+      "step": 10428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.053234819533276e-06,
+      "loss": 1.04,
+      "step": 10429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0523447339384293e-06,
+      "loss": 0.6246,
+      "step": 10430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.05145479149019e-06,
+      "loss": 0.5937,
+      "step": 10431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0505649922317766e-06,
+      "loss": 0.5363,
+      "step": 10432
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0496753362064003e-06,
+      "loss": 0.7535,
+      "step": 10433
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0487858234572653e-06,
+      "loss": 0.9267,
+      "step": 10434
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0478964540275696e-06,
+      "loss": 0.7626,
+      "step": 10435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0470072279605024e-06,
+      "loss": 0.6122,
+      "step": 10436
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0461181452992494e-06,
+      "loss": 0.871,
+      "step": 10437
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.045229206086982e-06,
+      "loss": 0.6924,
+      "step": 10438
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0443404103668744e-06,
+      "loss": 0.789,
+      "step": 10439
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0434517581820893e-06,
+      "loss": 0.5588,
+      "step": 10440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0425632495757776e-06,
+      "loss": 0.5351,
+      "step": 10441
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0416748845910933e-06,
+      "loss": 0.8381,
+      "step": 10442
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0407866632711742e-06,
+      "loss": 1.0421,
+      "step": 10443
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0398985856591543e-06,
+      "loss": 0.6338,
+      "step": 10444
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0390106517981656e-06,
+      "loss": 0.6438,
+      "step": 10445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0381228617313247e-06,
+      "loss": 0.6915,
+      "step": 10446
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.037235215501745e-06,
+      "loss": 0.7108,
+      "step": 10447
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0363477131525384e-06,
+      "loss": 0.752,
+      "step": 10448
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0354603547267985e-06,
+      "loss": 0.8037,
+      "step": 10449
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0345731402676187e-06,
+      "loss": 0.5128,
+      "step": 10450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0336860698180905e-06,
+      "loss": 0.7132,
+      "step": 10451
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0327991434212865e-06,
+      "loss": 0.8787,
+      "step": 10452
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.03191236112028e-06,
+      "loss": 0.705,
+      "step": 10453
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0310257229581358e-06,
+      "loss": 0.6302,
+      "step": 10454
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0301392289779125e-06,
+      "loss": 0.59,
+      "step": 10455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0292528792226596e-06,
+      "loss": 0.4133,
+      "step": 10456
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.028366673735421e-06,
+      "loss": 0.669,
+      "step": 10457
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0274806125592345e-06,
+      "loss": 0.9163,
+      "step": 10458
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0265946957371284e-06,
+      "loss": 0.8788,
+      "step": 10459
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0257089233121264e-06,
+      "loss": 0.7376,
+      "step": 10460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0248232953272435e-06,
+      "loss": 0.7055,
+      "step": 10461
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0239378118254906e-06,
+      "loss": 0.7182,
+      "step": 10462
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.023052472849863e-06,
+      "loss": 0.6366,
+      "step": 10463
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0221672784433616e-06,
+      "loss": 0.5941,
+      "step": 10464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0212822286489737e-06,
+      "loss": 1.0067,
+      "step": 10465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.020397323509675e-06,
+      "loss": 0.8452,
+      "step": 10466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.019512563068443e-06,
+      "loss": 0.4545,
+      "step": 10467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0186279473682458e-06,
+      "loss": 0.871,
+      "step": 10468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0177434764520364e-06,
+      "loss": 0.8853,
+      "step": 10469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.016859150362775e-06,
+      "loss": 0.6505,
+      "step": 10470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.015974969143402e-06,
+      "loss": 0.7491,
+      "step": 10471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0150909328368547e-06,
+      "loss": 0.7051,
+      "step": 10472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0142070414860704e-06,
+      "loss": 0.7044,
+      "step": 10473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.013323295133968e-06,
+      "loss": 1.1511,
+      "step": 10474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0124396938234663e-06,
+      "loss": 0.606,
+      "step": 10475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0115562375974766e-06,
+      "loss": 0.5576,
+      "step": 10476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0106729264989006e-06,
+      "loss": 0.7551,
+      "step": 10477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.009789760570636e-06,
+      "loss": 0.6595,
+      "step": 10478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0089067398555697e-06,
+      "loss": 0.6772,
+      "step": 10479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0080238643965853e-06,
+      "loss": 0.7472,
+      "step": 10480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0071411342365582e-06,
+      "loss": 0.7331,
+      "step": 10481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.006258549418355e-06,
+      "loss": 0.8415,
+      "step": 10482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0053761099848375e-06,
+      "loss": 0.635,
+      "step": 10483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0044938159788585e-06,
+      "loss": 0.7808,
+      "step": 10484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0036116674432653e-06,
+      "loss": 0.577,
+      "step": 10485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0027296644208982e-06,
+      "loss": 0.7574,
+      "step": 10486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.001847806954589e-06,
+      "loss": 0.7899,
+      "step": 10487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0009660950871634e-06,
+      "loss": 0.8183,
+      "step": 10488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0000845288614396e-06,
+      "loss": 0.7147,
+      "step": 10489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.999203108320232e-06,
+      "loss": 0.889,
+      "step": 10490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9983218335063376e-06,
+      "loss": 0.8244,
+      "step": 10491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9974407044625604e-06,
+      "loss": 0.5638,
+      "step": 10492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.99655972123169e-06,
+      "loss": 0.545,
+      "step": 10493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9956788838565065e-06,
+      "loss": 0.7555,
+      "step": 10494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9947981923797856e-06,
+      "loss": 0.9742,
+      "step": 10495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9939176468443004e-06,
+      "loss": 0.824,
+      "step": 10496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9930372472928095e-06,
+      "loss": 0.862,
+      "step": 10497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.992156993768066e-06,
+      "loss": 0.8282,
+      "step": 10498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.991276886312824e-06,
+      "loss": 0.7742,
+      "step": 10499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9903969249698175e-06,
+      "loss": 0.6811,
+      "step": 10500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.989517109781783e-06,
+      "loss": 0.9067,
+      "step": 10501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9886374407914453e-06,
+      "loss": 0.7327,
+      "step": 10502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9877579180415252e-06,
+      "loss": 0.629,
+      "step": 10503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9868785415747343e-06,
+      "loss": 0.7664,
+      "step": 10504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9859993114337773e-06,
+      "loss": 0.7035,
+      "step": 10505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9851202276613524e-06,
+      "loss": 0.9662,
+      "step": 10506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.984241290300151e-06,
+      "loss": 0.8048,
+      "step": 10507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.983362499392856e-06,
+      "loss": 0.905,
+      "step": 10508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9824838549821433e-06,
+      "loss": 0.6789,
+      "step": 10509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.981605357110685e-06,
+      "loss": 0.5948,
+      "step": 10510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.980727005821141e-06,
+      "loss": 0.9736,
+      "step": 10511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9798488011561674e-06,
+      "loss": 0.7782,
+      "step": 10512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9789707431584137e-06,
+      "loss": 0.8486,
+      "step": 10513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9780928318705166e-06,
+      "loss": 0.5588,
+      "step": 10514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9772150673351146e-06,
+      "loss": 0.6486,
+      "step": 10515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.976337449594835e-06,
+      "loss": 1.0317,
+      "step": 10516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9754599786922913e-06,
+      "loss": 0.3777,
+      "step": 10517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9745826546701035e-06,
+      "loss": 0.5312,
+      "step": 10518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.973705477570872e-06,
+      "loss": 0.6742,
+      "step": 10519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9728284474371943e-06,
+      "loss": 0.589,
+      "step": 10520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.971951564311668e-06,
+      "loss": 0.8337,
+      "step": 10521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.97107482823687e-06,
+      "loss": 0.5932,
+      "step": 10522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.970198239255379e-06,
+      "loss": 0.6097,
+      "step": 10523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.969321797409769e-06,
+      "loss": 0.5961,
+      "step": 10524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.968445502742597e-06,
+      "loss": 0.8145,
+      "step": 10525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9675693552964187e-06,
+      "loss": 0.9356,
+      "step": 10526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9666933551137877e-06,
+      "loss": 0.4472,
+      "step": 10527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.96581750223724e-06,
+      "loss": 1.0672,
+      "step": 10528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.964941796709311e-06,
+      "loss": 0.7386,
+      "step": 10529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9640662385725275e-06,
+      "loss": 0.6674,
+      "step": 10530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.963190827869409e-06,
+      "loss": 0.7755,
+      "step": 10531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.962315564642468e-06,
+      "loss": 0.6949,
+      "step": 10532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9614404489342097e-06,
+      "loss": 0.8376,
+      "step": 10533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9605654807871334e-06,
+      "loss": 0.947,
+      "step": 10534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9596906602437283e-06,
+      "loss": 0.8742,
+      "step": 10535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.958815987346479e-06,
+      "loss": 0.848,
+      "step": 10536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9579414621378624e-06,
+      "loss": 0.8201,
+      "step": 10537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.957067084660349e-06,
+      "loss": 0.9475,
+      "step": 10538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.956192854956397e-06,
+      "loss": 0.7233,
+      "step": 10539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.955318773068465e-06,
+      "loss": 0.7061,
+      "step": 10540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.954444839039003e-06,
+      "loss": 0.8791,
+      "step": 10541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9535710529104446e-06,
+      "loss": 0.6431,
+      "step": 10542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9526974147252297e-06,
+      "loss": 0.7118,
+      "step": 10543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9518239245257848e-06,
+      "loss": 0.7845,
+      "step": 10544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9509505823545232e-06,
+      "loss": 0.6106,
+      "step": 10545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.950077388253864e-06,
+      "loss": 0.4991,
+      "step": 10546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.949204342266208e-06,
+      "loss": 0.8197,
+      "step": 10547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.948331444433951e-06,
+      "loss": 0.9084,
+      "step": 10548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.947458694799489e-06,
+      "loss": 0.416,
+      "step": 10549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9465860934052006e-06,
+      "loss": 0.6651,
+      "step": 10550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.945713640293464e-06,
+      "loss": 0.6073,
+      "step": 10551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.944841335506647e-06,
+      "loss": 0.6245,
+      "step": 10552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.943969179087112e-06,
+      "loss": 0.7678,
+      "step": 10553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9430971710772122e-06,
+      "loss": 0.6265,
+      "step": 10554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9422253115192967e-06,
+      "loss": 0.8634,
+      "step": 10555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9413536004557037e-06,
+      "loss": 0.7766,
+      "step": 10556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9404820379287677e-06,
+      "loss": 0.5799,
+      "step": 10557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9396106239808126e-06,
+      "loss": 0.7688,
+      "step": 10558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.938739358654158e-06,
+      "loss": 0.843,
+      "step": 10559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9378682419911145e-06,
+      "loss": 0.7507,
+      "step": 10560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.936997274033986e-06,
+      "loss": 0.7451,
+      "step": 10561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.936126454825069e-06,
+      "loss": 0.763,
+      "step": 10562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9352557844066538e-06,
+      "loss": 0.5523,
+      "step": 10563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9343852628210224e-06,
+      "loss": 0.7596,
+      "step": 10564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.933514890110449e-06,
+      "loss": 0.7052,
+      "step": 10565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9326446663172036e-06,
+      "loss": 0.6914,
+      "step": 10566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.931774591483544e-06,
+      "loss": 0.8253,
+      "step": 10567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9309046656517227e-06,
+      "loss": 0.6078,
+      "step": 10568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9300348888639915e-06,
+      "loss": 0.7996,
+      "step": 10569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.929165261162584e-06,
+      "loss": 0.7744,
+      "step": 10570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.928295782589731e-06,
+      "loss": 0.6767,
+      "step": 10571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.927426453187663e-06,
+      "loss": 0.6975,
+      "step": 10572
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9265572729985916e-06,
+      "loss": 0.8054,
+      "step": 10573
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9256882420647264e-06,
+      "loss": 0.5415,
+      "step": 10574
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.924819360428276e-06,
+      "loss": 0.7296,
+      "step": 10575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.92395062813143e-06,
+      "loss": 0.6049,
+      "step": 10576
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9230820452163786e-06,
+      "loss": 0.7492,
+      "step": 10577
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.922213611725302e-06,
+      "loss": 0.713,
+      "step": 10578
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.921345327700375e-06,
+      "loss": 0.5108,
+      "step": 10579
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9204771931837628e-06,
+      "loss": 0.8493,
+      "step": 10580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9196092082176254e-06,
+      "loss": 0.6493,
+      "step": 10581
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.918741372844114e-06,
+      "loss": 0.8069,
+      "step": 10582
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9178736871053737e-06,
+      "loss": 0.9515,
+      "step": 10583
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9170061510435413e-06,
+      "loss": 0.7,
+      "step": 10584
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.916138764700747e-06,
+      "loss": 0.611,
+      "step": 10585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.915271528119116e-06,
+      "loss": 0.7978,
+      "step": 10586
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9144044413407577e-06,
+      "loss": 0.6697,
+      "step": 10587
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9135375044077854e-06,
+      "loss": 0.4017,
+      "step": 10588
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9126707173623004e-06,
+      "loss": 0.8107,
+      "step": 10589
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9118040802463918e-06,
+      "loss": 0.9099,
+      "step": 10590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.91093759310215e-06,
+      "loss": 0.7792,
+      "step": 10591
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9100712559716554e-06,
+      "loss": 0.8426,
+      "step": 10592
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9092050688969736e-06,
+      "loss": 0.7065,
+      "step": 10593
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9083390319201766e-06,
+      "loss": 0.7727,
+      "step": 10594
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.907473145083316e-06,
+      "loss": 0.6198,
+      "step": 10595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.906607408428442e-06,
+      "loss": 0.8034,
+      "step": 10596
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9057418219976026e-06,
+      "loss": 0.6184,
+      "step": 10597
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.904876385832828e-06,
+      "loss": 1.0433,
+      "step": 10598
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9040110999761458e-06,
+      "loss": 0.9024,
+      "step": 10599
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.903145964469582e-06,
+      "loss": 0.7542,
+      "step": 10600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.902280979355146e-06,
+      "loss": 0.4259,
+      "step": 10601
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9014161446748424e-06,
+      "loss": 0.8589,
+      "step": 10602
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9005514604706765e-06,
+      "loss": 0.8062,
+      "step": 10603
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8996869267846335e-06,
+      "loss": 0.5628,
+      "step": 10604
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8988225436587005e-06,
+      "loss": 0.6641,
+      "step": 10605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.897958311134853e-06,
+      "loss": 0.5436,
+      "step": 10606
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8970942292550625e-06,
+      "loss": 0.7988,
+      "step": 10607
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8962302980612902e-06,
+      "loss": 0.5435,
+      "step": 10608
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8953665175954905e-06,
+      "loss": 0.4508,
+      "step": 10609
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8945028878996125e-06,
+      "loss": 0.5499,
+      "step": 10610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8936394090155952e-06,
+      "loss": 0.6055,
+      "step": 10611
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8927760809853724e-06,
+      "loss": 0.4122,
+      "step": 10612
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.89191290385087e-06,
+      "loss": 0.8133,
+      "step": 10613
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.891049877654007e-06,
+      "loss": 0.6704,
+      "step": 10614
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8901870024366908e-06,
+      "loss": 0.7802,
+      "step": 10615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8893242782408289e-06,
+      "loss": 0.6343,
+      "step": 10616
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8884617051083183e-06,
+      "loss": 0.5661,
+      "step": 10617
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8875992830810424e-06,
+      "loss": 0.7244,
+      "step": 10618
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8867370122008888e-06,
+      "loss": 0.6818,
+      "step": 10619
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8858748925097309e-06,
+      "loss": 0.6359,
+      "step": 10620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8850129240494314e-06,
+      "loss": 0.5327,
+      "step": 10621
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8841511068618569e-06,
+      "loss": 0.741,
+      "step": 10622
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.883289440988854e-06,
+      "loss": 0.7019,
+      "step": 10623
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8824279264722694e-06,
+      "loss": 0.5757,
+      "step": 10624
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.881566563353941e-06,
+      "loss": 0.7797,
+      "step": 10625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.880705351675699e-06,
+      "loss": 0.9557,
+      "step": 10626
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8798442914793663e-06,
+      "loss": 0.8298,
+      "step": 10627
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8789833828067582e-06,
+      "loss": 0.5976,
+      "step": 10628
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8781226256996838e-06,
+      "loss": 0.5039,
+      "step": 10629
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8772620201999424e-06,
+      "loss": 0.4874,
+      "step": 10630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.876401566349329e-06,
+      "loss": 0.972,
+      "step": 10631
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8755412641896287e-06,
+      "loss": 0.7151,
+      "step": 10632
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8746811137626208e-06,
+      "loss": 0.8463,
+      "step": 10633
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8738211151100766e-06,
+      "loss": 0.6319,
+      "step": 10634
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8729612682737598e-06,
+      "loss": 0.4729,
+      "step": 10635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8721015732954272e-06,
+      "loss": 0.8946,
+      "step": 10636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8712420302168283e-06,
+      "loss": 0.8484,
+      "step": 10637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8703826390797047e-06,
+      "loss": 0.6638,
+      "step": 10638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.869523399925791e-06,
+      "loss": 0.879,
+      "step": 10639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.868664312796814e-06,
+      "loss": 0.9679,
+      "step": 10640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.867805377734494e-06,
+      "loss": 0.7107,
+      "step": 10641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8669465947805448e-06,
+      "loss": 0.629,
+      "step": 10642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8660879639766676e-06,
+      "loss": 0.7946,
+      "step": 10643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.865229485364561e-06,
+      "loss": 0.5527,
+      "step": 10644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8643711589859193e-06,
+      "loss": 0.6167,
+      "step": 10645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8635129848824202e-06,
+      "loss": 0.9044,
+      "step": 10646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8626549630957397e-06,
+      "loss": 0.7637,
+      "step": 10647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8617970936675505e-06,
+      "loss": 0.8796,
+      "step": 10648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8609393766395083e-06,
+      "loss": 0.5445,
+      "step": 10649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.860081812053266e-06,
+      "loss": 0.6224,
+      "step": 10650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.859224399950475e-06,
+      "loss": 0.7135,
+      "step": 10651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8583671403727683e-06,
+      "loss": 0.4288,
+      "step": 10652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8575100333617784e-06,
+      "loss": 0.5648,
+      "step": 10653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.856653078959129e-06,
+      "loss": 0.6611,
+      "step": 10654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.855796277206437e-06,
+      "loss": 0.6844,
+      "step": 10655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8549396281453098e-06,
+      "loss": 0.8321,
+      "step": 10656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8540831318173496e-06,
+      "loss": 0.5132,
+      "step": 10657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8532267882641503e-06,
+      "loss": 0.542,
+      "step": 10658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8523705975272981e-06,
+      "loss": 0.8544,
+      "step": 10659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8515145596483725e-06,
+      "loss": 0.622,
+      "step": 10660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8506586746689448e-06,
+      "loss": 0.8528,
+      "step": 10661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8498029426305814e-06,
+      "loss": 0.7261,
+      "step": 10662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.848947363574833e-06,
+      "loss": 0.6521,
+      "step": 10663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.848091937543255e-06,
+      "loss": 0.7639,
+      "step": 10664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8472366645773892e-06,
+      "loss": 0.5994,
+      "step": 10665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8463815447187644e-06,
+      "loss": 0.7921,
+      "step": 10666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8455265780089138e-06,
+      "loss": 0.7306,
+      "step": 10667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8446717644893563e-06,
+      "loss": 0.7968,
+      "step": 10668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8438171042015996e-06,
+      "loss": 0.9249,
+      "step": 10669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8429625971871546e-06,
+      "loss": 0.7327,
+      "step": 10670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8421082434875133e-06,
+      "loss": 0.3853,
+      "step": 10671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8412540431441666e-06,
+      "loss": 0.87,
+      "step": 10672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8403999961986012e-06,
+      "loss": 0.9646,
+      "step": 10673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8395461026922873e-06,
+      "loss": 0.7515,
+      "step": 10674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8386923626666925e-06,
+      "loss": 0.787,
+      "step": 10675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8378387761632815e-06,
+      "loss": 0.4571,
+      "step": 10676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8369853432235024e-06,
+      "loss": 0.9004,
+      "step": 10677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8361320638888002e-06,
+      "loss": 0.8721,
+      "step": 10678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8352789382006175e-06,
+      "loss": 0.6844,
+      "step": 10679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8344259662003793e-06,
+      "loss": 0.7186,
+      "step": 10680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8335731479295105e-06,
+      "loss": 0.7541,
+      "step": 10681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.832720483429426e-06,
+      "loss": 0.6021,
+      "step": 10682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.831867972741534e-06,
+      "loss": 0.7637,
+      "step": 10683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.831015615907235e-06,
+      "loss": 0.6743,
+      "step": 10684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8301634129679214e-06,
+      "loss": 1.1591,
+      "step": 10685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8293113639649785e-06,
+      "loss": 0.7302,
+      "step": 10686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.828459468939785e-06,
+      "loss": 0.8231,
+      "step": 10687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8276077279337107e-06,
+      "loss": 0.5854,
+      "step": 10688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.826756140988119e-06,
+      "loss": 0.9318,
+      "step": 10689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8259047081443671e-06,
+      "loss": 0.6222,
+      "step": 10690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8250534294437972e-06,
+      "loss": 0.5645,
+      "step": 10691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8242023049277558e-06,
+      "loss": 0.6829,
+      "step": 10692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.823351334637576e-06,
+      "loss": 0.8888,
+      "step": 10693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8225005186145778e-06,
+      "loss": 0.6057,
+      "step": 10694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.821649856900084e-06,
+      "loss": 0.6859,
+      "step": 10695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.820799349535406e-06,
+      "loss": 0.7552,
+      "step": 10696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8199489965618433e-06,
+      "loss": 0.7453,
+      "step": 10697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8190987980206932e-06,
+      "loss": 0.8451,
+      "step": 10698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8182487539532428e-06,
+      "loss": 0.7229,
+      "step": 10699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8173988644007746e-06,
+      "loss": 0.5484,
+      "step": 10700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8165491294045596e-06,
+      "loss": 0.8735,
+      "step": 10701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8156995490058648e-06,
+      "loss": 0.8373,
+      "step": 10702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8148501232459477e-06,
+      "loss": 0.6991,
+      "step": 10703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.814000852166059e-06,
+      "loss": 0.6908,
+      "step": 10704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8131517358074418e-06,
+      "loss": 0.4498,
+      "step": 10705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8123027742113314e-06,
+      "loss": 0.6591,
+      "step": 10706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8114539674189564e-06,
+      "loss": 0.5837,
+      "step": 10707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8106053154715364e-06,
+      "loss": 0.7224,
+      "step": 10708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8097568184102848e-06,
+      "loss": 0.474,
+      "step": 10709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8089084762764076e-06,
+      "loss": 0.6554,
+      "step": 10710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8080602891111014e-06,
+      "loss": 0.9633,
+      "step": 10711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8072122569555578e-06,
+      "loss": 0.6099,
+      "step": 10712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8063643798509594e-06,
+      "loss": 0.974,
+      "step": 10713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8055166578384814e-06,
+      "loss": 0.4566,
+      "step": 10714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8046690909592919e-06,
+      "loss": 0.7487,
+      "step": 10715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8038216792545526e-06,
+      "loss": 0.5653,
+      "step": 10716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.802974422765411e-06,
+      "loss": 0.8497,
+      "step": 10717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8021273215330194e-06,
+      "loss": 0.6271,
+      "step": 10718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8012803755985098e-06,
+      "loss": 0.7695,
+      "step": 10719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.8004335850030137e-06,
+      "loss": 0.6793,
+      "step": 10720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7995869497876573e-06,
+      "loss": 0.8157,
+      "step": 10721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7987404699935512e-06,
+      "loss": 0.6861,
+      "step": 10722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7978941456618027e-06,
+      "loss": 0.6063,
+      "step": 10723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.797047976833517e-06,
+      "loss": 0.9339,
+      "step": 10724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7962019635497812e-06,
+      "loss": 0.4834,
+      "step": 10725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7953561058516826e-06,
+      "loss": 0.9682,
+      "step": 10726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7945104037802974e-06,
+      "loss": 0.6539,
+      "step": 10727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7936648573766956e-06,
+      "loss": 0.6869,
+      "step": 10728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7928194666819398e-06,
+      "loss": 0.7017,
+      "step": 10729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7919742317370846e-06,
+      "loss": 1.0648,
+      "step": 10730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7911291525831764e-06,
+      "loss": 0.7049,
+      "step": 10731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7902842292612554e-06,
+      "loss": 0.8007,
+      "step": 10732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7894394618123527e-06,
+      "loss": 0.3482,
+      "step": 10733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7885948502774936e-06,
+      "loss": 0.7712,
+      "step": 10734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7877503946976954e-06,
+      "loss": 0.8023,
+      "step": 10735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7869060951139634e-06,
+      "loss": 0.8018,
+      "step": 10736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7860619515673034e-06,
+      "loss": 0.785,
+      "step": 10737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7852179640987099e-06,
+      "loss": 0.7528,
+      "step": 10738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7843741327491642e-06,
+      "loss": 0.8711,
+      "step": 10739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7835304575596501e-06,
+      "loss": 0.7611,
+      "step": 10740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7826869385711392e-06,
+      "loss": 0.9457,
+      "step": 10741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.78184357582459e-06,
+      "loss": 0.6515,
+      "step": 10742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7810003693609634e-06,
+      "loss": 0.7643,
+      "step": 10743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7801573192212086e-06,
+      "loss": 0.6868,
+      "step": 10744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7793144254462601e-06,
+      "loss": 0.8218,
+      "step": 10745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7784716880770597e-06,
+      "loss": 0.7418,
+      "step": 10746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7776291071545271e-06,
+      "loss": 0.6763,
+      "step": 10747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7767866827195806e-06,
+      "loss": 0.6662,
+      "step": 10748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7759444148131361e-06,
+      "loss": 0.6933,
+      "step": 10749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7751023034760911e-06,
+      "loss": 1.1287,
+      "step": 10750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7742603487493409e-06,
+      "loss": 0.6723,
+      "step": 10751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.773418550673779e-06,
+      "loss": 0.6171,
+      "step": 10752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7725769092902794e-06,
+      "loss": 0.6842,
+      "step": 10753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7717354246397173e-06,
+      "loss": 0.6332,
+      "step": 10754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7708940967629567e-06,
+      "loss": 0.7256,
+      "step": 10755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7700529257008558e-06,
+      "loss": 0.5389,
+      "step": 10756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.769211911494264e-06,
+      "loss": 0.6488,
+      "step": 10757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7683710541840233e-06,
+      "loss": 0.8416,
+      "step": 10758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7675303538109679e-06,
+      "loss": 0.8483,
+      "step": 10759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7666898104159248e-06,
+      "loss": 0.6123,
+      "step": 10760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7658494240397127e-06,
+      "loss": 1.0125,
+      "step": 10761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7650091947231446e-06,
+      "loss": 0.7913,
+      "step": 10762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7641691225070246e-06,
+      "loss": 0.7259,
+      "step": 10763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7633292074321444e-06,
+      "loss": 0.512,
+      "step": 10764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7624894495392985e-06,
+      "loss": 0.7451,
+      "step": 10765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7616498488692673e-06,
+      "loss": 0.5626,
+      "step": 10766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7608104054628195e-06,
+      "loss": 1.1318,
+      "step": 10767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.759971119360725e-06,
+      "loss": 0.9453,
+      "step": 10768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7591319906037429e-06,
+      "loss": 0.819,
+      "step": 10769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7582930192326187e-06,
+      "loss": 0.656,
+      "step": 10770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7574542052881e-06,
+      "loss": 0.7392,
+      "step": 10771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7566155488109215e-06,
+      "loss": 0.8712,
+      "step": 10772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7557770498418087e-06,
+      "loss": 0.8026,
+      "step": 10773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.754938708421482e-06,
+      "loss": 0.665,
+      "step": 10774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.754100524590654e-06,
+      "loss": 0.8671,
+      "step": 10775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7532624983900294e-06,
+      "loss": 0.8118,
+      "step": 10776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7524246298603053e-06,
+      "loss": 0.781,
+      "step": 10777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7515869190421708e-06,
+      "loss": 0.7633,
+      "step": 10778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7507493659763076e-06,
+      "loss": 0.8018,
+      "step": 10779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7499119707033897e-06,
+      "loss": 0.8804,
+      "step": 10780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7490747332640833e-06,
+      "loss": 0.881,
+      "step": 10781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7482376536990476e-06,
+      "loss": 0.7322,
+      "step": 10782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7474007320489328e-06,
+      "loss": 0.875,
+      "step": 10783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7465639683543822e-06,
+      "loss": 0.4263,
+      "step": 10784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7457273626560328e-06,
+      "loss": 0.9552,
+      "step": 10785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.744890914994511e-06,
+      "loss": 0.7171,
+      "step": 10786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7440546254104374e-06,
+      "loss": 0.965,
+      "step": 10787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7432184939444258e-06,
+      "loss": 0.7,
+      "step": 10788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7423825206370793e-06,
+      "loss": 1.048,
+      "step": 10789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.741546705528997e-06,
+      "loss": 0.9934,
+      "step": 10790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.740711048660767e-06,
+      "loss": 0.8576,
+      "step": 10791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7398755500729737e-06,
+      "loss": 0.8258,
+      "step": 10792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.739040209806186e-06,
+      "loss": 0.6819,
+      "step": 10793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7382050279009776e-06,
+      "loss": 0.7228,
+      "step": 10794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7373700043979015e-06,
+      "loss": 0.6545,
+      "step": 10795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7365351393375097e-06,
+      "loss": 0.8328,
+      "step": 10796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.73570043276035e-06,
+      "loss": 0.8235,
+      "step": 10797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7348658847069528e-06,
+      "loss": 0.548,
+      "step": 10798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7340314952178472e-06,
+      "loss": 0.7311,
+      "step": 10799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7331972643335582e-06,
+      "loss": 0.8571,
+      "step": 10800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7323631920945933e-06,
+      "loss": 0.8737,
+      "step": 10801
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.731529278541459e-06,
+      "loss": 0.5051,
+      "step": 10802
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7306955237146523e-06,
+      "loss": 0.518,
+      "step": 10803
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7298619276546635e-06,
+      "loss": 0.988,
+      "step": 10804
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7290284904019738e-06,
+      "loss": 0.9731,
+      "step": 10805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.728195211997058e-06,
+      "loss": 0.7469,
+      "step": 10806
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7273620924803813e-06,
+      "loss": 0.4583,
+      "step": 10807
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7265291318924032e-06,
+      "loss": 0.5167,
+      "step": 10808
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7256963302735752e-06,
+      "loss": 0.689,
+      "step": 10809
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7248636876643393e-06,
+      "loss": 0.7566,
+      "step": 10810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7240312041051333e-06,
+      "loss": 0.823,
+      "step": 10811
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7231988796363802e-06,
+      "loss": 0.717,
+      "step": 10812
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7223667142985046e-06,
+      "loss": 0.9425,
+      "step": 10813
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7215347081319189e-06,
+      "loss": 0.8751,
+      "step": 10814
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7207028611770233e-06,
+      "loss": 0.6215,
+      "step": 10815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.719871173474219e-06,
+      "loss": 0.6684,
+      "step": 10816
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7190396450638958e-06,
+      "loss": 0.7519,
+      "step": 10817
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7182082759864293e-06,
+      "loss": 0.5729,
+      "step": 10818
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7173770662821986e-06,
+      "loss": 0.7472,
+      "step": 10819
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7165460159915697e-06,
+      "loss": 0.7249,
+      "step": 10820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7157151251548953e-06,
+      "loss": 0.6856,
+      "step": 10821
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7148843938125336e-06,
+      "loss": 0.6464,
+      "step": 10822
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7140538220048213e-06,
+      "loss": 0.7429,
+      "step": 10823
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7132234097720935e-06,
+      "loss": 0.7477,
+      "step": 10824
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7123931571546826e-06,
+      "loss": 0.8183,
+      "step": 10825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7115630641929032e-06,
+      "loss": 0.6688,
+      "step": 10826
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7107331309270686e-06,
+      "loss": 0.801,
+      "step": 10827
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7099033573974822e-06,
+      "loss": 0.5971,
+      "step": 10828
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.709073743644441e-06,
+      "loss": 0.6185,
+      "step": 10829
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7082442897082329e-06,
+      "loss": 0.688,
+      "step": 10830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7074149956291386e-06,
+      "loss": 0.7659,
+      "step": 10831
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7065858614474311e-06,
+      "loss": 0.8027,
+      "step": 10832
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7057568872033758e-06,
+      "loss": 0.5801,
+      "step": 10833
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7049280729372296e-06,
+      "loss": 0.9212,
+      "step": 10834
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7040994186892423e-06,
+      "loss": 0.579,
+      "step": 10835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7032709244996559e-06,
+      "loss": 0.6308,
+      "step": 10836
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7024425904087043e-06,
+      "loss": 0.5569,
+      "step": 10837
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7016144164566134e-06,
+      "loss": 0.7192,
+      "step": 10838
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.7007864026836046e-06,
+      "loss": 0.6896,
+      "step": 10839
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.699958549129882e-06,
+      "loss": 0.7449,
+      "step": 10840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6991308558356545e-06,
+      "loss": 0.8388,
+      "step": 10841
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6983033228411172e-06,
+      "loss": 0.7819,
+      "step": 10842
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6974759501864523e-06,
+      "loss": 0.7586,
+      "step": 10843
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6966487379118452e-06,
+      "loss": 0.7547,
+      "step": 10844
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6958216860574661e-06,
+      "loss": 0.8559,
+      "step": 10845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6949947946634771e-06,
+      "loss": 0.8362,
+      "step": 10846
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6941680637700335e-06,
+      "loss": 0.8254,
+      "step": 10847
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6933414934172893e-06,
+      "loss": 0.7254,
+      "step": 10848
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6925150836453797e-06,
+      "loss": 0.6165,
+      "step": 10849
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6916888344944398e-06,
+      "loss": 0.5553,
+      "step": 10850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.690862746004594e-06,
+      "loss": 0.8431,
+      "step": 10851
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6900368182159598e-06,
+      "loss": 0.8093,
+      "step": 10852
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6892110511686466e-06,
+      "loss": 0.5168,
+      "step": 10853
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6883854449027564e-06,
+      "loss": 0.5953,
+      "step": 10854
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6875599994583825e-06,
+      "loss": 0.6503,
+      "step": 10855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6867347148756108e-06,
+      "loss": 0.6002,
+      "step": 10856
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.68590959119452e-06,
+      "loss": 0.7897,
+      "step": 10857
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6850846284551804e-06,
+      "loss": 0.6009,
+      "step": 10858
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.684259826697654e-06,
+      "loss": 1.0619,
+      "step": 10859
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6834351859619964e-06,
+      "loss": 0.6364,
+      "step": 10860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6826107062882534e-06,
+      "loss": 0.6991,
+      "step": 10861
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6817863877164654e-06,
+      "loss": 0.5895,
+      "step": 10862
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6809622302866623e-06,
+      "loss": 0.9299,
+      "step": 10863
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6801382340388689e-06,
+      "loss": 0.7542,
+      "step": 10864
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6793143990131017e-06,
+      "loss": 0.6016,
+      "step": 10865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6784907252493638e-06,
+      "loss": 0.8181,
+      "step": 10866
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6776672127876603e-06,
+      "loss": 0.8221,
+      "step": 10867
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6768438616679828e-06,
+      "loss": 0.8308,
+      "step": 10868
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6760206719303107e-06,
+      "loss": 0.8991,
+      "step": 10869
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6751976436146273e-06,
+      "loss": 0.6024,
+      "step": 10870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6743747767608965e-06,
+      "loss": 0.7093,
+      "step": 10871
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.673552071409078e-06,
+      "loss": 0.6436,
+      "step": 10872
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6727295275991311e-06,
+      "loss": 0.74,
+      "step": 10873
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6719071453709956e-06,
+      "loss": 0.5716,
+      "step": 10874
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6710849247646083e-06,
+      "loss": 0.8117,
+      "step": 10875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6702628658199037e-06,
+      "loss": 0.7217,
+      "step": 10876
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6694409685767981e-06,
+      "loss": 0.7659,
+      "step": 10877
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6686192330752076e-06,
+      "loss": 0.6922,
+      "step": 10878
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.667797659355037e-06,
+      "loss": 0.6478,
+      "step": 10879
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6669762474561857e-06,
+      "loss": 0.7289,
+      "step": 10880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6661549974185426e-06,
+      "loss": 0.8077,
+      "step": 10881
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6653339092819904e-06,
+      "loss": 0.7897,
+      "step": 10882
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6645129830864032e-06,
+      "loss": 0.7332,
+      "step": 10883
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6636922188716481e-06,
+      "loss": 0.5662,
+      "step": 10884
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6628716166775833e-06,
+      "loss": 0.6779,
+      "step": 10885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6620511765440595e-06,
+      "loss": 0.7562,
+      "step": 10886
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6612308985109222e-06,
+      "loss": 0.7214,
+      "step": 10887
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6604107826179999e-06,
+      "loss": 0.8712,
+      "step": 10888
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6595908289051266e-06,
+      "loss": 0.9129,
+      "step": 10889
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6587710374121203e-06,
+      "loss": 0.9332,
+      "step": 10890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6579514081787879e-06,
+      "loss": 0.56,
+      "step": 10891
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6571319412449377e-06,
+      "loss": 0.7506,
+      "step": 10892
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6563126366503657e-06,
+      "loss": 0.8672,
+      "step": 10893
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6554934944348544e-06,
+      "loss": 0.8909,
+      "step": 10894
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6546745146381892e-06,
+      "loss": 0.9825,
+      "step": 10895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6538556973001413e-06,
+      "loss": 0.8253,
+      "step": 10896
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6530370424604708e-06,
+      "loss": 0.4688,
+      "step": 10897
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6522185501589394e-06,
+      "loss": 0.5981,
+      "step": 10898
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6514002204352919e-06,
+      "loss": 0.8029,
+      "step": 10899
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.650582053329267e-06,
+      "loss": 0.5974,
+      "step": 10900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.649764048880604e-06,
+      "loss": 0.6183,
+      "step": 10901
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6489462071290213e-06,
+      "loss": 0.5304,
+      "step": 10902
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.648128528114238e-06,
+      "loss": 0.8875,
+      "step": 10903
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6473110118759628e-06,
+      "loss": 0.6336,
+      "step": 10904
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.646493658453896e-06,
+      "loss": 0.6959,
+      "step": 10905
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6456764678877319e-06,
+      "loss": 0.8184,
+      "step": 10906
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6448594402171542e-06,
+      "loss": 0.7471,
+      "step": 10907
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6440425754818413e-06,
+      "loss": 0.8588,
+      "step": 10908
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6432258737214613e-06,
+      "loss": 0.7951,
+      "step": 10909
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.642409334975677e-06,
+      "loss": 0.8858,
+      "step": 10910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6415929592841407e-06,
+      "loss": 0.7404,
+      "step": 10911
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6407767466864988e-06,
+      "loss": 0.6672,
+      "step": 10912
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.639960697222388e-06,
+      "loss": 0.9768,
+      "step": 10913
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.639144810931439e-06,
+      "loss": 0.7754,
+      "step": 10914
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6383290878532743e-06,
+      "loss": 1.036,
+      "step": 10915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6375135280275033e-06,
+      "loss": 0.8213,
+      "step": 10916
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6366981314937374e-06,
+      "loss": 0.4891,
+      "step": 10917
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6358828982915736e-06,
+      "loss": 0.5043,
+      "step": 10918
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6350678284605975e-06,
+      "loss": 0.7637,
+      "step": 10919
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6342529220403963e-06,
+      "loss": 0.9121,
+      "step": 10920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6334381790705439e-06,
+      "loss": 0.3949,
+      "step": 10921
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.632623599590603e-06,
+      "loss": 0.7776,
+      "step": 10922
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6318091836401328e-06,
+      "loss": 0.7651,
+      "step": 10923
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6309949312586882e-06,
+      "loss": 0.7768,
+      "step": 10924
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6301808424858067e-06,
+      "loss": 0.6161,
+      "step": 10925
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6293669173610238e-06,
+      "loss": 0.6708,
+      "step": 10926
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.628553155923867e-06,
+      "loss": 0.7148,
+      "step": 10927
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6277395582138545e-06,
+      "loss": 1.062,
+      "step": 10928
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.626926124270497e-06,
+      "loss": 0.5664,
+      "step": 10929
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6261128541332972e-06,
+      "loss": 0.8535,
+      "step": 10930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6252997478417497e-06,
+      "loss": 0.6828,
+      "step": 10931
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6244868054353412e-06,
+      "loss": 0.7945,
+      "step": 10932
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6236740269535505e-06,
+      "loss": 0.5275,
+      "step": 10933
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6228614124358488e-06,
+      "loss": 0.7705,
+      "step": 10934
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6220489619216988e-06,
+      "loss": 0.9433,
+      "step": 10935
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6212366754505559e-06,
+      "loss": 0.8021,
+      "step": 10936
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6204245530618662e-06,
+      "loss": 1.1432,
+      "step": 10937
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6196125947950692e-06,
+      "loss": 0.6632,
+      "step": 10938
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6188008006895956e-06,
+      "loss": 0.7923,
+      "step": 10939
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6179891707848689e-06,
+      "loss": 0.8156,
+      "step": 10940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6171777051203052e-06,
+      "loss": 0.708,
+      "step": 10941
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6163664037353077e-06,
+      "loss": 0.6927,
+      "step": 10942
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6155552666692797e-06,
+      "loss": 0.5785,
+      "step": 10943
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6147442939616126e-06,
+      "loss": 0.7499,
+      "step": 10944
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6139334856516842e-06,
+      "loss": 0.5742,
+      "step": 10945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6131228417788762e-06,
+      "loss": 0.7263,
+      "step": 10946
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.612312362382552e-06,
+      "loss": 0.7438,
+      "step": 10947
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6115020475020697e-06,
+      "loss": 0.6881,
+      "step": 10948
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.610691897176786e-06,
+      "loss": 0.6499,
+      "step": 10949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.609881911446039e-06,
+      "loss": 0.7064,
+      "step": 10950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6090720903491637e-06,
+      "loss": 0.7452,
+      "step": 10951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6082624339254927e-06,
+      "loss": 0.5328,
+      "step": 10952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6074529422143398e-06,
+      "loss": 0.7826,
+      "step": 10953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6066436152550185e-06,
+      "loss": 0.7844,
+      "step": 10954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6058344530868313e-06,
+      "loss": 0.8117,
+      "step": 10955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6050254557490746e-06,
+      "loss": 0.9398,
+      "step": 10956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6042166232810346e-06,
+      "loss": 0.5928,
+      "step": 10957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6034079557219907e-06,
+      "loss": 1.2403,
+      "step": 10958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6025994531112143e-06,
+      "loss": 0.698,
+      "step": 10959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6017911154879684e-06,
+      "loss": 0.6658,
+      "step": 10960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6009829428915085e-06,
+      "loss": 0.7902,
+      "step": 10961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.6001749353610819e-06,
+      "loss": 0.7036,
+      "step": 10962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.599367092935929e-06,
+      "loss": 0.6798,
+      "step": 10963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5985594156552764e-06,
+      "loss": 0.6837,
+      "step": 10964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5977519035583517e-06,
+      "loss": 0.7599,
+      "step": 10965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.59694455668437e-06,
+      "loss": 0.5333,
+      "step": 10966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.596137375072534e-06,
+      "loss": 0.6774,
+      "step": 10967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5953303587620472e-06,
+      "loss": 0.7892,
+      "step": 10968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5945235077921011e-06,
+      "loss": 0.8433,
+      "step": 10969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5937168222018735e-06,
+      "loss": 0.9012,
+      "step": 10970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5929103020305441e-06,
+      "loss": 0.7912,
+      "step": 10971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5921039473172801e-06,
+      "loss": 0.7521,
+      "step": 10972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.591297758101235e-06,
+      "loss": 0.5678,
+      "step": 10973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.590491734421567e-06,
+      "loss": 1.065,
+      "step": 10974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.589685876317414e-06,
+      "loss": 0.8178,
+      "step": 10975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5888801838279115e-06,
+      "loss": 0.8451,
+      "step": 10976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5880746569921867e-06,
+      "loss": 0.6713,
+      "step": 10977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.587269295849359e-06,
+      "loss": 0.6797,
+      "step": 10978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5864641004385384e-06,
+      "loss": 0.7937,
+      "step": 10979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5856590707988273e-06,
+      "loss": 0.6971,
+      "step": 10980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5848542069693201e-06,
+      "loss": 0.77,
+      "step": 10981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5840495089891045e-06,
+      "loss": 0.9709,
+      "step": 10982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5832449768972574e-06,
+      "loss": 0.8275,
+      "step": 10983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5824406107328506e-06,
+      "loss": 0.6223,
+      "step": 10984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5816364105349451e-06,
+      "loss": 0.7249,
+      "step": 10985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5808323763425965e-06,
+      "loss": 0.9572,
+      "step": 10986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5800285081948497e-06,
+      "loss": 0.6935,
+      "step": 10987
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5792248061307442e-06,
+      "loss": 0.6028,
+      "step": 10988
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5784212701893088e-06,
+      "loss": 0.769,
+      "step": 10989
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.577617900409566e-06,
+      "loss": 0.7651,
+      "step": 10990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.576814696830532e-06,
+      "loss": 0.6682,
+      "step": 10991
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5760116594912072e-06,
+      "loss": 0.4,
+      "step": 10992
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.575208788430594e-06,
+      "loss": 0.4978,
+      "step": 10993
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5744060836876828e-06,
+      "loss": 0.7808,
+      "step": 10994
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5736035453014515e-06,
+      "loss": 0.4593,
+      "step": 10995
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.572801173310874e-06,
+      "loss": 0.8854,
+      "step": 10996
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5719989677549197e-06,
+      "loss": 0.8086,
+      "step": 10997
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5711969286725426e-06,
+      "loss": 0.7645,
+      "step": 10998
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5703950561026914e-06,
+      "loss": 0.7879,
+      "step": 10999
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.569593350084312e-06,
+      "loss": 0.6857,
+      "step": 11000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5687918106563326e-06,
+      "loss": 1.0667,
+      "step": 11001
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5679904378576804e-06,
+      "loss": 0.6842,
+      "step": 11002
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5671892317272725e-06,
+      "loss": 0.7775,
+      "step": 11003
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.566388192304017e-06,
+      "loss": 0.8355,
+      "step": 11004
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5655873196268145e-06,
+      "loss": 0.8936,
+      "step": 11005
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5647866137345591e-06,
+      "loss": 0.5793,
+      "step": 11006
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5639860746661339e-06,
+      "loss": 0.6622,
+      "step": 11007
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5631857024604157e-06,
+      "loss": 0.7246,
+      "step": 11008
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5623854971562734e-06,
+      "loss": 0.7911,
+      "step": 11009
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5615854587925672e-06,
+      "loss": 0.4106,
+      "step": 11010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.560785587408149e-06,
+      "loss": 0.6004,
+      "step": 11011
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.559985883041863e-06,
+      "loss": 0.6116,
+      "step": 11012
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5591863457325452e-06,
+      "loss": 0.8774,
+      "step": 11013
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5583869755190246e-06,
+      "loss": 0.5503,
+      "step": 11014
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5575877724401162e-06,
+      "loss": 0.5904,
+      "step": 11015
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.556788736534637e-06,
+      "loss": 0.913,
+      "step": 11016
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5559898678413898e-06,
+      "loss": 0.6344,
+      "step": 11017
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5551911663991654e-06,
+      "loss": 1.0095,
+      "step": 11018
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5543926322467562e-06,
+      "loss": 0.8335,
+      "step": 11019
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5535942654229402e-06,
+      "loss": 0.8881,
+      "step": 11020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5527960659664853e-06,
+      "loss": 0.7312,
+      "step": 11021
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5519980339161594e-06,
+      "loss": 0.6955,
+      "step": 11022
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5512001693107126e-06,
+      "loss": 0.9165,
+      "step": 11023
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5504024721888922e-06,
+      "loss": 0.718,
+      "step": 11024
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.549604942589441e-06,
+      "loss": 0.731,
+      "step": 11025
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5488075805510844e-06,
+      "loss": 0.7839,
+      "step": 11026
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5480103861125444e-06,
+      "loss": 0.708,
+      "step": 11027
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.547213359312541e-06,
+      "loss": 0.7526,
+      "step": 11028
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.546416500189774e-06,
+      "loss": 0.7256,
+      "step": 11029
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5456198087829432e-06,
+      "loss": 0.8977,
+      "step": 11030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5448232851307388e-06,
+      "loss": 0.7464,
+      "step": 11031
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5440269292718414e-06,
+      "loss": 0.8293,
+      "step": 11032
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5432307412449244e-06,
+      "loss": 0.6105,
+      "step": 11033
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5424347210886538e-06,
+      "loss": 0.8821,
+      "step": 11034
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5416388688416856e-06,
+      "loss": 0.8199,
+      "step": 11035
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5408431845426691e-06,
+      "loss": 0.699,
+      "step": 11036
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5400476682302456e-06,
+      "loss": 0.7338,
+      "step": 11037
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.539252319943047e-06,
+      "loss": 0.8092,
+      "step": 11038
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5384571397196996e-06,
+      "loss": 0.7404,
+      "step": 11039
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.537662127598814e-06,
+      "loss": 0.7212,
+      "step": 11040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5368672836190045e-06,
+      "loss": 0.5892,
+      "step": 11041
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5360726078188698e-06,
+      "loss": 0.6716,
+      "step": 11042
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5352781002369976e-06,
+      "loss": 0.6815,
+      "step": 11043
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.534483760911976e-06,
+      "loss": 0.204,
+      "step": 11044
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5336895898823801e-06,
+      "loss": 0.7635,
+      "step": 11045
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5328955871867728e-06,
+      "loss": 0.6252,
+      "step": 11046
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5321017528637177e-06,
+      "loss": 0.5319,
+      "step": 11047
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5313080869517656e-06,
+      "loss": 0.3704,
+      "step": 11048
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5305145894894547e-06,
+      "loss": 0.5976,
+      "step": 11049
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5297212605153255e-06,
+      "loss": 0.738,
+      "step": 11050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5289281000678996e-06,
+      "loss": 0.7022,
+      "step": 11051
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5281351081856977e-06,
+      "loss": 0.632,
+      "step": 11052
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5273422849072284e-06,
+      "loss": 0.8131,
+      "step": 11053
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5265496302709942e-06,
+      "loss": 0.5014,
+      "step": 11054
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5257571443154885e-06,
+      "loss": 0.9929,
+      "step": 11055
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5249648270791967e-06,
+      "loss": 0.616,
+      "step": 11056
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.524172678600596e-06,
+      "loss": 0.6556,
+      "step": 11057
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5233806989181555e-06,
+      "loss": 0.7626,
+      "step": 11058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.522588888070336e-06,
+      "loss": 0.8385,
+      "step": 11059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.52179724609559e-06,
+      "loss": 0.7197,
+      "step": 11060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.521005773032362e-06,
+      "loss": 0.5607,
+      "step": 11061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5202144689190884e-06,
+      "loss": 0.7993,
+      "step": 11062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5194233337941966e-06,
+      "loss": 0.5183,
+      "step": 11063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5186323676961069e-06,
+      "loss": 0.7355,
+      "step": 11064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.517841570663231e-06,
+      "loss": 0.6831,
+      "step": 11065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5170509427339718e-06,
+      "loss": 0.9601,
+      "step": 11066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5162604839467265e-06,
+      "loss": 0.7706,
+      "step": 11067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5154701943398769e-06,
+      "loss": 0.7082,
+      "step": 11068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5146800739518064e-06,
+      "loss": 0.629,
+      "step": 11069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5138901228208863e-06,
+      "loss": 0.6629,
+      "step": 11070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.513100340985475e-06,
+      "loss": 0.6435,
+      "step": 11071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5123107284839267e-06,
+      "loss": 0.7065,
+      "step": 11072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5115212853545924e-06,
+      "loss": 0.6864,
+      "step": 11073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5107320116358043e-06,
+      "loss": 0.6867,
+      "step": 11074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5099429073658922e-06,
+      "loss": 0.9405,
+      "step": 11075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5091539725831822e-06,
+      "loss": 0.8597,
+      "step": 11076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5083652073259818e-06,
+      "loss": 0.9383,
+      "step": 11077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5075766116325984e-06,
+      "loss": 0.5804,
+      "step": 11078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5067881855413275e-06,
+      "loss": 0.5976,
+      "step": 11079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5059999290904575e-06,
+      "loss": 0.6235,
+      "step": 11080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5052118423182688e-06,
+      "loss": 0.8147,
+      "step": 11081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5044239252630332e-06,
+      "loss": 0.3971,
+      "step": 11082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5036361779630137e-06,
+      "loss": 0.9394,
+      "step": 11083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028486004564658e-06,
+      "loss": 0.7171,
+      "step": 11084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5020611927816365e-06,
+      "loss": 0.6423,
+      "step": 11085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012739549767646e-06,
+      "loss": 0.5554,
+      "step": 11086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5004868870800821e-06,
+      "loss": 0.5581,
+      "step": 11087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499699989129807e-06,
+      "loss": 0.7198,
+      "step": 11088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4989132611641576e-06,
+      "loss": 0.4507,
+      "step": 11089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4981267032213398e-06,
+      "loss": 0.6903,
+      "step": 11090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4973403153395461e-06,
+      "loss": 0.7735,
+      "step": 11091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.496554097556971e-06,
+      "loss": 0.6312,
+      "step": 11092
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4957680499117956e-06,
+      "loss": 0.7302,
+      "step": 11093
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.494982172442187e-06,
+      "loss": 0.6919,
+      "step": 11094
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4941964651863155e-06,
+      "loss": 0.7147,
+      "step": 11095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4934109281823361e-06,
+      "loss": 0.919,
+      "step": 11096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926255614683931e-06,
+      "loss": 0.9101,
+      "step": 11097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491840365082632e-06,
+      "loss": 0.8128,
+      "step": 11098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4910553390631794e-06,
+      "loss": 0.5298,
+      "step": 11099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4902704834481585e-06,
+      "loss": 0.7809,
+      "step": 11100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4894857982756894e-06,
+      "loss": 0.5504,
+      "step": 11101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4887012835838727e-06,
+      "loss": 0.5929,
+      "step": 11102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.487916939410808e-06,
+      "loss": 0.6427,
+      "step": 11103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4871327657945888e-06,
+      "loss": 0.8571,
+      "step": 11104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4863487627732936e-06,
+      "loss": 0.519,
+      "step": 11105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4855649303849962e-06,
+      "loss": 0.8232,
+      "step": 11106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4847812686677616e-06,
+      "loss": 0.5951,
+      "step": 11107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4839977776596476e-06,
+      "loss": 0.8469,
+      "step": 11108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4832144573987023e-06,
+      "loss": 0.7188,
+      "step": 11109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4824313079229662e-06,
+      "loss": 0.8207,
+      "step": 11110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4816483292704708e-06,
+      "loss": 0.5785,
+      "step": 11111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4808655214792405e-06,
+      "loss": 0.4674,
+      "step": 11112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.48008288458729e-06,
+      "loss": 0.8714,
+      "step": 11113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.479300418632627e-06,
+      "loss": 0.6153,
+      "step": 11114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4785181236532514e-06,
+      "loss": 0.9595,
+      "step": 11115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4777359996871492e-06,
+      "loss": 0.7234,
+      "step": 11116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.476954046772307e-06,
+      "loss": 0.6346,
+      "step": 11117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4761722649466992e-06,
+      "loss": 0.9098,
+      "step": 11118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4753906542482866e-06,
+      "loss": 0.7429,
+      "step": 11119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4746092147150303e-06,
+      "loss": 0.7899,
+      "step": 11120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4738279463848803e-06,
+      "loss": 0.6825,
+      "step": 11121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4730468492957722e-06,
+      "loss": 0.689,
+      "step": 11122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.472265923485643e-06,
+      "loss": 0.7771,
+      "step": 11123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4714851689924164e-06,
+      "loss": 1.0576,
+      "step": 11124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4707045858540053e-06,
+      "loss": 0.7301,
+      "step": 11125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4699241741083187e-06,
+      "loss": 0.7433,
+      "step": 11126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.469143933793255e-06,
+      "loss": 0.6862,
+      "step": 11127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4683638649467058e-06,
+      "loss": 0.8094,
+      "step": 11128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4675839676065534e-06,
+      "loss": 0.5989,
+      "step": 11129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.466804241810671e-06,
+      "loss": 0.6931,
+      "step": 11130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4660246875969248e-06,
+      "loss": 0.8955,
+      "step": 11131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4652453050031722e-06,
+      "loss": 0.5834,
+      "step": 11132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4644660940672628e-06,
+      "loss": 0.6962,
+      "step": 11133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4636870548270365e-06,
+      "loss": 0.6617,
+      "step": 11134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.462908187320326e-06,
+      "loss": 0.5261,
+      "step": 11135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4621294915849553e-06,
+      "loss": 1.0745,
+      "step": 11136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4613509676587407e-06,
+      "loss": 0.8767,
+      "step": 11137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.460572615579488e-06,
+      "loss": 0.8157,
+      "step": 11138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.459794435384998e-06,
+      "loss": 0.7448,
+      "step": 11139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4590164271130602e-06,
+      "loss": 0.8295,
+      "step": 11140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4582385908014574e-06,
+      "loss": 0.5489,
+      "step": 11141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4574609264879636e-06,
+      "loss": 0.9063,
+      "step": 11142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4566834342103458e-06,
+      "loss": 0.8325,
+      "step": 11143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4559061140063579e-06,
+      "loss": 0.5236,
+      "step": 11144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4551289659137497e-06,
+      "loss": 0.7106,
+      "step": 11145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4543519899702652e-06,
+      "loss": 0.9283,
+      "step": 11146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4535751862136326e-06,
+      "loss": 0.5437,
+      "step": 11147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.452798554681576e-06,
+      "loss": 0.7025,
+      "step": 11148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4520220954118147e-06,
+      "loss": 0.532,
+      "step": 11149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4512458084420516e-06,
+      "loss": 0.744,
+      "step": 11150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.450469693809986e-06,
+      "loss": 0.6005,
+      "step": 11151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4496937515533115e-06,
+      "loss": 0.5469,
+      "step": 11152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.448917981709707e-06,
+      "loss": 0.7782,
+      "step": 11153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4481423843168468e-06,
+      "loss": 0.7078,
+      "step": 11154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4473669594123958e-06,
+      "loss": 0.6897,
+      "step": 11155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.446591707034012e-06,
+      "loss": 1.1128,
+      "step": 11156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4458166272193424e-06,
+      "loss": 0.8357,
+      "step": 11157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4450417200060284e-06,
+      "loss": 0.7034,
+      "step": 11158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.444266985431701e-06,
+      "loss": 0.7386,
+      "step": 11159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4434924235339841e-06,
+      "loss": 0.7234,
+      "step": 11160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.442718034350492e-06,
+      "loss": 0.6283,
+      "step": 11161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4419438179188316e-06,
+      "loss": 0.8513,
+      "step": 11162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.441169774276603e-06,
+      "loss": 0.8284,
+      "step": 11163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4403959034613901e-06,
+      "loss": 0.6319,
+      "step": 11164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4396222055107805e-06,
+      "loss": 0.6774,
+      "step": 11165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4388486804623464e-06,
+      "loss": 0.4991,
+      "step": 11166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4380753283536474e-06,
+      "loss": 0.9246,
+      "step": 11167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.437302149222245e-06,
+      "loss": 0.6956,
+      "step": 11168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4365291431056871e-06,
+      "loss": 0.5384,
+      "step": 11169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4357563100415078e-06,
+      "loss": 0.697,
+      "step": 11170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4349836500672443e-06,
+      "loss": 0.5628,
+      "step": 11171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4342111632204152e-06,
+      "loss": 0.7136,
+      "step": 11172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4334388495385343e-06,
+      "loss": 0.7698,
+      "step": 11173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4326667090591122e-06,
+      "loss": 0.8933,
+      "step": 11174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.431894741819641e-06,
+      "loss": 0.6916,
+      "step": 11175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4311229478576106e-06,
+      "loss": 0.5836,
+      "step": 11176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4303513272105057e-06,
+      "loss": 0.529,
+      "step": 11177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4295798799157934e-06,
+      "loss": 0.4739,
+      "step": 11178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.428808606010938e-06,
+      "loss": 0.9263,
+      "step": 11179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.428037505533399e-06,
+      "loss": 0.8231,
+      "step": 11180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.427266578520619e-06,
+      "loss": 0.6977,
+      "step": 11181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4264958250100376e-06,
+      "loss": 0.6746,
+      "step": 11182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4257252450390846e-06,
+      "loss": 0.7749,
+      "step": 11183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4249548386451829e-06,
+      "loss": 0.8673,
+      "step": 11184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4241846058657437e-06,
+      "loss": 0.6096,
+      "step": 11185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4234145467381732e-06,
+      "loss": 0.5662,
+      "step": 11186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4226446612998673e-06,
+      "loss": 0.7705,
+      "step": 11187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4218749495882133e-06,
+      "loss": 0.8194,
+      "step": 11188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4211054116405914e-06,
+      "loss": 1.1967,
+      "step": 11189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4203360474943722e-06,
+      "loss": 0.6711,
+      "step": 11190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4195668571869197e-06,
+      "loss": 0.6493,
+      "step": 11191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4187978407555835e-06,
+      "loss": 0.663,
+      "step": 11192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4180289982377138e-06,
+      "loss": 0.7539,
+      "step": 11193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4172603296706478e-06,
+      "loss": 0.7421,
+      "step": 11194
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4164918350917096e-06,
+      "loss": 0.8282,
+      "step": 11195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4157235145382242e-06,
+      "loss": 0.5247,
+      "step": 11196
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4149553680475036e-06,
+      "loss": 0.7317,
+      "step": 11197
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.414187395656848e-06,
+      "loss": 0.5966,
+      "step": 11198
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4134195974035525e-06,
+      "loss": 0.7071,
+      "step": 11199
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4126519733249079e-06,
+      "loss": 0.9807,
+      "step": 11200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4118845234581873e-06,
+      "loss": 0.566,
+      "step": 11201
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4111172478406625e-06,
+      "loss": 0.4653,
+      "step": 11202
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4103501465095948e-06,
+      "loss": 0.5076,
+      "step": 11203
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4095832195022357e-06,
+      "loss": 0.5476,
+      "step": 11204
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4088164668558302e-06,
+      "loss": 0.6271,
+      "step": 11205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4080498886076145e-06,
+      "loss": 0.4033,
+      "step": 11206
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4072834847948147e-06,
+      "loss": 0.5988,
+      "step": 11207
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4065172554546502e-06,
+      "loss": 0.5159,
+      "step": 11208
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4057512006243312e-06,
+      "loss": 0.4928,
+      "step": 11209
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4049853203410595e-06,
+      "loss": 0.6118,
+      "step": 11210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4042196146420284e-06,
+      "loss": 0.6726,
+      "step": 11211
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4034540835644227e-06,
+      "loss": 0.6307,
+      "step": 11212
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4026887271454193e-06,
+      "loss": 0.7062,
+      "step": 11213
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4019235454221858e-06,
+      "loss": 0.8481,
+      "step": 11214
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.401158538431881e-06,
+      "loss": 0.7359,
+      "step": 11215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4003937062116568e-06,
+      "loss": 0.4576,
+      "step": 11216
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3996290487986568e-06,
+      "loss": 0.758,
+      "step": 11217
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.39886456623001e-06,
+      "loss": 0.7442,
+      "step": 11218
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.398100258542849e-06,
+      "loss": 0.699,
+      "step": 11219
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3973361257742845e-06,
+      "loss": 0.7397,
+      "step": 11220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.396572167961427e-06,
+      "loss": 0.6006,
+      "step": 11221
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3958083851413794e-06,
+      "loss": 0.5376,
+      "step": 11222
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.39504477735123e-06,
+      "loss": 0.5676,
+      "step": 11223
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3942813446280601e-06,
+      "loss": 0.7298,
+      "step": 11224
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3935180870089503e-06,
+      "loss": 0.5982,
+      "step": 11225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3927550045309612e-06,
+      "loss": 0.7598,
+      "step": 11226
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3919920972311506e-06,
+      "loss": 0.6019,
+      "step": 11227
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3912293651465725e-06,
+      "loss": 0.9011,
+      "step": 11228
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3904668083142615e-06,
+      "loss": 0.8128,
+      "step": 11229
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3897044267712523e-06,
+      "loss": 0.5599,
+      "step": 11230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3889422205545677e-06,
+      "loss": 0.5853,
+      "step": 11231
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3881801897012225e-06,
+      "loss": 0.9447,
+      "step": 11232
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.387418334248224e-06,
+      "loss": 0.7936,
+      "step": 11233
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3866566542325689e-06,
+      "loss": 0.7241,
+      "step": 11234
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.385895149691247e-06,
+      "loss": 0.7701,
+      "step": 11235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3851338206612402e-06,
+      "loss": 0.9346,
+      "step": 11236
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3843726671795194e-06,
+      "loss": 0.4488,
+      "step": 11237
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3836116892830487e-06,
+      "loss": 0.7994,
+      "step": 11238
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3828508870087854e-06,
+      "loss": 0.6707,
+      "step": 11239
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3820902603936714e-06,
+      "loss": 0.8877,
+      "step": 11240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3813298094746491e-06,
+      "loss": 0.8685,
+      "step": 11241
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.380569534288649e-06,
+      "loss": 0.5639,
+      "step": 11242
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.3798094348725871e-06,
+      "loss": 0.9709,
+      "step": 11243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3790495112633806e-06,
+      "loss": 0.691,
+      "step": 11244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3782897634979336e-06,
+      "loss": 0.5617,
+      "step": 11245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3775301916131374e-06,
+      "loss": 0.9984,
+      "step": 11246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3767707956458848e-06,
+      "loss": 0.6436,
+      "step": 11247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.37601157563305e-06,
+      "loss": 0.984,
+      "step": 11248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3752525316115024e-06,
+      "loss": 0.688,
+      "step": 11249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3744936636181078e-06,
+      "loss": 0.8805,
+      "step": 11250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3737349716897148e-06,
+      "loss": 0.9739,
+      "step": 11251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3729764558631681e-06,
+      "loss": 0.7849,
+      "step": 11252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3722181161753073e-06,
+      "loss": 0.5354,
+      "step": 11253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3714599526629552e-06,
+      "loss": 0.6604,
+      "step": 11254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3707019653629321e-06,
+      "loss": 0.6487,
+      "step": 11255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3699441543120479e-06,
+      "loss": 0.5352,
+      "step": 11256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3691865195471037e-06,
+      "loss": 0.6549,
+      "step": 11257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.368429061104893e-06,
+      "loss": 0.6847,
+      "step": 11258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3676717790222e-06,
+      "loss": 0.64,
+      "step": 11259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3669146733358002e-06,
+      "loss": 0.5915,
+      "step": 11260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3661577440824614e-06,
+      "loss": 0.7142,
+      "step": 11261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3654009912989418e-06,
+      "loss": 0.8097,
+      "step": 11262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3646444150219912e-06,
+      "loss": 0.8803,
+      "step": 11263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.363888015288352e-06,
+      "loss": 0.5395,
+      "step": 11264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3631317921347564e-06,
+      "loss": 0.8394,
+      "step": 11265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.362375745597929e-06,
+      "loss": 0.5518,
+      "step": 11266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3616198757145877e-06,
+      "loss": 0.8703,
+      "step": 11267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3608641825214342e-06,
+      "loss": 0.8456,
+      "step": 11268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3601086660551721e-06,
+      "loss": 0.6133,
+      "step": 11269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3593533263524917e-06,
+      "loss": 0.8136,
+      "step": 11270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3585981634500688e-06,
+      "loss": 0.7012,
+      "step": 11271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3578431773845824e-06,
+      "loss": 0.4878,
+      "step": 11272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.357088368192696e-06,
+      "loss": 0.6591,
+      "step": 11273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3563337359110624e-06,
+      "loss": 0.8267,
+      "step": 11274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3555792805763284e-06,
+      "loss": 0.5984,
+      "step": 11275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3548250022251375e-06,
+      "loss": 0.7432,
+      "step": 11276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.354070900894115e-06,
+      "loss": 0.5795,
+      "step": 11277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3533169766198835e-06,
+      "loss": 0.6678,
+      "step": 11278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3525632294390562e-06,
+      "loss": 0.7476,
+      "step": 11279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3518096593882362e-06,
+      "loss": 0.8479,
+      "step": 11280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.35105626650402e-06,
+      "loss": 0.5369,
+      "step": 11281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3503030508229946e-06,
+      "loss": 0.7482,
+      "step": 11282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.349550012381738e-06,
+      "loss": 1.0081,
+      "step": 11283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3487971512168197e-06,
+      "loss": 0.7025,
+      "step": 11284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3480444673648007e-06,
+      "loss": 0.6233,
+      "step": 11285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3472919608622336e-06,
+      "loss": 0.617,
+      "step": 11286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.346539631745663e-06,
+      "loss": 0.557,
+      "step": 11287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3457874800516235e-06,
+      "loss": 0.7466,
+      "step": 11288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.345035505816642e-06,
+      "loss": 0.7399,
+      "step": 11289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.344283709077236e-06,
+      "loss": 0.4077,
+      "step": 11290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3435320898699155e-06,
+      "loss": 0.7725,
+      "step": 11291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.342780648231181e-06,
+      "loss": 0.9675,
+      "step": 11292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3420293841975263e-06,
+      "loss": 0.7398,
+      "step": 11293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.34127829780543e-06,
+      "loss": 0.8846,
+      "step": 11294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.340527389091374e-06,
+      "loss": 0.7198,
+      "step": 11295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3397766580918193e-06,
+      "loss": 0.9546,
+      "step": 11296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.339026104843224e-06,
+      "loss": 0.6909,
+      "step": 11297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3382757293820408e-06,
+      "loss": 0.8517,
+      "step": 11298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3375255317447066e-06,
+      "loss": 0.6601,
+      "step": 11299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3367755119676524e-06,
+      "loss": 0.9159,
+      "step": 11300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3360256700873064e-06,
+      "loss": 1.0217,
+      "step": 11301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3352760061400782e-06,
+      "loss": 0.8304,
+      "step": 11302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3345265201623736e-06,
+      "loss": 0.4783,
+      "step": 11303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3337772121905945e-06,
+      "loss": 0.4776,
+      "step": 11304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3330280822611246e-06,
+      "loss": 0.7755,
+      "step": 11305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.332279130410346e-06,
+      "loss": 0.6854,
+      "step": 11306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.331530356674629e-06,
+      "loss": 0.7288,
+      "step": 11307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3307817610903374e-06,
+      "loss": 0.8765,
+      "step": 11308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.330033343693824e-06,
+      "loss": 0.7854,
+      "step": 11309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3292851045214344e-06,
+      "loss": 0.5841,
+      "step": 11310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3285370436095064e-06,
+      "loss": 0.7717,
+      "step": 11311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3277891609943662e-06,
+      "loss": 0.6408,
+      "step": 11312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3270414567123342e-06,
+      "loss": 0.6127,
+      "step": 11313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3262939307997201e-06,
+      "loss": 0.9641,
+      "step": 11314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3255465832928288e-06,
+      "loss": 0.6929,
+      "step": 11315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3247994142279486e-06,
+      "loss": 0.5591,
+      "step": 11316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3240524236413681e-06,
+      "loss": 0.7953,
+      "step": 11317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3233056115693637e-06,
+      "loss": 0.6339,
+      "step": 11318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.322558978048199e-06,
+      "loss": 0.8606,
+      "step": 11319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3218125231141354e-06,
+      "loss": 0.6725,
+      "step": 11320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3210662468034246e-06,
+      "loss": 0.4701,
+      "step": 11321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3203201491523027e-06,
+      "loss": 0.7484,
+      "step": 11322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3195742301970082e-06,
+      "loss": 0.8341,
+      "step": 11323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3188284899737612e-06,
+      "loss": 0.6373,
+      "step": 11324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3180829285187764e-06,
+      "loss": 0.8644,
+      "step": 11325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3173375458682647e-06,
+      "loss": 0.6479,
+      "step": 11326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3165923420584203e-06,
+      "loss": 0.7747,
+      "step": 11327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3158473171254333e-06,
+      "loss": 0.7075,
+      "step": 11328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3151024711054845e-06,
+      "loss": 0.7798,
+      "step": 11329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.314357804034746e-06,
+      "loss": 0.7266,
+      "step": 11330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3136133159493803e-06,
+      "loss": 0.5833,
+      "step": 11331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3128690068855426e-06,
+      "loss": 0.8358,
+      "step": 11332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.312124876879378e-06,
+      "loss": 0.8505,
+      "step": 11333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3113809259670247e-06,
+      "loss": 0.8805,
+      "step": 11334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3106371541846097e-06,
+      "loss": 0.5553,
+      "step": 11335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3098935615682534e-06,
+      "loss": 0.9105,
+      "step": 11336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3091501481540676e-06,
+      "loss": 0.7195,
+      "step": 11337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.308406913978153e-06,
+      "loss": 0.424,
+      "step": 11338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3076638590766038e-06,
+      "loss": 0.9009,
+      "step": 11339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.306920983485505e-06,
+      "loss": 0.8199,
+      "step": 11340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3061782872409334e-06,
+      "loss": 0.5439,
+      "step": 11341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3054357703789555e-06,
+      "loss": 0.5237,
+      "step": 11342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3046934329356314e-06,
+      "loss": 0.864,
+      "step": 11343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3039512749470078e-06,
+      "loss": 0.7078,
+      "step": 11344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.30320929644913e-06,
+      "loss": 0.5688,
+      "step": 11345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.30246749747803e-06,
+      "loss": 0.6308,
+      "step": 11346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3017258780697295e-06,
+      "loss": 0.6719,
+      "step": 11347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3009844382602426e-06,
+      "loss": 0.7251,
+      "step": 11348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.3002431780855817e-06,
+      "loss": 0.684,
+      "step": 11349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2995020975817392e-06,
+      "loss": 0.7656,
+      "step": 11350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.298761196784704e-06,
+      "loss": 0.8587,
+      "step": 11351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2980204757304615e-06,
+      "loss": 0.3852,
+      "step": 11352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.297279934454978e-06,
+      "loss": 0.6955,
+      "step": 11353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2965395729942182e-06,
+      "loss": 0.6059,
+      "step": 11354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.295799391384136e-06,
+      "loss": 0.5506,
+      "step": 11355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2950593896606767e-06,
+      "loss": 0.6727,
+      "step": 11356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2943195678597776e-06,
+      "loss": 0.8011,
+      "step": 11357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2935799260173655e-06,
+      "loss": 0.8829,
+      "step": 11358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2928404641693598e-06,
+      "loss": 0.6384,
+      "step": 11359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2921011823516715e-06,
+      "loss": 0.856,
+      "step": 11360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2913620806002026e-06,
+      "loss": 0.5869,
+      "step": 11361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2906231589508444e-06,
+      "loss": 0.6505,
+      "step": 11362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2898844174394827e-06,
+      "loss": 0.517,
+      "step": 11363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2891458561019914e-06,
+      "loss": 0.9566,
+      "step": 11364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.288407474974238e-06,
+      "loss": 0.83,
+      "step": 11365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2876692740920825e-06,
+      "loss": 0.5795,
+      "step": 11366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2869312534913686e-06,
+      "loss": 0.6068,
+      "step": 11367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2861934132079413e-06,
+      "loss": 0.6398,
+      "step": 11368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2854557532776323e-06,
+      "loss": 0.8054,
+      "step": 11369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2847182737362606e-06,
+      "loss": 0.8309,
+      "step": 11370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2839809746196458e-06,
+      "loss": 0.7255,
+      "step": 11371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2832438559635885e-06,
+      "loss": 0.5509,
+      "step": 11372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2825069178038857e-06,
+      "loss": 0.7752,
+      "step": 11373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.28177016017633e-06,
+      "loss": 0.6509,
+      "step": 11374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2810335831166954e-06,
+      "loss": 0.683,
+      "step": 11375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2802971866607522e-06,
+      "loss": 0.8449,
+      "step": 11376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2795609708442668e-06,
+      "loss": 0.8253,
+      "step": 11377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2788249357029874e-06,
+      "loss": 0.7725,
+      "step": 11378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2780890812726576e-06,
+      "loss": 0.6922,
+      "step": 11379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2773534075890181e-06,
+      "loss": 0.7312,
+      "step": 11380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2766179146877894e-06,
+      "loss": 0.6285,
+      "step": 11381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2758826026046917e-06,
+      "loss": 0.5889,
+      "step": 11382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.275147471375433e-06,
+      "loss": 0.8111,
+      "step": 11383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2744125210357145e-06,
+      "loss": 0.6072,
+      "step": 11384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2736777516212267e-06,
+      "loss": 0.7664,
+      "step": 11385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2729431631676525e-06,
+      "loss": 0.6631,
+      "step": 11386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2722087557106655e-06,
+      "loss": 0.4295,
+      "step": 11387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2714745292859304e-06,
+      "loss": 0.6347,
+      "step": 11388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2707404839291032e-06,
+      "loss": 0.5365,
+      "step": 11389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.270006619675832e-06,
+      "loss": 0.7359,
+      "step": 11390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2692729365617568e-06,
+      "loss": 0.6121,
+      "step": 11391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2685394346225016e-06,
+      "loss": 1.0149,
+      "step": 11392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2678061138936937e-06,
+      "loss": 0.7095,
+      "step": 11393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2670729744109432e-06,
+      "loss": 0.8018,
+      "step": 11394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.266340016209851e-06,
+      "loss": 0.8237,
+      "step": 11395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2656072393260148e-06,
+      "loss": 0.7963,
+      "step": 11396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.264874643795021e-06,
+      "loss": 0.5935,
+      "step": 11397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.264142229652442e-06,
+      "loss": 0.9014,
+      "step": 11398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2634099969338514e-06,
+      "loss": 0.9415,
+      "step": 11399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2626779456748046e-06,
+      "loss": 0.6327,
+      "step": 11400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2619460759108521e-06,
+      "loss": 0.5229,
+      "step": 11401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2612143876775396e-06,
+      "loss": 0.8186,
+      "step": 11402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260482881010396e-06,
+      "loss": 0.6688,
+      "step": 11403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2597515559449463e-06,
+      "loss": 0.6925,
+      "step": 11404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2590204125167067e-06,
+      "loss": 0.6996,
+      "step": 11405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.258289450761182e-06,
+      "loss": 0.6435,
+      "step": 11406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2575586707138716e-06,
+      "loss": 0.8408,
+      "step": 11407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2568280724102633e-06,
+      "loss": 0.8509,
+      "step": 11408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2560976558858372e-06,
+      "loss": 0.562,
+      "step": 11409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2553674211760647e-06,
+      "loss": 0.8737,
+      "step": 11410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2546373683164081e-06,
+      "loss": 0.6748,
+      "step": 11411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2539074973423205e-06,
+      "loss": 0.8954,
+      "step": 11412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.253177808289246e-06,
+      "loss": 0.7759,
+      "step": 11413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.252448301192622e-06,
+      "loss": 0.6265,
+      "step": 11414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.251718976087874e-06,
+      "loss": 0.9852,
+      "step": 11415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2509898330104202e-06,
+      "loss": 0.6277,
+      "step": 11416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.250260871995671e-06,
+      "loss": 0.7767,
+      "step": 11417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2495320930790262e-06,
+      "loss": 0.5863,
+      "step": 11418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.248803496295879e-06,
+      "loss": 0.8874,
+      "step": 11419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2480750816816067e-06,
+      "loss": 0.5764,
+      "step": 11420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2473468492715896e-06,
+      "loss": 0.6696,
+      "step": 11421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2466187991011908e-06,
+      "loss": 0.5805,
+      "step": 11422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2458909312057649e-06,
+      "loss": 0.5346,
+      "step": 11423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2451632456206587e-06,
+      "loss": 0.5398,
+      "step": 11424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2444357423812148e-06,
+      "loss": 0.5988,
+      "step": 11425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.243708421522759e-06,
+      "loss": 0.6734,
+      "step": 11426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2429812830806125e-06,
+      "loss": 0.8426,
+      "step": 11427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2422543270900906e-06,
+      "loss": 0.5691,
+      "step": 11428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2415275535864924e-06,
+      "loss": 0.4531,
+      "step": 11429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2408009626051137e-06,
+      "loss": 0.7688,
+      "step": 11430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2400745541812399e-06,
+      "loss": 0.5619,
+      "step": 11431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2393483283501468e-06,
+      "loss": 0.7733,
+      "step": 11432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.238622285147103e-06,
+      "loss": 0.804,
+      "step": 11433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2378964246073667e-06,
+      "loss": 0.6327,
+      "step": 11434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.237170746766188e-06,
+      "loss": 0.8151,
+      "step": 11435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2364452516588077e-06,
+      "loss": 0.4222,
+      "step": 11436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2357199393204577e-06,
+      "loss": 0.668,
+      "step": 11437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2349948097863618e-06,
+      "loss": 0.8311,
+      "step": 11438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2342698630917337e-06,
+      "loss": 0.8108,
+      "step": 11439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2335450992717802e-06,
+      "loss": 0.9297,
+      "step": 11440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2328205183616964e-06,
+      "loss": 0.6775,
+      "step": 11441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.232096120396672e-06,
+      "loss": 0.642,
+      "step": 11442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2313719054118817e-06,
+      "loss": 0.3683,
+      "step": 11443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2306478734424998e-06,
+      "loss": 0.9022,
+      "step": 11444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2299240245236876e-06,
+      "loss": 0.6823,
+      "step": 11445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2292003586905922e-06,
+      "loss": 0.7617,
+      "step": 11446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2284768759783633e-06,
+      "loss": 0.526,
+      "step": 11447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2277535764221304e-06,
+      "loss": 0.6226,
+      "step": 11448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2270304600570193e-06,
+      "loss": 0.696,
+      "step": 11449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2263075269181517e-06,
+      "loss": 0.8118,
+      "step": 11450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2255847770406297e-06,
+      "loss": 0.6638,
+      "step": 11451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2248622104595531e-06,
+      "loss": 0.74,
+      "step": 11452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2241398272100153e-06,
+      "loss": 0.8984,
+      "step": 11453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2234176273270937e-06,
+      "loss": 0.7097,
+      "step": 11454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.22269561084586e-06,
+      "loss": 0.3962,
+      "step": 11455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2219737778013819e-06,
+      "loss": 0.67,
+      "step": 11456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2212521282287093e-06,
+      "loss": 0.8217,
+      "step": 11457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2205306621628887e-06,
+      "loss": 0.6847,
+      "step": 11458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2198093796389572e-06,
+      "loss": 0.5527,
+      "step": 11459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2190882806919418e-06,
+      "loss": 0.7426,
+      "step": 11460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2183673653568617e-06,
+      "loss": 0.7946,
+      "step": 11461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.217646633668726e-06,
+      "loss": 0.8382,
+      "step": 11462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2169260856625358e-06,
+      "loss": 0.5737,
+      "step": 11463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.216205721373283e-06,
+      "loss": 0.6319,
+      "step": 11464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2154855408359507e-06,
+      "loss": 0.6638,
+      "step": 11465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2147655440855128e-06,
+      "loss": 0.7278,
+      "step": 11466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2140457311569354e-06,
+      "loss": 0.6634,
+      "step": 11467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2133261020851712e-06,
+      "loss": 0.6791,
+      "step": 11468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2126066569051708e-06,
+      "loss": 0.6831,
+      "step": 11469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2118873956518735e-06,
+      "loss": 0.6994,
+      "step": 11470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2111683183602041e-06,
+      "loss": 0.7441,
+      "step": 11471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2104494250650868e-06,
+      "loss": 0.6382,
+      "step": 11472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2097307158014337e-06,
+      "loss": 0.9996,
+      "step": 11473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.209012190604143e-06,
+      "loss": 0.6695,
+      "step": 11474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2082938495081143e-06,
+      "loss": 0.3702,
+      "step": 11475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2075756925482268e-06,
+      "loss": 0.6959,
+      "step": 11476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2068577197593594e-06,
+      "loss": 0.5968,
+      "step": 11477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2061399311763778e-06,
+      "loss": 0.8105,
+      "step": 11478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2054223268341403e-06,
+      "loss": 0.5801,
+      "step": 11479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.204704906767496e-06,
+      "loss": 0.9572,
+      "step": 11480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2039876710112847e-06,
+      "loss": 0.809,
+      "step": 11481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.203270619600337e-06,
+      "loss": 0.6094,
+      "step": 11482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2025537525694752e-06,
+      "loss": 0.8583,
+      "step": 11483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2018370699535126e-06,
+      "loss": 0.6014,
+      "step": 11484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2011205717872538e-06,
+      "loss": 1.0025,
+      "step": 11485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.200404258105493e-06,
+      "loss": 0.7021,
+      "step": 11486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1996881289430173e-06,
+      "loss": 0.8741,
+      "step": 11487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.198972184334603e-06,
+      "loss": 0.9419,
+      "step": 11488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1982564243150197e-06,
+      "loss": 1.0439,
+      "step": 11489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.197540848919026e-06,
+      "loss": 0.6723,
+      "step": 11490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1968254581813726e-06,
+      "loss": 0.3107,
+      "step": 11491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1961102521368007e-06,
+      "loss": 0.7389,
+      "step": 11492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1953952308200428e-06,
+      "loss": 0.8851,
+      "step": 11493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1946803942658224e-06,
+      "loss": 0.8876,
+      "step": 11494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1939657425088558e-06,
+      "loss": 0.6707,
+      "step": 11495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1932512755838448e-06,
+      "loss": 0.6751,
+      "step": 11496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1925369935254872e-06,
+      "loss": 0.6413,
+      "step": 11497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1918228963684735e-06,
+      "loss": 0.78,
+      "step": 11498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1911089841474793e-06,
+      "loss": 0.788,
+      "step": 11499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1903952568971733e-06,
+      "loss": 0.6462,
+      "step": 11500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.189681714652221e-06,
+      "loss": 0.5766,
+      "step": 11501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1889683574472695e-06,
+      "loss": 0.9078,
+      "step": 11502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.188255185316961e-06,
+      "loss": 0.7474,
+      "step": 11503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1875421982959345e-06,
+      "loss": 0.5719,
+      "step": 11504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1868293964188099e-06,
+      "loss": 0.6414,
+      "step": 11505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1861167797202045e-06,
+      "loss": 0.5032,
+      "step": 11506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.185404348234725e-06,
+      "loss": 0.7738,
+      "step": 11507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1846921019969683e-06,
+      "loss": 0.7474,
+      "step": 11508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1839800410415242e-06,
+      "loss": 0.8087,
+      "step": 11509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1832681654029721e-06,
+      "loss": 0.6445,
+      "step": 11510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1825564751158825e-06,
+      "loss": 0.6728,
+      "step": 11511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1818449702148176e-06,
+      "loss": 0.6154,
+      "step": 11512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1811336507343296e-06,
+      "loss": 1.0645,
+      "step": 11513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1804225167089622e-06,
+      "loss": 0.8085,
+      "step": 11514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1797115681732525e-06,
+      "loss": 0.6869,
+      "step": 11515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1790008051617203e-06,
+      "loss": 0.616,
+      "step": 11516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.178290227708888e-06,
+      "loss": 0.5062,
+      "step": 11517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.177579835849263e-06,
+      "loss": 0.7788,
+      "step": 11518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1768696296173387e-06,
+      "loss": 0.83,
+      "step": 11519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1761596090476102e-06,
+      "loss": 0.7882,
+      "step": 11520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1754497741745573e-06,
+      "loss": 0.6967,
+      "step": 11521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1747401250326484e-06,
+      "loss": 0.7467,
+      "step": 11522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1740306616563506e-06,
+      "loss": 0.6368,
+      "step": 11523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1733213840801144e-06,
+      "loss": 0.7861,
+      "step": 11524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1726122923383832e-06,
+      "loss": 0.8454,
+      "step": 11525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.171903386465597e-06,
+      "loss": 0.5913,
+      "step": 11526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1711946664961792e-06,
+      "loss": 0.5076,
+      "step": 11527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.170486132464546e-06,
+      "loss": 0.8551,
+      "step": 11528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1697777844051105e-06,
+      "loss": 0.8208,
+      "step": 11529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.169069622352268e-06,
+      "loss": 0.7041,
+      "step": 11530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1683616463404085e-06,
+      "loss": 0.9021,
+      "step": 11531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1676538564039186e-06,
+      "loss": 0.6945,
+      "step": 11532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1669462525771651e-06,
+      "loss": 0.7184,
+      "step": 11533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1662388348945132e-06,
+      "loss": 0.8332,
+      "step": 11534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1655316033903175e-06,
+      "loss": 0.7263,
+      "step": 11535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.1648245580989226e-06,
+      "loss": 0.8534,
+      "step": 11536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.164117699054665e-06,
+      "loss": 0.7152,
+      "step": 11537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1634110262918718e-06,
+      "loss": 0.5187,
+      "step": 11538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1627045398448605e-06,
+      "loss": 0.5957,
+      "step": 11539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1619982397479412e-06,
+      "loss": 0.8904,
+      "step": 11540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1612921260354132e-06,
+      "loss": 0.7529,
+      "step": 11541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1605861987415673e-06,
+      "loss": 0.7534,
+      "step": 11542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1598804579006872e-06,
+      "loss": 0.7252,
+      "step": 11543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1591749035470407e-06,
+      "loss": 0.9227,
+      "step": 11544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1584695357148968e-06,
+      "loss": 0.839,
+      "step": 11545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1577643544385092e-06,
+      "loss": 0.5869,
+      "step": 11546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1570593597521201e-06,
+      "loss": 0.621,
+      "step": 11547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.15635455168997e-06,
+      "loss": 0.467,
+      "step": 11548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1556499302862862e-06,
+      "loss": 0.7304,
+      "step": 11549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.154945495575283e-06,
+      "loss": 0.7693,
+      "step": 11550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1542412475911758e-06,
+      "loss": 0.9417,
+      "step": 11551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1535371863681604e-06,
+      "loss": 0.911,
+      "step": 11552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.152833311940429e-06,
+      "loss": 0.6078,
+      "step": 11553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1521296243421643e-06,
+      "loss": 0.951,
+      "step": 11554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1514261236075398e-06,
+      "loss": 0.7811,
+      "step": 11555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1507228097707186e-06,
+      "loss": 0.6374,
+      "step": 11556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1500196828658555e-06,
+      "loss": 0.628,
+      "step": 11557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.149316742927098e-06,
+      "loss": 0.7371,
+      "step": 11558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1486139899885805e-06,
+      "loss": 0.7231,
+      "step": 11559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.147911424084433e-06,
+      "loss": 0.8884,
+      "step": 11560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1472090452487728e-06,
+      "loss": 0.623,
+      "step": 11561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1465068535157098e-06,
+      "loss": 0.72,
+      "step": 11562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.145804848919344e-06,
+      "loss": 0.7058,
+      "step": 11563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1451030314937678e-06,
+      "loss": 0.7141,
+      "step": 11564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.144401401273062e-06,
+      "loss": 0.8052,
+      "step": 11565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.143699958291301e-06,
+      "loss": 0.6801,
+      "step": 11566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1429987025825485e-06,
+      "loss": 0.8041,
+      "step": 11567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1422976341808595e-06,
+      "loss": 0.8744,
+      "step": 11568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.14159675312028e-06,
+      "loss": 0.7076,
+      "step": 11569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1408960594348461e-06,
+      "loss": 0.5701,
+      "step": 11570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1401955531585878e-06,
+      "loss": 0.681,
+      "step": 11571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1394952343255206e-06,
+      "loss": 0.485,
+      "step": 11572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1387951029696543e-06,
+      "loss": 0.7024,
+      "step": 11573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1380951591249927e-06,
+      "loss": 0.7862,
+      "step": 11574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1373954028255236e-06,
+      "loss": 0.739,
+      "step": 11575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1366958341052282e-06,
+      "loss": 0.7489,
+      "step": 11576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.135996452998085e-06,
+      "loss": 0.8764,
+      "step": 11577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1352972595380534e-06,
+      "loss": 0.5419,
+      "step": 11578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1345982537590894e-06,
+      "loss": 0.9097,
+      "step": 11579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1338994356951384e-06,
+      "loss": 0.7872,
+      "step": 11580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1332008053801374e-06,
+      "loss": 0.5388,
+      "step": 11581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.132502362848014e-06,
+      "loss": 0.7177,
+      "step": 11582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1318041081326868e-06,
+      "loss": 0.3233,
+      "step": 11583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1311060412680646e-06,
+      "loss": 0.6744,
+      "step": 11584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1304081622880475e-06,
+      "loss": 0.7311,
+      "step": 11585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1297104712265272e-06,
+      "loss": 0.6313,
+      "step": 11586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1290129681173845e-06,
+      "loss": 0.8619,
+      "step": 11587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1283156529944945e-06,
+      "loss": 0.6422,
+      "step": 11588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1276185258917154e-06,
+      "loss": 0.8118,
+      "step": 11589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1269215868429078e-06,
+      "loss": 0.7262,
+      "step": 11590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1262248358819156e-06,
+      "loss": 0.5401,
+      "step": 11591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1255282730425708e-06,
+      "loss": 0.8048,
+      "step": 11592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1248318983587052e-06,
+      "loss": 0.4838,
+      "step": 11593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.124135711864137e-06,
+      "loss": 0.8715,
+      "step": 11594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1234397135926705e-06,
+      "loss": 0.8397,
+      "step": 11595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1227439035781095e-06,
+      "loss": 0.7901,
+      "step": 11596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1220482818542445e-06,
+      "loss": 0.685,
+      "step": 11597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1213528484548531e-06,
+      "loss": 0.9706,
+      "step": 11598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1206576034137122e-06,
+      "loss": 0.6641,
+      "step": 11599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1199625467645819e-06,
+      "loss": 0.7646,
+      "step": 11600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1192676785412154e-06,
+      "loss": 0.6769,
+      "step": 11601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.118572998777362e-06,
+      "loss": 0.7869,
+      "step": 11602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1178785075067527e-06,
+      "loss": 0.7321,
+      "step": 11603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1171842047631144e-06,
+      "loss": 0.8786,
+      "step": 11604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1164900905801685e-06,
+      "loss": 0.6425,
+      "step": 11605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1157961649916194e-06,
+      "loss": 0.8607,
+      "step": 11606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1151024280311668e-06,
+      "loss": 0.6767,
+      "step": 11607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.114408879732501e-06,
+      "loss": 0.4706,
+      "step": 11608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1137155201293021e-06,
+      "loss": 0.5257,
+      "step": 11609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1130223492552428e-06,
+      "loss": 0.4968,
+      "step": 11610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1123293671439844e-06,
+      "loss": 0.7106,
+      "step": 11611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1116365738291808e-06,
+      "loss": 0.8964,
+      "step": 11612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1109439693444758e-06,
+      "loss": 0.854,
+      "step": 11613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1102515537235037e-06,
+      "loss": 0.8056,
+      "step": 11614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.109559326999891e-06,
+      "loss": 0.4346,
+      "step": 11615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1088672892072555e-06,
+      "loss": 0.7672,
+      "step": 11616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1081754403792e-06,
+      "loss": 0.6862,
+      "step": 11617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1074837805493271e-06,
+      "loss": 0.8432,
+      "step": 11618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1067923097512256e-06,
+      "loss": 0.6444,
+      "step": 11619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1061010280184708e-06,
+      "loss": 0.6446,
+      "step": 11620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1054099353846387e-06,
+      "loss": 0.7461,
+      "step": 11621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1047190318832897e-06,
+      "loss": 0.9377,
+      "step": 11622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1040283175479721e-06,
+      "loss": 0.8712,
+      "step": 11623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.103337792412233e-06,
+      "loss": 0.7871,
+      "step": 11624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1026474565096068e-06,
+      "loss": 0.7377,
+      "step": 11625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1019573098736148e-06,
+      "loss": 0.9668,
+      "step": 11626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.1012673525377743e-06,
+      "loss": 0.7394,
+      "step": 11627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.100577584535592e-06,
+      "loss": 0.6703,
+      "step": 11628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.099888005900564e-06,
+      "loss": 0.9431,
+      "step": 11629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.099198616666179e-06,
+      "loss": 0.596,
+      "step": 11630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.098509416865915e-06,
+      "loss": 0.6704,
+      "step": 11631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0978204065332426e-06,
+      "loss": 0.5532,
+      "step": 11632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0971315857016213e-06,
+      "loss": 0.4825,
+      "step": 11633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0964429544045024e-06,
+      "loss": 0.6654,
+      "step": 11634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0957545126753276e-06,
+      "loss": 0.7156,
+      "step": 11635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0950662605475298e-06,
+      "loss": 0.6779,
+      "step": 11636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0943781980545332e-06,
+      "loss": 0.8398,
+      "step": 11637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0936903252297509e-06,
+      "loss": 0.6218,
+      "step": 11638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0930026421065887e-06,
+      "loss": 0.6395,
+      "step": 11639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0923151487184426e-06,
+      "loss": 0.5491,
+      "step": 11640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0916278450986983e-06,
+      "loss": 0.9507,
+      "step": 11641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.090940731280734e-06,
+      "loss": 0.6742,
+      "step": 11642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0902538072979184e-06,
+      "loss": 0.7064,
+      "step": 11643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0895670731836094e-06,
+      "loss": 0.8856,
+      "step": 11644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.088880528971159e-06,
+      "loss": 0.8033,
+      "step": 11645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0881941746939028e-06,
+      "loss": 0.4792,
+      "step": 11646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0875080103851787e-06,
+      "loss": 0.6846,
+      "step": 11647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0868220360783038e-06,
+      "loss": 0.7979,
+      "step": 11648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0861362518065915e-06,
+      "loss": 0.708,
+      "step": 11649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.085450657603349e-06,
+      "loss": 0.8712,
+      "step": 11650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.084765253501867e-06,
+      "loss": 0.7686,
+      "step": 11651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0840800395354312e-06,
+      "loss": 0.7226,
+      "step": 11652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0833950157373207e-06,
+      "loss": 0.8024,
+      "step": 11653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0827101821407988e-06,
+      "loss": 0.6787,
+      "step": 11654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.082025538779124e-06,
+      "loss": 0.7982,
+      "step": 11655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0813410856855444e-06,
+      "loss": 0.6161,
+      "step": 11656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0806568228932995e-06,
+      "loss": 0.8186,
+      "step": 11657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.079972750435619e-06,
+      "loss": 0.8038,
+      "step": 11658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0792888683457231e-06,
+      "loss": 0.7189,
+      "step": 11659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0786051766568233e-06,
+      "loss": 0.5746,
+      "step": 11660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0779216754021215e-06,
+      "loss": 0.6783,
+      "step": 11661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0772383646148111e-06,
+      "loss": 0.9844,
+      "step": 11662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0765552443280748e-06,
+      "loss": 0.7549,
+      "step": 11663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0758723145750882e-06,
+      "loss": 0.8238,
+      "step": 11664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0751895753890125e-06,
+      "loss": 0.709,
+      "step": 11665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0745070268030077e-06,
+      "loss": 0.9693,
+      "step": 11666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0738246688502207e-06,
+      "loss": 0.8927,
+      "step": 11667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0731425015637841e-06,
+      "loss": 0.5297,
+      "step": 11668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.07246052497683e-06,
+      "loss": 0.5923,
+      "step": 11669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.071778739122477e-06,
+      "loss": 0.5489,
+      "step": 11670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0710971440338314e-06,
+      "loss": 0.7412,
+      "step": 11671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0704157397439962e-06,
+      "loss": 0.5434,
+      "step": 11672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0697345262860638e-06,
+      "loss": 0.7353,
+      "step": 11673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0690535036931104e-06,
+      "loss": 0.4678,
+      "step": 11674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0683726719982146e-06,
+      "loss": 0.8012,
+      "step": 11675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0676920312344358e-06,
+      "loss": 0.9419,
+      "step": 11676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.067011581434827e-06,
+      "loss": 0.555,
+      "step": 11677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0663313226324373e-06,
+      "loss": 1.0361,
+      "step": 11678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0656512548602977e-06,
+      "loss": 0.5815,
+      "step": 11679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0649713781514348e-06,
+      "loss": 0.6081,
+      "step": 11680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.064291692538869e-06,
+      "loss": 0.8655,
+      "step": 11681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.0636121980556042e-06,
+      "loss": 0.7712,
+      "step": 11682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.062932894734639e-06,
+      "loss": 1.142,
+      "step": 11683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.062253782608963e-06,
+      "loss": 0.5344,
+      "step": 11684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0615748617115556e-06,
+      "loss": 0.693,
+      "step": 11685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.060896132075388e-06,
+      "loss": 0.7838,
+      "step": 11686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0602175937334202e-06,
+      "loss": 0.7511,
+      "step": 11687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.059539246718604e-06,
+      "loss": 0.9312,
+      "step": 11688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0588610910638825e-06,
+      "loss": 1.0178,
+      "step": 11689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0581831268021886e-06,
+      "loss": 1.0028,
+      "step": 11690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0575053539664465e-06,
+      "loss": 0.8146,
+      "step": 11691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0568277725895715e-06,
+      "loss": 0.8292,
+      "step": 11692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.056150382704465e-06,
+      "loss": 0.6506,
+      "step": 11693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0554731843440275e-06,
+      "loss": 0.5715,
+      "step": 11694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0547961775411453e-06,
+      "loss": 0.7056,
+      "step": 11695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0541193623286917e-06,
+      "loss": 0.5583,
+      "step": 11696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0534427387395391e-06,
+      "loss": 0.7076,
+      "step": 11697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0527663068065463e-06,
+      "loss": 0.6527,
+      "step": 11698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0520900665625589e-06,
+      "loss": 0.6509,
+      "step": 11699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0514140180404202e-06,
+      "loss": 0.7698,
+      "step": 11700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0507381612729623e-06,
+      "loss": 0.6061,
+      "step": 11701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0500624962930033e-06,
+      "loss": 0.721,
+      "step": 11702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.049387023133357e-06,
+      "loss": 0.6798,
+      "step": 11703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.048711741826826e-06,
+      "loss": 0.7757,
+      "step": 11704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0480366524062041e-06,
+      "loss": 0.7814,
+      "step": 11705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0473617549042757e-06,
+      "loss": 0.7647,
+      "step": 11706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0466870493538162e-06,
+      "loss": 0.7855,
+      "step": 11707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.04601253578759e-06,
+      "loss": 0.6404,
+      "step": 11708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0453382142383545e-06,
+      "loss": 0.6483,
+      "step": 11709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0446640847388561e-06,
+      "loss": 0.6862,
+      "step": 11710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0439901473218334e-06,
+      "loss": 0.7191,
+      "step": 11711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.043316402020013e-06,
+      "loss": 0.5728,
+      "step": 11712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0426428488661162e-06,
+      "loss": 0.8103,
+      "step": 11713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0419694878928505e-06,
+      "loss": 0.8757,
+      "step": 11714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0412963191329172e-06,
+      "loss": 0.6525,
+      "step": 11715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0406233426190077e-06,
+      "loss": 0.5685,
+      "step": 11716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0399505583838027e-06,
+      "loss": 0.8479,
+      "step": 11717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0392779664599766e-06,
+      "loss": 1.1,
+      "step": 11718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0386055668801887e-06,
+      "loss": 0.6753,
+      "step": 11719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0379333596770957e-06,
+      "loss": 0.5919,
+      "step": 11720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0372613448833429e-06,
+      "loss": 0.6214,
+      "step": 11721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0365895225315602e-06,
+      "loss": 0.5326,
+      "step": 11722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0359178926543795e-06,
+      "loss": 0.8927,
+      "step": 11723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0352464552844122e-06,
+      "loss": 0.7667,
+      "step": 11724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0345752104542662e-06,
+      "loss": 0.6184,
+      "step": 11725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.033904158196542e-06,
+      "loss": 0.7337,
+      "step": 11726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0332332985438248e-06,
+      "loss": 0.8876,
+      "step": 11727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0325626315286936e-06,
+      "loss": 0.6591,
+      "step": 11728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0318921571837204e-06,
+      "loss": 0.6,
+      "step": 11729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0312218755414633e-06,
+      "loss": 0.6891,
+      "step": 11730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0305517866344727e-06,
+      "loss": 0.8112,
+      "step": 11731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0298818904952911e-06,
+      "loss": 0.7178,
+      "step": 11732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0292121871564509e-06,
+      "loss": 0.6494,
+      "step": 11733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0285426766504736e-06,
+      "loss": 0.7587,
+      "step": 11734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0278733590098732e-06,
+      "loss": 1.1058,
+      "step": 11735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.027204234267154e-06,
+      "loss": 0.6662,
+      "step": 11736
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0265353024548103e-06,
+      "loss": 0.5829,
+      "step": 11737
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0258665636053272e-06,
+      "loss": 0.7703,
+      "step": 11738
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0251980177511806e-06,
+      "loss": 0.9316,
+      "step": 11739
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0245296649248388e-06,
+      "loss": 0.5636,
+      "step": 11740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0238615051587542e-06,
+      "loss": 0.7859,
+      "step": 11741
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0231935384853792e-06,
+      "loss": 0.6005,
+      "step": 11742
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0225257649371518e-06,
+      "loss": 0.7648,
+      "step": 11743
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0218581845464974e-06,
+      "loss": 0.9042,
+      "step": 11744
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0211907973458391e-06,
+      "loss": 0.5836,
+      "step": 11745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0205236033675875e-06,
+      "loss": 0.6306,
+      "step": 11746
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0198566026441398e-06,
+      "loss": 0.68,
+      "step": 11747
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0191897952078905e-06,
+      "loss": 0.58,
+      "step": 11748
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0185231810912223e-06,
+      "loss": 0.6342,
+      "step": 11749
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0178567603265038e-06,
+      "loss": 0.419,
+      "step": 11750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.017190532946104e-06,
+      "loss": 0.7296,
+      "step": 11751
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.016524498982372e-06,
+      "loss": 0.6576,
+      "step": 11752
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0158586584676533e-06,
+      "loss": 0.7994,
+      "step": 11753
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.015193011434286e-06,
+      "loss": 0.4995,
+      "step": 11754
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.014527557914593e-06,
+      "loss": 0.7842,
+      "step": 11755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0138622979408908e-06,
+      "loss": 0.8617,
+      "step": 11756
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0131972315454868e-06,
+      "loss": 0.8604,
+      "step": 11757
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.012532358760679e-06,
+      "loss": 0.5724,
+      "step": 11758
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0118676796187548e-06,
+      "loss": 0.7611,
+      "step": 11759
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0112031941519934e-06,
+      "loss": 0.7014,
+      "step": 11760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0105389023926643e-06,
+      "loss": 0.6816,
+      "step": 11761
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0098748043730267e-06,
+      "loss": 0.7479,
+      "step": 11762
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0092109001253314e-06,
+      "loss": 0.8818,
+      "step": 11763
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.00854718968182e-06,
+      "loss": 0.7404,
+      "step": 11764
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0078836730747232e-06,
+      "loss": 0.482,
+      "step": 11765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0072203503362642e-06,
+      "loss": 0.5642,
+      "step": 11766
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0065572214986557e-06,
+      "loss": 0.664,
+      "step": 11767
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.005894286594102e-06,
+      "loss": 0.6349,
+      "step": 11768
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0052315456547934e-06,
+      "loss": 0.7539,
+      "step": 11769
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0045689987129187e-06,
+      "loss": 0.7811,
+      "step": 11770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.003906645800653e-06,
+      "loss": 0.5464,
+      "step": 11771
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.003244486950158e-06,
+      "loss": 0.7113,
+      "step": 11772
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0025825221935937e-06,
+      "loss": 0.7658,
+      "step": 11773
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.0019207515631074e-06,
+      "loss": 0.6991,
+      "step": 11774
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.001259175090834e-06,
+      "loss": 0.7059,
+      "step": 11775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.000597792808901e-06,
+      "loss": 0.7463,
+      "step": 11776
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.999366047494318e-07,
+      "loss": 0.9076,
+      "step": 11777
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.992756109445313e-07,
+      "loss": 0.7693,
+      "step": 11778
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.986148114263e-07,
+      "loss": 0.5778,
+      "step": 11779
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.979542062268288e-07,
+      "loss": 0.7263,
+      "step": 11780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.972937953781985e-07,
+      "loss": 0.7003,
+      "step": 11781
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.96633578912481e-07,
+      "loss": 0.6988,
+      "step": 11782
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.959735568617373e-07,
+      "loss": 0.9164,
+      "step": 11783
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.953137292580206e-07,
+      "loss": 0.8072,
+      "step": 11784
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.94654096133374e-07,
+      "loss": 0.5998,
+      "step": 11785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.939946575198312e-07,
+      "loss": 0.5265,
+      "step": 11786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.933354134494167e-07,
+      "loss": 0.4938,
+      "step": 11787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.926763639541442e-07,
+      "loss": 0.5923,
+      "step": 11788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.920175090660206e-07,
+      "loss": 0.6462,
+      "step": 11789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.913588488170405e-07,
+      "loss": 0.7108,
+      "step": 11790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.90700383239191e-07,
+      "loss": 0.3489,
+      "step": 11791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.900421123644483e-07,
+      "loss": 0.6759,
+      "step": 11792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.893840362247809e-07,
+      "loss": 0.8833,
+      "step": 11793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.887261548521481e-07,
+      "loss": 0.6289,
+      "step": 11794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.880684682784935e-07,
+      "loss": 0.6888,
+      "step": 11795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.874109765357615e-07,
+      "loss": 0.4849,
+      "step": 11796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.867536796558814e-07,
+      "loss": 0.9016,
+      "step": 11797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.860965776707694e-07,
+      "loss": 0.8149,
+      "step": 11798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.85439670612341e-07,
+      "loss": 0.7969,
+      "step": 11799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.847829585124936e-07,
+      "loss": 0.6695,
+      "step": 11800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.841264414031198e-07,
+      "loss": 0.8071,
+      "step": 11801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.834701193161056e-07,
+      "loss": 0.6766,
+      "step": 11802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.8281399228332e-07,
+      "loss": 0.8316,
+      "step": 11803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.82158060336626e-07,
+      "loss": 0.5917,
+      "step": 11804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.81502323507882e-07,
+      "loss": 0.6922,
+      "step": 11805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.80846781828928e-07,
+      "loss": 0.9682,
+      "step": 11806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.801914353316006e-07,
+      "loss": 0.8476,
+      "step": 11807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.795362840477257e-07,
+      "loss": 0.7928,
+      "step": 11808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.788813280091186e-07,
+      "loss": 0.5205,
+      "step": 11809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.78226567247586e-07,
+      "loss": 0.6368,
+      "step": 11810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.775720017949253e-07,
+      "loss": 0.964,
+      "step": 11811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.769176316829237e-07,
+      "loss": 0.8043,
+      "step": 11812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.7626345694336e-07,
+      "loss": 0.9546,
+      "step": 11813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.756094776080017e-07,
+      "loss": 1.0315,
+      "step": 11814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.749556937086086e-07,
+      "loss": 0.6695,
+      "step": 11815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.74302105276932e-07,
+      "loss": 0.654,
+      "step": 11816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.73648712344707e-07,
+      "loss": 0.8635,
+      "step": 11817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.729955149436693e-07,
+      "loss": 0.6719,
+      "step": 11818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.723425131055392e-07,
+      "loss": 0.6678,
+      "step": 11819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.71689706862025e-07,
+      "loss": 0.4955,
+      "step": 11820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.710370962448323e-07,
+      "loss": 0.9026,
+      "step": 11821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.703846812856543e-07,
+      "loss": 0.5092,
+      "step": 11822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.697324620161703e-07,
+      "loss": 0.789,
+      "step": 11823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.690804384680568e-07,
+      "loss": 0.6174,
+      "step": 11824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.684286106729784e-07,
+      "loss": 0.7224,
+      "step": 11825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.677769786625869e-07,
+      "loss": 0.6414,
+      "step": 11826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.671255424685311e-07,
+      "loss": 0.6183,
+      "step": 11827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.664743021224438e-07,
+      "loss": 0.7631,
+      "step": 11828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.6582325765595e-07,
+      "loss": 0.7915,
+      "step": 11829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.651724091006714e-07,
+      "loss": 0.8765,
+      "step": 11830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.64521756488211e-07,
+      "loss": 0.8307,
+      "step": 11831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.638712998501665e-07,
+      "loss": 0.7259,
+      "step": 11832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.632210392181274e-07,
+      "loss": 0.8911,
+      "step": 11833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.62570974623671e-07,
+      "loss": 0.6438,
+      "step": 11834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.619211060983674e-07,
+      "loss": 0.871,
+      "step": 11835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.61271433673775e-07,
+      "loss": 0.7349,
+      "step": 11836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.606219573814447e-07,
+      "loss": 0.794,
+      "step": 11837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.599726772529166e-07,
+      "loss": 0.7812,
+      "step": 11838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.593235933197215e-07,
+      "loss": 0.3223,
+      "step": 11839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.586747056133805e-07,
+      "loss": 0.6927,
+      "step": 11840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.580260141654057e-07,
+      "loss": 0.8274,
+      "step": 11841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.573775190073004e-07,
+      "loss": 0.7047,
+      "step": 11842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.567292201705558e-07,
+      "loss": 0.6582,
+      "step": 11843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.560811176866568e-07,
+      "loss": 0.8487,
+      "step": 11844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.554332115870746e-07,
+      "loss": 0.7222,
+      "step": 11845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.547855019032753e-07,
+      "loss": 0.7849,
+      "step": 11846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.541379886667151e-07,
+      "loss": 0.6171,
+      "step": 11847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.534906719088355e-07,
+      "loss": 0.6544,
+      "step": 11848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.528435516610729e-07,
+      "loss": 0.7275,
+      "step": 11849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.521966279548567e-07,
+      "loss": 0.926,
+      "step": 11850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.51549900821599e-07,
+      "loss": 1.0013,
+      "step": 11851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.509033702927078e-07,
+      "loss": 0.6694,
+      "step": 11852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.502570363995839e-07,
+      "loss": 0.4605,
+      "step": 11853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.496108991736109e-07,
+      "loss": 0.5126,
+      "step": 11854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.489649586461686e-07,
+      "loss": 1.0221,
+      "step": 11855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.483192148486259e-07,
+      "loss": 0.817,
+      "step": 11856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.476736678123421e-07,
+      "loss": 0.8164,
+      "step": 11857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.470283175686662e-07,
+      "loss": 0.8908,
+      "step": 11858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.463831641489391e-07,
+      "loss": 0.8571,
+      "step": 11859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.457382075844906e-07,
+      "loss": 0.8075,
+      "step": 11860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.450934479066421e-07,
+      "loss": 0.7741,
+      "step": 11861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.444488851467043e-07,
+      "loss": 0.91,
+      "step": 11862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.438045193359796e-07,
+      "loss": 0.7649,
+      "step": 11863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.431603505057607e-07,
+      "loss": 0.679,
+      "step": 11864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.425163786873292e-07,
+      "loss": 0.5501,
+      "step": 11865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.418726039119592e-07,
+      "loss": 0.906,
+      "step": 11866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.412290262109147e-07,
+      "loss": 0.8626,
+      "step": 11867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.405856456154466e-07,
+      "loss": 0.6524,
+      "step": 11868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.399424621568026e-07,
+      "loss": 0.6916,
+      "step": 11869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.392994758662178e-07,
+      "loss": 0.8763,
+      "step": 11870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.386566867749131e-07,
+      "loss": 0.5948,
+      "step": 11871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.380140949141087e-07,
+      "loss": 0.8826,
+      "step": 11872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.373717003150107e-07,
+      "loss": 0.6247,
+      "step": 11873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.367295030088114e-07,
+      "loss": 0.699,
+      "step": 11874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.36087503026703e-07,
+      "loss": 0.6683,
+      "step": 11875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.35445700399859e-07,
+      "loss": 0.6039,
+      "step": 11876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.348040951594478e-07,
+      "loss": 0.8505,
+      "step": 11877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.341626873366305e-07,
+      "loss": 0.7652,
+      "step": 11878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.335214769625523e-07,
+      "loss": 0.458,
+      "step": 11879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.32880464068352e-07,
+      "loss": 0.6989,
+      "step": 11880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.322396486851626e-07,
+      "loss": 0.887,
+      "step": 11881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.315990308441015e-07,
+      "loss": 0.6649,
+      "step": 11882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.309586105762786e-07,
+      "loss": 0.5177,
+      "step": 11883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.303183879127953e-07,
+      "loss": 1.0257,
+      "step": 11884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.296783628847423e-07,
+      "loss": 0.696,
+      "step": 11885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.290385355232018e-07,
+      "loss": 0.7542,
+      "step": 11886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.283989058592446e-07,
+      "loss": 0.3834,
+      "step": 11887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.277594739239337e-07,
+      "loss": 0.7832,
+      "step": 11888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.271202397483214e-07,
+      "loss": 0.8105,
+      "step": 11889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.264812033634513e-07,
+      "loss": 0.8617,
+      "step": 11890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.258423648003562e-07,
+      "loss": 0.6049,
+      "step": 11891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.252037240900618e-07,
+      "loss": 0.8077,
+      "step": 11892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.245652812635774e-07,
+      "loss": 0.5437,
+      "step": 11893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.239270363519132e-07,
+      "loss": 0.5947,
+      "step": 11894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.232889893860625e-07,
+      "loss": 0.6882,
+      "step": 11895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.226511403970079e-07,
+      "loss": 0.7185,
+      "step": 11896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.220134894157285e-07,
+      "loss": 0.6731,
+      "step": 11897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.213760364731905e-07,
+      "loss": 0.6381,
+      "step": 11898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.207387816003471e-07,
+      "loss": 0.7668,
+      "step": 11899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.201017248281486e-07,
+      "loss": 0.8564,
+      "step": 11900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.194648661875322e-07,
+      "loss": 0.5998,
+      "step": 11901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.188282057094222e-07,
+      "loss": 0.9404,
+      "step": 11902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.181917434247417e-07,
+      "loss": 0.7187,
+      "step": 11903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.175554793643948e-07,
+      "loss": 0.6257,
+      "step": 11904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.169194135592824e-07,
+      "loss": 0.719,
+      "step": 11905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.162835460402925e-07,
+      "loss": 0.6864,
+      "step": 11906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.15647876838306e-07,
+      "loss": 0.8806,
+      "step": 11907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.15012405984192e-07,
+      "loss": 0.6113,
+      "step": 11908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.143771335088108e-07,
+      "loss": 0.4963,
+      "step": 11909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.137420594430135e-07,
+      "loss": 0.5458,
+      "step": 11910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.131071838176409e-07,
+      "loss": 0.8649,
+      "step": 11911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.124725066635243e-07,
+      "loss": 0.5846,
+      "step": 11912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.118380280114858e-07,
+      "loss": 0.4873,
+      "step": 11913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.112037478923364e-07,
+      "loss": 1.0124,
+      "step": 11914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.105696663368801e-07,
+      "loss": 0.827,
+      "step": 11915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.099357833759087e-07,
+      "loss": 0.6182,
+      "step": 11916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.093020990402057e-07,
+      "loss": 0.9509,
+      "step": 11917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.086686133605449e-07,
+      "loss": 0.5919,
+      "step": 11918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.080353263676894e-07,
+      "loss": 0.4775,
+      "step": 11919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.074022380923953e-07,
+      "loss": 0.8625,
+      "step": 11920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.067693485654033e-07,
+      "loss": 0.6754,
+      "step": 11921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.061366578174524e-07,
+      "loss": 0.5633,
+      "step": 11922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.055041658792673e-07,
+      "loss": 0.6656,
+      "step": 11923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.048718727815614e-07,
+      "loss": 0.5275,
+      "step": 11924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.042397785550405e-07,
+      "loss": 0.6579,
+      "step": 11925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.036078832304052e-07,
+      "loss": 0.8804,
+      "step": 11926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.02976186838338e-07,
+      "loss": 0.8157,
+      "step": 11927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.023446894095156e-07,
+      "loss": 0.7104,
+      "step": 11928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.017133909746095e-07,
+      "loss": 0.8827,
+      "step": 11929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.010822915642736e-07,
+      "loss": 0.5616,
+      "step": 11930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.004513912091573e-07,
+      "loss": 0.7604,
+      "step": 11931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.998206899398987e-07,
+      "loss": 0.7796,
+      "step": 11932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.991901877871262e-07,
+      "loss": 0.7521,
+      "step": 11933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.985598847814592e-07,
+      "loss": 0.6275,
+      "step": 11934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.979297809535065e-07,
+      "loss": 0.903,
+      "step": 11935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.972998763338686e-07,
+      "loss": 0.4839,
+      "step": 11936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.966701709531344e-07,
+      "loss": 0.6368,
+      "step": 11937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.960406648418856e-07,
+      "loss": 0.9301,
+      "step": 11938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.954113580306917e-07,
+      "loss": 0.575,
+      "step": 11939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.947822505501141e-07,
+      "loss": 0.9114,
+      "step": 11940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.941533424307037e-07,
+      "loss": 0.7585,
+      "step": 11941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.935246337030024e-07,
+      "loss": 0.7516,
+      "step": 11942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.928961243975437e-07,
+      "loss": 0.6702,
+      "step": 11943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.922678145448449e-07,
+      "loss": 0.8151,
+      "step": 11944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.916397041754238e-07,
+      "loss": 0.6488,
+      "step": 11945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.910117933197821e-07,
+      "loss": 0.7165,
+      "step": 11946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.903840820084097e-07,
+      "loss": 0.827,
+      "step": 11947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.897565702717936e-07,
+      "loss": 0.6422,
+      "step": 11948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.891292581404076e-07,
+      "loss": 0.7588,
+      "step": 11949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.885021456447123e-07,
+      "loss": 0.6591,
+      "step": 11950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.878752328151669e-07,
+      "loss": 0.7403,
+      "step": 11951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.872485196822123e-07,
+      "loss": 0.8411,
+      "step": 11952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.866220062762837e-07,
+      "loss": 0.6502,
+      "step": 11953
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.859956926278096e-07,
+      "loss": 0.6533,
+      "step": 11954
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.853695787672023e-07,
+      "loss": 0.7835,
+      "step": 11955
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.847436647248681e-07,
+      "loss": 0.7813,
+      "step": 11956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.841179505312058e-07,
+      "loss": 0.8923,
+      "step": 11957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.834924362165992e-07,
+      "loss": 0.5604,
+      "step": 11958
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.828671218114254e-07,
+      "loss": 0.7802,
+      "step": 11959
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.822420073460519e-07,
+      "loss": 0.5943,
+      "step": 11960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.816170928508367e-07,
+      "loss": 0.5649,
+      "step": 11961
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.809923783561264e-07,
+      "loss": 0.6905,
+      "step": 11962
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.803678638922592e-07,
+      "loss": 0.5251,
+      "step": 11963
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.79743549489564e-07,
+      "loss": 0.6251,
+      "step": 11964
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.791194351783583e-07,
+      "loss": 0.7097,
+      "step": 11965
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.784955209889517e-07,
+      "loss": 0.5044,
+      "step": 11966
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.778718069516429e-07,
+      "loss": 0.6524,
+      "step": 11967
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.772482930967225e-07,
+      "loss": 0.5278,
+      "step": 11968
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.766249794544662e-07,
+      "loss": 0.666,
+      "step": 11969
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.760018660551484e-07,
+      "loss": 0.4962,
+      "step": 11970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.75378952929028e-07,
+      "loss": 0.645,
+      "step": 11971
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.747562401063531e-07,
+      "loss": 1.0158,
+      "step": 11972
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.741337276173667e-07,
+      "loss": 0.6721,
+      "step": 11973
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.735114154923014e-07,
+      "loss": 0.6912,
+      "step": 11974
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.728893037613733e-07,
+      "loss": 0.8305,
+      "step": 11975
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.722673924547986e-07,
+      "loss": 0.8777,
+      "step": 11976
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.716456816027791e-07,
+      "loss": 0.835,
+      "step": 11977
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 8.710241712355039e-07,
+      "loss": 0.913,
+      "step": 11978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.704028613831561e-07,
+      "loss": 0.8089,
+      "step": 11979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.697817520759094e-07,
+      "loss": 0.4665,
+      "step": 11980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.691608433439263e-07,
+      "loss": 0.6003,
+      "step": 11981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.6854013521736e-07,
+      "loss": 0.5939,
+      "step": 11982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.679196277263535e-07,
+      "loss": 0.6428,
+      "step": 11983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.6729932090104e-07,
+      "loss": 0.9591,
+      "step": 11984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.666792147715447e-07,
+      "loss": 0.7134,
+      "step": 11985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.660593093679809e-07,
+      "loss": 0.9705,
+      "step": 11986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.654396047204528e-07,
+      "loss": 0.7988,
+      "step": 11987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.648201008590557e-07,
+      "loss": 0.8136,
+      "step": 11988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.642007978138739e-07,
+      "loss": 0.7742,
+      "step": 11989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.635816956149823e-07,
+      "loss": 0.8324,
+      "step": 11990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.629627942924473e-07,
+      "loss": 0.9791,
+      "step": 11991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.623440938763239e-07,
+      "loss": 0.8579,
+      "step": 11992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.617255943966579e-07,
+      "loss": 0.9672,
+      "step": 11993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.61107295883486e-07,
+      "loss": 0.6273,
+      "step": 11994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.604891983668334e-07,
+      "loss": 0.6728,
+      "step": 11995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.598713018767191e-07,
+      "loss": 0.9142,
+      "step": 11996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.592536064431467e-07,
+      "loss": 0.6312,
+      "step": 11997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.586361120961139e-07,
+      "loss": 0.772,
+      "step": 11998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.580188188656113e-07,
+      "loss": 0.7753,
+      "step": 11999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.574017267816126e-07,
+      "loss": 0.9119,
+      "step": 12000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.567848358740854e-07,
+      "loss": 0.6134,
+      "step": 12001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.561681461729926e-07,
+      "loss": 0.6809,
+      "step": 12002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.555516577082773e-07,
+      "loss": 0.7264,
+      "step": 12003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.549353705098784e-07,
+      "loss": 0.7097,
+      "step": 12004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.543192846077287e-07,
+      "loss": 0.5635,
+      "step": 12005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.537034000317429e-07,
+      "loss": 0.7063,
+      "step": 12006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.530877168118318e-07,
+      "loss": 0.7512,
+      "step": 12007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.524722349778941e-07,
+      "loss": 0.6784,
+      "step": 12008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.518569545598198e-07,
+      "loss": 0.7788,
+      "step": 12009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.512418755874891e-07,
+      "loss": 0.9347,
+      "step": 12010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.506269980907711e-07,
+      "loss": 0.8076,
+      "step": 12011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.500123220995271e-07,
+      "loss": 0.6951,
+      "step": 12012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.49397847643606e-07,
+      "loss": 0.7166,
+      "step": 12013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.4878357475285e-07,
+      "loss": 0.7167,
+      "step": 12014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.481695034570892e-07,
+      "loss": 0.7448,
+      "step": 12015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.475556337861457e-07,
+      "loss": 0.6811,
+      "step": 12016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.469419657698274e-07,
+      "loss": 0.8236,
+      "step": 12017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.463284994379401e-07,
+      "loss": 0.9689,
+      "step": 12018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.457152348202741e-07,
+      "loss": 0.8203,
+      "step": 12019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.451021719466085e-07,
+      "loss": 0.683,
+      "step": 12020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.444893108467194e-07,
+      "loss": 0.7734,
+      "step": 12021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.438766515503682e-07,
+      "loss": 0.7586,
+      "step": 12022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.432641940873043e-07,
+      "loss": 0.8725,
+      "step": 12023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.426519384872733e-07,
+      "loss": 0.6568,
+      "step": 12024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.420398847800093e-07,
+      "loss": 0.577,
+      "step": 12025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.414280329952313e-07,
+      "loss": 0.7227,
+      "step": 12026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.40816383162657e-07,
+      "loss": 0.7697,
+      "step": 12027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.402049353119862e-07,
+      "loss": 0.402,
+      "step": 12028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.395936894729134e-07,
+      "loss": 0.5712,
+      "step": 12029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.389826456751249e-07,
+      "loss": 0.6658,
+      "step": 12030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.383718039482924e-07,
+      "loss": 0.7634,
+      "step": 12031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.37761164322079e-07,
+      "loss": 0.8971,
+      "step": 12032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.371507268261436e-07,
+      "loss": 0.7571,
+      "step": 12033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.36540491490127e-07,
+      "loss": 0.6523,
+      "step": 12034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.359304583436645e-07,
+      "loss": 0.8514,
+      "step": 12035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.353206274163816e-07,
+      "loss": 0.6888,
+      "step": 12036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.347109987378937e-07,
+      "loss": 0.9206,
+      "step": 12037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.341015723378054e-07,
+      "loss": 0.5093,
+      "step": 12038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.334923482457131e-07,
+      "loss": 0.6904,
+      "step": 12039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.328833264912018e-07,
+      "loss": 0.9246,
+      "step": 12040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.322745071038474e-07,
+      "loss": 0.8506,
+      "step": 12041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.316658901132163e-07,
+      "loss": 0.9395,
+      "step": 12042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.310574755488643e-07,
+      "loss": 0.5725,
+      "step": 12043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.304492634403389e-07,
+      "loss": 0.5128,
+      "step": 12044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.298412538171735e-07,
+      "loss": 0.7767,
+      "step": 12045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.292334467088981e-07,
+      "loss": 0.7309,
+      "step": 12046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.286258421450294e-07,
+      "loss": 0.6594,
+      "step": 12047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.280184401550711e-07,
+      "loss": 0.8562,
+      "step": 12048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.274112407685242e-07,
+      "loss": 0.7371,
+      "step": 12049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.268042440148755e-07,
+      "loss": 0.7899,
+      "step": 12050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.261974499235987e-07,
+      "loss": 0.677,
+      "step": 12051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.255908585241662e-07,
+      "loss": 0.6191,
+      "step": 12052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.249844698460346e-07,
+      "loss": 0.2995,
+      "step": 12053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.243782839186504e-07,
+      "loss": 0.9628,
+      "step": 12054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.23772300771452e-07,
+      "loss": 0.5509,
+      "step": 12055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.23166520433869e-07,
+      "loss": 0.8546,
+      "step": 12056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.225609429353187e-07,
+      "loss": 0.7441,
+      "step": 12057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.219555683052099e-07,
+      "loss": 0.7936,
+      "step": 12058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.213503965729414e-07,
+      "loss": 0.7708,
+      "step": 12059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.207454277679023e-07,
+      "loss": 0.7088,
+      "step": 12060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.201406619194713e-07,
+      "loss": 0.8475,
+      "step": 12061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.195360990570178e-07,
+      "loss": 0.7083,
+      "step": 12062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.189317392099011e-07,
+      "loss": 0.6773,
+      "step": 12063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.18327582407471e-07,
+      "loss": 0.8645,
+      "step": 12064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.177236286790663e-07,
+      "loss": 0.944,
+      "step": 12065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.171198780540169e-07,
+      "loss": 0.8741,
+      "step": 12066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.165163305616431e-07,
+      "loss": 0.6192,
+      "step": 12067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.159129862312548e-07,
+      "loss": 0.4915,
+      "step": 12068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.153098450921514e-07,
+      "loss": 0.8136,
+      "step": 12069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.147069071736247e-07,
+      "loss": 0.5969,
+      "step": 12070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.141041725049537e-07,
+      "loss": 0.8085,
+      "step": 12071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.135016411154106e-07,
+      "loss": 0.7434,
+      "step": 12072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.128993130342538e-07,
+      "loss": 0.6513,
+      "step": 12073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.122971882907338e-07,
+      "loss": 0.9054,
+      "step": 12074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.116952669140949e-07,
+      "loss": 0.5717,
+      "step": 12075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.110935489335652e-07,
+      "loss": 0.6589,
+      "step": 12076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.104920343783656e-07,
+      "loss": 0.451,
+      "step": 12077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.098907232777109e-07,
+      "loss": 0.729,
+      "step": 12078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.092896156607988e-07,
+      "loss": 0.8224,
+      "step": 12079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.086887115568209e-07,
+      "loss": 0.7113,
+      "step": 12080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.08088010994963e-07,
+      "loss": 0.5906,
+      "step": 12081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.07487514004392e-07,
+      "loss": 0.8195,
+      "step": 12082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.06887220614272e-07,
+      "loss": 1.0488,
+      "step": 12083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.062871308537545e-07,
+      "loss": 0.7511,
+      "step": 12084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.056872447519814e-07,
+      "loss": 0.8556,
+      "step": 12085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.050875623380855e-07,
+      "loss": 0.5886,
+      "step": 12086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.044880836411889e-07,
+      "loss": 0.5614,
+      "step": 12087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.038888086904034e-07,
+      "loss": 0.436,
+      "step": 12088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.032897375148324e-07,
+      "loss": 0.6254,
+      "step": 12089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.026908701435681e-07,
+      "loss": 0.8372,
+      "step": 12090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.020922066056936e-07,
+      "loss": 0.5227,
+      "step": 12091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.014937469302819e-07,
+      "loss": 0.6008,
+      "step": 12092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.008954911463934e-07,
+      "loss": 0.8782,
+      "step": 12093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.002974392830842e-07,
+      "loss": 0.7562,
+      "step": 12094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.996995913693978e-07,
+      "loss": 0.6386,
+      "step": 12095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.991019474343631e-07,
+      "loss": 0.6293,
+      "step": 12096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.98504507507008e-07,
+      "loss": 0.5677,
+      "step": 12097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.979072716163449e-07,
+      "loss": 0.907,
+      "step": 12098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.97310239791374e-07,
+      "loss": 0.903,
+      "step": 12099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.967134120610942e-07,
+      "loss": 0.9063,
+      "step": 12100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.961167884544852e-07,
+      "loss": 0.5714,
+      "step": 12101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.955203690005203e-07,
+      "loss": 0.8594,
+      "step": 12102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.949241537281677e-07,
+      "loss": 0.7166,
+      "step": 12103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.943281426663774e-07,
+      "loss": 0.6341,
+      "step": 12104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7318,
+      "step": 12105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.931367332902534e-07,
+      "loss": 0.7907,
+      "step": 12106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.925413350337785e-07,
+      "loss": 0.6977,
+      "step": 12107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.919461411035828e-07,
+      "loss": 0.6724,
+      "step": 12108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.913511515285722e-07,
+      "loss": 0.6801,
+      "step": 12109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.9075636633764e-07,
+      "loss": 0.6208,
+      "step": 12110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.901617855596705e-07,
+      "loss": 0.8715,
+      "step": 12111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.895674092235395e-07,
+      "loss": 0.7348,
+      "step": 12112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.88973237358111e-07,
+      "loss": 0.8196,
+      "step": 12113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.883792699922394e-07,
+      "loss": 0.7989,
+      "step": 12114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.877855071547691e-07,
+      "loss": 0.6451,
+      "step": 12115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.871919488745361e-07,
+      "loss": 0.7462,
+      "step": 12116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.865985951803646e-07,
+      "loss": 0.686,
+      "step": 12117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.860054461010697e-07,
+      "loss": 0.6442,
+      "step": 12118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.854125016654563e-07,
+      "loss": 0.7461,
+      "step": 12119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.848197619023207e-07,
+      "loss": 0.8639,
+      "step": 12120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.84227226840445e-07,
+      "loss": 0.6064,
+      "step": 12121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.836348965086071e-07,
+      "loss": 0.9193,
+      "step": 12122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.830427709355726e-07,
+      "loss": 0.628,
+      "step": 12123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.824508501500932e-07,
+      "loss": 0.6596,
+      "step": 12124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.818591341809184e-07,
+      "loss": 0.5996,
+      "step": 12125
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.812676230567829e-07,
+      "loss": 1.026,
+      "step": 12126
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.806763168064103e-07,
+      "loss": 0.6151,
+      "step": 12127
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.800852154585153e-07,
+      "loss": 0.6696,
+      "step": 12128
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.79494319041808e-07,
+      "loss": 0.7429,
+      "step": 12129
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.789036275849804e-07,
+      "loss": 0.9921,
+      "step": 12130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.783131411167189e-07,
+      "loss": 0.7635,
+      "step": 12131
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.777228596656994e-07,
+      "loss": 0.7811,
+      "step": 12132
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.77132783260588e-07,
+      "loss": 0.7419,
+      "step": 12133
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.765429119300394e-07,
+      "loss": 0.752,
+      "step": 12134
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.759532457027008e-07,
+      "loss": 0.6866,
+      "step": 12135
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.753637846072076e-07,
+      "loss": 0.8399,
+      "step": 12136
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.747745286721852e-07,
+      "loss": 0.7047,
+      "step": 12137
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.741854779262503e-07,
+      "loss": 0.7343,
+      "step": 12138
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.735966323980088e-07,
+      "loss": 0.7994,
+      "step": 12139
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.73007992116056e-07,
+      "loss": 0.5578,
+      "step": 12140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.724195571089787e-07,
+      "loss": 0.7972,
+      "step": 12141
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.718313274053529e-07,
+      "loss": 0.8278,
+      "step": 12142
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.712433030337446e-07,
+      "loss": 0.7585,
+      "step": 12143
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.706554840227099e-07,
+      "loss": 0.4854,
+      "step": 12144
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.700678704007947e-07,
+      "loss": 0.7647,
+      "step": 12145
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.694804621965368e-07,
+      "loss": 0.5146,
+      "step": 12146
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.68893259438459e-07,
+      "loss": 0.6005,
+      "step": 12147
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.683062621550824e-07,
+      "loss": 0.5838,
+      "step": 12148
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.677194703749091e-07,
+      "loss": 0.7413,
+      "step": 12149
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.671328841264352e-07,
+      "loss": 0.5514,
+      "step": 12150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.665465034381519e-07,
+      "loss": 0.799,
+      "step": 12151
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.659603283385308e-07,
+      "loss": 0.8014,
+      "step": 12152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.653743588560387e-07,
+      "loss": 0.8146,
+      "step": 12153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.647885950191353e-07,
+      "loss": 0.5926,
+      "step": 12154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.642030368562636e-07,
+      "loss": 0.6508,
+      "step": 12155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.636176843958599e-07,
+      "loss": 0.8253,
+      "step": 12156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.630325376663538e-07,
+      "loss": 0.3525,
+      "step": 12157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.624475966961592e-07,
+      "loss": 0.7686,
+      "step": 12158
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.618628615136825e-07,
+      "loss": 0.7127,
+      "step": 12159
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.61278332147321e-07,
+      "loss": 0.5252,
+      "step": 12160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.606940086254605e-07,
+      "loss": 0.6909,
+      "step": 12161
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.601098909764781e-07,
+      "loss": 0.6342,
+      "step": 12162
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.595259792287397e-07,
+      "loss": 0.7921,
+      "step": 12163
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.589422734106011e-07,
+      "loss": 0.7571,
+      "step": 12164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.583587735504106e-07,
+      "loss": 0.9095,
+      "step": 12165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.57775479676503e-07,
+      "loss": 0.685,
+      "step": 12166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.571923918172053e-07,
+      "loss": 1.022,
+      "step": 12167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.566095100008347e-07,
+      "loss": 0.6611,
+      "step": 12168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.560268342556948e-07,
+      "loss": 0.7548,
+      "step": 12169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.554443646100851e-07,
+      "loss": 0.816,
+      "step": 12170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.548621010922919e-07,
+      "loss": 0.9708,
+      "step": 12171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.542800437305886e-07,
+      "loss": 0.8302,
+      "step": 12172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.536981925532444e-07,
+      "loss": 0.6555,
+      "step": 12173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.531165475885161e-07,
+      "loss": 0.7042,
+      "step": 12174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.525351088646466e-07,
+      "loss": 0.7869,
+      "step": 12175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.519538764098766e-07,
+      "loss": 0.7215,
+      "step": 12176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.513728502524286e-07,
+      "loss": 0.7346,
+      "step": 12177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.507920304205202e-07,
+      "loss": 0.5877,
+      "step": 12178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.502114169423597e-07,
+      "loss": 0.8979,
+      "step": 12179
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.496310098461406e-07,
+      "loss": 0.8777,
+      "step": 12180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.490508091600495e-07,
+      "loss": 0.8221,
+      "step": 12181
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.484708149122655e-07,
+      "loss": 0.859,
+      "step": 12182
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.478910271309509e-07,
+      "loss": 0.7285,
+      "step": 12183
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.473114458442642e-07,
+      "loss": 0.7836,
+      "step": 12184
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.467320710803505e-07,
+      "loss": 0.735,
+      "step": 12185
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.461529028673465e-07,
+      "loss": 0.5969,
+      "step": 12186
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.455739412333779e-07,
+      "loss": 0.7242,
+      "step": 12187
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.449951862065607e-07,
+      "loss": 0.849,
+      "step": 12188
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.444166378150014e-07,
+      "loss": 0.4041,
+      "step": 12189
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.438382960867957e-07,
+      "loss": 0.7786,
+      "step": 12190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.432601610500295e-07,
+      "loss": 0.6945,
+      "step": 12191
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.426822327327787e-07,
+      "loss": 0.5815,
+      "step": 12192
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.421045111631092e-07,
+      "loss": 0.7206,
+      "step": 12193
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.415269963690769e-07,
+      "loss": 0.7654,
+      "step": 12194
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.409496883787276e-07,
+      "loss": 0.4547,
+      "step": 12195
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.403725872200978e-07,
+      "loss": 0.7529,
+      "step": 12196
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.397956929212097e-07,
+      "loss": 0.5938,
+      "step": 12197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.392190055100834e-07,
+      "loss": 0.7872,
+      "step": 12198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.386425250147233e-07,
+      "loss": 0.5695,
+      "step": 12199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.380662514631221e-07,
+      "loss": 0.8587,
+      "step": 12200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.374901848832683e-07,
+      "loss": 0.7812,
+      "step": 12201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.36914325303138e-07,
+      "loss": 0.6159,
+      "step": 12202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.363386727506943e-07,
+      "loss": 0.9688,
+      "step": 12203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.357632272538917e-07,
+      "loss": 0.8621,
+      "step": 12204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.351879888406799e-07,
+      "loss": 0.7159,
+      "step": 12205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.346129575389898e-07,
+      "loss": 0.6593,
+      "step": 12206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.340381333767487e-07,
+      "loss": 0.6294,
+      "step": 12207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.33463516381871e-07,
+      "loss": 0.7615,
+      "step": 12208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.328891065822613e-07,
+      "loss": 0.4118,
+      "step": 12209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.323149040058159e-07,
+      "loss": 0.7947,
+      "step": 12210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.317409086804189e-07,
+      "loss": 0.4971,
+      "step": 12211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.311671206339454e-07,
+      "loss": 0.5653,
+      "step": 12212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.305935398942598e-07,
+      "loss": 0.7502,
+      "step": 12213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.30020166489217e-07,
+      "loss": 0.4355,
+      "step": 12214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.294470004466625e-07,
+      "loss": 0.706,
+      "step": 12215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.288740417944295e-07,
+      "loss": 0.8292,
+      "step": 12216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.283012905603437e-07,
+      "loss": 0.9081,
+      "step": 12217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.277287467722188e-07,
+      "loss": 0.6138,
+      "step": 12218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.271564104578604e-07,
+      "loss": 0.9042,
+      "step": 12219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.265842816450596e-07,
+      "loss": 0.4662,
+      "step": 12220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.260123603616043e-07,
+      "loss": 0.5618,
+      "step": 12221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.254406466352682e-07,
+      "loss": 0.6591,
+      "step": 12222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.248691404938119e-07,
+      "loss": 0.7489,
+      "step": 12223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.242978419649937e-07,
+      "loss": 0.6554,
+      "step": 12224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.237267510765549e-07,
+      "loss": 0.638,
+      "step": 12225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.231558678562289e-07,
+      "loss": 1.0148,
+      "step": 12226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.225851923317428e-07,
+      "loss": 0.9424,
+      "step": 12227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.220147245308074e-07,
+      "loss": 0.8022,
+      "step": 12228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.214444644811247e-07,
+      "loss": 0.8123,
+      "step": 12229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.208744122103933e-07,
+      "loss": 0.9559,
+      "step": 12230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.203045677462922e-07,
+      "loss": 0.4581,
+      "step": 12231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.197349311164947e-07,
+      "loss": 0.6473,
+      "step": 12232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.191655023486682e-07,
+      "loss": 0.954,
+      "step": 12233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.185962814704616e-07,
+      "loss": 0.7203,
+      "step": 12234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.180272685095191e-07,
+      "loss": 0.8081,
+      "step": 12235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.174584634934739e-07,
+      "loss": 0.7096,
+      "step": 12236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.168898664499485e-07,
+      "loss": 0.6108,
+      "step": 12237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.163214774065558e-07,
+      "loss": 0.6885,
+      "step": 12238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.157532963908986e-07,
+      "loss": 0.7025,
+      "step": 12239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.151853234305695e-07,
+      "loss": 0.8004,
+      "step": 12240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.146175585531501e-07,
+      "loss": 0.8164,
+      "step": 12241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.140500017862129e-07,
+      "loss": 0.9747,
+      "step": 12242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.134826531573208e-07,
+      "loss": 0.6082,
+      "step": 12243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.129155126940269e-07,
+      "loss": 0.8429,
+      "step": 12244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.123485804238689e-07,
+      "loss": 0.6319,
+      "step": 12245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.117818563743828e-07,
+      "loss": 0.7205,
+      "step": 12246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.112153405730898e-07,
+      "loss": 0.9914,
+      "step": 12247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.106490330474991e-07,
+      "loss": 0.8503,
+      "step": 12248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.100829338251147e-07,
+      "loss": 0.6653,
+      "step": 12249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.09517042933428e-07,
+      "loss": 0.818,
+      "step": 12250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.089513603999177e-07,
+      "loss": 0.9683,
+      "step": 12251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.083858862520593e-07,
+      "loss": 0.7933,
+      "step": 12252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.078206205173094e-07,
+      "loss": 0.6756,
+      "step": 12253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.072555632231199e-07,
+      "loss": 0.4894,
+      "step": 12254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.066907143969353e-07,
+      "loss": 0.6561,
+      "step": 12255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.061260740661818e-07,
+      "loss": 0.8729,
+      "step": 12256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.055616422582817e-07,
+      "loss": 0.7022,
+      "step": 12257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.049974190006458e-07,
+      "loss": 0.756,
+      "step": 12258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.044334043206736e-07,
+      "loss": 0.68,
+      "step": 12259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.038695982457561e-07,
+      "loss": 0.5711,
+      "step": 12260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.033060008032721e-07,
+      "loss": 0.9054,
+      "step": 12261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.027426120205932e-07,
+      "loss": 0.9165,
+      "step": 12262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.02179431925078e-07,
+      "loss": 0.8792,
+      "step": 12263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.016164605440761e-07,
+      "loss": 0.6936,
+      "step": 12264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.010536979049277e-07,
+      "loss": 0.7954,
+      "step": 12265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.004911440349616e-07,
+      "loss": 0.444,
+      "step": 12266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.999287989614972e-07,
+      "loss": 0.7391,
+      "step": 12267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.99366662711844e-07,
+      "loss": 0.8295,
+      "step": 12268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.988047353133004e-07,
+      "loss": 0.707,
+      "step": 12269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.982430167931558e-07,
+      "loss": 0.7474,
+      "step": 12270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.97681507178688e-07,
+      "loss": 0.6207,
+      "step": 12271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.971202064971671e-07,
+      "loss": 0.5394,
+      "step": 12272
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.965591147758482e-07,
+      "loss": 0.8453,
+      "step": 12273
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.959982320419828e-07,
+      "loss": 0.7897,
+      "step": 12274
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.95437558322809e-07,
+      "loss": 0.8355,
+      "step": 12275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.948770936455528e-07,
+      "loss": 0.5261,
+      "step": 12276
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.94316838037431e-07,
+      "loss": 0.7655,
+      "step": 12277
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.93756791525656e-07,
+      "loss": 0.755,
+      "step": 12278
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.931969541374206e-07,
+      "loss": 0.5524,
+      "step": 12279
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.926373258999125e-07,
+      "loss": 0.7524,
+      "step": 12280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.920779068403127e-07,
+      "loss": 0.5482,
+      "step": 12281
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.915186969857846e-07,
+      "loss": 0.9846,
+      "step": 12282
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.909596963634857e-07,
+      "loss": 0.75,
+      "step": 12283
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.904009050005634e-07,
+      "loss": 0.6309,
+      "step": 12284
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.898423229241535e-07,
+      "loss": 0.4579,
+      "step": 12285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.89283950161383e-07,
+      "loss": 0.489,
+      "step": 12286
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.887257867393677e-07,
+      "loss": 0.6959,
+      "step": 12287
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.881678326852137e-07,
+      "loss": 0.5638,
+      "step": 12288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.876100880260167e-07,
+      "loss": 0.7863,
+      "step": 12289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.870525527888627e-07,
+      "loss": 0.5895,
+      "step": 12290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.864952270008274e-07,
+      "loss": 0.78,
+      "step": 12291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.85938110688975e-07,
+      "loss": 0.5767,
+      "step": 12292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.85381203880362e-07,
+      "loss": 0.7607,
+      "step": 12293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.848245066020326e-07,
+      "loss": 0.7205,
+      "step": 12294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.84268018881023e-07,
+      "loss": 1.0094,
+      "step": 12295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.83711740744355e-07,
+      "loss": 0.7212,
+      "step": 12296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.831556722190453e-07,
+      "loss": 0.8075,
+      "step": 12297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.825998133320987e-07,
+      "loss": 0.4278,
+      "step": 12298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.820441641105064e-07,
+      "loss": 0.9152,
+      "step": 12299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.814887245812563e-07,
+      "loss": 0.6796,
+      "step": 12300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.809334947713192e-07,
+      "loss": 0.6981,
+      "step": 12301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.80378474707658e-07,
+      "loss": 0.6778,
+      "step": 12302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.798236644172301e-07,
+      "loss": 0.6479,
+      "step": 12303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.792690639269744e-07,
+      "loss": 0.8062,
+      "step": 12304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.78714673263825e-07,
+      "loss": 0.7483,
+      "step": 12305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.781604924547075e-07,
+      "loss": 0.5223,
+      "step": 12306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.776065215265315e-07,
+      "loss": 0.8238,
+      "step": 12307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.770527605061989e-07,
+      "loss": 0.8363,
+      "step": 12308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.764992094206052e-07,
+      "loss": 0.8744,
+      "step": 12309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.759458682966297e-07,
+      "loss": 0.7655,
+      "step": 12310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.753927371611452e-07,
+      "loss": 0.6834,
+      "step": 12311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.74839816041013e-07,
+      "loss": 0.7082,
+      "step": 12312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.74287104963085e-07,
+      "loss": 0.6522,
+      "step": 12313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.737346039542019e-07,
+      "loss": 0.6216,
+      "step": 12314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.731823130411952e-07,
+      "loss": 0.7091,
+      "step": 12315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.726302322508854e-07,
+      "loss": 0.7966,
+      "step": 12316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.720783616100828e-07,
+      "loss": 0.8049,
+      "step": 12317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.715267011455889e-07,
+      "loss": 0.8192,
+      "step": 12318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.709752508841933e-07,
+      "loss": 0.5834,
+      "step": 12319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.704240108526772e-07,
+      "loss": 0.8149,
+      "step": 12320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.698729810778065e-07,
+      "loss": 0.8777,
+      "step": 12321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.693221615863449e-07,
+      "loss": 0.6325,
+      "step": 12322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.68771552405042e-07,
+      "loss": 0.9428,
+      "step": 12323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.682211535606326e-07,
+      "loss": 0.7156,
+      "step": 12324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.676709650798491e-07,
+      "loss": 0.7905,
+      "step": 12325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.671209869894113e-07,
+      "loss": 0.593,
+      "step": 12326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.665712193160234e-07,
+      "loss": 0.7176,
+      "step": 12327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.660216620863885e-07,
+      "loss": 0.7494,
+      "step": 12328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.654723153271913e-07,
+      "loss": 0.7047,
+      "step": 12329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.6492317906511e-07,
+      "loss": 0.5994,
+      "step": 12330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.643742533268149e-07,
+      "loss": 0.6824,
+      "step": 12331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.638255381389607e-07,
+      "loss": 0.7343,
+      "step": 12332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.632770335281952e-07,
+      "loss": 0.7219,
+      "step": 12333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.627287395211556e-07,
+      "loss": 0.6877,
+      "step": 12334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.621806561444683e-07,
+      "loss": 0.6256,
+      "step": 12335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.616327834247505e-07,
+      "loss": 0.5178,
+      "step": 12336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.610851213886077e-07,
+      "loss": 0.6673,
+      "step": 12337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.605376700626371e-07,
+      "loss": 0.6228,
+      "step": 12338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.599904294734227e-07,
+      "loss": 0.7867,
+      "step": 12339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.594433996475418e-07,
+      "loss": 0.892,
+      "step": 12340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.588965806115588e-07,
+      "loss": 0.5813,
+      "step": 12341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.583499723920289e-07,
+      "loss": 0.5691,
+      "step": 12342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.578035750154977e-07,
+      "loss": 0.8544,
+      "step": 12343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.572573885084987e-07,
+      "loss": 0.6014,
+      "step": 12344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.567114128975571e-07,
+      "loss": 0.3742,
+      "step": 12345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.561656482091866e-07,
+      "loss": 0.3325,
+      "step": 12346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.55620094469891e-07,
+      "loss": 0.7798,
+      "step": 12347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.550747517061656e-07,
+      "loss": 0.8704,
+      "step": 12348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.545296199444917e-07,
+      "loss": 0.7844,
+      "step": 12349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.539846992113419e-07,
+      "loss": 0.5379,
+      "step": 12350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.534399895331822e-07,
+      "loss": 0.9233,
+      "step": 12351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.528954909364627e-07,
+      "loss": 0.983,
+      "step": 12352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.523512034476254e-07,
+      "loss": 0.6004,
+      "step": 12353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.518071270931059e-07,
+      "loss": 0.849,
+      "step": 12354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.512632618993231e-07,
+      "loss": 0.7145,
+      "step": 12355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.507196078926881e-07,
+      "loss": 0.799,
+      "step": 12356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.501761650996052e-07,
+      "loss": 0.6552,
+      "step": 12357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.496329335464635e-07,
+      "loss": 0.7286,
+      "step": 12358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.490899132596445e-07,
+      "loss": 0.7618,
+      "step": 12359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.485471042655184e-07,
+      "loss": 0.6871,
+      "step": 12360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.480045065904461e-07,
+      "loss": 0.7964,
+      "step": 12361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.474621202607773e-07,
+      "loss": 0.7628,
+      "step": 12362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.469199453028518e-07,
+      "loss": 0.8524,
+      "step": 12363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.46377981743e-07,
+      "loss": 0.9416,
+      "step": 12364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.458362296075399e-07,
+      "loss": 0.8442,
+      "step": 12365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.452946889227813e-07,
+      "loss": 0.8412,
+      "step": 12366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.447533597150229e-07,
+      "loss": 0.7697,
+      "step": 12367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.442122420105545e-07,
+      "loss": 0.5038,
+      "step": 12368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.436713358356506e-07,
+      "loss": 0.8266,
+      "step": 12369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.431306412165827e-07,
+      "loss": 0.7258,
+      "step": 12370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.42590158179609e-07,
+      "loss": 0.8153,
+      "step": 12371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.420498867509723e-07,
+      "loss": 0.6545,
+      "step": 12372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.415098269569142e-07,
+      "loss": 0.8888,
+      "step": 12373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.409699788236612e-07,
+      "loss": 0.6526,
+      "step": 12374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.404303423774261e-07,
+      "loss": 0.6083,
+      "step": 12375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.398909176444206e-07,
+      "loss": 0.4954,
+      "step": 12376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.393517046508363e-07,
+      "loss": 0.9231,
+      "step": 12377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.388127034228598e-07,
+      "loss": 0.5975,
+      "step": 12378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.382739139866695e-07,
+      "loss": 0.4196,
+      "step": 12379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.377353363684268e-07,
+      "loss": 0.4849,
+      "step": 12380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.371969705942877e-07,
+      "loss": 0.8437,
+      "step": 12381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.366588166903992e-07,
+      "loss": 0.8966,
+      "step": 12382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.361208746828928e-07,
+      "loss": 0.6998,
+      "step": 12383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.355831445978921e-07,
+      "loss": 0.8785,
+      "step": 12384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.35045626461514e-07,
+      "loss": 0.4682,
+      "step": 12385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.345083202998592e-07,
+      "loss": 0.4576,
+      "step": 12386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.339712261390213e-07,
+      "loss": 0.6317,
+      "step": 12387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.334343440050839e-07,
+      "loss": 0.7916,
+      "step": 12388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.328976739241188e-07,
+      "loss": 0.5688,
+      "step": 12389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.323612159221887e-07,
+      "loss": 0.6299,
+      "step": 12390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.318249700253453e-07,
+      "loss": 0.7532,
+      "step": 12391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.312889362596303e-07,
+      "loss": 0.6363,
+      "step": 12392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.307531146510754e-07,
+      "loss": 0.5927,
+      "step": 12393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.302175052257009e-07,
+      "loss": 0.81,
+      "step": 12394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.296821080095183e-07,
+      "loss": 0.9601,
+      "step": 12395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.291469230285285e-07,
+      "loss": 0.6791,
+      "step": 12396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.286119503087185e-07,
+      "loss": 0.8825,
+      "step": 12397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.280771898760718e-07,
+      "loss": 0.7418,
+      "step": 12398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.275426417565572e-07,
+      "loss": 0.6593,
+      "step": 12399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.270083059761312e-07,
+      "loss": 0.7503,
+      "step": 12400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.264741825607457e-07,
+      "loss": 0.9925,
+      "step": 12401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.259402715363394e-07,
+      "loss": 0.5362,
+      "step": 12402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.254065729288373e-07,
+      "loss": 0.9966,
+      "step": 12403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.248730867641617e-07,
+      "loss": 0.571,
+      "step": 12404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.243398130682165e-07,
+      "loss": 0.4384,
+      "step": 12405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.238067518669005e-07,
+      "loss": 0.9585,
+      "step": 12406
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.232739031861007e-07,
+      "loss": 0.4801,
+      "step": 12407
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.227412670516942e-07,
+      "loss": 0.7089,
+      "step": 12408
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.222088434895462e-07,
+      "loss": 0.6973,
+      "step": 12409
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.216766325255136e-07,
+      "loss": 0.6839,
+      "step": 12410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.211446341854415e-07,
+      "loss": 0.8624,
+      "step": 12411
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.206128484951662e-07,
+      "loss": 0.5417,
+      "step": 12412
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.200812754805119e-07,
+      "loss": 0.7212,
+      "step": 12413
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.195499151672934e-07,
+      "loss": 0.7327,
+      "step": 12414
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.190187675813159e-07,
+      "loss": 0.7027,
+      "step": 12415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.184878327483723e-07,
+      "loss": 0.517,
+      "step": 12416
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.179571106942466e-07,
+      "loss": 1.0822,
+      "step": 12417
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.174266014447128e-07,
+      "loss": 0.552,
+      "step": 12418
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.168963050255333e-07,
+      "loss": 0.6772,
+      "step": 12419
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.163662214624616e-07,
+      "loss": 0.7402,
+      "step": 12420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.158363507812393e-07,
+      "loss": 0.7226,
+      "step": 12421
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.153066930075985e-07,
+      "loss": 0.9365,
+      "step": 12422
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.147772481672609e-07,
+      "loss": 0.7985,
+      "step": 12423
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.142480162859399e-07,
+      "loss": 0.7462,
+      "step": 12424
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.137189973893331e-07,
+      "loss": 0.9405,
+      "step": 12425
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.131901915031319e-07,
+      "loss": 0.8197,
+      "step": 12426
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.126615986530193e-07,
+      "loss": 0.7862,
+      "step": 12427
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.121332188646628e-07,
+      "loss": 0.6623,
+      "step": 12428
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.116050521637218e-07,
+      "loss": 0.8519,
+      "step": 12429
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.110770985758479e-07,
+      "loss": 0.9145,
+      "step": 12430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.105493581266786e-07,
+      "loss": 0.7929,
+      "step": 12431
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.100218308418409e-07,
+      "loss": 0.7028,
+      "step": 12432
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.094945167469568e-07,
+      "loss": 0.8659,
+      "step": 12433
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.089674158676312e-07,
+      "loss": 0.8617,
+      "step": 12434
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.084405282294625e-07,
+      "loss": 0.5683,
+      "step": 12435
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.079138538580381e-07,
+      "loss": 0.6694,
+      "step": 12436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.073873927789343e-07,
+      "loss": 0.7331,
+      "step": 12437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.068611450177181e-07,
+      "loss": 0.7898,
+      "step": 12438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.063351105999454e-07,
+      "loss": 0.4768,
+      "step": 12439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.058092895511614e-07,
+      "loss": 0.4764,
+      "step": 12440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.052836818969027e-07,
+      "loss": 0.7907,
+      "step": 12441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.047582876626934e-07,
+      "loss": 0.7043,
+      "step": 12442
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.042331068740481e-07,
+      "loss": 0.7143,
+      "step": 12443
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.037081395564731e-07,
+      "loss": 0.5213,
+      "step": 12444
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.031833857354579e-07,
+      "loss": 0.8255,
+      "step": 12445
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.026588454364901e-07,
+      "loss": 0.7077,
+      "step": 12446
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.021345186850419e-07,
+      "loss": 0.786,
+      "step": 12447
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.016104055065747e-07,
+      "loss": 0.6168,
+      "step": 12448
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.010865059265425e-07,
+      "loss": 0.7474,
+      "step": 12449
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.005628199703878e-07,
+      "loss": 0.6219,
+      "step": 12450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.000393476635391e-07,
+      "loss": 0.7674,
+      "step": 12451
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.995160890314222e-07,
+      "loss": 0.5994,
+      "step": 12452
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.989930440994451e-07,
+      "loss": 0.6922,
+      "step": 12453
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.984702128930076e-07,
+      "loss": 0.4922,
+      "step": 12454
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.979475954375036e-07,
+      "loss": 0.623,
+      "step": 12455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.974251917583096e-07,
+      "loss": 0.6621,
+      "step": 12456
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.969030018807953e-07,
+      "loss": 0.5655,
+      "step": 12457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.963810258303227e-07,
+      "loss": 0.7608,
+      "step": 12458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.958592636322368e-07,
+      "loss": 0.3933,
+      "step": 12459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.953377153118761e-07,
+      "loss": 0.7849,
+      "step": 12460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.948163808945723e-07,
+      "loss": 0.5809,
+      "step": 12461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.942952604056396e-07,
+      "loss": 0.8169,
+      "step": 12462
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.93774353870385e-07,
+      "loss": 0.7805,
+      "step": 12463
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.932536613141065e-07,
+      "loss": 0.5756,
+      "step": 12464
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.927331827620902e-07,
+      "loss": 0.7754,
+      "step": 12465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.922129182396113e-07,
+      "loss": 0.6113,
+      "step": 12466
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.916928677719363e-07,
+      "loss": 0.7412,
+      "step": 12467
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.911730313843201e-07,
+      "loss": 0.4966,
+      "step": 12468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.906534091020067e-07,
+      "loss": 0.8667,
+      "step": 12469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.90134000950231e-07,
+      "loss": 0.621,
+      "step": 12470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.896148069542173e-07,
+      "loss": 0.8272,
+      "step": 12471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.890958271391795e-07,
+      "loss": 0.6807,
+      "step": 12472
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.885770615303182e-07,
+      "loss": 0.5727,
+      "step": 12473
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.880585101528291e-07,
+      "loss": 0.675,
+      "step": 12474
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.875401730318947e-07,
+      "loss": 0.7639,
+      "step": 12475
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.870220501926833e-07,
+      "loss": 0.8005,
+      "step": 12476
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.8650414166036e-07,
+      "loss": 0.6387,
+      "step": 12477
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.859864474600757e-07,
+      "loss": 0.7192,
+      "step": 12478
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.854689676169689e-07,
+      "loss": 0.8821,
+      "step": 12479
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.849517021561712e-07,
+      "loss": 0.5432,
+      "step": 12480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.844346511028032e-07,
+      "loss": 0.6442,
+      "step": 12481
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.839178144819729e-07,
+      "loss": 0.9086,
+      "step": 12482
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.834011923187804e-07,
+      "loss": 0.451,
+      "step": 12483
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.82884784638314e-07,
+      "loss": 0.6564,
+      "step": 12484
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.823685914656524e-07,
+      "loss": 0.5553,
+      "step": 12485
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.818526128258622e-07,
+      "loss": 0.5603,
+      "step": 12486
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.813368487440024e-07,
+      "loss": 0.837,
+      "step": 12487
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.80821299245119e-07,
+      "loss": 0.6699,
+      "step": 12488
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.803059643542491e-07,
+      "loss": 0.8153,
+      "step": 12489
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.797908440964184e-07,
+      "loss": 0.9711,
+      "step": 12490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.792759384966423e-07,
+      "loss": 0.8009,
+      "step": 12491
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.78761247579927e-07,
+      "loss": 0.6086,
+      "step": 12492
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.782467713712669e-07,
+      "loss": 0.6773,
+      "step": 12493
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.777325098956466e-07,
+      "loss": 0.7155,
+      "step": 12494
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.772184631780398e-07,
+      "loss": 0.5615,
+      "step": 12495
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.767046312434099e-07,
+      "loss": 0.6784,
+      "step": 12496
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.76191014116711e-07,
+      "loss": 0.8838,
+      "step": 12497
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.756776118228863e-07,
+      "loss": 0.7653,
+      "step": 12498
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.751644243868649e-07,
+      "loss": 0.7882,
+      "step": 12499
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.746514518335727e-07,
+      "loss": 0.7289,
+      "step": 12500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.741386941879179e-07,
+      "loss": 0.7102,
+      "step": 12501
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.736261514748021e-07,
+      "loss": 0.8235,
+      "step": 12502
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.731138237191186e-07,
+      "loss": 0.6842,
+      "step": 12503
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.726017109457449e-07,
+      "loss": 0.4311,
+      "step": 12504
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.720898131795494e-07,
+      "loss": 0.6767,
+      "step": 12505
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.715781304453955e-07,
+      "loss": 0.5292,
+      "step": 12506
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.710666627681289e-07,
+      "loss": 0.7942,
+      "step": 12507
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.705554101725885e-07,
+      "loss": 0.6396,
+      "step": 12508
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.700443726836025e-07,
+      "loss": 0.6915,
+      "step": 12509
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.69533550325988e-07,
+      "loss": 0.7838,
+      "step": 12510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.690229431245519e-07,
+      "loss": 0.935,
+      "step": 12511
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.685125511040917e-07,
+      "loss": 0.5328,
+      "step": 12512
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.680023742893926e-07,
+      "loss": 0.7481,
+      "step": 12513
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.674924127052306e-07,
+      "loss": 0.6506,
+      "step": 12514
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.669826663763706e-07,
+      "loss": 0.622,
+      "step": 12515
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.664731353275676e-07,
+      "loss": 0.5868,
+      "step": 12516
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.659638195835671e-07,
+      "loss": 0.896,
+      "step": 12517
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.654547191690995e-07,
+      "loss": 0.579,
+      "step": 12518
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.649458341088915e-07,
+      "loss": 0.5007,
+      "step": 12519
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.644371644276558e-07,
+      "loss": 0.7632,
+      "step": 12520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.639287101500923e-07,
+      "loss": 0.7712,
+      "step": 12521
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.63420471300895e-07,
+      "loss": 0.6848,
+      "step": 12522
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.629124479047466e-07,
+      "loss": 1.0424,
+      "step": 12523
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.624046399863148e-07,
+      "loss": 0.6656,
+      "step": 12524
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.618970475702629e-07,
+      "loss": 0.7556,
+      "step": 12525
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.61389670681241e-07,
+      "loss": 0.7848,
+      "step": 12526
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.608825093438863e-07,
+      "loss": 0.7938,
+      "step": 12527
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.603755635828318e-07,
+      "loss": 0.6676,
+      "step": 12528
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.598688334226926e-07,
+      "loss": 0.6519,
+      "step": 12529
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.593623188880781e-07,
+      "loss": 0.7743,
+      "step": 12530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.588560200035886e-07,
+      "loss": 0.7138,
+      "step": 12531
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.583499367938077e-07,
+      "loss": 0.638,
+      "step": 12532
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.578440692833137e-07,
+      "loss": 0.6199,
+      "step": 12533
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.573384174966751e-07,
+      "loss": 0.7265,
+      "step": 12534
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.568329814584444e-07,
+      "loss": 0.8119,
+      "step": 12535
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.563277611931689e-07,
+      "loss": 0.6013,
+      "step": 12536
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.558227567253832e-07,
+      "loss": 0.5607,
+      "step": 12537
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.553179680796117e-07,
+      "loss": 0.5804,
+      "step": 12538
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.548133952803681e-07,
+      "loss": 0.7274,
+      "step": 12539
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.543090383521566e-07,
+      "loss": 0.6176,
+      "step": 12540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.538048973194699e-07,
+      "loss": 0.7528,
+      "step": 12541
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.533009722067901e-07,
+      "loss": 0.5734,
+      "step": 12542
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.527972630385897e-07,
+      "loss": 0.8383,
+      "step": 12543
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.522937698393305e-07,
+      "loss": 1.0086,
+      "step": 12544
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.517904926334644e-07,
+      "loss": 0.4914,
+      "step": 12545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.512874314454286e-07,
+      "loss": 0.7367,
+      "step": 12546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.507845862996564e-07,
+      "loss": 0.657,
+      "step": 12547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.502819572205675e-07,
+      "loss": 0.6687,
+      "step": 12548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.497795442325682e-07,
+      "loss": 0.5707,
+      "step": 12549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.492773473600599e-07,
+      "loss": 0.4824,
+      "step": 12550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.487753666274309e-07,
+      "loss": 0.9002,
+      "step": 12551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.482736020590551e-07,
+      "loss": 0.5106,
+      "step": 12552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.477720536793035e-07,
+      "loss": 0.7893,
+      "step": 12553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.472707215125328e-07,
+      "loss": 0.8683,
+      "step": 12554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.467696055830862e-07,
+      "loss": 0.9234,
+      "step": 12555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.46268705915301e-07,
+      "loss": 0.7832,
+      "step": 12556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.457680225335022e-07,
+      "loss": 0.7292,
+      "step": 12557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.452675554620046e-07,
+      "loss": 0.5681,
+      "step": 12558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.447673047251117e-07,
+      "loss": 0.7246,
+      "step": 12559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.44267270347118e-07,
+      "loss": 0.6368,
+      "step": 12560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.437674523523056e-07,
+      "loss": 0.8478,
+      "step": 12561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.432678507649486e-07,
+      "loss": 0.5093,
+      "step": 12562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.427684656093074e-07,
+      "loss": 0.878,
+      "step": 12563
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.422692969096344e-07,
+      "loss": 0.4295,
+      "step": 12564
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.417703446901706e-07,
+      "loss": 0.8246,
+      "step": 12565
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.412716089751468e-07,
+      "loss": 0.786,
+      "step": 12566
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.407730897887831e-07,
+      "loss": 0.7282,
+      "step": 12567
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.402747871552877e-07,
+      "loss": 0.8495,
+      "step": 12568
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.397767010988614e-07,
+      "loss": 0.8464,
+      "step": 12569
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.39278831643692e-07,
+      "loss": 0.7638,
+      "step": 12570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.387811788139569e-07,
+      "loss": 0.9566,
+      "step": 12571
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.382837426338238e-07,
+      "loss": 0.7092,
+      "step": 12572
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.377865231274504e-07,
+      "loss": 0.5006,
+      "step": 12573
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.372895203189837e-07,
+      "loss": 0.8941,
+      "step": 12574
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.367927342325558e-07,
+      "loss": 0.4735,
+      "step": 12575
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.362961648922971e-07,
+      "loss": 0.3697,
+      "step": 12576
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.35799812322319e-07,
+      "loss": 0.5834,
+      "step": 12577
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.353036765467262e-07,
+      "loss": 0.524,
+      "step": 12578
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.348077575896149e-07,
+      "loss": 0.4399,
+      "step": 12579
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.343120554750652e-07,
+      "loss": 0.5005,
+      "step": 12580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.338165702271503e-07,
+      "loss": 0.458,
+      "step": 12581
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.333213018699357e-07,
+      "loss": 0.7133,
+      "step": 12582
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.328262504274695e-07,
+      "loss": 0.8069,
+      "step": 12583
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.323314159237935e-07,
+      "loss": 0.7563,
+      "step": 12584
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.318367983829393e-07,
+      "loss": 0.8807,
+      "step": 12585
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.313423978289262e-07,
+      "loss": 0.7153,
+      "step": 12586
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.308482142857635e-07,
+      "loss": 0.6026,
+      "step": 12587
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.30354247777451e-07,
+      "loss": 0.6849,
+      "step": 12588
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.29860498327977e-07,
+      "loss": 0.7152,
+      "step": 12589
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.293669659613187e-07,
+      "loss": 0.8756,
+      "step": 12590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.288736507014436e-07,
+      "loss": 0.9359,
+      "step": 12591
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.28380552572309e-07,
+      "loss": 0.7207,
+      "step": 12592
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.27887671597862e-07,
+      "loss": 0.7427,
+      "step": 12593
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.273950078020357e-07,
+      "loss": 0.7385,
+      "step": 12594
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.269025612087569e-07,
+      "loss": 0.9145,
+      "step": 12595
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.264103318419417e-07,
+      "loss": 0.6669,
+      "step": 12596
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.259183197254903e-07,
+      "loss": 0.6088,
+      "step": 12597
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.254265248832996e-07,
+      "loss": 0.6558,
+      "step": 12598
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.249349473392529e-07,
+      "loss": 0.7176,
+      "step": 12599
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.244435871172188e-07,
+      "loss": 0.7008,
+      "step": 12600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.239524442410627e-07,
+      "loss": 0.6559,
+      "step": 12601
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.234615187346354e-07,
+      "loss": 0.7865,
+      "step": 12602
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.229708106217751e-07,
+      "loss": 0.8433,
+      "step": 12603
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.224803199263162e-07,
+      "loss": 0.7014,
+      "step": 12604
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.21990046672075e-07,
+      "loss": 0.4983,
+      "step": 12605
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.214999908828605e-07,
+      "loss": 0.7698,
+      "step": 12606
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.210101525824745e-07,
+      "loss": 0.6664,
+      "step": 12607
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.205205317947021e-07,
+      "loss": 0.7183,
+      "step": 12608
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.200311285433213e-07,
+      "loss": 0.7027,
+      "step": 12609
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.195419428520987e-07,
+      "loss": 0.5919,
+      "step": 12610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.190529747447909e-07,
+      "loss": 0.6431,
+      "step": 12611
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.185642242451439e-07,
+      "loss": 0.9305,
+      "step": 12612
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.180756913768925e-07,
+      "loss": 1.0273,
+      "step": 12613
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.175873761637612e-07,
+      "loss": 0.633,
+      "step": 12614
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.170992786294638e-07,
+      "loss": 0.679,
+      "step": 12615
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.16611398797704e-07,
+      "loss": 0.5638,
+      "step": 12616
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.16123736692175e-07,
+      "loss": 0.8661,
+      "step": 12617
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.156362923365587e-07,
+      "loss": 0.6999,
+      "step": 12618
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.151490657545266e-07,
+      "loss": 0.7488,
+      "step": 12619
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.146620569697403e-07,
+      "loss": 0.6954,
+      "step": 12620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.141752660058508e-07,
+      "loss": 0.7986,
+      "step": 12621
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.136886928864954e-07,
+      "loss": 0.5362,
+      "step": 12622
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.132023376353063e-07,
+      "loss": 0.6014,
+      "step": 12623
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.127162002759023e-07,
+      "loss": 0.6301,
+      "step": 12624
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.12230280831889e-07,
+      "loss": 0.6851,
+      "step": 12625
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.117445793268672e-07,
+      "loss": 0.6121,
+      "step": 12626
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.112590957844233e-07,
+      "loss": 0.7083,
+      "step": 12627
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.107738302281318e-07,
+      "loss": 0.7528,
+      "step": 12628
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.102887826815589e-07,
+      "loss": 0.6473,
+      "step": 12629
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.09803953168263e-07,
+      "loss": 0.5631,
+      "step": 12630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.093193417117853e-07,
+      "loss": 0.4963,
+      "step": 12631
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.088349483356619e-07,
+      "loss": 0.5521,
+      "step": 12632
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.083507730634152e-07,
+      "loss": 0.7535,
+      "step": 12633
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.07866815918559e-07,
+      "loss": 0.5503,
+      "step": 12634
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.073830769245946e-07,
+      "loss": 0.8818,
+      "step": 12635
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.068995561050155e-07,
+      "loss": 0.8311,
+      "step": 12636
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.064162534833011e-07,
+      "loss": 0.7001,
+      "step": 12637
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.059331690829228e-07,
+      "loss": 0.7664,
+      "step": 12638
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.054503029273406e-07,
+      "loss": 0.6499,
+      "step": 12639
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.049676550400035e-07,
+      "loss": 0.6609,
+      "step": 12640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.044852254443505e-07,
+      "loss": 0.7598,
+      "step": 12641
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.040030141638102e-07,
+      "loss": 0.5746,
+      "step": 12642
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.035210212217989e-07,
+      "loss": 0.9431,
+      "step": 12643
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.03039246641725e-07,
+      "loss": 0.8922,
+      "step": 12644
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.025576904469842e-07,
+      "loss": 0.6372,
+      "step": 12645
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.020763526609624e-07,
+      "loss": 0.6259,
+      "step": 12646
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.015952333070351e-07,
+      "loss": 0.7183,
+      "step": 12647
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.011143324085649e-07,
+      "loss": 0.7068,
+      "step": 12648
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.006336499889075e-07,
+      "loss": 0.6438,
+      "step": 12649
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.001531860714076e-07,
+      "loss": 0.6604,
+      "step": 12650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.996729406793943e-07,
+      "loss": 0.4574,
+      "step": 12651
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.991929138361929e-07,
+      "loss": 0.7371,
+      "step": 12652
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.987131055651135e-07,
+      "loss": 0.7613,
+      "step": 12653
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.982335158894552e-07,
+      "loss": 0.9111,
+      "step": 12654
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.977541448325124e-07,
+      "loss": 0.5479,
+      "step": 12655
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.972749924175618e-07,
+      "loss": 1.0121,
+      "step": 12656
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.967960586678722e-07,
+      "loss": 0.6562,
+      "step": 12657
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.963173436067042e-07,
+      "loss": 0.4831,
+      "step": 12658
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.958388472573034e-07,
+      "loss": 0.6766,
+      "step": 12659
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.953605696429081e-07,
+      "loss": 0.701,
+      "step": 12660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.948825107867444e-07,
+      "loss": 0.8789,
+      "step": 12661
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.944046707120287e-07,
+      "loss": 0.3943,
+      "step": 12662
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.939270494419657e-07,
+      "loss": 0.7469,
+      "step": 12663
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.934496469997501e-07,
+      "loss": 0.8425,
+      "step": 12664
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.929724634085664e-07,
+      "loss": 0.6077,
+      "step": 12665
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.924954986915875e-07,
+      "loss": 0.7563,
+      "step": 12666
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.920187528719766e-07,
+      "loss": 0.6244,
+      "step": 12667
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.915422259728858e-07,
+      "loss": 0.6191,
+      "step": 12668
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.910659180174571e-07,
+      "loss": 0.9629,
+      "step": 12669
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.905898290288191e-07,
+      "loss": 0.5694,
+      "step": 12670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.901139590300947e-07,
+      "loss": 0.7618,
+      "step": 12671
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.896383080443934e-07,
+      "loss": 0.7376,
+      "step": 12672
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.891628760948114e-07,
+      "loss": 0.8033,
+      "step": 12673
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.886876632044402e-07,
+      "loss": 0.6634,
+      "step": 12674
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.882126693963568e-07,
+      "loss": 0.7721,
+      "step": 12675
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.877378946936262e-07,
+      "loss": 0.4162,
+      "step": 12676
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.872633391193066e-07,
+      "loss": 0.7214,
+      "step": 12677
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.867890026964451e-07,
+      "loss": 0.7364,
+      "step": 12678
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.863148854480731e-07,
+      "loss": 0.4592,
+      "step": 12679
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.858409873972192e-07,
+      "loss": 0.9182,
+      "step": 12680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.853673085668947e-07,
+      "loss": 0.5875,
+      "step": 12681
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.84893848980102e-07,
+      "loss": 0.8871,
+      "step": 12682
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.844206086598368e-07,
+      "loss": 0.7876,
+      "step": 12683
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.839475876290788e-07,
+      "loss": 0.7046,
+      "step": 12684
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.834747859107996e-07,
+      "loss": 0.5235,
+      "step": 12685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.830022035279602e-07,
+      "loss": 0.6861,
+      "step": 12686
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.8252984050351e-07,
+      "loss": 0.7074,
+      "step": 12687
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.820576968603885e-07,
+      "loss": 0.7407,
+      "step": 12688
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.815857726215251e-07,
+      "loss": 0.5627,
+      "step": 12689
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.811140678098364e-07,
+      "loss": 0.7514,
+      "step": 12690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.806425824482308e-07,
+      "loss": 0.5573,
+      "step": 12691
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.80171316559605e-07,
+      "loss": 0.6609,
+      "step": 12692
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797002701668441e-07,
+      "loss": 0.6453,
+      "step": 12693
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.792294432928247e-07,
+      "loss": 0.6108,
+      "step": 12694
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.787588359604106e-07,
+      "loss": 0.912,
+      "step": 12695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.78288448192456e-07,
+      "loss": 0.5121,
+      "step": 12696
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.778182800118053e-07,
+      "loss": 1.0081,
+      "step": 12697
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.773483314412886e-07,
+      "loss": 0.8299,
+      "step": 12698
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.768786025037308e-07,
+      "loss": 0.7363,
+      "step": 12699
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7640909322194286e-07,
+      "loss": 0.568,
+      "step": 12700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7593980361872293e-07,
+      "loss": 0.7654,
+      "step": 12701
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.75470733716864e-07,
+      "loss": 0.5495,
+      "step": 12702
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.750018835391451e-07,
+      "loss": 1.0807,
+      "step": 12703
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7453325310833355e-07,
+      "loss": 0.9747,
+      "step": 12704
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.740648424471872e-07,
+      "loss": 0.813,
+      "step": 12705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7359665157845625e-07,
+      "loss": 0.731,
+      "step": 12706
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7312868052487416e-07,
+      "loss": 0.7424,
+      "step": 12707
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.726609293091683e-07,
+      "loss": 0.5485,
+      "step": 12708
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7219339795405385e-07,
+      "loss": 0.9297,
+      "step": 12709
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7172608648223597e-07,
+      "loss": 1.0066,
+      "step": 12710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.712589949164076e-07,
+      "loss": 0.7199,
+      "step": 12711
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.707921232792534e-07,
+      "loss": 0.5608,
+      "step": 12712
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.7032547159344466e-07,
+      "loss": 0.6902,
+      "step": 12713
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6985903988164436e-07,
+      "loss": 0.5617,
+      "step": 12714
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.693928281665033e-07,
+      "loss": 0.4276,
+      "step": 12715
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.689268364706617e-07,
+      "loss": 0.4383,
+      "step": 12716
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6846106481675035e-07,
+      "loss": 0.7776,
+      "step": 12717
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.679955132273878e-07,
+      "loss": 0.6958,
+      "step": 12718
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.675301817251826e-07,
+      "loss": 0.4966,
+      "step": 12719
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6706507033273283e-07,
+      "loss": 0.9291,
+      "step": 12720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.66600179072626e-07,
+      "loss": 0.908,
+      "step": 12721
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.661355079674379e-07,
+      "loss": 0.5215,
+      "step": 12722
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6567105703973493e-07,
+      "loss": 0.7442,
+      "step": 12723
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.652068263120702e-07,
+      "loss": 0.6212,
+      "step": 12724
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6474281580699054e-07,
+      "loss": 0.4919,
+      "step": 12725
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6427902554702974e-07,
+      "loss": 0.7783,
+      "step": 12726
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6381545555470807e-07,
+      "loss": 0.8198,
+      "step": 12727
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.633521058525409e-07,
+      "loss": 0.4892,
+      "step": 12728
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.628889764630279e-07,
+      "loss": 0.945,
+      "step": 12729
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.624260674086589e-07,
+      "loss": 0.5833,
+      "step": 12730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.6196337871191824e-07,
+      "loss": 0.563,
+      "step": 12731
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.615009103952717e-07,
+      "loss": 0.7481,
+      "step": 12732
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.610386624811786e-07,
+      "loss": 0.8375,
+      "step": 12733
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.605766349920893e-07,
+      "loss": 0.7406,
+      "step": 12734
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.601148279504386e-07,
+      "loss": 0.8317,
+      "step": 12735
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5965324137865366e-07,
+      "loss": 0.9909,
+      "step": 12736
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5919187529915144e-07,
+      "loss": 0.7698,
+      "step": 12737
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5873072973433683e-07,
+      "loss": 0.8658,
+      "step": 12738
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5826980470660364e-07,
+      "loss": 0.6672,
+      "step": 12739
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.578091002383367e-07,
+      "loss": 0.8715,
+      "step": 12740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5734861635190864e-07,
+      "loss": 0.9389,
+      "step": 12741
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5688835306968214e-07,
+      "loss": 0.7659,
+      "step": 12742
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.564283104140088e-07,
+      "loss": 0.9415,
+      "step": 12743
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.559684884072291e-07,
+      "loss": 0.5336,
+      "step": 12744
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5550888707167505e-07,
+      "loss": 0.8204,
+      "step": 12745
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5504950642966286e-07,
+      "loss": 0.6668,
+      "step": 12746
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5459034650350453e-07,
+      "loss": 0.5532,
+      "step": 12747
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.541314073154984e-07,
+      "loss": 0.9499,
+      "step": 12748
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5367268888792835e-07,
+      "loss": 0.8722,
+      "step": 12749
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.532141912430743e-07,
+      "loss": 0.8988,
+      "step": 12750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5275591440320175e-07,
+      "loss": 0.6121,
+      "step": 12751
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5229785839056404e-07,
+      "loss": 0.784,
+      "step": 12752
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5184002322740784e-07,
+      "loss": 0.431,
+      "step": 12753
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5138240893596697e-07,
+      "loss": 0.5533,
+      "step": 12754
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.50925015538462e-07,
+      "loss": 0.4871,
+      "step": 12755
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5046784305710854e-07,
+      "loss": 0.9119,
+      "step": 12756
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.5001089151410606e-07,
+      "loss": 1.0775,
+      "step": 12757
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.495541609316456e-07,
+      "loss": 0.7128,
+      "step": 12758
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4909765133190787e-07,
+      "loss": 0.7128,
+      "step": 12759
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.486413627370623e-07,
+      "loss": 0.8143,
+      "step": 12760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.481852951692672e-07,
+      "loss": 0.9355,
+      "step": 12761
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.477294486506706e-07,
+      "loss": 0.8101,
+      "step": 12762
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.472738232034096e-07,
+      "loss": 0.6283,
+      "step": 12763
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4681841884961164e-07,
+      "loss": 0.6742,
+      "step": 12764
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4636323561139115e-07,
+      "loss": 0.6018,
+      "step": 12765
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4590827351085386e-07,
+      "loss": 0.6801,
+      "step": 12766
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4545353257009373e-07,
+      "loss": 0.8721,
+      "step": 12767
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.449990128111947e-07,
+      "loss": 0.8832,
+      "step": 12768
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4454471425622925e-07,
+      "loss": 0.5088,
+      "step": 12769
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4409063692725907e-07,
+      "loss": 0.8677,
+      "step": 12770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.43636780846336e-07,
+      "loss": 0.6573,
+      "step": 12771
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4318314603550074e-07,
+      "loss": 0.8081,
+      "step": 12772
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.42729732516784e-07,
+      "loss": 0.7972,
+      "step": 12773
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.422765403122015e-07,
+      "loss": 0.8591,
+      "step": 12774
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.418235694437645e-07,
+      "loss": 0.7963,
+      "step": 12775
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4137081993347043e-07,
+      "loss": 0.6622,
+      "step": 12776
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4091829180330503e-07,
+      "loss": 0.5168,
+      "step": 12777
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.4046598507524354e-07,
+      "loss": 0.6751,
+      "step": 12778
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.400138997712544e-07,
+      "loss": 0.6759,
+      "step": 12779
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3956203591328963e-07,
+      "loss": 0.8673,
+      "step": 12780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3911039352329263e-07,
+      "loss": 0.6539,
+      "step": 12781
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.386589726231999e-07,
+      "loss": 0.565,
+      "step": 12782
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.382077732349299e-07,
+      "loss": 0.7402,
+      "step": 12783
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.377567953803957e-07,
+      "loss": 0.8208,
+      "step": 12784
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.373060390814976e-07,
+      "loss": 0.7857,
+      "step": 12785
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.368555043601269e-07,
+      "loss": 0.7764,
+      "step": 12786
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3640519123816115e-07,
+      "loss": 0.5161,
+      "step": 12787
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.359550997374701e-07,
+      "loss": 0.7206,
+      "step": 12788
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.355052298799112e-07,
+      "loss": 0.7988,
+      "step": 12789
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3505558168733086e-07,
+      "loss": 0.5349,
+      "step": 12790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3460615518156614e-07,
+      "loss": 0.648,
+      "step": 12791
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.341569503844423e-07,
+      "loss": 0.5859,
+      "step": 12792
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.33707967317773e-07,
+      "loss": 0.8576,
+      "step": 12793
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3325920600336355e-07,
+      "loss": 0.6496,
+      "step": 12794
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3281066646300605e-07,
+      "loss": 0.6722,
+      "step": 12795
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.323623487184847e-07,
+      "loss": 0.4373,
+      "step": 12796
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.319142527915676e-07,
+      "loss": 0.7628,
+      "step": 12797
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.314663787040185e-07,
+      "loss": 0.5064,
+      "step": 12798
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.310187264775878e-07,
+      "loss": 0.7409,
+      "step": 12799
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3057129613401137e-07,
+      "loss": 0.7845,
+      "step": 12800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.3012408769502134e-07,
+      "loss": 0.614,
+      "step": 12801
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2967710118233473e-07,
+      "loss": 0.6149,
+      "step": 12802
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.292303366176559e-07,
+      "loss": 0.4256,
+      "step": 12803
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2878379402268466e-07,
+      "loss": 0.5837,
+      "step": 12804
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.283374734191037e-07,
+      "loss": 0.8658,
+      "step": 12805
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.278913748285879e-07,
+      "loss": 0.63,
+      "step": 12806
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.274454982728032e-07,
+      "loss": 0.73,
+      "step": 12807
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2699984377340076e-07,
+      "loss": 0.7202,
+      "step": 12808
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.26554411352022e-07,
+      "loss": 1.0115,
+      "step": 12809
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.26109201030302e-07,
+      "loss": 0.4448,
+      "step": 12810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2566421282985824e-07,
+      "loss": 1.0013,
+      "step": 12811
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2521944677230087e-07,
+      "loss": 0.5674,
+      "step": 12812
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2477490287923026e-07,
+      "loss": 0.7345,
+      "step": 12813
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2433058117223423e-07,
+      "loss": 0.5353,
+      "step": 12814
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.238864816728899e-07,
+      "loss": 0.5801,
+      "step": 12815
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2344260440276455e-07,
+      "loss": 0.5467,
+      "step": 12816
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2299894938341355e-07,
+      "loss": 1.0424,
+      "step": 12817
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.225555166363832e-07,
+      "loss": 0.7113,
+      "step": 12818
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.221123061832066e-07,
+      "loss": 0.591,
+      "step": 12819
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.216693180454079e-07,
+      "loss": 0.9051,
+      "step": 12820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2122655224450083e-07,
+      "loss": 0.7651,
+      "step": 12821
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.207840088019843e-07,
+      "loss": 0.7972,
+      "step": 12822
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.2034168773935293e-07,
+      "loss": 0.5612,
+      "step": 12823
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.198995890780866e-07,
+      "loss": 0.7066,
+      "step": 12824
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.194577128396521e-07,
+      "loss": 0.9657,
+      "step": 12825
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1901605904551055e-07,
+      "loss": 0.7053,
+      "step": 12826
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.18574627717111e-07,
+      "loss": 0.8138,
+      "step": 12827
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1813341887588735e-07,
+      "loss": 0.7757,
+      "step": 12828
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.17692432543268e-07,
+      "loss": 0.7972,
+      "step": 12829
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1725166874066967e-07,
+      "loss": 0.7308,
+      "step": 12830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.168111274894937e-07,
+      "loss": 0.8222,
+      "step": 12831
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.163708088111379e-07,
+      "loss": 0.6345,
+      "step": 12832
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1593071272698295e-07,
+      "loss": 0.8366,
+      "step": 12833
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1549083925840174e-07,
+      "loss": 0.968,
+      "step": 12834
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.150511884267555e-07,
+      "loss": 0.6125,
+      "step": 12835
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.146117602533956e-07,
+      "loss": 0.6677,
+      "step": 12836
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1417255475966145e-07,
+      "loss": 0.7686,
+      "step": 12837
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1373357196688226e-07,
+      "loss": 0.8218,
+      "step": 12838
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1329481189637643e-07,
+      "loss": 0.6173,
+      "step": 12839
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1285627456945087e-07,
+      "loss": 0.6248,
+      "step": 12840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1241796000740296e-07,
+      "loss": 0.4861,
+      "step": 12841
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1197986823151847e-07,
+      "loss": 0.5893,
+      "step": 12842
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.11541999263072e-07,
+      "loss": 0.5959,
+      "step": 12843
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1110435312332777e-07,
+      "loss": 0.9248,
+      "step": 12844
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1066692983353917e-07,
+      "loss": 0.648,
+      "step": 12845
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.1022972941494875e-07,
+      "loss": 1.0316,
+      "step": 12846
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0979275188878843e-07,
+      "loss": 0.5926,
+      "step": 12847
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0935599727627837e-07,
+      "loss": 0.7408,
+      "step": 12848
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.089194655986306e-07,
+      "loss": 0.7755,
+      "step": 12849
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0848315687704087e-07,
+      "loss": 0.8907,
+      "step": 12850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.080470711327011e-07,
+      "loss": 0.7235,
+      "step": 12851
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.076112083867878e-07,
+      "loss": 0.6641,
+      "step": 12852
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0717556866046615e-07,
+      "loss": 0.8022,
+      "step": 12853
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.067401519748931e-07,
+      "loss": 0.5533,
+      "step": 12854
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.063049583512152e-07,
+      "loss": 0.8842,
+      "step": 12855
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.058699878105643e-07,
+      "loss": 0.8317,
+      "step": 12856
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.054352403740641e-07,
+      "loss": 0.476,
+      "step": 12857
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0500071606283e-07,
+      "loss": 0.9291,
+      "step": 12858
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.045664148979606e-07,
+      "loss": 0.9615,
+      "step": 12859
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.0413233690054786e-07,
+      "loss": 0.7793,
+      "step": 12860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.036984820916723e-07,
+      "loss": 0.8821,
+      "step": 12861
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.032648504924025e-07,
+      "loss": 0.6398,
+      "step": 12862
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.028314421237972e-07,
+      "loss": 0.7477,
+      "step": 12863
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0239825700690404e-07,
+      "loss": 0.3494,
+      "step": 12864
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.019652951627595e-07,
+      "loss": 0.4136,
+      "step": 12865
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.0153255661238955e-07,
+      "loss": 0.6003,
+      "step": 12866
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.011000413768096e-07,
+      "loss": 0.787,
+      "step": 12867
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.006677494770228e-07,
+      "loss": 0.8959,
+      "step": 12868
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.002356809340241e-07,
+      "loss": 0.5703,
+      "step": 12869
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.998038357687933e-07,
+      "loss": 0.6218,
+      "step": 12870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.993722140023049e-07,
+      "loss": 0.9051,
+      "step": 12871
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.989408156555197e-07,
+      "loss": 0.6732,
+      "step": 12872
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.985096407493838e-07,
+      "loss": 0.7106,
+      "step": 12873
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.980786893048405e-07,
+      "loss": 0.7909,
+      "step": 12874
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9764796134281745e-07,
+      "loss": 0.7535,
+      "step": 12875
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9721745688422906e-07,
+      "loss": 0.5905,
+      "step": 12876
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9678717594998527e-07,
+      "loss": 0.514,
+      "step": 12877
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9635711856098104e-07,
+      "loss": 0.627,
+      "step": 12878
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.959272847380985e-07,
+      "loss": 0.4113,
+      "step": 12879
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.95497674502216e-07,
+      "loss": 0.9136,
+      "step": 12880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9506828787419294e-07,
+      "loss": 0.509,
+      "step": 12881
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.946391248748821e-07,
+      "loss": 0.7461,
+      "step": 12882
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.942101855251279e-07,
+      "loss": 0.498,
+      "step": 12883
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.937814698457576e-07,
+      "loss": 0.8658,
+      "step": 12884
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9335297785759064e-07,
+      "loss": 0.5198,
+      "step": 12885
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9292470958143924e-07,
+      "loss": 0.8064,
+      "step": 12886
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9249666503809846e-07,
+      "loss": 0.573,
+      "step": 12887
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9206884424835613e-07,
+      "loss": 0.7529,
+      "step": 12888
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.916412472329884e-07,
+      "loss": 0.876,
+      "step": 12889
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.912138740127608e-07,
+      "loss": 0.7455,
+      "step": 12890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9078672460842736e-07,
+      "loss": 0.7752,
+      "step": 12891
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.903597990407326e-07,
+      "loss": 0.6712,
+      "step": 12892
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.899330973304083e-07,
+      "loss": 0.8103,
+      "step": 12893
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8950661949817614e-07,
+      "loss": 0.8168,
+      "step": 12894
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.890803655647485e-07,
+      "loss": 0.8184,
+      "step": 12895
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8865433555082434e-07,
+      "loss": 0.9115,
+      "step": 12896
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.882285294770938e-07,
+      "loss": 0.7391,
+      "step": 12897
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.878029473642331e-07,
+      "loss": 1.014,
+      "step": 12898
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8737758923291236e-07,
+      "loss": 0.5892,
+      "step": 12899
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8695245510378787e-07,
+      "loss": 0.5153,
+      "step": 12900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8652754499750257e-07,
+      "loss": 0.6409,
+      "step": 12901
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8610285893469433e-07,
+      "loss": 0.8076,
+      "step": 12902
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.856783969359873e-07,
+      "loss": 0.8824,
+      "step": 12903
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.852541590219916e-07,
+      "loss": 0.7139,
+      "step": 12904
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8483014521331184e-07,
+      "loss": 0.7444,
+      "step": 12905
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8440635553053993e-07,
+      "loss": 0.8538,
+      "step": 12906
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.839827899942544e-07,
+      "loss": 0.7165,
+      "step": 12907
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.835594486250255e-07,
+      "loss": 0.7764,
+      "step": 12908
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8313633144341164e-07,
+      "loss": 0.7176,
+      "step": 12909
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8271343846996156e-07,
+      "loss": 0.8427,
+      "step": 12910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.82290769725211e-07,
+      "loss": 0.5621,
+      "step": 12911
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.818683252296868e-07,
+      "loss": 0.6391,
+      "step": 12912
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.814461050039031e-07,
+      "loss": 0.7792,
+      "step": 12913
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.810241090683653e-07,
+      "loss": 0.767,
+      "step": 12914
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8060233744356634e-07,
+      "loss": 0.8314,
+      "step": 12915
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.8018079014998874e-07,
+      "loss": 0.8955,
+      "step": 12916
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7975946720810343e-07,
+      "loss": 0.4343,
+      "step": 12917
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.793383686383717e-07,
+      "loss": 0.6085,
+      "step": 12918
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.789174944612434e-07,
+      "loss": 0.5199,
+      "step": 12919
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7849684469715655e-07,
+      "loss": 0.8864,
+      "step": 12920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7807641936653984e-07,
+      "loss": 0.6153,
+      "step": 12921
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7765621848981025e-07,
+      "loss": 0.8784,
+      "step": 12922
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7723624208737376e-07,
+      "loss": 0.7141,
+      "step": 12923
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7681649017962565e-07,
+      "loss": 0.6447,
+      "step": 12924
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7639696278695126e-07,
+      "loss": 0.6411,
+      "step": 12925
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7597765992972267e-07,
+      "loss": 0.755,
+      "step": 12926
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.755585816283014e-07,
+      "loss": 0.811,
+      "step": 12927
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7513972790304276e-07,
+      "loss": 0.6005,
+      "step": 12928
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7472109877428443e-07,
+      "loss": 0.6763,
+      "step": 12929
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.743026942623557e-07,
+      "loss": 0.623,
+      "step": 12930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.738845143875791e-07,
+      "loss": 0.6484,
+      "step": 12931
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7346655917025964e-07,
+      "loss": 0.832,
+      "step": 12932
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.730488286306944e-07,
+      "loss": 0.562,
+      "step": 12933
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.72631322789172e-07,
+      "loss": 0.9071,
+      "step": 12934
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.722140416659653e-07,
+      "loss": 0.738,
+      "step": 12935
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7179698528133967e-07,
+      "loss": 1.1318,
+      "step": 12936
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7138015365554834e-07,
+      "loss": 0.5733,
+      "step": 12937
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7096354680883407e-07,
+      "loss": 0.8422,
+      "step": 12938
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.705471647614284e-07,
+      "loss": 0.6842,
+      "step": 12939
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7013100753355123e-07,
+      "loss": 0.8181,
+      "step": 12940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.697150751454137e-07,
+      "loss": 0.9573,
+      "step": 12941
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.692993676172141e-07,
+      "loss": 0.6585,
+      "step": 12942
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6888388496914065e-07,
+      "loss": 0.7738,
+      "step": 12943
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.684686272213694e-07,
+      "loss": 0.704,
+      "step": 12944
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.680535943940688e-07,
+      "loss": 0.5704,
+      "step": 12945
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6763878650738996e-07,
+      "loss": 0.7593,
+      "step": 12946
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6722420358148105e-07,
+      "loss": 0.5984,
+      "step": 12947
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.668098456364744e-07,
+      "loss": 0.4691,
+      "step": 12948
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.663957126924905e-07,
+      "loss": 0.4529,
+      "step": 12949
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6598180476964286e-07,
+      "loss": 0.7376,
+      "step": 12950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.655681218880325e-07,
+      "loss": 0.6809,
+      "step": 12951
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6515466406774616e-07,
+      "loss": 0.6548,
+      "step": 12952
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6474143132886607e-07,
+      "loss": 0.8732,
+      "step": 12953
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.643284236914579e-07,
+      "loss": 0.7536,
+      "step": 12954
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6391564117557787e-07,
+      "loss": 0.856,
+      "step": 12955
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6350308380127433e-07,
+      "loss": 0.823,
+      "step": 12956
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.630907515885801e-07,
+      "loss": 0.835,
+      "step": 12957
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.626786445575192e-07,
+      "loss": 0.8248,
+      "step": 12958
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6226676272810733e-07,
+      "loss": 0.4211,
+      "step": 12959
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6185510612034403e-07,
+      "loss": 0.6475,
+      "step": 12960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.614436747542205e-07,
+      "loss": 0.8239,
+      "step": 12961
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6103246864971906e-07,
+      "loss": 0.5931,
+      "step": 12962
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.606214878268077e-07,
+      "loss": 0.7409,
+      "step": 12963
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.6021073230544434e-07,
+      "loss": 0.7796,
+      "step": 12964
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5980020210557744e-07,
+      "loss": 0.9876,
+      "step": 12965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5938989724714324e-07,
+      "loss": 0.7318,
+      "step": 12966
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5897981775006697e-07,
+      "loss": 0.7502,
+      "step": 12967
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5856996363426376e-07,
+      "loss": 0.5852,
+      "step": 12968
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.581603349196372e-07,
+      "loss": 0.6382,
+      "step": 12969
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5775093162607966e-07,
+      "loss": 0.5847,
+      "step": 12970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.573417537734736e-07,
+      "loss": 0.8321,
+      "step": 12971
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.569328013816892e-07,
+      "loss": 0.9778,
+      "step": 12972
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5652407447058723e-07,
+      "loss": 0.6868,
+      "step": 12973
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5611557306001464e-07,
+      "loss": 0.8569,
+      "step": 12974
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5570729716981167e-07,
+      "loss": 0.6499,
+      "step": 12975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.552992468198058e-07,
+      "loss": 0.5866,
+      "step": 12976
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5489142202981e-07,
+      "loss": 0.6515,
+      "step": 12977
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.544838228196318e-07,
+      "loss": 0.579,
+      "step": 12978
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5407644920906593e-07,
+      "loss": 0.7411,
+      "step": 12979
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.536693012178938e-07,
+      "loss": 0.5894,
+      "step": 12980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5326237886588734e-07,
+      "loss": 0.8052,
+      "step": 12981
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5285568217281077e-07,
+      "loss": 0.5657,
+      "step": 12982
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.524492111584121e-07,
+      "loss": 0.7181,
+      "step": 12983
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5204296584243126e-07,
+      "loss": 0.5458,
+      "step": 12984
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.516369462445968e-07,
+      "loss": 0.5681,
+      "step": 12985
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.512311523846257e-07,
+      "loss": 0.3539,
+      "step": 12986
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5082558428222555e-07,
+      "loss": 0.8791,
+      "step": 12987
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.504202419570907e-07,
+      "loss": 0.6638,
+      "step": 12988
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.5001512542890636e-07,
+      "loss": 0.675,
+      "step": 12989
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4961023471734635e-07,
+      "loss": 0.6996,
+      "step": 12990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.492055698420732e-07,
+      "loss": 0.7416,
+      "step": 12991
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4880113082273795e-07,
+      "loss": 0.8403,
+      "step": 12992
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.48396917678982e-07,
+      "loss": 0.7855,
+      "step": 12993
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4799293043043527e-07,
+      "loss": 0.6633,
+      "step": 12994
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.475891690967159e-07,
+      "loss": 0.8757,
+      "step": 12995
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4718563369743217e-07,
+      "loss": 0.8271,
+      "step": 12996
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.467823242521806e-07,
+      "loss": 0.6682,
+      "step": 12997
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4637924078054717e-07,
+      "loss": 0.8014,
+      "step": 12998
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4597638330210795e-07,
+      "loss": 0.4032,
+      "step": 12999
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4557375183642385e-07,
+      "loss": 0.5618,
+      "step": 13000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4517134640305097e-07,
+      "loss": 0.6897,
+      "step": 13001
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4476916702152976e-07,
+      "loss": 0.9243,
+      "step": 13002
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4436721371139016e-07,
+      "loss": 0.7174,
+      "step": 13003
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.4396548649215543e-07,
+      "loss": 0.6559,
+      "step": 13004
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.435639853833317e-07,
+      "loss": 0.4508,
+      "step": 13005
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.431627104044172e-07,
+      "loss": 0.6957,
+      "step": 13006
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.427616615749013e-07,
+      "loss": 0.687,
+      "step": 13007
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.423608389142574e-07,
+      "loss": 1.0737,
+      "step": 13008
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4196024244195157e-07,
+      "loss": 0.9445,
+      "step": 13009
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.415598721774388e-07,
+      "loss": 0.6102,
+      "step": 13010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.411597281401613e-07,
+      "loss": 0.6424,
+      "step": 13011
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.407598103495513e-07,
+      "loss": 0.7937,
+      "step": 13012
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.4036011882502996e-07,
+      "loss": 0.6366,
+      "step": 13013
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.399606535860078e-07,
+      "loss": 0.6276,
+      "step": 13014
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.395614146518833e-07,
+      "loss": 0.7916,
+      "step": 13015
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.391624020420453e-07,
+      "loss": 0.7811,
+      "step": 13016
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3876361577587115e-07,
+      "loss": 0.6235,
+      "step": 13017
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.383650558727264e-07,
+      "loss": 0.6746,
+      "step": 13018
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.379667223519667e-07,
+      "loss": 0.6475,
+      "step": 13019
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3756861523293605e-07,
+      "loss": 0.7028,
+      "step": 13020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.371707345349684e-07,
+      "loss": 0.7543,
+      "step": 13021
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.367730802773839e-07,
+      "loss": 0.6883,
+      "step": 13022
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.363756524794959e-07,
+      "loss": 0.9803,
+      "step": 13023
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3597845116060455e-07,
+      "loss": 0.6539,
+      "step": 13024
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.355814763399973e-07,
+      "loss": 0.7815,
+      "step": 13025
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.351847280369541e-07,
+      "loss": 0.6946,
+      "step": 13026
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3478820627074307e-07,
+      "loss": 1.0615,
+      "step": 13027
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.34391911060617e-07,
+      "loss": 0.7332,
+      "step": 13028
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3399584242582496e-07,
+      "loss": 0.8752,
+      "step": 13029
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3360000038559827e-07,
+      "loss": 0.4685,
+      "step": 13030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3320438495916043e-07,
+      "loss": 1.0494,
+      "step": 13031
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.328089961657266e-07,
+      "loss": 0.4801,
+      "step": 13032
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.324138340244948e-07,
+      "loss": 0.7059,
+      "step": 13033
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.320188985546552e-07,
+      "loss": 0.8782,
+      "step": 13034
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.316241897753902e-07,
+      "loss": 0.5053,
+      "step": 13035
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3122970770586514e-07,
+      "loss": 0.8382,
+      "step": 13036
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3083545236523794e-07,
+      "loss": 0.6106,
+      "step": 13037
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.30441423772655e-07,
+      "loss": 0.6771,
+      "step": 13038
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.3004762194725094e-07,
+      "loss": 0.8434,
+      "step": 13039
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.296540469081505e-07,
+      "loss": 0.6844,
+      "step": 13040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2926069867446673e-07,
+      "loss": 0.6826,
+      "step": 13041
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2886757726530163e-07,
+      "loss": 0.7738,
+      "step": 13042
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.284746826997459e-07,
+      "loss": 0.7775,
+      "step": 13043
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2808201499687997e-07,
+      "loss": 0.7859,
+      "step": 13044
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.276895741757735e-07,
+      "loss": 0.779,
+      "step": 13045
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2729736025548354e-07,
+      "loss": 0.509,
+      "step": 13046
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.269053732550581e-07,
+      "loss": 0.7001,
+      "step": 13047
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2651361319353205e-07,
+      "loss": 0.7214,
+      "step": 13048
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.261220800899323e-07,
+      "loss": 0.7229,
+      "step": 13049
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.257307739632698e-07,
+      "loss": 0.7298,
+      "step": 13050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.253396948325499e-07,
+      "loss": 0.6517,
+      "step": 13051
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.249488427167646e-07,
+      "loss": 0.7002,
+      "step": 13052
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.245582176348927e-07,
+      "loss": 0.9287,
+      "step": 13053
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.241678196059067e-07,
+      "loss": 0.705,
+      "step": 13054
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.237776486487643e-07,
+      "loss": 0.8328,
+      "step": 13055
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2338770478241245e-07,
+      "loss": 0.8911,
+      "step": 13056
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.229979880257883e-07,
+      "loss": 0.5455,
+      "step": 13057
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.226084983978195e-07,
+      "loss": 0.7869,
+      "step": 13058
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.222192359174181e-07,
+      "loss": 0.7935,
+      "step": 13059
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.218302006034896e-07,
+      "loss": 0.7479,
+      "step": 13060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.21441392474926e-07,
+      "loss": 0.5459,
+      "step": 13061
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2105281155060843e-07,
+      "loss": 0.7207,
+      "step": 13062
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.206644578494089e-07,
+      "loss": 0.7264,
+      "step": 13063
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.2027633139018577e-07,
+      "loss": 0.7878,
+      "step": 13064
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1988843219178776e-07,
+      "loss": 0.5891,
+      "step": 13065
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.195007602730532e-07,
+      "loss": 0.6546,
+      "step": 13066
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1911331565280756e-07,
+      "loss": 0.7379,
+      "step": 13067
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1872609834986635e-07,
+      "loss": 0.7904,
+      "step": 13068
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.183391083830345e-07,
+      "loss": 0.919,
+      "step": 13069
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1795234577110533e-07,
+      "loss": 0.6093,
+      "step": 13070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1756581053286107e-07,
+      "loss": 0.7784,
+      "step": 13071
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.171795026870722e-07,
+      "loss": 0.7093,
+      "step": 13072
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.167934222524999e-07,
+      "loss": 1.0036,
+      "step": 13073
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1640756924789296e-07,
+      "loss": 0.7591,
+      "step": 13074
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.160219436919909e-07,
+      "loss": 0.6977,
+      "step": 13075
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1563654560351706e-07,
+      "loss": 0.6816,
+      "step": 13076
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1525137500119207e-07,
+      "loss": 0.8427,
+      "step": 13077
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.148664319037176e-07,
+      "loss": 0.5896,
+      "step": 13078
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1448171632978807e-07,
+      "loss": 0.7578,
+      "step": 13079
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.140972282980886e-07,
+      "loss": 0.8494,
+      "step": 13080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1371296782728875e-07,
+      "loss": 0.7706,
+      "step": 13081
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.133289349360491e-07,
+      "loss": 0.6369,
+      "step": 13082
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.129451296430219e-07,
+      "loss": 0.7538,
+      "step": 13083
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.125615519668435e-07,
+      "loss": 0.5599,
+      "step": 13084
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1217820192614156e-07,
+      "loss": 0.4106,
+      "step": 13085
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1179507953953523e-07,
+      "loss": 0.5898,
+      "step": 13086
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.114121848256268e-07,
+      "loss": 0.9624,
+      "step": 13087
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.110295178030126e-07,
+      "loss": 0.7573,
+      "step": 13088
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.106470784902754e-07,
+      "loss": 0.8766,
+      "step": 13089
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.1026486690598765e-07,
+      "loss": 0.7937,
+      "step": 13090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0988288306871115e-07,
+      "loss": 0.8047,
+      "step": 13091
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0950112699699554e-07,
+      "loss": 0.5903,
+      "step": 13092
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.091195987093798e-07,
+      "loss": 0.4694,
+      "step": 13093
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.087382982243925e-07,
+      "loss": 0.7528,
+      "step": 13094
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0835722556055103e-07,
+      "loss": 0.7423,
+      "step": 13095
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.079763807363606e-07,
+      "loss": 0.8038,
+      "step": 13096
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0759576377031697e-07,
+      "loss": 0.7907,
+      "step": 13097
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.07215374680902e-07,
+      "loss": 0.5452,
+      "step": 13098
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0683521348659095e-07,
+      "loss": 0.5533,
+      "step": 13099
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.064552802058457e-07,
+      "loss": 0.4827,
+      "step": 13100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.060755748571137e-07,
+      "loss": 0.7385,
+      "step": 13101
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0569609745883745e-07,
+      "loss": 0.7559,
+      "step": 13102
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0531684802944497e-07,
+      "loss": 0.7676,
+      "step": 13103
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0493782658735205e-07,
+      "loss": 0.788,
+      "step": 13104
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.045590331509679e-07,
+      "loss": 0.6195,
+      "step": 13105
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.041804677386856e-07,
+      "loss": 0.7953,
+      "step": 13106
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0380213036888926e-07,
+      "loss": 0.7261,
+      "step": 13107
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.034240210599543e-07,
+      "loss": 0.7456,
+      "step": 13108
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0304613983024033e-07,
+      "loss": 0.8856,
+      "step": 13109
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0266848669809947e-07,
+      "loss": 0.5583,
+      "step": 13110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.022910616818714e-07,
+      "loss": 0.6494,
+      "step": 13111
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0191386479988537e-07,
+      "loss": 0.7161,
+      "step": 13112
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.015368960704584e-07,
+      "loss": 0.744,
+      "step": 13113
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.011601555118976e-07,
+      "loss": 0.8155,
+      "step": 13114
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0078364314249874e-07,
+      "loss": 0.626,
+      "step": 13115
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0040735898054617e-07,
+      "loss": 1.0869,
+      "step": 13116
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.0003130304431304e-07,
+      "loss": 0.5569,
+      "step": 13117
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9965547535206254e-07,
+      "loss": 0.5347,
+      "step": 13118
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.99279875922045e-07,
+      "loss": 0.7735,
+      "step": 13119
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9890450477250087e-07,
+      "loss": 0.5673,
+      "step": 13120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.985293619216595e-07,
+      "loss": 0.7915,
+      "step": 13121
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.98154447387739e-07,
+      "loss": 0.8627,
+      "step": 13122
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9777976118894604e-07,
+      "loss": 0.7315,
+      "step": 13123
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9740530334347596e-07,
+      "loss": 0.4263,
+      "step": 13124
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9703107386951535e-07,
+      "loss": 0.778,
+      "step": 13125
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9665707278523525e-07,
+      "loss": 0.5991,
+      "step": 13126
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.962833001088e-07,
+      "loss": 0.7569,
+      "step": 13127
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9590975585836124e-07,
+      "loss": 0.6806,
+      "step": 13128
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.955364400520583e-07,
+      "loss": 0.7741,
+      "step": 13129
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.951633527080194e-07,
+      "loss": 0.7121,
+      "step": 13130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.947904938443663e-07,
+      "loss": 0.995,
+      "step": 13131
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.944178634792033e-07,
+      "loss": 0.5698,
+      "step": 13132
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.940454616306265e-07,
+      "loss": 0.8007,
+      "step": 13133
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.936732883167226e-07,
+      "loss": 0.5492,
+      "step": 13134
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9330134355556374e-07,
+      "loss": 0.453,
+      "step": 13135
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.929296273652132e-07,
+      "loss": 0.5275,
+      "step": 13136
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9255813976372227e-07,
+      "loss": 1.0267,
+      "step": 13137
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9218688076913195e-07,
+      "loss": 0.6707,
+      "step": 13138
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.918158503994711e-07,
+      "loss": 1.0622,
+      "step": 13139
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9144504867275826e-07,
+      "loss": 0.5951,
+      "step": 13140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.910744756070011e-07,
+      "loss": 0.8118,
+      "step": 13141
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9070413122019525e-07,
+      "loss": 0.9212,
+      "step": 13142
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9033401553032525e-07,
+      "loss": 0.7989,
+      "step": 13143
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.899641285553656e-07,
+      "loss": 0.8082,
+      "step": 13144
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8959447031327916e-07,
+      "loss": 0.8415,
+      "step": 13145
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8922504082201764e-07,
+      "loss": 0.8283,
+      "step": 13146
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.888558400995212e-07,
+      "loss": 0.9852,
+      "step": 13147
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8848686816371985e-07,
+      "loss": 0.7878,
+      "step": 13148
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.881181250325299e-07,
+      "loss": 0.9484,
+      "step": 13149
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.877496107238614e-07,
+      "loss": 0.5564,
+      "step": 13150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8738132525560957e-07,
+      "loss": 0.7766,
+      "step": 13151
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.870132686456578e-07,
+      "loss": 0.7496,
+      "step": 13152
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8664544091188297e-07,
+      "loss": 0.8213,
+      "step": 13153
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.8627784207214513e-07,
+      "loss": 0.5334,
+      "step": 13154
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8591047214429626e-07,
+      "loss": 0.9261,
+      "step": 13155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.855433311461786e-07,
+      "loss": 0.7432,
+      "step": 13156
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8517641909562075e-07,
+      "loss": 0.9335,
+      "step": 13157
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.848097360104396e-07,
+      "loss": 0.8071,
+      "step": 13158
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.844432819084447e-07,
+      "loss": 0.4224,
+      "step": 13159
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.840770568074308e-07,
+      "loss": 0.923,
+      "step": 13160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8371106072518194e-07,
+      "loss": 0.725,
+      "step": 13161
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.833452936794745e-07,
+      "loss": 0.5751,
+      "step": 13162
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8297975568806924e-07,
+      "loss": 1.064,
+      "step": 13163
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8261444676871755e-07,
+      "loss": 0.5886,
+      "step": 13164
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.822493669391613e-07,
+      "loss": 0.6186,
+      "step": 13165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8188451621712854e-07,
+      "loss": 0.5065,
+      "step": 13166
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.815198946203379e-07,
+      "loss": 0.8341,
+      "step": 13167
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8115550216649687e-07,
+      "loss": 0.8174,
+      "step": 13168
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.807913388733008e-07,
+      "loss": 0.8293,
+      "step": 13169
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.8042740475843487e-07,
+      "loss": 0.3895,
+      "step": 13170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.800636998395723e-07,
+      "loss": 0.7223,
+      "step": 13171
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.797002241343766e-07,
+      "loss": 0.7986,
+      "step": 13172
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.793369776604987e-07,
+      "loss": 0.5397,
+      "step": 13173
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.789739604355779e-07,
+      "loss": 0.9569,
+      "step": 13174
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7861117247724446e-07,
+      "loss": 0.9012,
+      "step": 13175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.782486138031171e-07,
+      "loss": 0.6391,
+      "step": 13176
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7788628443080003e-07,
+      "loss": 0.7522,
+      "step": 13177
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.77524184377892e-07,
+      "loss": 0.8196,
+      "step": 13178
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7716231366197663e-07,
+      "loss": 0.7407,
+      "step": 13179
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.768006723006261e-07,
+      "loss": 1.1548,
+      "step": 13180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.764392603114052e-07,
+      "loss": 0.761,
+      "step": 13181
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7607807771186325e-07,
+      "loss": 0.5966,
+      "step": 13182
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7571712451953944e-07,
+      "loss": 0.535,
+      "step": 13183
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.75356400751966e-07,
+      "loss": 0.5595,
+      "step": 13184
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7499590642665773e-07,
+      "loss": 0.6272,
+      "step": 13185
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.746356415611229e-07,
+      "loss": 1.0276,
+      "step": 13186
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7427560617285575e-07,
+      "loss": 0.7919,
+      "step": 13187
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7391580027934185e-07,
+      "loss": 0.5648,
+      "step": 13188
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.735562238980538e-07,
+      "loss": 0.745,
+      "step": 13189
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7319687704645323e-07,
+      "loss": 0.922,
+      "step": 13190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.728377597419918e-07,
+      "loss": 0.8213,
+      "step": 13191
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.724788720021093e-07,
+      "loss": 0.4475,
+      "step": 13192
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7212021384423415e-07,
+      "loss": 0.6428,
+      "step": 13193
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7176178528578347e-07,
+      "loss": 0.7204,
+      "step": 13194
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.714035863441639e-07,
+      "loss": 0.5689,
+      "step": 13195
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7104561703677035e-07,
+      "loss": 0.6784,
+      "step": 13196
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7068787738098735e-07,
+      "loss": 0.5903,
+      "step": 13197
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.703303673941876e-07,
+      "loss": 0.6029,
+      "step": 13198
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6997308709373225e-07,
+      "loss": 0.8663,
+      "step": 13199
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.696160364969724e-07,
+      "loss": 0.4824,
+      "step": 13200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6925921562124867e-07,
+      "loss": 0.8226,
+      "step": 13201
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6890262448388607e-07,
+      "loss": 0.5293,
+      "step": 13202
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.68546263102204e-07,
+      "loss": 0.5526,
+      "step": 13203
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6819013149350926e-07,
+      "loss": 0.7537,
+      "step": 13204
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6783422967509465e-07,
+      "loss": 0.7214,
+      "step": 13205
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6747855766424357e-07,
+      "loss": 0.9985,
+      "step": 13206
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6712311547823056e-07,
+      "loss": 0.8989,
+      "step": 13207
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6676790313431567e-07,
+      "loss": 0.5326,
+      "step": 13208
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.664129206497479e-07,
+      "loss": 0.7618,
+      "step": 13209
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.66058168041769e-07,
+      "loss": 0.8814,
+      "step": 13210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.657036453276046e-07,
+      "loss": 0.6874,
+      "step": 13211
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.653493525244721e-07,
+      "loss": 0.6398,
+      "step": 13212
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6499528964957666e-07,
+      "loss": 0.5265,
+      "step": 13213
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6464145672011274e-07,
+      "loss": 0.3745,
+      "step": 13214
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6428785375326337e-07,
+      "loss": 0.7337,
+      "step": 13215
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6393448076620086e-07,
+      "loss": 0.6974,
+      "step": 13216
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6358133777608596e-07,
+      "loss": 0.4668,
+      "step": 13217
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6322842480006773e-07,
+      "loss": 0.7146,
+      "step": 13218
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.628757418552852e-07,
+      "loss": 0.6958,
+      "step": 13219
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.625232889588658e-07,
+      "loss": 0.6115,
+      "step": 13220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.621710661279253e-07,
+      "loss": 0.4446,
+      "step": 13221
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6181907337956835e-07,
+      "loss": 0.934,
+      "step": 13222
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.614673107308896e-07,
+      "loss": 0.4491,
+      "step": 13223
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6111577819897205e-07,
+      "loss": 0.5194,
+      "step": 13224
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6076447580088426e-07,
+      "loss": 0.6987,
+      "step": 13225
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.604134035536898e-07,
+      "loss": 0.6979,
+      "step": 13226
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.6006256147443675e-07,
+      "loss": 0.7548,
+      "step": 13227
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.597119495801614e-07,
+      "loss": 0.6363,
+      "step": 13228
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.59361567887893e-07,
+      "loss": 0.7677,
+      "step": 13229
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.590114164146451e-07,
+      "loss": 0.6213,
+      "step": 13230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5866149517742247e-07,
+      "loss": 0.3479,
+      "step": 13231
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.583118041932198e-07,
+      "loss": 0.8405,
+      "step": 13232
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.579623434790174e-07,
+      "loss": 1.0128,
+      "step": 13233
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5761311305178617e-07,
+      "loss": 0.866,
+      "step": 13234
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5726411292848697e-07,
+      "loss": 0.7036,
+      "step": 13235
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.569153431260668e-07,
+      "loss": 0.584,
+      "step": 13236
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5656680366146325e-07,
+      "loss": 0.6395,
+      "step": 13237
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5621849455160383e-07,
+      "loss": 0.5154,
+      "step": 13238
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5587041581340235e-07,
+      "loss": 0.8649,
+      "step": 13239
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.555225674637618e-07,
+      "loss": 0.7126,
+      "step": 13240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5517494951957544e-07,
+      "loss": 0.9619,
+      "step": 13241
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.548275619977247e-07,
+      "loss": 0.7858,
+      "step": 13242
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.544804049150795e-07,
+      "loss": 0.5936,
+      "step": 13243
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.541334782884991e-07,
+      "loss": 0.8013,
+      "step": 13244
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5378678213483057e-07,
+      "loss": 0.8676,
+      "step": 13245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5344031647091094e-07,
+      "loss": 0.8651,
+      "step": 13246
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.530940813135657e-07,
+      "loss": 1.104,
+      "step": 13247
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.527480766796092e-07,
+      "loss": 0.899,
+      "step": 13248
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.524023025858446e-07,
+      "loss": 0.3703,
+      "step": 13249
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5205675904906136e-07,
+      "loss": 0.5862,
+      "step": 13250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.5171144608604317e-07,
+      "loss": 0.6252,
+      "step": 13251
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.513663637135583e-07,
+      "loss": 0.6386,
+      "step": 13252
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.510215119483639e-07,
+      "loss": 0.974,
+      "step": 13253
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.506768908072077e-07,
+      "loss": 0.505,
+      "step": 13254
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.503325003068269e-07,
+      "loss": 0.8737,
+      "step": 13255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.499883404639436e-07,
+      "loss": 0.8982,
+      "step": 13256
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4964441129527337e-07,
+      "loss": 0.6027,
+      "step": 13257
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.493007128175173e-07,
+      "loss": 0.7494,
+      "step": 13258
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4895724504736596e-07,
+      "loss": 0.9492,
+      "step": 13259
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.486140080014998e-07,
+      "loss": 0.6777,
+      "step": 13260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4827100169658726e-07,
+      "loss": 0.6086,
+      "step": 13261
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.47928226149286e-07,
+      "loss": 0.6365,
+      "step": 13262
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.475856813762417e-07,
+      "loss": 0.6596,
+      "step": 13263
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.472433673940894e-07,
+      "loss": 0.5761,
+      "step": 13264
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.469012842194535e-07,
+      "loss": 0.4273,
+      "step": 13265
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.465594318689457e-07,
+      "loss": 0.5389,
+      "step": 13266
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.462178103591678e-07,
+      "loss": 0.8897,
+      "step": 13267
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.458764197067093e-07,
+      "loss": 0.777,
+      "step": 13268
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.455352599281502e-07,
+      "loss": 0.4507,
+      "step": 13269
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.451943310400573e-07,
+      "loss": 0.5431,
+      "step": 13270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4485363305898734e-07,
+      "loss": 0.8459,
+      "step": 13271
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4451316600148544e-07,
+      "loss": 0.727,
+      "step": 13272
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.441729298840861e-07,
+      "loss": 0.8829,
+      "step": 13273
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4383292472331177e-07,
+      "loss": 0.8699,
+      "step": 13274
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4349315053567466e-07,
+      "loss": 0.3729,
+      "step": 13275
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4315360733767446e-07,
+      "loss": 0.6122,
+      "step": 13276
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.428142951458012e-07,
+      "loss": 0.8563,
+      "step": 13277
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4247521397653183e-07,
+      "loss": 0.6552,
+      "step": 13278
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4213636384633257e-07,
+      "loss": 0.7627,
+      "step": 13279
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.417977447716613e-07,
+      "loss": 0.5094,
+      "step": 13280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4145935676896106e-07,
+      "loss": 0.7586,
+      "step": 13281
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4112119985466364e-07,
+      "loss": 0.679,
+      "step": 13282
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.407832740451932e-07,
+      "loss": 0.6743,
+      "step": 13283
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4044557935695934e-07,
+      "loss": 0.6646,
+      "step": 13284
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.4010811580636006e-07,
+      "loss": 0.818,
+      "step": 13285
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.397708834097867e-07,
+      "loss": 0.724,
+      "step": 13286
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3943388218361395e-07,
+      "loss": 0.7956,
+      "step": 13287
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.390971121442082e-07,
+      "loss": 0.8691,
+      "step": 13288
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3876057330792344e-07,
+      "loss": 0.7505,
+      "step": 13289
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.38424265691104e-07,
+      "loss": 0.6237,
+      "step": 13290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3808818931008116e-07,
+      "loss": 0.7809,
+      "step": 13291
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3775234418117577e-07,
+      "loss": 0.9076,
+      "step": 13292
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3741673032069757e-07,
+      "loss": 0.8138,
+      "step": 13293
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3708134774494517e-07,
+      "loss": 0.7779,
+      "step": 13294
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3674619647020558e-07,
+      "loss": 0.7626,
+      "step": 13295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3641127651275463e-07,
+      "loss": 0.7267,
+      "step": 13296
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3607658788885712e-07,
+      "loss": 0.6048,
+      "step": 13297
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.357421306147656e-07,
+      "loss": 0.5792,
+      "step": 13298
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.354079047067237e-07,
+      "loss": 0.8028,
+      "step": 13299
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3507391018096236e-07,
+      "loss": 0.9089,
+      "step": 13300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.3474014705369974e-07,
+      "loss": 0.7694,
+      "step": 13301
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.344066153411456e-07,
+      "loss": 0.7684,
+      "step": 13302
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.340733150594976e-07,
+      "loss": 0.6698,
+      "step": 13303
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.337402462249394e-07,
+      "loss": 0.7605,
+      "step": 13304
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3340740885364922e-07,
+      "loss": 0.5195,
+      "step": 13305
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3307480296178798e-07,
+      "loss": 0.9237,
+      "step": 13306
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3274242856550834e-07,
+      "loss": 0.5096,
+      "step": 13307
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3241028568095292e-07,
+      "loss": 0.8372,
+      "step": 13308
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3207837432424996e-07,
+      "loss": 0.7675,
+      "step": 13309
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3174669451151766e-07,
+      "loss": 0.6982,
+      "step": 13310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.314152462588659e-07,
+      "loss": 0.6061,
+      "step": 13311
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3108402958238796e-07,
+      "loss": 0.5107,
+      "step": 13312
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3075304449816927e-07,
+      "loss": 0.5284,
+      "step": 13313
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.304222910222853e-07,
+      "loss": 0.6535,
+      "step": 13314
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.3009176917079602e-07,
+      "loss": 0.8892,
+      "step": 13315
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2976147895975408e-07,
+      "loss": 0.7835,
+      "step": 13316
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2943142040519838e-07,
+      "loss": 0.9557,
+      "step": 13317
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2910159352315774e-07,
+      "loss": 0.6349,
+      "step": 13318
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2877199832964992e-07,
+      "loss": 0.6202,
+      "step": 13319
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2844263484068097e-07,
+      "loss": 0.7238,
+      "step": 13320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2811350307224534e-07,
+      "loss": 0.8096,
+      "step": 13321
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2778460304032635e-07,
+      "loss": 0.8504,
+      "step": 13322
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2745593476089734e-07,
+      "loss": 0.6793,
+      "step": 13323
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2712749824991831e-07,
+      "loss": 0.7715,
+      "step": 13324
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2679929352334041e-07,
+      "loss": 0.732,
+      "step": 13325
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2647132059709974e-07,
+      "loss": 0.6995,
+      "step": 13326
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.261435794871264e-07,
+      "loss": 0.926,
+      "step": 13327
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2581607020933538e-07,
+      "loss": 0.6725,
+      "step": 13328
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2548879277963065e-07,
+      "loss": 0.5906,
+      "step": 13329
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.251617472139067e-07,
+      "loss": 0.6253,
+      "step": 13330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2483493352804585e-07,
+      "loss": 0.596,
+      "step": 13331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2450835173791818e-07,
+      "loss": 0.7441,
+      "step": 13332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2418200185938488e-07,
+      "loss": 0.6492,
+      "step": 13333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2385588390829327e-07,
+      "loss": 0.7195,
+      "step": 13334
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2352999790048124e-07,
+      "loss": 0.5423,
+      "step": 13335
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2320434385177391e-07,
+      "loss": 0.8768,
+      "step": 13336
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2287892177798642e-07,
+      "loss": 0.367,
+      "step": 13337
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2255373169492277e-07,
+      "loss": 0.7694,
+      "step": 13338
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2222877361837425e-07,
+      "loss": 0.7979,
+      "step": 13339
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2190404756412266e-07,
+      "loss": 0.8366,
+      "step": 13340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2157955354793648e-07,
+      "loss": 0.6365,
+      "step": 13341
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.212552915855748e-07,
+      "loss": 0.7571,
+      "step": 13342
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.20931261692785e-07,
+      "loss": 0.7849,
+      "step": 13343
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.206074638853023e-07,
+      "loss": 0.7529,
+      "step": 13344
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.2028389817885187e-07,
+      "loss": 0.7825,
+      "step": 13345
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1996056458914617e-07,
+      "loss": 0.7453,
+      "step": 13346
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1963746313188762e-07,
+      "loss": 0.9606,
+      "step": 13347
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.19314593822767e-07,
+      "loss": 0.5481,
+      "step": 13348
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1899195667746397e-07,
+      "loss": 0.9559,
+      "step": 13349
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1866955171164606e-07,
+      "loss": 0.7073,
+      "step": 13350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1834737894097124e-07,
+      "loss": 0.8117,
+      "step": 13351
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1802543838108425e-07,
+      "loss": 0.784,
+      "step": 13352
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1770373004762035e-07,
+      "loss": 0.7955,
+      "step": 13353
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1738225395620093e-07,
+      "loss": 0.7776,
+      "step": 13354
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1706101012243853e-07,
+      "loss": 0.8897,
+      "step": 13355
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1673999856193562e-07,
+      "loss": 0.8878,
+      "step": 13356
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1641921929027864e-07,
+      "loss": 0.55,
+      "step": 13357
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1609867232304628e-07,
+      "loss": 0.8215,
+      "step": 13358
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1577835767580713e-07,
+      "loss": 0.7312,
+      "step": 13359
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1545827536411433e-07,
+      "loss": 0.6955,
+      "step": 13360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1513842540351272e-07,
+      "loss": 0.7561,
+      "step": 13361
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1481880780953536e-07,
+      "loss": 0.9397,
+      "step": 13362
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1449942259770373e-07,
+      "loss": 0.7348,
+      "step": 13363
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1418026978352768e-07,
+      "loss": 0.5269,
+      "step": 13364
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1386134938250645e-07,
+      "loss": 0.5154,
+      "step": 13365
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1354266141012825e-07,
+      "loss": 0.9168,
+      "step": 13366
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1322420588186898e-07,
+      "loss": 0.9276,
+      "step": 13367
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1290598281319353e-07,
+      "loss": 0.5856,
+      "step": 13368
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1258799221955618e-07,
+      "loss": 0.9158,
+      "step": 13369
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1227023411639957e-07,
+      "loss": 0.6682,
+      "step": 13370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1195270851915416e-07,
+      "loss": 0.7291,
+      "step": 13371
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.116354154432404e-07,
+      "loss": 0.7551,
+      "step": 13372
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.113183549040676e-07,
+      "loss": 0.9824,
+      "step": 13373
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.110015269170318e-07,
+      "loss": 0.6216,
+      "step": 13374
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1068493149752013e-07,
+      "loss": 0.8882,
+      "step": 13375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1036856866090748e-07,
+      "loss": 0.8456,
+      "step": 13376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.1005243842255552e-07,
+      "loss": 0.8153,
+      "step": 13377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0973654079781858e-07,
+      "loss": 0.6755,
+      "step": 13378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0942087580203774e-07,
+      "loss": 0.7329,
+      "step": 13379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0910544345053961e-07,
+      "loss": 0.6565,
+      "step": 13380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0879024375864642e-07,
+      "loss": 0.8136,
+      "step": 13381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0847527674166257e-07,
+      "loss": 0.5589,
+      "step": 13382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0816054241488358e-07,
+      "loss": 0.7252,
+      "step": 13383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0784604079359616e-07,
+      "loss": 0.545,
+      "step": 13384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0753177189307138e-07,
+      "loss": 0.7629,
+      "step": 13385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0721773572857096e-07,
+      "loss": 0.8111,
+      "step": 13386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.069039323153471e-07,
+      "loss": 0.6863,
+      "step": 13387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0659036166863766e-07,
+      "loss": 0.9605,
+      "step": 13388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.06277023803671e-07,
+      "loss": 0.6888,
+      "step": 13389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0596391873566323e-07,
+      "loss": 0.7881,
+      "step": 13390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0565104647982003e-07,
+      "loss": 0.9717,
+      "step": 13391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0533840705133533e-07,
+      "loss": 0.8243,
+      "step": 13392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.05026000465392e-07,
+      "loss": 0.4891,
+      "step": 13393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0471382673716067e-07,
+      "loss": 0.8285,
+      "step": 13394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.04401885881802e-07,
+      "loss": 0.8094,
+      "step": 13395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0409017791446495e-07,
+      "loss": 0.5907,
+      "step": 13396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0377870285028634e-07,
+      "loss": 0.6162,
+      "step": 13397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.034674607043935e-07,
+      "loss": 0.8366,
+      "step": 13398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0315645149189934e-07,
+      "loss": 0.5722,
+      "step": 13399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0284567522790843e-07,
+      "loss": 0.5583,
+      "step": 13400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0253513192751374e-07,
+      "loss": 0.6166,
+      "step": 13401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0222482160579371e-07,
+      "loss": 0.6271,
+      "step": 13402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0191474427782076e-07,
+      "loss": 0.844,
+      "step": 13403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0160489995865173e-07,
+      "loss": 0.7612,
+      "step": 13404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0129528866333293e-07,
+      "loss": 0.6503,
+      "step": 13405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.009859104069012e-07,
+      "loss": 0.5686,
+      "step": 13406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0067676520438117e-07,
+      "loss": 0.5524,
+      "step": 13407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0036785307078422e-07,
+      "loss": 0.8788,
+      "step": 13408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0005917402111275e-07,
+      "loss": 0.4589,
+      "step": 13409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.99750728070357e-07,
+      "loss": 0.8239,
+      "step": 13410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.994425152334961e-07,
+      "loss": 0.8411,
+      "step": 13411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.991345355254981e-07,
+      "loss": 0.7684,
+      "step": 13412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.988267889613188e-07,
+      "loss": 0.5435,
+      "step": 13413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.98519275555904e-07,
+      "loss": 0.9578,
+      "step": 13414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9821199532418678e-07,
+      "loss": 0.6706,
+      "step": 13415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9790494828108962e-07,
+      "loss": 0.4208,
+      "step": 13416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9759813444152342e-07,
+      "loss": 0.9181,
+      "step": 13417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9729155382038845e-07,
+      "loss": 0.8872,
+      "step": 13418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9698520643257334e-07,
+      "loss": 0.7149,
+      "step": 13419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9667909229295402e-07,
+      "loss": 0.8114,
+      "step": 13420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9637321141639743e-07,
+      "loss": 0.8888,
+      "step": 13421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9606756381775783e-07,
+      "loss": 0.7184,
+      "step": 13422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9576214951187833e-07,
+      "loss": 0.7219,
+      "step": 13423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9545696851358987e-07,
+      "loss": 0.8418,
+      "step": 13424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9515202083771388e-07,
+      "loss": 0.6992,
+      "step": 13425
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9484730649905969e-07,
+      "loss": 0.4909,
+      "step": 13426
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9454282551242486e-07,
+      "loss": 0.6618,
+      "step": 13427
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9423857789259426e-07,
+      "loss": 0.8964,
+      "step": 13428
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.939345636543455e-07,
+      "loss": 0.639,
+      "step": 13429
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9363078281244118e-07,
+      "loss": 0.8167,
+      "step": 13430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.933272353816329e-07,
+      "loss": 0.6944,
+      "step": 13431
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9302392137666383e-07,
+      "loss": 0.697,
+      "step": 13432
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9272084081226272e-07,
+      "loss": 0.6736,
+      "step": 13433
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9241799370314674e-07,
+      "loss": 0.5991,
+      "step": 13434
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9211538006402574e-07,
+      "loss": 0.8712,
+      "step": 13435
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9181299990959358e-07,
+      "loss": 0.5373,
+      "step": 13436
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9151085325453512e-07,
+      "loss": 0.5549,
+      "step": 13437
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9120894011352364e-07,
+      "loss": 0.7085,
+      "step": 13438
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9090726050122078e-07,
+      "loss": 0.7093,
+      "step": 13439
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9060581443227698e-07,
+      "loss": 0.5436,
+      "step": 13440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9030460192133115e-07,
+      "loss": 0.7345,
+      "step": 13441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9000362298301156e-07,
+      "loss": 0.7113,
+      "step": 13442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.897028776319343e-07,
+      "loss": 0.8718,
+      "step": 13443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8940236588270433e-07,
+      "loss": 0.9042,
+      "step": 13444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.89102087749915e-07,
+      "loss": 0.5723,
+      "step": 13445
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.888020432481502e-07,
+      "loss": 0.47,
+      "step": 13446
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8850223239197828e-07,
+      "loss": 0.6533,
+      "step": 13447
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8820265519596147e-07,
+      "loss": 0.7572,
+      "step": 13448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8790331167464758e-07,
+      "loss": 0.5732,
+      "step": 13449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.876042018425722e-07,
+      "loss": 0.6639,
+      "step": 13450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8730532571426208e-07,
+      "loss": 0.7525,
+      "step": 13451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8700668330423223e-07,
+      "loss": 0.6766,
+      "step": 13452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8670827462698381e-07,
+      "loss": 0.6129,
+      "step": 13453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8641009969700917e-07,
+      "loss": 0.6903,
+      "step": 13454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8611215852879005e-07,
+      "loss": 0.7696,
+      "step": 13455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8581445113679208e-07,
+      "loss": 0.6725,
+      "step": 13456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8551697753547594e-07,
+      "loss": 0.8552,
+      "step": 13457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8521973773928613e-07,
+      "loss": 0.4689,
+      "step": 13458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.849227317626573e-07,
+      "loss": 0.817,
+      "step": 13459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.846259596200145e-07,
+      "loss": 0.66,
+      "step": 13460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8432942132576792e-07,
+      "loss": 0.7822,
+      "step": 13461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8403311689431934e-07,
+      "loss": 0.7931,
+      "step": 13462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.837370463400584e-07,
+      "loss": 0.9627,
+      "step": 13463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8344120967736245e-07,
+      "loss": 0.8436,
+      "step": 13464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8314560692059836e-07,
+      "loss": 0.4945,
+      "step": 13465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8285023808412182e-07,
+      "loss": 0.7948,
+      "step": 13466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8255510318227642e-07,
+      "loss": 0.7162,
+      "step": 13467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8226020222939456e-07,
+      "loss": 0.7545,
+      "step": 13468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.819655352397981e-07,
+      "loss": 0.7483,
+      "step": 13469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.816711022277967e-07,
+      "loss": 0.5593,
+      "step": 13470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8137690320768842e-07,
+      "loss": 0.6427,
+      "step": 13471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8108293819376066e-07,
+      "loss": 0.7511,
+      "step": 13472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.807892072002898e-07,
+      "loss": 0.6999,
+      "step": 13473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8049571024154e-07,
+      "loss": 0.6023,
+      "step": 13474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8020244733176263e-07,
+      "loss": 0.7612,
+      "step": 13475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7990941848520181e-07,
+      "loss": 0.7805,
+      "step": 13476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7961662371608734e-07,
+      "loss": 0.5445,
+      "step": 13477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.793240630386367e-07,
+      "loss": 0.8389,
+      "step": 13478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7903173646705908e-07,
+      "loss": 0.904,
+      "step": 13479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7873964401555034e-07,
+      "loss": 0.7146,
+      "step": 13480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7844778569829412e-07,
+      "loss": 0.5506,
+      "step": 13481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7815616152946523e-07,
+      "loss": 0.484,
+      "step": 13482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7786477152322623e-07,
+      "loss": 0.7824,
+      "step": 13483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.775736156937269e-07,
+      "loss": 0.6357,
+      "step": 13484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.772826940551059e-07,
+      "loss": 0.5926,
+      "step": 13485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7699200662149253e-07,
+      "loss": 0.8447,
+      "step": 13486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7670155340700323e-07,
+      "loss": 0.6811,
+      "step": 13487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7641133442574232e-07,
+      "loss": 0.6403,
+      "step": 13488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7612134969180462e-07,
+      "loss": 0.6488,
+      "step": 13489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7583159921927217e-07,
+      "loss": 0.59,
+      "step": 13490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7554208302221654e-07,
+      "loss": 0.9734,
+      "step": 13491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.75252801114697e-07,
+      "loss": 0.6157,
+      "step": 13492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7496375351076178e-07,
+      "loss": 0.5697,
+      "step": 13493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7467494022444853e-07,
+      "loss": 0.6416,
+      "step": 13494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7438636126978213e-07,
+      "loss": 0.5758,
+      "step": 13495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7409801666077697e-07,
+      "loss": 0.7517,
+      "step": 13496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.738099064114368e-07,
+      "loss": 0.5666,
+      "step": 13497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7352203053575156e-07,
+      "loss": 0.7568,
+      "step": 13498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7323438904770284e-07,
+      "loss": 0.5998,
+      "step": 13499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7294698196125836e-07,
+      "loss": 0.5742,
+      "step": 13500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.726598092903753e-07,
+      "loss": 0.7345,
+      "step": 13501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7237287104900023e-07,
+      "loss": 0.8279,
+      "step": 13502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7208616725106818e-07,
+      "loss": 0.9305,
+      "step": 13503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7179969791050023e-07,
+      "loss": 0.6064,
+      "step": 13504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7151346304121019e-07,
+      "loss": 1.1368,
+      "step": 13505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.712274626570981e-07,
+      "loss": 0.775,
+      "step": 13506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.709416967720512e-07,
+      "loss": 0.8875,
+      "step": 13507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7065616539995e-07,
+      "loss": 0.7733,
+      "step": 13508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7037086855465902e-07,
+      "loss": 0.5595,
+      "step": 13509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7008580625003212e-07,
+      "loss": 0.9023,
+      "step": 13510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.698009784999155e-07,
+      "loss": 0.5018,
+      "step": 13511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6951638531813918e-07,
+      "loss": 0.6526,
+      "step": 13512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6923202671852379e-07,
+      "loss": 0.6807,
+      "step": 13513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6894790271487937e-07,
+      "loss": 0.6911,
+      "step": 13514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6866401332100378e-07,
+      "loss": 0.9174,
+      "step": 13515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.683803585506827e-07,
+      "loss": 0.8064,
+      "step": 13516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6809693841769226e-07,
+      "loss": 0.8043,
+      "step": 13517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6781375293579538e-07,
+      "loss": 0.4793,
+      "step": 13518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6753080211874496e-07,
+      "loss": 0.5983,
+      "step": 13519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6724808598028108e-07,
+      "loss": 0.7329,
+      "step": 13520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6696560453413446e-07,
+      "loss": 0.6209,
+      "step": 13521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6668335779402246e-07,
+      "loss": 0.902,
+      "step": 13522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6640134577365074e-07,
+      "loss": 1.0992,
+      "step": 13523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6611956848671673e-07,
+      "loss": 0.3965,
+      "step": 13524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6583802594690334e-07,
+      "loss": 0.6964,
+      "step": 13525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6555671816788245e-07,
+      "loss": 0.9741,
+      "step": 13526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.652756451633164e-07,
+      "loss": 0.682,
+      "step": 13527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.649948069468543e-07,
+      "loss": 0.7066,
+      "step": 13528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6471420353213362e-07,
+      "loss": 0.9182,
+      "step": 13529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6443383493278287e-07,
+      "loss": 0.6822,
+      "step": 13530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6415370116241725e-07,
+      "loss": 0.6609,
+      "step": 13531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6387380223463922e-07,
+      "loss": 0.6542,
+      "step": 13532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6359413816304405e-07,
+      "loss": 0.7796,
+      "step": 13533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6331470896121082e-07,
+      "loss": 0.4767,
+      "step": 13534
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6303551464270985e-07,
+      "loss": 0.9832,
+      "step": 13535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6275655522110078e-07,
+      "loss": 0.7288,
+      "step": 13536
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.624778307099295e-07,
+      "loss": 0.9184,
+      "step": 13537
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.621993411227324e-07,
+      "loss": 0.7191,
+      "step": 13538
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6192108647303307e-07,
+      "loss": 0.6324,
+      "step": 13539
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.616430667743446e-07,
+      "loss": 0.6358,
+      "step": 13540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6136528204016898e-07,
+      "loss": 0.6156,
+      "step": 13541
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6108773228399543e-07,
+      "loss": 0.7125,
+      "step": 13542
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6081041751930314e-07,
+      "loss": 0.8479,
+      "step": 13543
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6053333775955914e-07,
+      "loss": 0.6689,
+      "step": 13544
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.6025649301821877e-07,
+      "loss": 0.6033,
+      "step": 13545
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.599798833087268e-07,
+      "loss": 0.6655,
+      "step": 13546
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5970350864451645e-07,
+      "loss": 0.6617,
+      "step": 13547
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5942736903900913e-07,
+      "loss": 0.8355,
+      "step": 13548
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.591514645056147e-07,
+      "loss": 0.945,
+      "step": 13549
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5887579505773244e-07,
+      "loss": 0.7623,
+      "step": 13550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.586003607087483e-07,
+      "loss": 0.8099,
+      "step": 13551
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.583251614720399e-07,
+      "loss": 0.8967,
+      "step": 13552
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5805019736097105e-07,
+      "loss": 0.6879,
+      "step": 13553
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.577754683888938e-07,
+      "loss": 1.1417,
+      "step": 13554
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.575009745691508e-07,
+      "loss": 0.9734,
+      "step": 13555
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.572267159150731e-07,
+      "loss": 0.6501,
+      "step": 13556
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.569526924399778e-07,
+      "loss": 0.6894,
+      "step": 13557
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5667890415717257e-07,
+      "loss": 0.9924,
+      "step": 13558
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5640535107995457e-07,
+      "loss": 0.7039,
+      "step": 13559
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5613203322160708e-07,
+      "loss": 0.4789,
+      "step": 13560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5585895059540336e-07,
+      "loss": 0.7678,
+      "step": 13561
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5558610321460554e-07,
+      "loss": 0.7769,
+      "step": 13562
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5531349109246364e-07,
+      "loss": 0.5682,
+      "step": 13563
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5504111424221646e-07,
+      "loss": 0.8599,
+      "step": 13564
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5476897267709178e-07,
+      "loss": 0.4811,
+      "step": 13565
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.544970664103046e-07,
+      "loss": 0.696,
+      "step": 13566
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5422539545506043e-07,
+      "loss": 0.7461,
+      "step": 13567
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5395395982455263e-07,
+      "loss": 0.5996,
+      "step": 13568
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5368275953196177e-07,
+      "loss": 0.921,
+      "step": 13569
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.534117945904584e-07,
+      "loss": 0.8143,
+      "step": 13570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5314106501320258e-07,
+      "loss": 0.564,
+      "step": 13571
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.528705708133399e-07,
+      "loss": 0.8268,
+      "step": 13572
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.526003120040076e-07,
+      "loss": 0.8169,
+      "step": 13573
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5233028859832965e-07,
+      "loss": 0.6582,
+      "step": 13574
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5206050060942002e-07,
+      "loss": 0.8425,
+      "step": 13575
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5179094805037986e-07,
+      "loss": 0.6139,
+      "step": 13576
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5152163093429762e-07,
+      "loss": 0.7754,
+      "step": 13577
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5125254927425504e-07,
+      "loss": 0.7911,
+      "step": 13578
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5098370308331888e-07,
+      "loss": 0.7748,
+      "step": 13579
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5071509237454262e-07,
+      "loss": 0.8001,
+      "step": 13580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5044671716097414e-07,
+      "loss": 0.6736,
+      "step": 13581
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.5017857745564468e-07,
+      "loss": 0.8034,
+      "step": 13582
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4991067327157494e-07,
+      "loss": 0.7343,
+      "step": 13583
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4964300462177782e-07,
+      "loss": 0.7212,
+      "step": 13584
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.493755715192502e-07,
+      "loss": 1.1079,
+      "step": 13585
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4910837397697886e-07,
+      "loss": 0.9971,
+      "step": 13586
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4884141200794233e-07,
+      "loss": 0.5434,
+      "step": 13587
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.485746856251019e-07,
+      "loss": 0.5963,
+      "step": 13588
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.483081948414128e-07,
+      "loss": 0.7294,
+      "step": 13589
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4804193966981517e-07,
+      "loss": 0.548,
+      "step": 13590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4777592012324037e-07,
+      "loss": 0.7617,
+      "step": 13591
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4751013621460585e-07,
+      "loss": 0.8367,
+      "step": 13592
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4724458795681962e-07,
+      "loss": 0.5556,
+      "step": 13593
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.4697927536277746e-07,
+      "loss": 0.562,
+      "step": 13594
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.467141984453635e-07,
+      "loss": 0.7174,
+      "step": 13595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4644935721745023e-07,
+      "loss": 0.6948,
+      "step": 13596
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4618475169190017e-07,
+      "loss": 0.6647,
+      "step": 13597
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.45920381881563e-07,
+      "loss": 0.7967,
+      "step": 13598
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4565624779927568e-07,
+      "loss": 0.7069,
+      "step": 13599
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.453923494578674e-07,
+      "loss": 0.4467,
+      "step": 13600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4512868687015348e-07,
+      "loss": 0.692,
+      "step": 13601
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.448652600489364e-07,
+      "loss": 0.62,
+      "step": 13602
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4460206900701102e-07,
+      "loss": 0.5966,
+      "step": 13603
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.443391137571587e-07,
+      "loss": 0.7003,
+      "step": 13604
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4407639431214705e-07,
+      "loss": 0.6337,
+      "step": 13605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.438139106847364e-07,
+      "loss": 0.6954,
+      "step": 13606
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.435516628876732e-07,
+      "loss": 0.6779,
+      "step": 13607
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4328965093369284e-07,
+      "loss": 0.6836,
+      "step": 13608
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4302787483551962e-07,
+      "loss": 0.6063,
+      "step": 13609
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.427663346058661e-07,
+      "loss": 0.7437,
+      "step": 13610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4250503025743268e-07,
+      "loss": 0.8444,
+      "step": 13611
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.422439618029109e-07,
+      "loss": 0.4226,
+      "step": 13612
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4198312925497672e-07,
+      "loss": 0.7283,
+      "step": 13613
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.417225326262983e-07,
+      "loss": 0.7691,
+      "step": 13614
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.414621719295306e-07,
+      "loss": 0.7588,
+      "step": 13615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4120204717731788e-07,
+      "loss": 0.7952,
+      "step": 13616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4094215838229176e-07,
+      "loss": 0.5387,
+      "step": 13617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4068250555707374e-07,
+      "loss": 0.6172,
+      "step": 13618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4042308871427323e-07,
+      "loss": 0.4187,
+      "step": 13619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.4016390786648793e-07,
+      "loss": 0.5189,
+      "step": 13620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3990496302630496e-07,
+      "loss": 0.4108,
+      "step": 13621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.396462542062993e-07,
+      "loss": 0.8374,
+      "step": 13622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.393877814190342e-07,
+      "loss": 0.426,
+      "step": 13623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.391295446770624e-07,
+      "loss": 0.8263,
+      "step": 13624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.388715439929239e-07,
+      "loss": 0.7882,
+      "step": 13625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3861377937914922e-07,
+      "loss": 0.9142,
+      "step": 13626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3835625084825387e-07,
+      "loss": 0.7455,
+      "step": 13627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3809895841274624e-07,
+      "loss": 0.7334,
+      "step": 13628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3784190208512127e-07,
+      "loss": 0.7712,
+      "step": 13629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3758508187786069e-07,
+      "loss": 0.8397,
+      "step": 13630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3732849780343727e-07,
+      "loss": 0.8232,
+      "step": 13631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.370721498743127e-07,
+      "loss": 0.758,
+      "step": 13632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3681603810293375e-07,
+      "loss": 0.5973,
+      "step": 13633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.365601625017393e-07,
+      "loss": 0.7131,
+      "step": 13634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3630452308315557e-07,
+      "loss": 0.5536,
+      "step": 13635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3604911985959647e-07,
+      "loss": 1.0164,
+      "step": 13636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3579395284346542e-07,
+      "loss": 0.8043,
+      "step": 13637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3553902204715364e-07,
+      "loss": 0.7191,
+      "step": 13638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3528432748304177e-07,
+      "loss": 0.5452,
+      "step": 13639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.350298691634988e-07,
+      "loss": 0.7481,
+      "step": 13640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3477564710088097e-07,
+      "loss": 0.6521,
+      "step": 13641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.345216613075351e-07,
+      "loss": 0.6299,
+      "step": 13642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3426791179579456e-07,
+      "loss": 0.7523,
+      "step": 13643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3401439857798292e-07,
+      "loss": 0.8362,
+      "step": 13644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3376112166641087e-07,
+      "loss": 0.8214,
+      "step": 13645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3350808107337908e-07,
+      "loss": 0.8673,
+      "step": 13646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.33255276811175e-07,
+      "loss": 0.6353,
+      "step": 13647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.33002708892076e-07,
+      "loss": 0.6842,
+      "step": 13648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3275037732834783e-07,
+      "loss": 0.6813,
+      "step": 13649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3249828213224346e-07,
+      "loss": 0.8232,
+      "step": 13650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3224642331600645e-07,
+      "loss": 0.5603,
+      "step": 13651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3199480089186757e-07,
+      "loss": 0.746,
+      "step": 13652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3174341487204478e-07,
+      "loss": 0.9067,
+      "step": 13653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3149226526874837e-07,
+      "loss": 0.5765,
+      "step": 13654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.312413520941741e-07,
+      "loss": 0.6006,
+      "step": 13655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3099067536050614e-07,
+      "loss": 0.7225,
+      "step": 13656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3074023507991917e-07,
+      "loss": 0.7568,
+      "step": 13657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3049003126457517e-07,
+      "loss": 0.8335,
+      "step": 13658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3024006392662326e-07,
+      "loss": 0.3985,
+      "step": 13659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2999033307820542e-07,
+      "loss": 0.501,
+      "step": 13660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2974083873144637e-07,
+      "loss": 0.7787,
+      "step": 13661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.294915808984637e-07,
+      "loss": 0.5315,
+      "step": 13662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2924255959136267e-07,
+      "loss": 0.5514,
+      "step": 13663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.289937748222353e-07,
+      "loss": 0.4598,
+      "step": 13664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2874522660316412e-07,
+      "loss": 0.6406,
+      "step": 13665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.284969149462184e-07,
+      "loss": 0.6913,
+      "step": 13666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2824883986345736e-07,
+      "loss": 0.8077,
+      "step": 13667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2800100136692862e-07,
+      "loss": 0.8523,
+      "step": 13668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2775339946866805e-07,
+      "loss": 0.7985,
+      "step": 13669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2750603418069886e-07,
+      "loss": 0.7859,
+      "step": 13670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2725890551503472e-07,
+      "loss": 0.6224,
+      "step": 13671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2701201348367608e-07,
+      "loss": 0.5936,
+      "step": 13672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.267653580986139e-07,
+      "loss": 0.6909,
+      "step": 13673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2651893937182635e-07,
+      "loss": 0.6522,
+      "step": 13674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2627275731527832e-07,
+      "loss": 0.686,
+      "step": 13675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2602681194092747e-07,
+      "loss": 0.8013,
+      "step": 13676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2578110326071702e-07,
+      "loss": 0.4452,
+      "step": 13677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2553563128657743e-07,
+      "loss": 0.533,
+      "step": 13678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2529039603043248e-07,
+      "loss": 0.6405,
+      "step": 13679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2504539750418988e-07,
+      "loss": 0.8219,
+      "step": 13680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2480063571974677e-07,
+      "loss": 0.7513,
+      "step": 13681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2455611068899086e-07,
+      "loss": 0.7249,
+      "step": 13682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2431182242379713e-07,
+      "loss": 1.033,
+      "step": 13683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.240677709360272e-07,
+      "loss": 0.6473,
+      "step": 13684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2382395623753484e-07,
+      "loss": 0.8576,
+      "step": 13685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.23580378340159e-07,
+      "loss": 0.7828,
+      "step": 13686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.233370372557291e-07,
+      "loss": 0.5408,
+      "step": 13687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2309393299606233e-07,
+      "loss": 0.7802,
+      "step": 13688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2285106557296479e-07,
+      "loss": 0.5023,
+      "step": 13689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2260843499823094e-07,
+      "loss": 0.5671,
+      "step": 13690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2236604128364305e-07,
+      "loss": 0.7941,
+      "step": 13691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2212388444097278e-07,
+      "loss": 0.6988,
+      "step": 13692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.218819644819802e-07,
+      "loss": 0.54,
+      "step": 13693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.216402814184131e-07,
+      "loss": 0.4898,
+      "step": 13694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2139883526200825e-07,
+      "loss": 0.6968,
+      "step": 13695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2115762602449177e-07,
+      "loss": 0.6321,
+      "step": 13696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2091665371757654e-07,
+      "loss": 0.4926,
+      "step": 13697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.20675918352966e-07,
+      "loss": 0.7361,
+      "step": 13698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.2043541994234963e-07,
+      "loss": 0.7997,
+      "step": 13699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.201951584974076e-07,
+      "loss": 0.9814,
+      "step": 13700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1995513402980775e-07,
+      "loss": 0.78,
+      "step": 13701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1971534655120631e-07,
+      "loss": 0.9426,
+      "step": 13702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1947579607324678e-07,
+      "loss": 0.565,
+      "step": 13703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1923648260756426e-07,
+      "loss": 0.7632,
+      "step": 13704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1899740616578004e-07,
+      "loss": 0.8197,
+      "step": 13705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1875856675950314e-07,
+      "loss": 0.6741,
+      "step": 13706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.185199644003332e-07,
+      "loss": 0.7278,
+      "step": 13707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1828159909985814e-07,
+      "loss": 0.78,
+      "step": 13708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1804347086965207e-07,
+      "loss": 0.93,
+      "step": 13709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1780557972128014e-07,
+      "loss": 0.5624,
+      "step": 13710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1756792566629538e-07,
+      "loss": 0.8194,
+      "step": 13711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1733050871623852e-07,
+      "loss": 0.7704,
+      "step": 13712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1709332888263869e-07,
+      "loss": 0.5362,
+      "step": 13713
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.16856386177015e-07,
+      "loss": 0.7889,
+      "step": 13714
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1661968061087325e-07,
+      "loss": 0.708,
+      "step": 13715
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1638321219570869e-07,
+      "loss": 0.7134,
+      "step": 13716
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1614698094300547e-07,
+      "loss": 0.5489,
+      "step": 13717
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1591098686423497e-07,
+      "loss": 0.7429,
+      "step": 13718
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1567522997085856e-07,
+      "loss": 0.6791,
+      "step": 13719
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1543971027432433e-07,
+      "loss": 0.4782,
+      "step": 13720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1520442778607032e-07,
+      "loss": 0.6291,
+      "step": 13721
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1496938251752244e-07,
+      "loss": 0.4202,
+      "step": 13722
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1473457448009484e-07,
+      "loss": 0.7235,
+      "step": 13723
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.145000036851912e-07,
+      "loss": 0.6785,
+      "step": 13724
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1426567014420297e-07,
+      "loss": 0.732,
+      "step": 13725
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1403157386850827e-07,
+      "loss": 0.6606,
+      "step": 13726
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1379771486947799e-07,
+      "loss": 0.5208,
+      "step": 13727
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1356409315846806e-07,
+      "loss": 0.544,
+      "step": 13728
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1333070874682217e-07,
+      "loss": 0.5469,
+      "step": 13729
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1309756164587682e-07,
+      "loss": 0.7413,
+      "step": 13730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1286465186695294e-07,
+      "loss": 0.7669,
+      "step": 13731
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1263197942136039e-07,
+      "loss": 0.7941,
+      "step": 13732
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1239954432040068e-07,
+      "loss": 0.9563,
+      "step": 13733
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1216734657536032e-07,
+      "loss": 0.6003,
+      "step": 13734
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.119353861975142e-07,
+      "loss": 0.8111,
+      "step": 13735
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1170366319812942e-07,
+      "loss": 0.9071,
+      "step": 13736
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1147217758845752e-07,
+      "loss": 0.6423,
+      "step": 13737
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1124092937974007e-07,
+      "loss": 0.7523,
+      "step": 13738
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1100991858320865e-07,
+      "loss": 0.8587,
+      "step": 13739
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1077914521007981e-07,
+      "loss": 0.7406,
+      "step": 13740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.105486092715624e-07,
+      "loss": 0.8747,
+      "step": 13741
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1031831077885024e-07,
+      "loss": 0.7481,
+      "step": 13742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1008824974312827e-07,
+      "loss": 0.4853,
+      "step": 13743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0985842617556919e-07,
+      "loss": 0.6771,
+      "step": 13744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0962884008733299e-07,
+      "loss": 0.4416,
+      "step": 13745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0939949148956908e-07,
+      "loss": 0.5441,
+      "step": 13746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0917038039341632e-07,
+      "loss": 0.662,
+      "step": 13747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0894150681000026e-07,
+      "loss": 0.7351,
+      "step": 13748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0871287075043534e-07,
+      "loss": 0.87,
+      "step": 13749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0848447222582604e-07,
+      "loss": 0.7645,
+      "step": 13750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0825631124726233e-07,
+      "loss": 0.7249,
+      "step": 13751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0802838782582537e-07,
+      "loss": 0.5886,
+      "step": 13752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0780070197258408e-07,
+      "loss": 0.9479,
+      "step": 13753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0757325369859462e-07,
+      "loss": 0.7245,
+      "step": 13754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0734604301490315e-07,
+      "loss": 0.5961,
+      "step": 13755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0711906993254417e-07,
+      "loss": 0.4877,
+      "step": 13756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0689233446253832e-07,
+      "loss": 0.2953,
+      "step": 13757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0666583661589847e-07,
+      "loss": 0.5562,
+      "step": 13758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0643957640362356e-07,
+      "loss": 0.72,
+      "step": 13759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0621355383670095e-07,
+      "loss": 0.6682,
+      "step": 13760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0598776892610685e-07,
+      "loss": 0.6067,
+      "step": 13761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0576222168280636e-07,
+      "loss": 0.8323,
+      "step": 13762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0553691211775242e-07,
+      "loss": 0.7241,
+      "step": 13763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0531184024188679e-07,
+      "loss": 0.957,
+      "step": 13764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.050870060661402e-07,
+      "loss": 0.6071,
+      "step": 13765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0486240960143057e-07,
+      "loss": 0.737,
+      "step": 13766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.046380508586653e-07,
+      "loss": 0.6145,
+      "step": 13767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0441392984874011e-07,
+      "loss": 0.6753,
+      "step": 13768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0419004658253795e-07,
+      "loss": 0.6168,
+      "step": 13769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0396640107093237e-07,
+      "loss": 0.7675,
+      "step": 13770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0374299332478355e-07,
+      "loss": 0.6336,
+      "step": 13771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0351982335494115e-07,
+      "loss": 0.8116,
+      "step": 13772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0329689117224262e-07,
+      "loss": 0.6973,
+      "step": 13773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.030741967875143e-07,
+      "loss": 0.547,
+      "step": 13774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0285174021157141e-07,
+      "loss": 0.9166,
+      "step": 13775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0262952145521643e-07,
+      "loss": 0.61,
+      "step": 13776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0240754052924073e-07,
+      "loss": 0.8183,
+      "step": 13777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.021857974444257e-07,
+      "loss": 0.5998,
+      "step": 13778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0196429221153825e-07,
+      "loss": 0.6155,
+      "step": 13779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0174302484133592e-07,
+      "loss": 0.7154,
+      "step": 13780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0152199534456453e-07,
+      "loss": 0.5896,
+      "step": 13781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0130120373195718e-07,
+      "loss": 0.7344,
+      "step": 13782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0108065001423584e-07,
+      "loss": 0.2646,
+      "step": 13783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.008603342021125e-07,
+      "loss": 0.6567,
+      "step": 13784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0064025630628583e-07,
+      "loss": 0.5837,
+      "step": 13785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0042041633744226e-07,
+      "loss": 0.575,
+      "step": 13786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.002008143062605e-07,
+      "loss": 0.9041,
+      "step": 13787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.998145022340255e-08,
+      "loss": 0.9594,
+      "step": 13788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.976232409952214e-08,
+      "loss": 0.7973,
+      "step": 13789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.954343594526072e-08,
+      "loss": 0.6352,
+      "step": 13790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.932478577124815e-08,
+      "loss": 0.7629,
+      "step": 13791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.910637358810316e-08,
+      "loss": 0.8433,
+      "step": 13792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.888819940643169e-08,
+      "loss": 0.3339,
+      "step": 13793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.867026323682971e-08,
+      "loss": 0.7574,
+      "step": 13794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.845256508987988e-08,
+      "loss": 0.859,
+      "step": 13795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.823510497615485e-08,
+      "loss": 0.7936,
+      "step": 13796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.801788290621505e-08,
+      "loss": 0.728,
+      "step": 13797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.780089889060984e-08,
+      "loss": 0.44,
+      "step": 13798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.758415293987523e-08,
+      "loss": 0.6322,
+      "step": 13799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.736764506453832e-08,
+      "loss": 0.7583,
+      "step": 13800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.715137527511298e-08,
+      "loss": 0.8232,
+      "step": 13801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.693534358210133e-08,
+      "loss": 0.8563,
+      "step": 13802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.671954999599554e-08,
+      "loss": 0.4745,
+      "step": 13803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.650399452727499e-08,
+      "loss": 0.6617,
+      "step": 13804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.628867718640633e-08,
+      "loss": 0.6938,
+      "step": 13805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.607359798384785e-08,
+      "loss": 0.8403,
+      "step": 13806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.585875693004343e-08,
+      "loss": 0.8198,
+      "step": 13807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.564415403542581e-08,
+      "loss": 0.6788,
+      "step": 13808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.542978931041835e-08,
+      "loss": 0.4985,
+      "step": 13809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.521566276542993e-08,
+      "loss": 0.5774,
+      "step": 13810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.50017744108589e-08,
+      "loss": 0.8288,
+      "step": 13811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.478812425709361e-08,
+      "loss": 0.6501,
+      "step": 13812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.457471231450855e-08,
+      "loss": 0.7263,
+      "step": 13813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.436153859346708e-08,
+      "loss": 0.6868,
+      "step": 13814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.414860310432317e-08,
+      "loss": 0.5926,
+      "step": 13815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.393590585741686e-08,
+      "loss": 0.515,
+      "step": 13816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.372344686307655e-08,
+      "loss": 0.7877,
+      "step": 13817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.351122613162123e-08,
+      "loss": 0.4546,
+      "step": 13818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.329924367335541e-08,
+      "loss": 0.6831,
+      "step": 13819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.308749949857476e-08,
+      "loss": 1.019,
+      "step": 13820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.287599361756217e-08,
+      "loss": 0.5566,
+      "step": 13821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.26647260405883e-08,
+      "loss": 0.7475,
+      "step": 13822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.245369677791271e-08,
+      "loss": 0.6303,
+      "step": 13823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.224290583978446e-08,
+      "loss": 0.6842,
+      "step": 13824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.203235323644034e-08,
+      "loss": 0.6684,
+      "step": 13825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.182203897810438e-08,
+      "loss": 0.5324,
+      "step": 13826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.16119630749901e-08,
+      "loss": 0.6971,
+      "step": 13827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.140212553730044e-08,
+      "loss": 0.7876,
+      "step": 13828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.119252637522502e-08,
+      "loss": 0.6914,
+      "step": 13829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.098316559894237e-08,
+      "loss": 0.5742,
+      "step": 13830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.077404321861993e-08,
+      "loss": 0.9357,
+      "step": 13831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.056515924441345e-08,
+      "loss": 0.4946,
+      "step": 13832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.035651368646647e-08,
+      "loss": 0.5699,
+      "step": 13833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.014810655491202e-08,
+      "loss": 0.6892,
+      "step": 13834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.993993785987088e-08,
+      "loss": 0.8741,
+      "step": 13835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.973200761145162e-08,
+      "loss": 0.6101,
+      "step": 13836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.95243158197523e-08,
+      "loss": 0.7299,
+      "step": 13837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.931686249485927e-08,
+      "loss": 0.8944,
+      "step": 13838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.910964764684671e-08,
+      "loss": 0.6923,
+      "step": 13839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.890267128577768e-08,
+      "loss": 0.8533,
+      "step": 13840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.869593342170412e-08,
+      "loss": 0.6746,
+      "step": 13841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.84894340646647e-08,
+      "loss": 0.7097,
+      "step": 13842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.82831732246886e-08,
+      "loss": 0.4369,
+      "step": 13843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.807715091179226e-08,
+      "loss": 0.741,
+      "step": 13844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.787136713598043e-08,
+      "loss": 0.8253,
+      "step": 13845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.766582190724627e-08,
+      "loss": 0.7126,
+      "step": 13846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.746051523557231e-08,
+      "loss": 1.1977,
+      "step": 13847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.725544713092837e-08,
+      "loss": 0.4551,
+      "step": 13848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.705061760327372e-08,
+      "loss": 0.6239,
+      "step": 13849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.684602666255482e-08,
+      "loss": 0.443,
+      "step": 13850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.664167431870762e-08,
+      "loss": 0.3997,
+      "step": 13851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.643756058165587e-08,
+      "loss": 0.6918,
+      "step": 13852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.623368546131217e-08,
+      "loss": 0.7079,
+      "step": 13853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.603004896757694e-08,
+      "loss": 1.0789,
+      "step": 13854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.58266511103395e-08,
+      "loss": 0.6854,
+      "step": 13855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.562349189947695e-08,
+      "loss": 0.6827,
+      "step": 13856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.542057134485638e-08,
+      "loss": 0.7601,
+      "step": 13857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.521788945633103e-08,
+      "loss": 0.8055,
+      "step": 13858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.501544624374413e-08,
+      "loss": 0.727,
+      "step": 13859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.481324171692785e-08,
+      "loss": 0.5935,
+      "step": 13860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.461127588570039e-08,
+      "loss": 0.6462,
+      "step": 13861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.440954875987007e-08,
+      "loss": 0.6988,
+      "step": 13862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.420806034923401e-08,
+      "loss": 0.542,
+      "step": 13863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.400681066357718e-08,
+      "loss": 0.5691,
+      "step": 13864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.380579971267178e-08,
+      "loss": 0.7736,
+      "step": 13865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.360502750627996e-08,
+      "loss": 0.6063,
+      "step": 13866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.34044940541523e-08,
+      "loss": 0.5184,
+      "step": 13867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.320419936602708e-08,
+      "loss": 0.7712,
+      "step": 13868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.300414345163043e-08,
+      "loss": 0.5982,
+      "step": 13869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.280432632067848e-08,
+      "loss": 0.8248,
+      "step": 13870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.26047479828751e-08,
+      "loss": 0.8361,
+      "step": 13871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.240540844791145e-08,
+      "loss": 0.7039,
+      "step": 13872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.220630772546812e-08,
+      "loss": 0.7395,
+      "step": 13873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.20074458252157e-08,
+      "loss": 0.6393,
+      "step": 13874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.180882275680868e-08,
+      "loss": 0.4549,
+      "step": 13875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.161043852989492e-08,
+      "loss": 0.5602,
+      "step": 13876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.141229315410837e-08,
+      "loss": 0.5818,
+      "step": 13877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.121438663907022e-08,
+      "loss": 0.8655,
+      "step": 13878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.101671899439279e-08,
+      "loss": 0.6907,
+      "step": 13879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.081929022967505e-08,
+      "loss": 0.5788,
+      "step": 13880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.06221003545038e-08,
+      "loss": 0.6937,
+      "step": 13881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.042514937845691e-08,
+      "loss": 0.7281,
+      "step": 13882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.022843731109676e-08,
+      "loss": 0.781,
+      "step": 13883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 8.003196416197734e-08,
+      "loss": 0.7569,
+      "step": 13884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.983572994064049e-08,
+      "loss": 0.547,
+      "step": 13885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.963973465661523e-08,
+      "loss": 0.6916,
+      "step": 13886
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.944397831941952e-08,
+      "loss": 0.9438,
+      "step": 13887
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.92484609385602e-08,
+      "loss": 0.7109,
+      "step": 13888
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 7.905318252353189e-08,
+      "loss": 0.7,
+      "step": 13889
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.885814308381811e-08,
+      "loss": 0.5523,
+      "step": 13890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.866334262889075e-08,
+      "loss": 0.9139,
+      "step": 13891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.846878116820944e-08,
+      "loss": 0.7188,
+      "step": 13892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.82744587112222e-08,
+      "loss": 0.6805,
+      "step": 13893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.8080375267367e-08,
+      "loss": 0.8981,
+      "step": 13894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.788653084606856e-08,
+      "loss": 0.805,
+      "step": 13895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.769292545674045e-08,
+      "loss": 0.7015,
+      "step": 13896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.749955910878459e-08,
+      "loss": 0.8302,
+      "step": 13897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.730643181159181e-08,
+      "loss": 0.7947,
+      "step": 13898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.711354357454071e-08,
+      "loss": 0.8193,
+      "step": 13899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.692089440699824e-08,
+      "loss": 0.6818,
+      "step": 13900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.672848431832081e-08,
+      "loss": 0.5066,
+      "step": 13901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.65363133178515e-08,
+      "loss": 0.7619,
+      "step": 13902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.634438141492284e-08,
+      "loss": 0.9436,
+      "step": 13903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.615268861885572e-08,
+      "loss": 0.6343,
+      "step": 13904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.59612349389599e-08,
+      "loss": 0.589,
+      "step": 13905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.577002038453185e-08,
+      "loss": 0.7792,
+      "step": 13906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.557904496485801e-08,
+      "loss": 0.6633,
+      "step": 13907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.53883086892132e-08,
+      "loss": 0.6302,
+      "step": 13908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.519781156685946e-08,
+      "loss": 0.6383,
+      "step": 13909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.500755360704771e-08,
+      "loss": 0.8652,
+      "step": 13910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.481753481901832e-08,
+      "loss": 1.0095,
+      "step": 13911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.46277552119984e-08,
+      "loss": 0.8124,
+      "step": 13912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.443821479520441e-08,
+      "loss": 0.6538,
+      "step": 13913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.42489135778407e-08,
+      "loss": 0.79,
+      "step": 13914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.405985156910045e-08,
+      "loss": 0.6306,
+      "step": 13915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.387102877816521e-08,
+      "loss": 0.5176,
+      "step": 13916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.368244521420486e-08,
+      "loss": 0.772,
+      "step": 13917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.349410088637766e-08,
+      "loss": 0.8071,
+      "step": 13918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.330599580382957e-08,
+      "loss": 0.8732,
+      "step": 13919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.311812997569556e-08,
+      "loss": 0.7986,
+      "step": 13920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.293050341109942e-08,
+      "loss": 0.6288,
+      "step": 13921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.274311611915275e-08,
+      "loss": 0.5093,
+      "step": 13922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.255596810895548e-08,
+      "loss": 0.7157,
+      "step": 13923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.236905938959593e-08,
+      "loss": 0.9264,
+      "step": 13924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.218238997015071e-08,
+      "loss": 0.583,
+      "step": 13925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.199595985968589e-08,
+      "loss": 0.7046,
+      "step": 13926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.180976906725423e-08,
+      "loss": 0.7996,
+      "step": 13927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.162381760189852e-08,
+      "loss": 0.7181,
+      "step": 13928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.143810547264762e-08,
+      "loss": 0.9356,
+      "step": 13929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.125263268852211e-08,
+      "loss": 0.9234,
+      "step": 13930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.106739925852757e-08,
+      "loss": 0.6928,
+      "step": 13931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.088240519165956e-08,
+      "loss": 0.806,
+      "step": 13932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.069765049690369e-08,
+      "loss": 0.4309,
+      "step": 13933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.051313518323e-08,
+      "loss": 0.5452,
+      "step": 13934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.03288592596002e-08,
+      "loss": 0.6934,
+      "step": 13935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 7.014482273496326e-08,
+      "loss": 0.7183,
+      "step": 13936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.996102561825591e-08,
+      "loss": 0.8918,
+      "step": 13937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.977746791840434e-08,
+      "loss": 0.6451,
+      "step": 13938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.959414964432309e-08,
+      "loss": 0.6296,
+      "step": 13939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.941107080491339e-08,
+      "loss": 0.579,
+      "step": 13940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.922823140906754e-08,
+      "loss": 0.6855,
+      "step": 13941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.904563146566345e-08,
+      "loss": 0.4571,
+      "step": 13942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.886327098356905e-08,
+      "loss": 0.7191,
+      "step": 13943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.868114997164055e-08,
+      "loss": 0.8252,
+      "step": 13944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.849926843872257e-08,
+      "loss": 0.6309,
+      "step": 13945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.831762639364691e-08,
+      "loss": 0.6703,
+      "step": 13946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.813622384523543e-08,
+      "loss": 0.8195,
+      "step": 13947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.795506080229774e-08,
+      "loss": 0.8169,
+      "step": 13948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.777413727363069e-08,
+      "loss": 0.5578,
+      "step": 13949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.759345326802113e-08,
+      "loss": 0.5247,
+      "step": 13950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.741300879424262e-08,
+      "loss": 0.7933,
+      "step": 13951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.723280386105979e-08,
+      "loss": 0.7818,
+      "step": 13952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.705283847722288e-08,
+      "loss": 0.6211,
+      "step": 13953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.687311265147045e-08,
+      "loss": 0.6481,
+      "step": 13954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.669362639253274e-08,
+      "loss": 0.5062,
+      "step": 13955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.651437970912499e-08,
+      "loss": 0.8744,
+      "step": 13956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.633537260995138e-08,
+      "loss": 0.5941,
+      "step": 13957
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.615660510370658e-08,
+      "loss": 0.6695,
+      "step": 13958
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.597807719907034e-08,
+      "loss": 0.7774,
+      "step": 13959
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.579978890471295e-08,
+      "loss": 0.8441,
+      "step": 13960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.562174022929358e-08,
+      "loss": 0.6644,
+      "step": 13961
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.544393118145753e-08,
+      "loss": 0.768,
+      "step": 13962
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.526636176984013e-08,
+      "loss": 0.7762,
+      "step": 13963
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.50890320030656e-08,
+      "loss": 0.6402,
+      "step": 13964
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.491194188974426e-08,
+      "loss": 0.5191,
+      "step": 13965
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.473509143847645e-08,
+      "loss": 0.6363,
+      "step": 13966
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.455848065785086e-08,
+      "loss": 0.7844,
+      "step": 13967
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.438210955644453e-08,
+      "loss": 0.7072,
+      "step": 13968
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.42059781428217e-08,
+      "loss": 0.7088,
+      "step": 13969
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.40300864255361e-08,
+      "loss": 0.4973,
+      "step": 13970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.385443441312977e-08,
+      "loss": 0.734,
+      "step": 13971
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.367902211413257e-08,
+      "loss": 0.7824,
+      "step": 13972
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.350384953706379e-08,
+      "loss": 0.6631,
+      "step": 13973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.332891669042885e-08,
+      "loss": 0.9111,
+      "step": 13974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.315422358272483e-08,
+      "loss": 0.7505,
+      "step": 13975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.297977022243384e-08,
+      "loss": 0.8295,
+      "step": 13976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.280555661802857e-08,
+      "loss": 0.7883,
+      "step": 13977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.263158277796888e-08,
+      "loss": 0.6441,
+      "step": 13978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.24578487107036e-08,
+      "loss": 0.807,
+      "step": 13979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.228435442466985e-08,
+      "loss": 0.7648,
+      "step": 13980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.211109992829367e-08,
+      "loss": 0.6361,
+      "step": 13981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.193808522998723e-08,
+      "loss": 0.6528,
+      "step": 13982
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.176531033815436e-08,
+      "loss": 0.8259,
+      "step": 13983
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.159277526118445e-08,
+      "loss": 0.6988,
+      "step": 13984
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.142048000745638e-08,
+      "loss": 0.5623,
+      "step": 13985
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.124842458533731e-08,
+      "loss": 0.7428,
+      "step": 13986
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.107660900318336e-08,
+      "loss": 0.637,
+      "step": 13987
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.090503326933728e-08,
+      "loss": 0.8288,
+      "step": 13988
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.073369739213241e-08,
+      "loss": 0.7489,
+      "step": 13989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.056260137988823e-08,
+      "loss": 0.8731,
+      "step": 13990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.039174524091473e-08,
+      "loss": 0.7697,
+      "step": 13991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.022112898350862e-08,
+      "loss": 0.6509,
+      "step": 13992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.005075261595495e-08,
+      "loss": 0.6546,
+      "step": 13993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.988061614652929e-08,
+      "loss": 0.9671,
+      "step": 13994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.971071958349229e-08,
+      "loss": 0.668,
+      "step": 13995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.954106293509565e-08,
+      "loss": 0.8263,
+      "step": 13996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9371646209577805e-08,
+      "loss": 0.7935,
+      "step": 13997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.9202469415166605e-08,
+      "loss": 0.7657,
+      "step": 13998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.903353256007716e-08,
+      "loss": 0.4939,
+      "step": 13999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8864835652513997e-08,
+      "loss": 0.6712,
+      "step": 14000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.869637870067002e-08,
+      "loss": 0.5424,
+      "step": 14001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.852816171272424e-08,
+      "loss": 0.754,
+      "step": 14002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.836018469684734e-08,
+      "loss": 0.7047,
+      "step": 14003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.8192447661196694e-08,
+      "loss": 0.5199,
+      "step": 14004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.802495061391688e-08,
+      "loss": 0.8181,
+      "step": 14005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.785769356314419e-08,
+      "loss": 0.892,
+      "step": 14006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.769067651699878e-08,
+      "loss": 0.6529,
+      "step": 14007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7523899483592494e-08,
+      "loss": 0.745,
+      "step": 14008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.735736247102497e-08,
+      "loss": 0.6317,
+      "step": 14009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7191065487382516e-08,
+      "loss": 0.6058,
+      "step": 14010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.7025008540742e-08,
+      "loss": 0.7198,
+      "step": 14011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.685919163916753e-08,
+      "loss": 0.5983,
+      "step": 14012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.669361479071156e-08,
+      "loss": 0.6174,
+      "step": 14013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6528278003414315e-08,
+      "loss": 0.8063,
+      "step": 14014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.63631812853066e-08,
+      "loss": 0.6271,
+      "step": 14015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6198324644404226e-08,
+      "loss": 0.8209,
+      "step": 14016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.6033708088713556e-08,
+      "loss": 0.7408,
+      "step": 14017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.586933162622987e-08,
+      "loss": 0.6801,
+      "step": 14018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.570519526493456e-08,
+      "loss": 0.7615,
+      "step": 14019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.554129901279903e-08,
+      "loss": 0.8966,
+      "step": 14020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.537764287778302e-08,
+      "loss": 0.4351,
+      "step": 14021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.521422686783295e-08,
+      "loss": 0.9809,
+      "step": 14022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.505105099088637e-08,
+      "loss": 0.6437,
+      "step": 14023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4888115254865817e-08,
+      "loss": 0.6295,
+      "step": 14024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.472541966768552e-08,
+      "loss": 0.6166,
+      "step": 14025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.456296423724528e-08,
+      "loss": 0.6362,
+      "step": 14026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.440074897143488e-08,
+      "loss": 0.4533,
+      "step": 14027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4238773878131925e-08,
+      "loss": 0.4678,
+      "step": 14028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.4077038965202885e-08,
+      "loss": 0.692,
+      "step": 14029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.391554424050094e-08,
+      "loss": 0.6312,
+      "step": 14030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.375428971186925e-08,
+      "loss": 0.8205,
+      "step": 14031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.359327538713932e-08,
+      "loss": 0.773,
+      "step": 14032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3432501274129914e-08,
+      "loss": 0.7099,
+      "step": 14033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3271967380649216e-08,
+      "loss": 0.7239,
+      "step": 14034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.3111673714492664e-08,
+      "loss": 0.8402,
+      "step": 14035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2951620283444026e-08,
+      "loss": 0.7128,
+      "step": 14036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.279180709527765e-08,
+      "loss": 0.5936,
+      "step": 14037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.263223415775287e-08,
+      "loss": 0.3523,
+      "step": 14038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.247290147862016e-08,
+      "loss": 0.7135,
+      "step": 14039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2313809065616674e-08,
+      "loss": 0.7914,
+      "step": 14040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.215495692646788e-08,
+      "loss": 0.3844,
+      "step": 14041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.199634506888929e-08,
+      "loss": 0.9439,
+      "step": 14042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.183797350058251e-08,
+      "loss": 0.5395,
+      "step": 14043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.167984222923916e-08,
+      "loss": 0.3743,
+      "step": 14044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.152195126253812e-08,
+      "loss": 0.6664,
+      "step": 14045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.136430060814712e-08,
+      "loss": 0.6604,
+      "step": 14046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.120689027372228e-08,
+      "loss": 0.5754,
+      "step": 14047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.104972026690802e-08,
+      "loss": 0.6267,
+      "step": 14048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.089279059533658e-08,
+      "loss": 0.9032,
+      "step": 14049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.073610126662909e-08,
+      "loss": 0.6568,
+      "step": 14050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.057965228839501e-08,
+      "loss": 0.9489,
+      "step": 14051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.04234436682316e-08,
+      "loss": 0.6187,
+      "step": 14052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.026747541372501e-08,
+      "loss": 0.9349,
+      "step": 14053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.011174753244974e-08,
+      "loss": 0.6256,
+      "step": 14054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.995626003196807e-08,
+      "loss": 0.8629,
+      "step": 14055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.980101291983064e-08,
+      "loss": 0.9492,
+      "step": 14056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9646006203577515e-08,
+      "loss": 0.9005,
+      "step": 14057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9491239890735456e-08,
+      "loss": 0.7003,
+      "step": 14058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.9336713988820676e-08,
+      "loss": 0.7771,
+      "step": 14059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.918242850533772e-08,
+      "loss": 0.7843,
+      "step": 14060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.902838344777894e-08,
+      "loss": 0.736,
+      "step": 14061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8874578823624454e-08,
+      "loss": 0.7698,
+      "step": 14062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.872101464034495e-08,
+      "loss": 0.7124,
+      "step": 14063
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.85676909053967e-08,
+      "loss": 0.942,
+      "step": 14064
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.841460762622596e-08,
+      "loss": 0.7836,
+      "step": 14065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.8261764810266784e-08,
+      "loss": 0.858,
+      "step": 14066
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.810916246494157e-08,
+      "loss": 0.628,
+      "step": 14067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.795680059766161e-08,
+      "loss": 0.663,
+      "step": 14068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7804679215825435e-08,
+      "loss": 0.8918,
+      "step": 14069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.765279832682102e-08,
+      "loss": 0.6341,
+      "step": 14070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.750115793802412e-08,
+      "loss": 0.5938,
+      "step": 14071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.734975805679831e-08,
+      "loss": 0.5774,
+      "step": 14072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.719859869049659e-08,
+      "loss": 0.5745,
+      "step": 14073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.7047679846459194e-08,
+      "loss": 0.5927,
+      "step": 14074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.689700153201526e-08,
+      "loss": 0.7896,
+      "step": 14075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6746563754482274e-08,
+      "loss": 0.6332,
+      "step": 14076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6596366521166614e-08,
+      "loss": 0.4537,
+      "step": 14077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6446409839360773e-08,
+      "loss": 0.869,
+      "step": 14078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.6296693716348375e-08,
+      "loss": 0.96,
+      "step": 14079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.614721815939971e-08,
+      "loss": 0.5673,
+      "step": 14080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.599798317577342e-08,
+      "loss": 0.5881,
+      "step": 14081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.58489887727176e-08,
+      "loss": 0.5213,
+      "step": 14082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.570023495746645e-08,
+      "loss": 1.0172,
+      "step": 14083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5551721737244756e-08,
+      "loss": 0.518,
+      "step": 14084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5403449119265085e-08,
+      "loss": 0.7544,
+      "step": 14085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.525541711072723e-08,
+      "loss": 0.7946,
+      "step": 14086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.5107625718820434e-08,
+      "loss": 0.8411,
+      "step": 14087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.49600749507223e-08,
+      "loss": 0.8849,
+      "step": 14088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.481276481359764e-08,
+      "loss": 0.5725,
+      "step": 14089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.466569531460019e-08,
+      "loss": 0.9763,
+      "step": 14090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.451886646087256e-08,
+      "loss": 0.9392,
+      "step": 14091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.4372278259545153e-08,
+      "loss": 0.5621,
+      "step": 14092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.422593071773618e-08,
+      "loss": 0.6823,
+      "step": 14093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.407982384255327e-08,
+      "loss": 0.6413,
+      "step": 14094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3933957641091317e-08,
+      "loss": 0.9293,
+      "step": 14095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.378833212043465e-08,
+      "loss": 0.9405,
+      "step": 14096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3642947287654284e-08,
+      "loss": 0.8203,
+      "step": 14097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3497803149811776e-08,
+      "loss": 0.673,
+      "step": 14098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.335289971395484e-08,
+      "loss": 0.666,
+      "step": 14099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.3208236987120624e-08,
+      "loss": 0.6744,
+      "step": 14100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.306381497633405e-08,
+      "loss": 0.8635,
+      "step": 14101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.291963368860952e-08,
+      "loss": 0.4964,
+      "step": 14102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2775693130948094e-08,
+      "loss": 0.7188,
+      "step": 14103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2631993310340294e-08,
+      "loss": 0.7516,
+      "step": 14104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2488534233764425e-08,
+      "loss": 0.802,
+      "step": 14105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2345315908187135e-08,
+      "loss": 0.6799,
+      "step": 14106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2202338340563974e-08,
+      "loss": 0.8839,
+      "step": 14107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.2059601537838276e-08,
+      "loss": 0.8627,
+      "step": 14108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.191710550694172e-08,
+      "loss": 0.6899,
+      "step": 14109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.177485025479378e-08,
+      "loss": 0.7026,
+      "step": 14110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.163283578830335e-08,
+      "loss": 0.6678,
+      "step": 14111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.14910621143666e-08,
+      "loss": 0.7745,
+      "step": 14112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.134952923986968e-08,
+      "loss": 0.7522,
+      "step": 14113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.120823717168432e-08,
+      "loss": 0.5978,
+      "step": 14114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.1067185916672247e-08,
+      "loss": 0.6873,
+      "step": 14115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.092637548168466e-08,
+      "loss": 0.8287,
+      "step": 14116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.078580587355774e-08,
+      "loss": 0.9213,
+      "step": 14117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.064547709911992e-08,
+      "loss": 0.6877,
+      "step": 14118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.050538916518465e-08,
+      "loss": 0.836,
+      "step": 14119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.036554207855536e-08,
+      "loss": 0.8059,
+      "step": 14120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.02259358460233e-08,
+      "loss": 0.8034,
+      "step": 14121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.008657047436859e-08,
+      "loss": 0.7754,
+      "step": 14122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.994744597035916e-08,
+      "loss": 0.5734,
+      "step": 14123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.98085623407507e-08,
+      "loss": 0.7151,
+      "step": 14124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9669919592288385e-08,
+      "loss": 0.8039,
+      "step": 14125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.953151773170516e-08,
+      "loss": 0.9909,
+      "step": 14126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.9393356765721757e-08,
+      "loss": 0.6558,
+      "step": 14127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.925543670104726e-08,
+      "loss": 0.8334,
+      "step": 14128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.911775754438074e-08,
+      "loss": 0.6209,
+      "step": 14129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8980319302407976e-08,
+      "loss": 1.0013,
+      "step": 14130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8843121981802515e-08,
+      "loss": 0.4424,
+      "step": 14131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.870616558922735e-08,
+      "loss": 0.5274,
+      "step": 14132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.856945013133384e-08,
+      "loss": 0.9496,
+      "step": 14133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8432975614761114e-08,
+      "loss": 0.9411,
+      "step": 14134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8296742046136646e-08,
+      "loss": 0.8024,
+      "step": 14135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8160749432076816e-08,
+      "loss": 0.713,
+      "step": 14136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8024997779185225e-08,
+      "loss": 0.5927,
+      "step": 14137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7889487094054375e-08,
+      "loss": 0.6997,
+      "step": 14138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.775421738326568e-08,
+      "loss": 0.7992,
+      "step": 14139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.7619188653387764e-08,
+      "loss": 0.6021,
+      "step": 14140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.748440091097816e-08,
+      "loss": 0.8866,
+      "step": 14141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.734985416258219e-08,
+      "loss": 0.724,
+      "step": 14142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.721554841473407e-08,
+      "loss": 0.7198,
+      "step": 14143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.708148367395637e-08,
+      "loss": 0.466,
+      "step": 14144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.694765994675886e-08,
+      "loss": 0.8381,
+      "step": 14145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6814077239641366e-08,
+      "loss": 0.9497,
+      "step": 14146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.668073555909091e-08,
+      "loss": 0.6313,
+      "step": 14147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6547634911581755e-08,
+      "loss": 0.939,
+      "step": 14148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6414775303579285e-08,
+      "loss": 0.4491,
+      "step": 14149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.628215674153501e-08,
+      "loss": 0.6703,
+      "step": 14150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.614977923188823e-08,
+      "loss": 0.6272,
+      "step": 14151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6017642781068784e-08,
+      "loss": 0.6656,
+      "step": 14152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.588574739549322e-08,
+      "loss": 0.6954,
+      "step": 14153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.575409308156697e-08,
+      "loss": 1.0197,
+      "step": 14154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5622679845682685e-08,
+      "loss": 0.8976,
+      "step": 14155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.54915076942236e-08,
+      "loss": 0.8593,
+      "step": 14156
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.536057663355852e-08,
+      "loss": 0.8142,
+      "step": 14157
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.5229886670046785e-08,
+      "loss": 0.5735,
+      "step": 14158
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.509943781003444e-08,
+      "loss": 0.7883,
+      "step": 14159
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.496923005985642e-08,
+      "loss": 1.0803,
+      "step": 14160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.483926342583654e-08,
+      "loss": 0.6718,
+      "step": 14161
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4709537914285887e-08,
+      "loss": 0.5981,
+      "step": 14162
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4580053531504396e-08,
+      "loss": 0.6081,
+      "step": 14163
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.445081028378039e-08,
+      "loss": 0.7498,
+      "step": 14164
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.432180817738995e-08,
+      "loss": 0.6757,
+      "step": 14165
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.4193047218598067e-08,
+      "loss": 0.7206,
+      "step": 14166
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.406452741365807e-08,
+      "loss": 0.6595,
+      "step": 14167
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.393624876881052e-08,
+      "loss": 0.6241,
+      "step": 14168
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3808211290284886e-08,
+      "loss": 0.6406,
+      "step": 14169
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.368041498430008e-08,
+      "loss": 0.5296,
+      "step": 14170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.355285985706169e-08,
+      "loss": 0.4521,
+      "step": 14171
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.3425545914763655e-08,
+      "loss": 0.4268,
+      "step": 14172
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.329847316358936e-08,
+      "loss": 0.7563,
+      "step": 14173
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.317164160970998e-08,
+      "loss": 0.7043,
+      "step": 14174
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.304505125928392e-08,
+      "loss": 0.6291,
+      "step": 14175
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.29187021184596e-08,
+      "loss": 0.8567,
+      "step": 14176
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.279259419337266e-08,
+      "loss": 0.6802,
+      "step": 14177
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.266672749014765e-08,
+      "loss": 0.7134,
+      "step": 14178
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2541102014895795e-08,
+      "loss": 0.9088,
+      "step": 14179
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.241571777371888e-08,
+      "loss": 0.8599,
+      "step": 14180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.229057477270647e-08,
+      "loss": 0.5064,
+      "step": 14181
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.2165673017933716e-08,
+      "loss": 0.5185,
+      "step": 14182
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.204101251546854e-08,
+      "loss": 0.8587,
+      "step": 14183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.191659327136387e-08,
+      "loss": 0.646,
+      "step": 14184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.179241529166099e-08,
+      "loss": 0.8197,
+      "step": 14185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.16684785823923e-08,
+      "loss": 0.6432,
+      "step": 14186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.154478314957465e-08,
+      "loss": 0.779,
+      "step": 14187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1421328999216014e-08,
+      "loss": 0.9133,
+      "step": 14188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.129811613731104e-08,
+      "loss": 0.8244,
+      "step": 14189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.1175144569844386e-08,
+      "loss": 0.6088,
+      "step": 14190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.105241430278683e-08,
+      "loss": 0.4145,
+      "step": 14191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0929925342099156e-08,
+      "loss": 0.9647,
+      "step": 14192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.080767769372939e-08,
+      "loss": 0.553,
+      "step": 14193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.068567136361444e-08,
+      "loss": 0.9355,
+      "step": 14194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.056390635767903e-08,
+      "loss": 0.8152,
+      "step": 14195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0442382681836745e-08,
+      "loss": 0.7604,
+      "step": 14196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0321100341988986e-08,
+      "loss": 0.7288,
+      "step": 14197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.020005934402603e-08,
+      "loss": 0.8122,
+      "step": 14198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.007925969382486e-08,
+      "loss": 0.5917,
+      "step": 14199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9958701397252984e-08,
+      "loss": 0.4877,
+      "step": 14200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.983838446016407e-08,
+      "loss": 0.7101,
+      "step": 14201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.971830888840177e-08,
+      "loss": 1.0157,
+      "step": 14202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.959847468779753e-08,
+      "loss": 0.8985,
+      "step": 14203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.947888186417003e-08,
+      "loss": 0.9956,
+      "step": 14204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.93595304233274e-08,
+      "loss": 0.5733,
+      "step": 14205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9240420371066113e-08,
+      "loss": 0.9196,
+      "step": 14206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9121551713169882e-08,
+      "loss": 0.6173,
+      "step": 14207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9002924455411308e-08,
+      "loss": 0.7819,
+      "step": 14208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8884538603551892e-08,
+      "loss": 0.9022,
+      "step": 14209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8766394163339816e-08,
+      "loss": 0.9455,
+      "step": 14210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8648491140513267e-08,
+      "loss": 0.6933,
+      "step": 14211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.853082954079822e-08,
+      "loss": 0.7334,
+      "step": 14212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8413409369907887e-08,
+      "loss": 0.8067,
+      "step": 14213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8296230633544918e-08,
+      "loss": 0.5442,
+      "step": 14214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.817929333739977e-08,
+      "loss": 0.8936,
+      "step": 14215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.8062597487151232e-08,
+      "loss": 0.7335,
+      "step": 14216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7946143088466437e-08,
+      "loss": 0.7617,
+      "step": 14217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.782993014700086e-08,
+      "loss": 0.6597,
+      "step": 14218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7713958668397767e-08,
+      "loss": 0.662,
+      "step": 14219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7598228658289873e-08,
+      "loss": 0.6027,
+      "step": 14220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7482740122296015e-08,
+      "loss": 0.5131,
+      "step": 14221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7367493066026153e-08,
+      "loss": 0.5422,
+      "step": 14222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7252487495075807e-08,
+      "loss": 0.7341,
+      "step": 14223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7137723415030516e-08,
+      "loss": 0.7533,
+      "step": 14224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7023200831463592e-08,
+      "loss": 0.6867,
+      "step": 14225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.69089197499367e-08,
+      "loss": 0.7221,
+      "step": 14226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6794880175999293e-08,
+      "loss": 0.7761,
+      "step": 14227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6681082115189717e-08,
+      "loss": 0.6237,
+      "step": 14228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6567525573034102e-08,
+      "loss": 0.7139,
+      "step": 14229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6454210555046933e-08,
+      "loss": 0.8652,
+      "step": 14230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.634113706673158e-08,
+      "loss": 0.7133,
+      "step": 14231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6228305113579766e-08,
+      "loss": 0.6899,
+      "step": 14232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6115714701069327e-08,
+      "loss": 0.588,
+      "step": 14233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.6003365834669226e-08,
+      "loss": 0.7004,
+      "step": 14234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5891258519835094e-08,
+      "loss": 0.5573,
+      "step": 14235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.577939276201036e-08,
+      "loss": 0.6146,
+      "step": 14236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.566776856662956e-08,
+      "loss": 0.5891,
+      "step": 14237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5556385939111693e-08,
+      "loss": 0.8628,
+      "step": 14238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5445244884865772e-08,
+      "loss": 0.6557,
+      "step": 14239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5334345409290805e-08,
+      "loss": 0.6488,
+      "step": 14240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5223687517770823e-08,
+      "loss": 0.7051,
+      "step": 14241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.511327121567986e-08,
+      "loss": 0.5119,
+      "step": 14242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.5003096508381396e-08,
+      "loss": 0.7443,
+      "step": 14243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4893163401223942e-08,
+      "loss": 0.8501,
+      "step": 14244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4783471899547662e-08,
+      "loss": 0.6719,
+      "step": 14245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.46740220086783e-08,
+      "loss": 0.5095,
+      "step": 14246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4564813733932157e-08,
+      "loss": 0.7422,
+      "step": 14247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.445584708061166e-08,
+      "loss": 0.3945,
+      "step": 14248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.434712205400924e-08,
+      "loss": 0.5422,
+      "step": 14249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4238638659405124e-08,
+      "loss": 0.817,
+      "step": 14250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4130396902066755e-08,
+      "loss": 0.6629,
+      "step": 14251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.4022396787251045e-08,
+      "loss": 0.7174,
+      "step": 14252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3914638320203243e-08,
+      "loss": 0.3924,
+      "step": 14253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.380712150615583e-08,
+      "loss": 0.6949,
+      "step": 14254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.369984635033018e-08,
+      "loss": 0.7885,
+      "step": 14255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3592812857935467e-08,
+      "loss": 0.6666,
+      "step": 14256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3486021034170857e-08,
+      "loss": 0.7815,
+      "step": 14257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3379470884221656e-08,
+      "loss": 0.5518,
+      "step": 14258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3273162413262053e-08,
+      "loss": 0.779,
+      "step": 14259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.3167095626455137e-08,
+      "loss": 0.6454,
+      "step": 14260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.306127052895124e-08,
+      "loss": 0.6948,
+      "step": 14261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2955687125890136e-08,
+      "loss": 0.5884,
+      "step": 14262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2850345422398834e-08,
+      "loss": 0.57,
+      "step": 14263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2745245423593244e-08,
+      "loss": 0.8166,
+      "step": 14264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.264038713457706e-08,
+      "loss": 0.7132,
+      "step": 14265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.253577056044287e-08,
+      "loss": 0.6162,
+      "step": 14266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2431395706271064e-08,
+      "loss": 0.6769,
+      "step": 14267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2327262577130358e-08,
+      "loss": 0.7931,
+      "step": 14268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2223371178077824e-08,
+      "loss": 0.6481,
+      "step": 14269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2119721514158865e-08,
+      "loss": 0.7644,
+      "step": 14270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.2016313590406125e-08,
+      "loss": 0.7126,
+      "step": 14271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.191314741184225e-08,
+      "loss": 0.752,
+      "step": 14272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1810222983477127e-08,
+      "loss": 0.621,
+      "step": 14273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1707540310309528e-08,
+      "loss": 0.4263,
+      "step": 14274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1605099397324914e-08,
+      "loss": 0.6977,
+      "step": 14275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1502900249498748e-08,
+      "loss": 1.0601,
+      "step": 14276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1400942871794283e-08,
+      "loss": 0.6642,
+      "step": 14277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.129922726916256e-08,
+      "loss": 0.7857,
+      "step": 14278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1197753446542958e-08,
+      "loss": 0.8715,
+      "step": 14279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.1096521408864312e-08,
+      "loss": 0.9045,
+      "step": 14280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0995531161041028e-08,
+      "loss": 0.4933,
+      "step": 14281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0894782707979733e-08,
+      "loss": 0.5962,
+      "step": 14282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.079427605457096e-08,
+      "loss": 0.6957,
+      "step": 14283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.069401120569692e-08,
+      "loss": 0.6929,
+      "step": 14284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0593988166226485e-08,
+      "loss": 0.8069,
+      "step": 14285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0494206941016336e-08,
+      "loss": 0.6451,
+      "step": 14286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.039466753491315e-08,
+      "loss": 0.4114,
+      "step": 14287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0295369952750278e-08,
+      "loss": 0.6561,
+      "step": 14288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.0196314199349977e-08,
+      "loss": 0.4738,
+      "step": 14289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.009750027952284e-08,
+      "loss": 0.9821,
+      "step": 14290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.999892819806726e-08,
+      "loss": 0.7617,
+      "step": 14291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9900597959770506e-08,
+      "loss": 0.3614,
+      "step": 14292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9802509569407656e-08,
+      "loss": 0.5467,
+      "step": 14293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.970466303174212e-08,
+      "loss": 0.7262,
+      "step": 14294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9607058351525655e-08,
+      "loss": 0.6355,
+      "step": 14295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.950969553349835e-08,
+      "loss": 0.5876,
+      "step": 14296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.94125745823881e-08,
+      "loss": 0.7398,
+      "step": 14297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9315695502911126e-08,
+      "loss": 0.699,
+      "step": 14298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9219058299773106e-08,
+      "loss": 0.8595,
+      "step": 14299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9122662977666405e-08,
+      "loss": 0.422,
+      "step": 14300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.9026509541272276e-08,
+      "loss": 1.0233,
+      "step": 14301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8930597995260315e-08,
+      "loss": 0.5215,
+      "step": 14302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8834928344288463e-08,
+      "loss": 0.7916,
+      "step": 14303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.873950059300189e-08,
+      "loss": 0.4856,
+      "step": 14304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.864431474603634e-08,
+      "loss": 0.721,
+      "step": 14305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8549370808012556e-08,
+      "loss": 0.4665,
+      "step": 14306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8454668783542407e-08,
+      "loss": 0.7414,
+      "step": 14307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8360208677224434e-08,
+      "loss": 0.6237,
+      "step": 14308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.826599049364608e-08,
+      "loss": 0.8735,
+      "step": 14309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8172014237383128e-08,
+      "loss": 0.7463,
+      "step": 14310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8078279912999153e-08,
+      "loss": 0.8956,
+      "step": 14311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7984787525045512e-08,
+      "loss": 0.668,
+      "step": 14312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.789153707806357e-08,
+      "loss": 0.6068,
+      "step": 14313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.779852857658082e-08,
+      "loss": 0.5807,
+      "step": 14314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.770576202511476e-08,
+      "loss": 0.7749,
+      "step": 14315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7613237428170115e-08,
+      "loss": 0.8807,
+      "step": 14316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7520954790239964e-08,
+      "loss": 0.6187,
+      "step": 14317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.742891411580627e-08,
+      "loss": 0.8942,
+      "step": 14318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7337115409338246e-08,
+      "loss": 0.9076,
+      "step": 14319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7245558675294538e-08,
+      "loss": 0.9472,
+      "step": 14320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7154243918121038e-08,
+      "loss": 0.8437,
+      "step": 14321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7063171142251977e-08,
+      "loss": 1.0103,
+      "step": 14322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6972340352110485e-08,
+      "loss": 0.9569,
+      "step": 14323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6881751552107472e-08,
+      "loss": 0.8856,
+      "step": 14324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6791404746642203e-08,
+      "loss": 0.9695,
+      "step": 14325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.670129994010228e-08,
+      "loss": 1.0107,
+      "step": 14326
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6611437136863083e-08,
+      "loss": 0.8833,
+      "step": 14327
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.652181634128891e-08,
+      "loss": 0.6776,
+      "step": 14328
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6432437557732383e-08,
+      "loss": 0.7297,
+      "step": 14329
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.6343300790533367e-08,
+      "loss": 0.9385,
+      "step": 14330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.625440604402062e-08,
+      "loss": 0.672,
+      "step": 14331
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6165753322511247e-08,
+      "loss": 0.6926,
+      "step": 14332
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.607734263031069e-08,
+      "loss": 0.7666,
+      "step": 14333
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.598917397171218e-08,
+      "loss": 0.7984,
+      "step": 14334
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.590124735099785e-08,
+      "loss": 0.9359,
+      "step": 14335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5813562772437063e-08,
+      "loss": 0.6322,
+      "step": 14336
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5726120240288632e-08,
+      "loss": 0.6132,
+      "step": 14337
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5638919758798054e-08,
+      "loss": 0.8749,
+      "step": 14338
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5551961332201382e-08,
+      "loss": 0.7329,
+      "step": 14339
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5465244964720795e-08,
+      "loss": 0.5182,
+      "step": 14340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5378770660567367e-08,
+      "loss": 0.9295,
+      "step": 14341
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5292538423940518e-08,
+      "loss": 0.5337,
+      "step": 14342
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5206548259028565e-08,
+      "loss": 0.5816,
+      "step": 14343
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5120800170007056e-08,
+      "loss": 0.5996,
+      "step": 14344
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5035294161039882e-08,
+      "loss": 0.7439,
+      "step": 14345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4950030236279832e-08,
+      "loss": 0.7247,
+      "step": 14346
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4865008399867486e-08,
+      "loss": 0.7635,
+      "step": 14347
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4780228655931206e-08,
+      "loss": 0.5874,
+      "step": 14348
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4695691008589363e-08,
+      "loss": 0.4307,
+      "step": 14349
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4611395461946454e-08,
+      "loss": 0.7517,
+      "step": 14350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4527342020096424e-08,
+      "loss": 0.8731,
+      "step": 14351
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4443530687120455e-08,
+      "loss": 0.7837,
+      "step": 14352
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4359961467089734e-08,
+      "loss": 0.9425,
+      "step": 14353
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4276634364062124e-08,
+      "loss": 0.7313,
+      "step": 14354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4193549382083837e-08,
+      "loss": 0.3781,
+      "step": 14355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4110706525190532e-08,
+      "loss": 0.8229,
+      "step": 14356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4028105797404546e-08,
+      "loss": 0.7631,
+      "step": 14357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3945747202738224e-08,
+      "loss": 0.7449,
+      "step": 14358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3863630745189483e-08,
+      "loss": 0.817,
+      "step": 14359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3781756428747351e-08,
+      "loss": 0.7089,
+      "step": 14360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3700124257388092e-08,
+      "loss": 0.5949,
+      "step": 14361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3618734235075203e-08,
+      "loss": 0.7821,
+      "step": 14362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3537586365761635e-08,
+      "loss": 0.711,
+      "step": 14363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3456680653388121e-08,
+      "loss": 0.9195,
+      "step": 14364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3376017101883742e-08,
+      "loss": 0.6798,
+      "step": 14365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3295595715165366e-08,
+      "loss": 0.8141,
+      "step": 14366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3215416497138756e-08,
+      "loss": 0.7577,
+      "step": 14367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3135479451697464e-08,
+      "loss": 0.7163,
+      "step": 14368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3055784582723941e-08,
+      "loss": 0.8232,
+      "step": 14369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2976331894088423e-08,
+      "loss": 0.6831,
+      "step": 14370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2897121389648381e-08,
+      "loss": 0.6376,
+      "step": 14371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2818153073251294e-08,
+      "loss": 0.8071,
+      "step": 14372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2739426948732426e-08,
+      "loss": 0.6565,
+      "step": 14373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2660943019914274e-08,
+      "loss": 0.5967,
+      "step": 14374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2582701290608234e-08,
+      "loss": 0.7117,
+      "step": 14375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2504701764614047e-08,
+      "loss": 0.8515,
+      "step": 14376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2426944445719791e-08,
+      "loss": 0.7328,
+      "step": 14377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.234942933770189e-08,
+      "loss": 0.8775,
+      "step": 14378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2272156444324001e-08,
+      "loss": 0.7546,
+      "step": 14379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2195125769339233e-08,
+      "loss": 0.5715,
+      "step": 14380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2118337316487928e-08,
+      "loss": 0.8742,
+      "step": 14381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2041791089499878e-08,
+      "loss": 0.6094,
+      "step": 14382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1965487092091554e-08,
+      "loss": 0.8097,
+      "step": 14383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1889425327968884e-08,
+      "loss": 0.7716,
+      "step": 14384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1813605800825578e-08,
+      "loss": 0.8828,
+      "step": 14385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1738028514344246e-08,
+      "loss": 0.7981,
+      "step": 14386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1662693472194175e-08,
+      "loss": 0.6355,
+      "step": 14387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.158760067803466e-08,
+      "loss": 1.0869,
+      "step": 14388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1512750135511674e-08,
+      "loss": 0.6463,
+      "step": 14389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1438141848260642e-08,
+      "loss": 0.6566,
+      "step": 14390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1363775819904776e-08,
+      "loss": 0.8249,
+      "step": 14391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1289652054055633e-08,
+      "loss": 0.9069,
+      "step": 14392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1215770554312e-08,
+      "loss": 0.5839,
+      "step": 14393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1142131324262673e-08,
+      "loss": 0.6341,
+      "step": 14394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.106873436748368e-08,
+      "loss": 0.5804,
+      "step": 14395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0995579687539392e-08,
+      "loss": 0.5894,
+      "step": 14396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.092266728798197e-08,
+      "loss": 0.7095,
+      "step": 14397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0849997172352466e-08,
+      "loss": 0.6758,
+      "step": 14398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0777569344180283e-08,
+      "loss": 0.6538,
+      "step": 14399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0705383806982606e-08,
+      "loss": 0.729,
+      "step": 14400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0633440564264408e-08,
+      "loss": 1.0054,
+      "step": 14401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0561739619520118e-08,
+      "loss": 0.6978,
+      "step": 14402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0490280976231392e-08,
+      "loss": 0.7236,
+      "step": 14403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0419064637868791e-08,
+      "loss": 0.9545,
+      "step": 14404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0348090607890104e-08,
+      "loss": 0.7218,
+      "step": 14405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0277358889743128e-08,
+      "loss": 0.6725,
+      "step": 14406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0206869486861781e-08,
+      "loss": 0.6222,
+      "step": 14407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0136622402669438e-08,
+      "loss": 0.7143,
+      "step": 14408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.006661764057837e-08,
+      "loss": 0.5834,
+      "step": 14409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.996855203986966e-09,
+      "loss": 0.532,
+      "step": 14410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.92733509628363e-09,
+      "loss": 0.8006,
+      "step": 14411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.858057320844549e-09,
+      "loss": 0.7579,
+      "step": 14412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.789021881034255e-09,
+      "loss": 0.8389,
+      "step": 14413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.72022878020451e-09,
+      "loss": 0.4926,
+      "step": 14414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.651678021697086e-09,
+      "loss": 0.3716,
+      "step": 14415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.58336960884043e-09,
+      "loss": 0.57,
+      "step": 14416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.515303544951337e-09,
+      "loss": 0.6919,
+      "step": 14417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.447479833336604e-09,
+      "loss": 0.721,
+      "step": 14418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.379898477289152e-09,
+      "loss": 0.9465,
+      "step": 14419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.312559480090799e-09,
+      "loss": 0.6652,
+      "step": 14420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.24546284501171e-09,
+      "loss": 0.8746,
+      "step": 14421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.178608575310387e-09,
+      "loss": 0.5705,
+      "step": 14422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.111996674234235e-09,
+      "loss": 0.7533,
+      "step": 14423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.045627145016778e-09,
+      "loss": 0.8585,
+      "step": 14424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.979499990882102e-09,
+      "loss": 0.7787,
+      "step": 14425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.913615215040972e-09,
+      "loss": 0.9318,
+      "step": 14426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.847972820693052e-09,
+      "loss": 0.569,
+      "step": 14427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.782572811026901e-09,
+      "loss": 0.7051,
+      "step": 14428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.717415189217204e-09,
+      "loss": 0.7847,
+      "step": 14429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.652499958429205e-09,
+      "loss": 0.9647,
+      "step": 14430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.587827121814829e-09,
+      "loss": 0.8009,
+      "step": 14431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.52339668251545e-09,
+      "loss": 0.5458,
+      "step": 14432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.459208643659122e-09,
+      "loss": 0.6183,
+      "step": 14433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.395263008363353e-09,
+      "loss": 0.5925,
+      "step": 14434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.331559779733989e-09,
+      "loss": 0.6744,
+      "step": 14435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.268098960863557e-09,
+      "loss": 0.7271,
+      "step": 14436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.204880554835149e-09,
+      "loss": 0.8724,
+      "step": 14437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.141904564717973e-09,
+      "loss": 0.5679,
+      "step": 14438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.079170993570695e-09,
+      "loss": 0.7059,
+      "step": 14439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.016679844439768e-09,
+      "loss": 0.8838,
+      "step": 14440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.954431120359985e-09,
+      "loss": 0.6235,
+      "step": 14441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.892424824354484e-09,
+      "loss": 0.6835,
+      "step": 14442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.83066095943419e-09,
+      "loss": 0.6961,
+      "step": 14443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.769139528598368e-09,
+      "loss": 0.8426,
+      "step": 14444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.707860534834632e-09,
+      "loss": 0.5645,
+      "step": 14445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.646823981119489e-09,
+      "loss": 0.7428,
+      "step": 14446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.586029870416678e-09,
+      "loss": 0.6193,
+      "step": 14447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.525478205678838e-09,
+      "loss": 0.6745,
+      "step": 14448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.465168989845839e-09,
+      "loss": 0.8417,
+      "step": 14449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.405102225847005e-09,
+      "loss": 0.6941,
+      "step": 14450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.345277916599447e-09,
+      "loss": 0.6665,
+      "step": 14451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.285696065007508e-09,
+      "loss": 0.6079,
+      "step": 14452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.226356673966095e-09,
+      "loss": 0.8757,
+      "step": 14453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.167259746355681e-09,
+      "loss": 0.6529,
+      "step": 14454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.108405285046749e-09,
+      "loss": 0.7104,
+      "step": 14455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.049793292897011e-09,
+      "loss": 0.742,
+      "step": 14456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.991423772753636e-09,
+      "loss": 0.7543,
+      "step": 14457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.933296727450467e-09,
+      "loss": 0.396,
+      "step": 14458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.8754121598108015e-09,
+      "loss": 0.8584,
+      "step": 14459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.817770072645169e-09,
+      "loss": 0.7792,
+      "step": 14460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.760370468753552e-09,
+      "loss": 0.412,
+      "step": 14461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.703213350922611e-09,
+      "loss": 0.486,
+      "step": 14462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.646298721928457e-09,
+      "loss": 0.8646,
+      "step": 14463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.5896265845355465e-09,
+      "loss": 0.671,
+      "step": 14464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.533196941495013e-09,
+      "loss": 0.6159,
+      "step": 14465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.4770097955479945e-09,
+      "loss": 0.9524,
+      "step": 14466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.421065149422867e-09,
+      "loss": 0.6742,
+      "step": 14467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.365363005836345e-09,
+      "loss": 0.7258,
+      "step": 14468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.309903367493486e-09,
+      "loss": 0.6572,
+      "step": 14469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.254686237087693e-09,
+      "loss": 0.7787,
+      "step": 14470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.1997116173007075e-09,
+      "loss": 0.5312,
+      "step": 14471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.144979510802063e-09,
+      "loss": 0.6655,
+      "step": 14472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.090489920249076e-09,
+      "loss": 0.6807,
+      "step": 14473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.036242848289075e-09,
+      "loss": 0.8133,
+      "step": 14474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.982238297555509e-09,
+      "loss": 0.9164,
+      "step": 14475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.928476270671835e-09,
+      "loss": 0.5401,
+      "step": 14476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.874956770248186e-09,
+      "loss": 0.7984,
+      "step": 14477
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.821679798883595e-09,
+      "loss": 0.7223,
+      "step": 14478
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.768645359165992e-09,
+      "loss": 0.6471,
+      "step": 14479
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.715853453669984e-09,
+      "loss": 0.6866,
+      "step": 14480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.6633040849601865e-09,
+      "loss": 0.7791,
+      "step": 14481
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.610997255587891e-09,
+      "loss": 0.8216,
+      "step": 14482
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.558932968093844e-09,
+      "loss": 0.7884,
+      "step": 14483
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.507111225006023e-09,
+      "loss": 0.8398,
+      "step": 14484
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.455532028841304e-09,
+      "loss": 0.7322,
+      "step": 14485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.404195382104349e-09,
+      "loss": 0.8851,
+      "step": 14486
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.35310128728761e-09,
+      "loss": 1.0672,
+      "step": 14487
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.302249746873545e-09,
+      "loss": 0.4709,
+      "step": 14488
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2516407633312895e-09,
+      "loss": 0.3988,
+      "step": 14489
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.201274339117768e-09,
+      "loss": 0.7068,
+      "step": 14490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.151150476679912e-09,
+      "loss": 0.7178,
+      "step": 14491
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.101269178451329e-09,
+      "loss": 0.7677,
+      "step": 14492
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0516304468545274e-09,
+      "loss": 0.6847,
+      "step": 14493
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.002234284300356e-09,
+      "loss": 0.8738,
+      "step": 14494
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.953080693187451e-09,
+      "loss": 0.5958,
+      "step": 14495
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.904169675902237e-09,
+      "loss": 0.8427,
+      "step": 14496
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.855501234821147e-09,
+      "loss": 1.0356,
+      "step": 14497
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.8070753723067356e-09,
+      "loss": 0.8744,
+      "step": 14498
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.758892090711009e-09,
+      "loss": 0.8677,
+      "step": 14499
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.710951392374319e-09,
+      "loss": 0.5703,
+      "step": 14500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.663253279624247e-09,
+      "loss": 0.865,
+      "step": 14501
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.615797754776719e-09,
+      "loss": 0.9825,
+      "step": 14502
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.568584820137667e-09,
+      "loss": 0.3988,
+      "step": 14503
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5216144779985924e-09,
+      "loss": 1.0617,
+      "step": 14504
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.474886730641004e-09,
+      "loss": 0.5637,
+      "step": 14505
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.428401580334196e-09,
+      "loss": 0.7031,
+      "step": 14506
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.382159029335808e-09,
+      "loss": 0.852,
+      "step": 14507
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.336159079891267e-09,
+      "loss": 0.9505,
+      "step": 14508
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.290401734234339e-09,
+      "loss": 0.7971,
+      "step": 14509
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.244886994587138e-09,
+      "loss": 0.6822,
+      "step": 14510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.199614863160673e-09,
+      "loss": 0.6396,
+      "step": 14511
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1545853421520735e-09,
+      "loss": 0.7764,
+      "step": 14512
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.10979843374959e-09,
+      "loss": 0.5031,
+      "step": 14513
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.065254140127595e-09,
+      "loss": 0.6613,
+      "step": 14514
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.020952463448802e-09,
+      "loss": 0.6615,
+      "step": 14515
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.976893405865379e-09,
+      "loss": 0.4937,
+      "step": 14516
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9330769695167245e-09,
+      "loss": 0.7982,
+      "step": 14517
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.889503156530583e-09,
+      "loss": 0.5319,
+      "step": 14518
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8461719690235936e-09,
+      "loss": 0.6377,
+      "step": 14519
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.803083409098518e-09,
+      "loss": 0.8003,
+      "step": 14520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.760237478849793e-09,
+      "loss": 0.6567,
+      "step": 14521
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.717634180356866e-09,
+      "loss": 0.7781,
+      "step": 14522
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6752735156886377e-09,
+      "loss": 1.1321,
+      "step": 14523
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6331554869034612e-09,
+      "loss": 0.7096,
+      "step": 14524
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.591280096045258e-09,
+      "loss": 1.0821,
+      "step": 14525
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5496473451485105e-09,
+      "loss": 0.5744,
+      "step": 14526
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5082572362349354e-09,
+      "loss": 0.7739,
+      "step": 14527
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.467109771314037e-09,
+      "loss": 0.8048,
+      "step": 14528
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4262049523847707e-09,
+      "loss": 0.77,
+      "step": 14529
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3855427814327714e-09,
+      "loss": 0.7229,
+      "step": 14530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3451232604336802e-09,
+      "loss": 0.4525,
+      "step": 14531
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.304946391349817e-09,
+      "loss": 0.7786,
+      "step": 14532
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.265012176131843e-09,
+      "loss": 0.7642,
+      "step": 14533
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2253206167198737e-09,
+      "loss": 0.8525,
+      "step": 14534
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1858717150412554e-09,
+      "loss": 0.7638,
+      "step": 14535
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1466654730122338e-09,
+      "loss": 0.5466,
+      "step": 14536
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1077018925351753e-09,
+      "loss": 0.7607,
+      "step": 14537
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0689809755041213e-09,
+      "loss": 0.7579,
+      "step": 14538
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0305027237981234e-09,
+      "loss": 0.821,
+      "step": 14539
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9922671392862425e-09,
+      "loss": 0.8336,
+      "step": 14540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9542742238253263e-09,
+      "loss": 0.6289,
+      "step": 14541
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.916523979260566e-09,
+      "loss": 0.882,
+      "step": 14542
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8790164074254944e-09,
+      "loss": 0.6285,
+      "step": 14543
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.841751510140878e-09,
+      "loss": 0.8845,
+      "step": 14544
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.804729289216379e-09,
+      "loss": 0.748,
+      "step": 14545
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7679497464511152e-09,
+      "loss": 0.6111,
+      "step": 14546
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7314128836297692e-09,
+      "loss": 0.8169,
+      "step": 14547
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.695118702527033e-09,
+      "loss": 0.5344,
+      "step": 14548
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6590672049059407e-09,
+      "loss": 0.6395,
+      "step": 14549
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.623258392517314e-09,
+      "loss": 0.6411,
+      "step": 14550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.587692267099762e-09,
+      "loss": 0.7853,
+      "step": 14551
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5523688303802362e-09,
+      "loss": 0.6162,
+      "step": 14552
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5172880840745873e-09,
+      "loss": 0.5745,
+      "step": 14553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.482450029885897e-09,
+      "loss": 0.6315,
+      "step": 14554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4478546695066996e-09,
+      "loss": 0.8416,
+      "step": 14555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4135020046167636e-09,
+      "loss": 0.8372,
+      "step": 14556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3793920368841984e-09,
+      "loss": 0.524,
+      "step": 14557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.345524767966012e-09,
+      "loss": 0.7998,
+      "step": 14558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.311900199506445e-09,
+      "loss": 0.694,
+      "step": 14559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.27851833313808e-09,
+      "loss": 0.747,
+      "step": 14560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.245379170482398e-09,
+      "loss": 0.9112,
+      "step": 14561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2124827131492222e-09,
+      "loss": 0.5225,
+      "step": 14562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1798289627350535e-09,
+      "loss": 0.7238,
+      "step": 14563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1474179208264e-09,
+      "loss": 0.5037,
+      "step": 14564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1152495889970035e-09,
+      "loss": 0.6381,
+      "step": 14565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.083323968809503e-09,
+      "loss": 0.893,
+      "step": 14566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.051641061813214e-09,
+      "loss": 0.6312,
+      "step": 14567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.020200869548017e-09,
+      "loss": 0.4677,
+      "step": 14568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.989003393539912e-09,
+      "loss": 0.9037,
+      "step": 14569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.958048635304355e-09,
+      "loss": 0.9859,
+      "step": 14570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9273365963440315e-09,
+      "loss": 0.6995,
+      "step": 14571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8968672781510823e-09,
+      "loss": 0.9315,
+      "step": 14572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8666406822048787e-09,
+      "loss": 0.591,
+      "step": 14573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8366568099736915e-09,
+      "loss": 0.6519,
+      "step": 14574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8069156629130225e-09,
+      "loss": 0.7102,
+      "step": 14575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7774172424672719e-09,
+      "loss": 0.6347,
+      "step": 14576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7481615500691829e-09,
+      "loss": 0.6239,
+      "step": 14577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7191485871392855e-09,
+      "loss": 0.5753,
+      "step": 14578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6903783550864527e-09,
+      "loss": 0.7677,
+      "step": 14579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6618508553084556e-09,
+      "loss": 0.7122,
+      "step": 14580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6335660891902972e-09,
+      "loss": 0.7683,
+      "step": 14581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6055240581058784e-09,
+      "loss": 0.5797,
+      "step": 14582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5777247634163328e-09,
+      "loss": 0.8092,
+      "step": 14583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5501682064722467e-09,
+      "loss": 0.8065,
+      "step": 14584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5228543886114389e-09,
+      "loss": 0.6673,
+      "step": 14585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4957833111600706e-09,
+      "loss": 0.7105,
+      "step": 14586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4689549754337563e-09,
+      "loss": 0.4862,
+      "step": 14587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4423693827347873e-09,
+      "loss": 0.6049,
+      "step": 14588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4160265343549084e-09,
+      "loss": 0.4359,
+      "step": 14589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3899264315719862e-09,
+      "loss": 0.7611,
+      "step": 14590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3640690756544506e-09,
+      "loss": 0.5624,
+      "step": 14591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3384544678579636e-09,
+      "loss": 0.9561,
+      "step": 14592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3130826094265304e-09,
+      "loss": 0.5025,
+      "step": 14593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2879535015919432e-09,
+      "loss": 0.5718,
+      "step": 14594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2630671455748922e-09,
+      "loss": 0.6513,
+      "step": 14595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2384235425838554e-09,
+      "loss": 0.8669,
+      "step": 14596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.214022693815653e-09,
+      "loss": 0.9699,
+      "step": 14597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.189864600454338e-09,
+      "loss": 0.6606,
+      "step": 14598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1659492636745262e-09,
+      "loss": 0.9893,
+      "step": 14599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.142276684636956e-09,
+      "loss": 0.711,
+      "step": 14600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.118846864490708e-09,
+      "loss": 0.4191,
+      "step": 14601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0956598043748713e-09,
+      "loss": 0.7727,
+      "step": 14602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0727155054141014e-09,
+      "loss": 0.5142,
+      "step": 14603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0500139687236178e-09,
+      "loss": 0.6779,
+      "step": 14604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0275551954053166e-09,
+      "loss": 0.6964,
+      "step": 14605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.005339186550547e-09,
+      "loss": 0.9185,
+      "step": 14606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.833659432367803e-10,
+      "loss": 0.9238,
+      "step": 14607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.616354665326066e-10,
+      "loss": 0.7163,
+      "step": 14608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.401477574932927e-10,
+      "loss": 0.7913,
+      "step": 14609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.189028171607827e-10,
+      "loss": 0.5813,
+      "step": 14610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.979006465681395e-10,
+      "loss": 0.8938,
+      "step": 14611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.771412467356577e-10,
+      "loss": 0.7231,
+      "step": 14612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.566246186697547e-10,
+      "loss": 0.6086,
+      "step": 14613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.363507633685209e-10,
+      "loss": 0.62,
+      "step": 14614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.163196818150587e-10,
+      "loss": 0.7843,
+      "step": 14615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.96531374983589e-10,
+      "loss": 0.653,
+      "step": 14616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.769858438338995e-10,
+      "loss": 0.4763,
+      "step": 14617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.57683089315786e-10,
+      "loss": 0.8266,
+      "step": 14618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.386231123668319e-10,
+      "loss": 0.7616,
+      "step": 14619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.198059139118529e-10,
+      "loss": 0.8976,
+      "step": 14620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.012314948651178e-10,
+      "loss": 0.6169,
+      "step": 14621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.828998561286826e-10,
+      "loss": 0.6042,
+      "step": 14622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.648109985929462e-10,
+      "loss": 0.9308,
+      "step": 14623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.469649231360952e-10,
+      "loss": 0.7455,
+      "step": 14624
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.293616306246586e-10,
+      "loss": 0.6775,
+      "step": 14625
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.120011219140631e-10,
+      "loss": 0.9081,
+      "step": 14626
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.948833978469682e-10,
+      "loss": 0.4099,
+      "step": 14627
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.780084592543756e-10,
+      "loss": 0.7088,
+      "step": 14628
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.61376306956185e-10,
+      "loss": 0.5942,
+      "step": 14629
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.449869417606391e-10,
+      "loss": 0.8748,
+      "step": 14630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.288403644626571e-10,
+      "loss": 0.476,
+      "step": 14631
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.12936575846612e-10,
+      "loss": 0.892,
+      "step": 14632
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.972755766846637e-10,
+      "loss": 0.9108,
+      "step": 14633
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.818573677384253e-10,
+      "loss": 0.7872,
+      "step": 14634
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.666819497556318e-10,
+      "loss": 0.6706,
+      "step": 14635
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.5174932347347153e-10,
+      "loss": 0.5647,
+      "step": 14636
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3705948961692003e-10,
+      "loss": 0.7972,
+      "step": 14637
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2261244889985067e-10,
+      "loss": 0.6229,
+      "step": 14638
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.0840820202336927e-10,
+      "loss": 0.8749,
+      "step": 14639
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9444674967803465e-10,
+      "loss": 0.7884,
+      "step": 14640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8072809254108277e-10,
+      "loss": 0.8244,
+      "step": 14641
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.672522312786475e-10,
+      "loss": 0.8907,
+      "step": 14642
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.540191665457604e-10,
+      "loss": 0.9555,
+      "step": 14643
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.410288989846855e-10,
+      "loss": 0.953,
+      "step": 14644
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2828142922658456e-10,
+      "loss": 0.8872,
+      "step": 14645
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.157767578898518e-10,
+      "loss": 0.7558,
+      "step": 14646
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0351488558233444e-10,
+      "loss": 0.5081,
+      "step": 14647
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.91495812899667e-10,
+      "loss": 0.3492,
+      "step": 14648
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.797195404247166e-10,
+      "loss": 0.9498,
+      "step": 14649
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6818606873035836e-10,
+      "loss": 0.7156,
+      "step": 14650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.568953983761446e-10,
+      "loss": 0.5016,
+      "step": 14651
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4584752990997053e-10,
+      "loss": 0.5426,
+      "step": 14652
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3504246386918394e-10,
+      "loss": 0.8729,
+      "step": 14653
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.244802007783653e-10,
+      "loss": 0.5982,
+      "step": 14654
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.141607411498825e-10,
+      "loss": 0.8579,
+      "step": 14655
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0408408548500126e-10,
+      "loss": 0.6209,
+      "step": 14656
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.94250234273885e-10,
+      "loss": 0.8989,
+      "step": 14657
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8465918799337458e-10,
+      "loss": 0.7172,
+      "step": 14658
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.753109471092085e-10,
+      "loss": 0.7205,
+      "step": 14659
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6620551207546799e-10,
+      "loss": 0.6617,
+      "step": 14660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5734288333457692e-10,
+      "loss": 0.661,
+      "step": 14661
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4872306131674675e-10,
+      "loss": 0.6341,
+      "step": 14662
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4034604644053152e-10,
+      "loss": 0.7274,
+      "step": 14663
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.32211839112828e-10,
+      "loss": 0.5881,
+      "step": 14664
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2432043972832042e-10,
+      "loss": 0.6472,
+      "step": 14665
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.16671848671146e-10,
+      "loss": 0.5528,
+      "step": 14666
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0926606631156411e-10,
+      "loss": 0.7137,
+      "step": 14667
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0210309300984212e-10,
+      "loss": 0.6932,
+      "step": 14668
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.518292911403493e-11,
+      "loss": 0.6259,
+      "step": 14669
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.850557495998502e-11,
+      "loss": 0.6382,
+      "step": 14670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.20710308718775e-11,
+      "loss": 0.8995,
+      "step": 14671
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.587929716224018e-11,
+      "loss": 0.8106,
+      "step": 14672
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.993037413194348e-11,
+      "loss": 0.7224,
+      "step": 14673
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.422426206964538e-11,
+      "loss": 0.8834,
+      "step": 14674
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.876096125234653e-11,
+      "loss": 0.7777,
+      "step": 14675
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.354047194539025e-11,
+      "loss": 0.7487,
+      "step": 14676
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.856279440301759e-11,
+      "loss": 0.743,
+      "step": 14677
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.382792886614695e-11,
+      "loss": 0.7295,
+      "step": 14678
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9335875564594505e-11,
+      "loss": 0.6531,
+      "step": 14679
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.508663471707419e-11,
+      "loss": 0.9108,
+      "step": 14680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.108020653008748e-11,
+      "loss": 0.8436,
+      "step": 14681
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7316591197368292e-11,
+      "loss": 0.7091,
+      "step": 14682
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3795788902658546e-11,
+      "loss": 0.983,
+      "step": 14683
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.051779981582236e-11,
+      "loss": 0.733,
+      "step": 14684
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.748262409728696e-11,
+      "loss": 0.7099,
+      "step": 14685
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4690261893601786e-11,
+      "loss": 0.6306,
+      "step": 14686
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.214071334021405e-11,
+      "loss": 0.6659,
+      "step": 14687
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.833978562023838e-12,
+      "loss": 0.5789,
+      "step": 14688
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.770057670053455e-12,
+      "loss": 0.6789,
+      "step": 14689
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.948950764778083e-12,
+      "loss": 0.78,
+      "step": 14690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.370657934460454e-12,
+      "loss": 0.5813,
+      "step": 14691
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.035179256261067e-12,
+      "loss": 0.6589,
+      "step": 14692
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9425147945728582e-12,
+      "loss": 0.8774,
+      "step": 14693
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0926646032416444e-12,
+      "loss": 1.0412,
+      "step": 14694
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.856287222354538e-13,
+      "loss": 0.5988,
+      "step": 14695
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2140718208542013e-13,
+      "loss": 1.0653,
+      "step": 14696
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 0.541,
+      "step": 14697
+    },
+    {
+      "epoch": 1.0,
+      "step": 14697,
+      "total_flos": 1.6050245966611087e+18,
+      "train_loss": 0.7725475761118556,
+      "train_runtime": 502738.5976,
+      "train_samples_per_second": 0.468,
+      "train_steps_per_second": 0.029
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 14697,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "total_flos": 1.6050245966611087e+18,
+  "trial_name": null,
+  "trial_params": null
+}