diff --git "a/checkpoint-17578/trainer_state.json" "b/checkpoint-17578/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-17578/trainer_state.json"
@@ -0,0 +1,105793 @@
+{
+  "best_metric": 0.14592407643795013,
+  "best_model_checkpoint": "sms-spam-detection-distilbert/checkpoint-17578",
+  "epoch": 34.0,
+  "eval_steps": 1,
+  "global_step": 17578,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-09,
+      "loss": 0.6077,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-08,
+      "loss": 0.7147,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8e-08,
+      "loss": 0.577,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4e-08,
+      "loss": 0.5799,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-08,
+      "loss": 0.6932,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6e-08,
+      "loss": 0.7667,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2e-08,
+      "loss": 0.697,
+      "step": 7
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.8e-08,
+      "loss": 0.641,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3999999999999994e-08,
+      "loss": 0.6376,
+      "step": 9
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-08,
+      "loss": 0.7103,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.6e-08,
+      "loss": 0.6982,
+      "step": 11
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2e-08,
+      "loss": 0.7186,
+      "step": 12
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.8e-08,
+      "loss": 0.576,
+      "step": 13
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.4e-08,
+      "loss": 0.7489,
+      "step": 14
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9e-08,
+      "loss": 0.5871,
+      "step": 15
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.6e-08,
+      "loss": 0.7269,
+      "step": 16
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.02e-07,
+      "loss": 0.6255,
+      "step": 17
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0799999999999999e-07,
+      "loss": 0.684,
+      "step": 18
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.14e-07,
+      "loss": 0.6615,
+      "step": 19
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2e-07,
+      "loss": 0.684,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.26e-07,
+      "loss": 0.609,
+      "step": 21
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.32e-07,
+      "loss": 0.7892,
+      "step": 22
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.38e-07,
+      "loss": 0.6788,
+      "step": 23
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.44e-07,
+      "loss": 0.6952,
+      "step": 24
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5e-07,
+      "loss": 0.5913,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.56e-07,
+      "loss": 0.6831,
+      "step": 26
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.62e-07,
+      "loss": 0.6776,
+      "step": 27
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.68e-07,
+      "loss": 0.6629,
+      "step": 28
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.7399999999999997e-07,
+      "loss": 0.6976,
+      "step": 29
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8e-07,
+      "loss": 0.5984,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.86e-07,
+      "loss": 0.7765,
+      "step": 31
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.92e-07,
+      "loss": 0.5947,
+      "step": 32
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.98e-07,
+      "loss": 0.7051,
+      "step": 33
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.04e-07,
+      "loss": 0.6681,
+      "step": 34
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.0999999999999997e-07,
+      "loss": 0.6943,
+      "step": 35
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1599999999999998e-07,
+      "loss": 0.6083,
+      "step": 36
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.2199999999999998e-07,
+      "loss": 0.7168,
+      "step": 37
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.28e-07,
+      "loss": 0.7269,
+      "step": 38
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.34e-07,
+      "loss": 0.7264,
+      "step": 39
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.4e-07,
+      "loss": 0.6886,
+      "step": 40
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.4599999999999995e-07,
+      "loss": 0.7161,
+      "step": 41
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.52e-07,
+      "loss": 0.6032,
+      "step": 42
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.58e-07,
+      "loss": 0.7654,
+      "step": 43
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.64e-07,
+      "loss": 0.6059,
+      "step": 44
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.7e-07,
+      "loss": 0.6628,
+      "step": 45
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.76e-07,
+      "loss": 0.7566,
+      "step": 46
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.8199999999999996e-07,
+      "loss": 0.7187,
+      "step": 47
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.88e-07,
+      "loss": 0.7424,
+      "step": 48
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9399999999999996e-07,
+      "loss": 0.6537,
+      "step": 49
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3e-07,
+      "loss": 0.6786,
+      "step": 50
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.999883720930232e-07,
+      "loss": 0.7453,
+      "step": 51
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9997674418604646e-07,
+      "loss": 0.7568,
+      "step": 52
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9996511627906975e-07,
+      "loss": 0.724,
+      "step": 53
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9995348837209303e-07,
+      "loss": 0.7181,
+      "step": 54
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9994186046511627e-07,
+      "loss": 0.6034,
+      "step": 55
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.999302325581395e-07,
+      "loss": 0.711,
+      "step": 56
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.999186046511628e-07,
+      "loss": 0.7415,
+      "step": 57
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9990697674418603e-07,
+      "loss": 0.6255,
+      "step": 58
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9989534883720926e-07,
+      "loss": 0.768,
+      "step": 59
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9988372093023255e-07,
+      "loss": 0.6595,
+      "step": 60
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.998720930232558e-07,
+      "loss": 0.7033,
+      "step": 61
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.998604651162791e-07,
+      "loss": 0.7185,
+      "step": 62
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.998488372093023e-07,
+      "loss": 0.7143,
+      "step": 63
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9983720930232554e-07,
+      "loss": 0.7505,
+      "step": 64
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9982558139534883e-07,
+      "loss": 0.6988,
+      "step": 65
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9981395348837207e-07,
+      "loss": 0.6044,
+      "step": 66
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9980232558139536e-07,
+      "loss": 0.5671,
+      "step": 67
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.997906976744186e-07,
+      "loss": 0.6981,
+      "step": 68
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.997790697674419e-07,
+      "loss": 0.7368,
+      "step": 69
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.997674418604651e-07,
+      "loss": 0.615,
+      "step": 70
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9975581395348835e-07,
+      "loss": 0.6954,
+      "step": 71
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.997441860465116e-07,
+      "loss": 0.6998,
+      "step": 72
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9973255813953487e-07,
+      "loss": 0.6352,
+      "step": 73
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.997209302325581e-07,
+      "loss": 0.7396,
+      "step": 74
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.997093023255814e-07,
+      "loss": 0.6193,
+      "step": 75
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9969767441860463e-07,
+      "loss": 0.5944,
+      "step": 76
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9968604651162787e-07,
+      "loss": 0.6107,
+      "step": 77
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9967441860465116e-07,
+      "loss": 0.6894,
+      "step": 78
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.996627906976744e-07,
+      "loss": 0.6443,
+      "step": 79
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.996511627906977e-07,
+      "loss": 0.5871,
+      "step": 80
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.996395348837209e-07,
+      "loss": 0.6081,
+      "step": 81
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.996279069767442e-07,
+      "loss": 0.6833,
+      "step": 82
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9961627906976744e-07,
+      "loss": 0.6181,
+      "step": 83
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9960465116279067e-07,
+      "loss": 0.6874,
+      "step": 84
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.995930232558139e-07,
+      "loss": 0.6549,
+      "step": 85
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.995813953488372e-07,
+      "loss": 0.6763,
+      "step": 86
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9956976744186043e-07,
+      "loss": 0.6803,
+      "step": 87
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.995581395348837e-07,
+      "loss": 0.7286,
+      "step": 88
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9954651162790695e-07,
+      "loss": 0.701,
+      "step": 89
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9953488372093024e-07,
+      "loss": 0.6598,
+      "step": 90
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.995232558139535e-07,
+      "loss": 0.7093,
+      "step": 91
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.995116279069767e-07,
+      "loss": 0.7018,
+      "step": 92
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9949999999999995e-07,
+      "loss": 0.6719,
+      "step": 93
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9948837209302324e-07,
+      "loss": 0.6035,
+      "step": 94
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.994767441860465e-07,
+      "loss": 0.7296,
+      "step": 95
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9946511627906976e-07,
+      "loss": 0.6615,
+      "step": 96
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.99453488372093e-07,
+      "loss": 0.589,
+      "step": 97
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.994418604651163e-07,
+      "loss": 0.7383,
+      "step": 98
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.994302325581395e-07,
+      "loss": 0.6227,
+      "step": 99
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9941860465116275e-07,
+      "loss": 0.6802,
+      "step": 100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9940697674418604e-07,
+      "loss": 0.5848,
+      "step": 101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9939534883720933e-07,
+      "loss": 0.6897,
+      "step": 102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9938372093023257e-07,
+      "loss": 0.7244,
+      "step": 103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.993720930232558e-07,
+      "loss": 0.6597,
+      "step": 104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9936046511627904e-07,
+      "loss": 0.5964,
+      "step": 105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9934883720930227e-07,
+      "loss": 0.6221,
+      "step": 106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9933720930232556e-07,
+      "loss": 0.706,
+      "step": 107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9932558139534885e-07,
+      "loss": 0.6645,
+      "step": 108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.993139534883721e-07,
+      "loss": 0.7437,
+      "step": 109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.993023255813953e-07,
+      "loss": 0.6739,
+      "step": 110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.992906976744186e-07,
+      "loss": 0.6874,
+      "step": 111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9927906976744184e-07,
+      "loss": 0.7442,
+      "step": 112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.992674418604651e-07,
+      "loss": 0.6059,
+      "step": 113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9925581395348837e-07,
+      "loss": 0.6536,
+      "step": 114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.992441860465116e-07,
+      "loss": 0.7726,
+      "step": 115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.992325581395349e-07,
+      "loss": 0.5927,
+      "step": 116
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.992209302325581e-07,
+      "loss": 0.7632,
+      "step": 117
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9920930232558136e-07,
+      "loss": 0.7488,
+      "step": 118
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9919767441860465e-07,
+      "loss": 0.7257,
+      "step": 119
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.991860465116279e-07,
+      "loss": 0.5875,
+      "step": 120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9917441860465117e-07,
+      "loss": 0.6151,
+      "step": 121
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.991627906976744e-07,
+      "loss": 0.6033,
+      "step": 122
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.991511627906977e-07,
+      "loss": 0.6606,
+      "step": 123
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9913953488372093e-07,
+      "loss": 0.5826,
+      "step": 124
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9912790697674416e-07,
+      "loss": 0.6101,
+      "step": 125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.991162790697674e-07,
+      "loss": 0.579,
+      "step": 126
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.991046511627907e-07,
+      "loss": 0.6848,
+      "step": 127
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.990930232558139e-07,
+      "loss": 0.5633,
+      "step": 128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.990813953488372e-07,
+      "loss": 0.766,
+      "step": 129
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9906976744186045e-07,
+      "loss": 0.7422,
+      "step": 130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.990581395348837e-07,
+      "loss": 0.7083,
+      "step": 131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9904651162790697e-07,
+      "loss": 0.5907,
+      "step": 132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.990348837209302e-07,
+      "loss": 0.6804,
+      "step": 133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.990232558139535e-07,
+      "loss": 0.7069,
+      "step": 134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9901162790697673e-07,
+      "loss": 0.6175,
+      "step": 135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.99e-07,
+      "loss": 0.6991,
+      "step": 136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9898837209302325e-07,
+      "loss": 0.6139,
+      "step": 137
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.989767441860465e-07,
+      "loss": 0.5777,
+      "step": 138
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.989651162790697e-07,
+      "loss": 0.5965,
+      "step": 139
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.98953488372093e-07,
+      "loss": 0.5791,
+      "step": 140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9894186046511625e-07,
+      "loss": 0.5867,
+      "step": 141
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9893023255813953e-07,
+      "loss": 0.673,
+      "step": 142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9891860465116277e-07,
+      "loss": 0.7482,
+      "step": 143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9890697674418606e-07,
+      "loss": 0.6975,
+      "step": 144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.988953488372093e-07,
+      "loss": 0.6776,
+      "step": 145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9888372093023253e-07,
+      "loss": 0.735,
+      "step": 146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9887209302325576e-07,
+      "loss": 0.6682,
+      "step": 147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9886046511627905e-07,
+      "loss": 0.7265,
+      "step": 148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9884883720930234e-07,
+      "loss": 0.6844,
+      "step": 149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.988372093023256e-07,
+      "loss": 0.5953,
+      "step": 150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.988255813953488e-07,
+      "loss": 0.5939,
+      "step": 151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.988139534883721e-07,
+      "loss": 0.7235,
+      "step": 152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9880232558139533e-07,
+      "loss": 0.5932,
+      "step": 153
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9879069767441857e-07,
+      "loss": 0.7399,
+      "step": 154
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9877906976744186e-07,
+      "loss": 0.6919,
+      "step": 155
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9876744186046515e-07,
+      "loss": 0.5957,
+      "step": 156
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.987558139534884e-07,
+      "loss": 0.5882,
+      "step": 157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.987441860465116e-07,
+      "loss": 0.5845,
+      "step": 158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9873255813953485e-07,
+      "loss": 0.5777,
+      "step": 159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.987209302325581e-07,
+      "loss": 0.6922,
+      "step": 160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.987093023255814e-07,
+      "loss": 0.591,
+      "step": 161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9869767441860466e-07,
+      "loss": 0.6011,
+      "step": 162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.986860465116279e-07,
+      "loss": 0.6179,
+      "step": 163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9867441860465113e-07,
+      "loss": 0.5808,
+      "step": 164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.986627906976744e-07,
+      "loss": 0.5943,
+      "step": 165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9865116279069766e-07,
+      "loss": 0.6645,
+      "step": 166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.986395348837209e-07,
+      "loss": 0.7018,
+      "step": 167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.986279069767442e-07,
+      "loss": 0.6012,
+      "step": 168
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.986162790697674e-07,
+      "loss": 0.7126,
+      "step": 169
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.986046511627907e-07,
+      "loss": 0.743,
+      "step": 170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9859302325581394e-07,
+      "loss": 0.7256,
+      "step": 171
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.985813953488372e-07,
+      "loss": 0.709,
+      "step": 172
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9856976744186046e-07,
+      "loss": 0.7198,
+      "step": 173
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.985581395348837e-07,
+      "loss": 0.7259,
+      "step": 174
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.98546511627907e-07,
+      "loss": 0.662,
+      "step": 175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.985348837209302e-07,
+      "loss": 0.7085,
+      "step": 176
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.985232558139535e-07,
+      "loss": 0.6937,
+      "step": 177
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9851162790697674e-07,
+      "loss": 0.7071,
+      "step": 178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.985e-07,
+      "loss": 0.7145,
+      "step": 179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.984883720930232e-07,
+      "loss": 0.6935,
+      "step": 180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.984767441860465e-07,
+      "loss": 0.7342,
+      "step": 181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.9846511627906974e-07,
+      "loss": 0.7067,
+      "step": 182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.98453488372093e-07,
+      "loss": 0.6318,
+      "step": 183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.9844186046511626e-07,
+      "loss": 0.6961,
+      "step": 184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.984302325581395e-07,
+      "loss": 0.6011,
+      "step": 185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.984186046511628e-07,
+      "loss": 0.6672,
+      "step": 186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.98406976744186e-07,
+      "loss": 0.7037,
+      "step": 187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.9839534883720926e-07,
+      "loss": 0.7736,
+      "step": 188
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9838372093023254e-07,
+      "loss": 0.6809,
+      "step": 189
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9837209302325583e-07,
+      "loss": 0.5929,
+      "step": 190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9836046511627907e-07,
+      "loss": 0.6906,
+      "step": 191
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.983488372093023e-07,
+      "loss": 0.5894,
+      "step": 192
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9833720930232554e-07,
+      "loss": 0.5778,
+      "step": 193
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.983255813953488e-07,
+      "loss": 0.5915,
+      "step": 194
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.9831395348837206e-07,
+      "loss": 0.687,
+      "step": 195
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.9830232558139535e-07,
+      "loss": 0.5901,
+      "step": 196
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.982906976744186e-07,
+      "loss": 0.6727,
+      "step": 197
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.9827906976744187e-07,
+      "loss": 0.691,
+      "step": 198
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.982674418604651e-07,
+      "loss": 0.7081,
+      "step": 199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9825581395348834e-07,
+      "loss": 0.7036,
+      "step": 200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.982441860465116e-07,
+      "loss": 0.6154,
+      "step": 201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9823255813953487e-07,
+      "loss": 0.7048,
+      "step": 202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9822093023255815e-07,
+      "loss": 0.6736,
+      "step": 203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.982093023255814e-07,
+      "loss": 0.586,
+      "step": 204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.981976744186046e-07,
+      "loss": 0.6839,
+      "step": 205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.981860465116279e-07,
+      "loss": 0.5951,
+      "step": 206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.9817441860465115e-07,
+      "loss": 0.6001,
+      "step": 207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.981627906976744e-07,
+      "loss": 0.6267,
+      "step": 208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.9815116279069767e-07,
+      "loss": 0.6957,
+      "step": 209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.981395348837209e-07,
+      "loss": 0.716,
+      "step": 210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.981279069767442e-07,
+      "loss": 0.5956,
+      "step": 211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9811627906976743e-07,
+      "loss": 0.7075,
+      "step": 212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.9810465116279067e-07,
+      "loss": 0.63,
+      "step": 213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.980930232558139e-07,
+      "loss": 0.719,
+      "step": 214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.980813953488372e-07,
+      "loss": 0.6956,
+      "step": 215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.980697674418605e-07,
+      "loss": 0.6689,
+      "step": 216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.980581395348837e-07,
+      "loss": 0.7106,
+      "step": 217
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9804651162790695e-07,
+      "loss": 0.6994,
+      "step": 218
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.9803488372093024e-07,
+      "loss": 0.6698,
+      "step": 219
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.9802325581395347e-07,
+      "loss": 0.6045,
+      "step": 220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.980116279069767e-07,
+      "loss": 0.6749,
+      "step": 221
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.98e-07,
+      "loss": 0.6471,
+      "step": 222
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.9798837209302323e-07,
+      "loss": 0.5687,
+      "step": 223
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.979767441860465e-07,
+      "loss": 0.7129,
+      "step": 224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.9796511627906975e-07,
+      "loss": 0.6802,
+      "step": 225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.97953488372093e-07,
+      "loss": 0.6256,
+      "step": 226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.979418604651163e-07,
+      "loss": 0.7108,
+      "step": 227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.979302325581395e-07,
+      "loss": 0.6194,
+      "step": 228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.979186046511628e-07,
+      "loss": 0.7325,
+      "step": 229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.9790697674418604e-07,
+      "loss": 0.5843,
+      "step": 230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.978953488372093e-07,
+      "loss": 0.7095,
+      "step": 231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9788372093023256e-07,
+      "loss": 0.6631,
+      "step": 232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.978720930232558e-07,
+      "loss": 0.6811,
+      "step": 233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9786046511627903e-07,
+      "loss": 0.5892,
+      "step": 234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.978488372093023e-07,
+      "loss": 0.5965,
+      "step": 235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9783720930232555e-07,
+      "loss": 0.7604,
+      "step": 236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9782558139534884e-07,
+      "loss": 0.6629,
+      "step": 237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.978139534883721e-07,
+      "loss": 0.6557,
+      "step": 238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.978023255813953e-07,
+      "loss": 0.7204,
+      "step": 239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.977906976744186e-07,
+      "loss": 0.5715,
+      "step": 240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9777906976744184e-07,
+      "loss": 0.7016,
+      "step": 241
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9776744186046507e-07,
+      "loss": 0.7467,
+      "step": 242
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9775581395348836e-07,
+      "loss": 0.6733,
+      "step": 243
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9774418604651165e-07,
+      "loss": 0.6178,
+      "step": 244
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.977325581395349e-07,
+      "loss": 0.6407,
+      "step": 245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.977209302325581e-07,
+      "loss": 0.6438,
+      "step": 246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.9770930232558135e-07,
+      "loss": 0.7366,
+      "step": 247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.9769767441860464e-07,
+      "loss": 0.6702,
+      "step": 248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.976860465116279e-07,
+      "loss": 0.6802,
+      "step": 249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.9767441860465116e-07,
+      "loss": 0.6066,
+      "step": 250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.976627906976744e-07,
+      "loss": 0.6577,
+      "step": 251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.976511627906977e-07,
+      "loss": 0.5794,
+      "step": 252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.976395348837209e-07,
+      "loss": 0.6839,
+      "step": 253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.9762790697674416e-07,
+      "loss": 0.7469,
+      "step": 254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.976162790697674e-07,
+      "loss": 0.6168,
+      "step": 255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.976046511627907e-07,
+      "loss": 0.6762,
+      "step": 256
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.9759302325581397e-07,
+      "loss": 0.6043,
+      "step": 257
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.975813953488372e-07,
+      "loss": 0.6818,
+      "step": 258
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.9756976744186044e-07,
+      "loss": 0.7309,
+      "step": 259
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.9755813953488373e-07,
+      "loss": 0.6654,
+      "step": 260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.9754651162790696e-07,
+      "loss": 0.6773,
+      "step": 261
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.975348837209302e-07,
+      "loss": 0.7118,
+      "step": 262
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.975232558139535e-07,
+      "loss": 0.6431,
+      "step": 263
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.975116279069767e-07,
+      "loss": 0.7325,
+      "step": 264
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.975e-07,
+      "loss": 0.7084,
+      "step": 265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.9748837209302325e-07,
+      "loss": 0.6711,
+      "step": 266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.974767441860465e-07,
+      "loss": 0.6158,
+      "step": 267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.974651162790697e-07,
+      "loss": 0.6719,
+      "step": 268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.97453488372093e-07,
+      "loss": 0.6373,
+      "step": 269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.974418604651163e-07,
+      "loss": 0.6777,
+      "step": 270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.9743023255813953e-07,
+      "loss": 0.5759,
+      "step": 271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.9741860465116276e-07,
+      "loss": 0.6704,
+      "step": 272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.9740697674418605e-07,
+      "loss": 0.6377,
+      "step": 273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.973953488372093e-07,
+      "loss": 0.7404,
+      "step": 274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.973837209302325e-07,
+      "loss": 0.6042,
+      "step": 275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.973720930232558e-07,
+      "loss": 0.7493,
+      "step": 276
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.9736046511627905e-07,
+      "loss": 0.6668,
+      "step": 277
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.9734883720930233e-07,
+      "loss": 0.5995,
+      "step": 278
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.9733720930232557e-07,
+      "loss": 0.5588,
+      "step": 279
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.973255813953488e-07,
+      "loss": 0.6172,
+      "step": 280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.973139534883721e-07,
+      "loss": 0.6379,
+      "step": 281
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.9730232558139533e-07,
+      "loss": 0.585,
+      "step": 282
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.972906976744186e-07,
+      "loss": 0.6848,
+      "step": 283
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.9727906976744185e-07,
+      "loss": 0.7272,
+      "step": 284
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.9726744186046514e-07,
+      "loss": 0.6968,
+      "step": 285
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.972558139534884e-07,
+      "loss": 0.6796,
+      "step": 286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.972441860465116e-07,
+      "loss": 0.5757,
+      "step": 287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.9723255813953484e-07,
+      "loss": 0.6773,
+      "step": 288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.9722093023255813e-07,
+      "loss": 0.6538,
+      "step": 289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.9720930232558137e-07,
+      "loss": 0.6623,
+      "step": 290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.9719767441860466e-07,
+      "loss": 0.5933,
+      "step": 291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.971860465116279e-07,
+      "loss": 0.6281,
+      "step": 292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9717441860465113e-07,
+      "loss": 0.7215,
+      "step": 293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.971627906976744e-07,
+      "loss": 0.673,
+      "step": 294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9715116279069765e-07,
+      "loss": 0.686,
+      "step": 295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.971395348837209e-07,
+      "loss": 0.569,
+      "step": 296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9712790697674417e-07,
+      "loss": 0.5943,
+      "step": 297
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.9711627906976746e-07,
+      "loss": 0.7511,
+      "step": 298
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.971046511627907e-07,
+      "loss": 0.5853,
+      "step": 299
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.9709302325581393e-07,
+      "loss": 0.5808,
+      "step": 300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.9708139534883717e-07,
+      "loss": 0.6786,
+      "step": 301
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.9706976744186046e-07,
+      "loss": 0.6519,
+      "step": 302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.970581395348837e-07,
+      "loss": 0.6863,
+      "step": 303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.97046511627907e-07,
+      "loss": 0.5704,
+      "step": 304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.970348837209302e-07,
+      "loss": 0.6597,
+      "step": 305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.970232558139535e-07,
+      "loss": 0.6766,
+      "step": 306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.9701162790697674e-07,
+      "loss": 0.5955,
+      "step": 307
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.9699999999999997e-07,
+      "loss": 0.7168,
+      "step": 308
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.969883720930232e-07,
+      "loss": 0.6062,
+      "step": 309
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.969767441860465e-07,
+      "loss": 0.6441,
+      "step": 310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.969651162790698e-07,
+      "loss": 0.6278,
+      "step": 311
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.96953488372093e-07,
+      "loss": 0.587,
+      "step": 312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9694186046511625e-07,
+      "loss": 0.5758,
+      "step": 313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9693023255813954e-07,
+      "loss": 0.5684,
+      "step": 314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.969186046511628e-07,
+      "loss": 0.6393,
+      "step": 315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.96906976744186e-07,
+      "loss": 0.6879,
+      "step": 316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.968953488372093e-07,
+      "loss": 0.6973,
+      "step": 317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9688372093023254e-07,
+      "loss": 0.6462,
+      "step": 318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.968720930232558e-07,
+      "loss": 0.6785,
+      "step": 319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9686046511627906e-07,
+      "loss": 0.6366,
+      "step": 320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.968488372093023e-07,
+      "loss": 0.6116,
+      "step": 321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9683720930232553e-07,
+      "loss": 0.6565,
+      "step": 322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.968255813953488e-07,
+      "loss": 0.6917,
+      "step": 323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.968139534883721e-07,
+      "loss": 0.554,
+      "step": 324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9680232558139534e-07,
+      "loss": 0.6649,
+      "step": 325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.967906976744186e-07,
+      "loss": 0.5917,
+      "step": 326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9677906976744187e-07,
+      "loss": 0.677,
+      "step": 327
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.967674418604651e-07,
+      "loss": 0.5769,
+      "step": 328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9675581395348834e-07,
+      "loss": 0.6652,
+      "step": 329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.967441860465116e-07,
+      "loss": 0.6786,
+      "step": 330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9673255813953486e-07,
+      "loss": 0.6715,
+      "step": 331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9672093023255815e-07,
+      "loss": 0.6446,
+      "step": 332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.967093023255814e-07,
+      "loss": 0.5746,
+      "step": 333
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.966976744186046e-07,
+      "loss": 0.6572,
+      "step": 334
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.966860465116279e-07,
+      "loss": 0.6704,
+      "step": 335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9667441860465114e-07,
+      "loss": 0.6781,
+      "step": 336
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.966627906976744e-07,
+      "loss": 0.6836,
+      "step": 337
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.9665116279069767e-07,
+      "loss": 0.6001,
+      "step": 338
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9663953488372095e-07,
+      "loss": 0.5849,
+      "step": 339
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.966279069767442e-07,
+      "loss": 0.6282,
+      "step": 340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.966162790697674e-07,
+      "loss": 0.6721,
+      "step": 341
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9660465116279066e-07,
+      "loss": 0.7138,
+      "step": 342
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.9659302325581395e-07,
+      "loss": 0.5946,
+      "step": 343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.965813953488372e-07,
+      "loss": 0.5768,
+      "step": 344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.9656976744186047e-07,
+      "loss": 0.705,
+      "step": 345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.965581395348837e-07,
+      "loss": 0.5694,
+      "step": 346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.9654651162790694e-07,
+      "loss": 0.6731,
+      "step": 347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.9653488372093023e-07,
+      "loss": 0.5935,
+      "step": 348
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.9652325581395346e-07,
+      "loss": 0.6807,
+      "step": 349
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.965116279069767e-07,
+      "loss": 0.6892,
+      "step": 350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.965e-07,
+      "loss": 0.6657,
+      "step": 351
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.964883720930233e-07,
+      "loss": 0.7094,
+      "step": 352
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.964767441860465e-07,
+      "loss": 0.6999,
+      "step": 353
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.9646511627906975e-07,
+      "loss": 0.6784,
+      "step": 354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.96453488372093e-07,
+      "loss": 0.5895,
+      "step": 355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.9644186046511627e-07,
+      "loss": 0.6579,
+      "step": 356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.964302325581395e-07,
+      "loss": 0.6925,
+      "step": 357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.964186046511628e-07,
+      "loss": 0.6309,
+      "step": 358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.9640697674418603e-07,
+      "loss": 0.7292,
+      "step": 359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.963953488372093e-07,
+      "loss": 0.704,
+      "step": 360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.9638372093023255e-07,
+      "loss": 0.6593,
+      "step": 361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.963720930232558e-07,
+      "loss": 0.6907,
+      "step": 362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.96360465116279e-07,
+      "loss": 0.6968,
+      "step": 363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.963488372093023e-07,
+      "loss": 0.6034,
+      "step": 364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.963372093023256e-07,
+      "loss": 0.6725,
+      "step": 365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9632558139534883e-07,
+      "loss": 0.6499,
+      "step": 366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9631395348837207e-07,
+      "loss": 0.6435,
+      "step": 367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.9630232558139536e-07,
+      "loss": 0.6743,
+      "step": 368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.962906976744186e-07,
+      "loss": 0.7023,
+      "step": 369
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9627906976744183e-07,
+      "loss": 0.6846,
+      "step": 370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.962674418604651e-07,
+      "loss": 0.5676,
+      "step": 371
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9625581395348835e-07,
+      "loss": 0.5704,
+      "step": 372
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9624418604651164e-07,
+      "loss": 0.5703,
+      "step": 373
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.962325581395349e-07,
+      "loss": 0.6227,
+      "step": 374
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.962209302325581e-07,
+      "loss": 0.6465,
+      "step": 375
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.9620930232558135e-07,
+      "loss": 0.6804,
+      "step": 376
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.9619767441860463e-07,
+      "loss": 0.6652,
+      "step": 377
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.961860465116279e-07,
+      "loss": 0.5508,
+      "step": 378
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.9617441860465116e-07,
+      "loss": 0.6082,
+      "step": 379
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.961627906976744e-07,
+      "loss": 0.6304,
+      "step": 380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.961511627906977e-07,
+      "loss": 0.6659,
+      "step": 381
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.961395348837209e-07,
+      "loss": 0.6501,
+      "step": 382
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.9612790697674415e-07,
+      "loss": 0.6721,
+      "step": 383
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.9611627906976744e-07,
+      "loss": 0.6441,
+      "step": 384
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.961046511627907e-07,
+      "loss": 0.6736,
+      "step": 385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9609302325581396e-07,
+      "loss": 0.6618,
+      "step": 386
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.960813953488372e-07,
+      "loss": 0.5977,
+      "step": 387
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9606976744186043e-07,
+      "loss": 0.6804,
+      "step": 388
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.960581395348837e-07,
+      "loss": 0.7043,
+      "step": 389
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9604651162790696e-07,
+      "loss": 0.6692,
+      "step": 390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.960348837209302e-07,
+      "loss": 0.705,
+      "step": 391
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.960232558139535e-07,
+      "loss": 0.574,
+      "step": 392
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9601162790697677e-07,
+      "loss": 0.6024,
+      "step": 393
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.96e-07,
+      "loss": 0.655,
+      "step": 394
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9598837209302324e-07,
+      "loss": 0.6975,
+      "step": 395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.959767441860465e-07,
+      "loss": 0.6735,
+      "step": 396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.9596511627906976e-07,
+      "loss": 0.5832,
+      "step": 397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.95953488372093e-07,
+      "loss": 0.5853,
+      "step": 398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.959418604651163e-07,
+      "loss": 0.6465,
+      "step": 399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.959302325581395e-07,
+      "loss": 0.7059,
+      "step": 400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.9591860465116276e-07,
+      "loss": 0.6718,
+      "step": 401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.9590697674418604e-07,
+      "loss": 0.66,
+      "step": 402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.958953488372093e-07,
+      "loss": 0.6314,
+      "step": 403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.958837209302325e-07,
+      "loss": 0.6475,
+      "step": 404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.958720930232558e-07,
+      "loss": 0.6874,
+      "step": 405
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.958604651162791e-07,
+      "loss": 0.6113,
+      "step": 406
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.958488372093023e-07,
+      "loss": 0.6381,
+      "step": 407
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.9583720930232556e-07,
+      "loss": 0.5966,
+      "step": 408
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.958255813953488e-07,
+      "loss": 0.5898,
+      "step": 409
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.958139534883721e-07,
+      "loss": 0.6283,
+      "step": 410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.958023255813953e-07,
+      "loss": 0.6253,
+      "step": 411
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.957906976744186e-07,
+      "loss": 0.669,
+      "step": 412
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9577906976744184e-07,
+      "loss": 0.6488,
+      "step": 413
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9576744186046513e-07,
+      "loss": 0.6449,
+      "step": 414
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9575581395348837e-07,
+      "loss": 0.5461,
+      "step": 415
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.957441860465116e-07,
+      "loss": 0.6448,
+      "step": 416
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9573255813953484e-07,
+      "loss": 0.5709,
+      "step": 417
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.957209302325581e-07,
+      "loss": 0.5874,
+      "step": 418
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.957093023255814e-07,
+      "loss": 0.7094,
+      "step": 419
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9569767441860465e-07,
+      "loss": 0.6198,
+      "step": 420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.956860465116279e-07,
+      "loss": 0.639,
+      "step": 421
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.9567441860465117e-07,
+      "loss": 0.6406,
+      "step": 422
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.956627906976744e-07,
+      "loss": 0.5953,
+      "step": 423
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.9565116279069764e-07,
+      "loss": 0.6528,
+      "step": 424
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.9563953488372093e-07,
+      "loss": 0.5762,
+      "step": 425
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.9562790697674417e-07,
+      "loss": 0.6289,
+      "step": 426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9561627906976745e-07,
+      "loss": 0.5831,
+      "step": 427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.956046511627907e-07,
+      "loss": 0.5929,
+      "step": 428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.955930232558139e-07,
+      "loss": 0.6578,
+      "step": 429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9558139534883716e-07,
+      "loss": 0.7299,
+      "step": 430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9556976744186045e-07,
+      "loss": 0.6837,
+      "step": 431
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.955581395348837e-07,
+      "loss": 0.6021,
+      "step": 432
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.9554651162790697e-07,
+      "loss": 0.6293,
+      "step": 433
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.955348837209302e-07,
+      "loss": 0.5627,
+      "step": 434
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.955232558139535e-07,
+      "loss": 0.5681,
+      "step": 435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.9551162790697673e-07,
+      "loss": 0.6815,
+      "step": 436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9549999999999997e-07,
+      "loss": 0.6809,
+      "step": 437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9548837209302325e-07,
+      "loss": 0.6752,
+      "step": 438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.954767441860465e-07,
+      "loss": 0.6497,
+      "step": 439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.954651162790698e-07,
+      "loss": 0.5833,
+      "step": 440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.95453488372093e-07,
+      "loss": 0.6359,
+      "step": 441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9544186046511625e-07,
+      "loss": 0.6382,
+      "step": 442
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9543023255813954e-07,
+      "loss": 0.5873,
+      "step": 443
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9541860465116277e-07,
+      "loss": 0.6479,
+      "step": 444
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.95406976744186e-07,
+      "loss": 0.6265,
+      "step": 445
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.953953488372093e-07,
+      "loss": 0.6835,
+      "step": 446
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.953837209302326e-07,
+      "loss": 0.6328,
+      "step": 447
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.953720930232558e-07,
+      "loss": 0.586,
+      "step": 448
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.9536046511627905e-07,
+      "loss": 0.6524,
+      "step": 449
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.953488372093023e-07,
+      "loss": 0.6458,
+      "step": 450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.953372093023256e-07,
+      "loss": 0.5651,
+      "step": 451
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.953255813953488e-07,
+      "loss": 0.6157,
+      "step": 452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.953139534883721e-07,
+      "loss": 0.6413,
+      "step": 453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.9530232558139534e-07,
+      "loss": 0.5493,
+      "step": 454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.9529069767441857e-07,
+      "loss": 0.5891,
+      "step": 455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.9527906976744186e-07,
+      "loss": 0.6265,
+      "step": 456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.952674418604651e-07,
+      "loss": 0.6631,
+      "step": 457
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9525581395348833e-07,
+      "loss": 0.5967,
+      "step": 458
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.952441860465116e-07,
+      "loss": 0.6722,
+      "step": 459
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.952325581395349e-07,
+      "loss": 0.6713,
+      "step": 460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.9522093023255814e-07,
+      "loss": 0.6673,
+      "step": 461
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.952093023255814e-07,
+      "loss": 0.6403,
+      "step": 462
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.951976744186046e-07,
+      "loss": 0.6583,
+      "step": 463
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.951860465116279e-07,
+      "loss": 0.6462,
+      "step": 464
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9517441860465114e-07,
+      "loss": 0.5379,
+      "step": 465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.951627906976744e-07,
+      "loss": 0.6914,
+      "step": 466
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.9515116279069766e-07,
+      "loss": 0.6265,
+      "step": 467
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.9513953488372095e-07,
+      "loss": 0.6842,
+      "step": 468
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.951279069767442e-07,
+      "loss": 0.5861,
+      "step": 469
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.951162790697674e-07,
+      "loss": 0.5311,
+      "step": 470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.9510465116279065e-07,
+      "loss": 0.6053,
+      "step": 471
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.9509302325581394e-07,
+      "loss": 0.6676,
+      "step": 472
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.9508139534883723e-07,
+      "loss": 0.6622,
+      "step": 473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9506976744186046e-07,
+      "loss": 0.5842,
+      "step": 474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.950581395348837e-07,
+      "loss": 0.6925,
+      "step": 475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.95046511627907e-07,
+      "loss": 0.6562,
+      "step": 476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.950348837209302e-07,
+      "loss": 0.5865,
+      "step": 477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9502325581395346e-07,
+      "loss": 0.5932,
+      "step": 478
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9501162790697675e-07,
+      "loss": 0.6316,
+      "step": 479
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.95e-07,
+      "loss": 0.6477,
+      "step": 480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9498837209302327e-07,
+      "loss": 0.6007,
+      "step": 481
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.949767441860465e-07,
+      "loss": 0.6563,
+      "step": 482
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9496511627906974e-07,
+      "loss": 0.5772,
+      "step": 483
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.94953488372093e-07,
+      "loss": 0.5489,
+      "step": 484
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9494186046511626e-07,
+      "loss": 0.6767,
+      "step": 485
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.949302325581395e-07,
+      "loss": 0.6651,
+      "step": 486
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.949186046511628e-07,
+      "loss": 0.6046,
+      "step": 487
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.94906976744186e-07,
+      "loss": 0.6592,
+      "step": 488
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.948953488372093e-07,
+      "loss": 0.6509,
+      "step": 489
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9488372093023255e-07,
+      "loss": 0.5846,
+      "step": 490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.948720930232558e-07,
+      "loss": 0.6799,
+      "step": 491
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9486046511627907e-07,
+      "loss": 0.5752,
+      "step": 492
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.948488372093023e-07,
+      "loss": 0.5539,
+      "step": 493
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.948372093023256e-07,
+      "loss": 0.5572,
+      "step": 494
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.9482558139534883e-07,
+      "loss": 0.6099,
+      "step": 495
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.9481395348837206e-07,
+      "loss": 0.5864,
+      "step": 496
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.9480232558139535e-07,
+      "loss": 0.5917,
+      "step": 497
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.947906976744186e-07,
+      "loss": 0.5714,
+      "step": 498
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.947790697674418e-07,
+      "loss": 0.5459,
+      "step": 499
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.947674418604651e-07,
+      "loss": 0.579,
+      "step": 500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.947558139534884e-07,
+      "loss": 0.6167,
+      "step": 501
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9474418604651163e-07,
+      "loss": 0.6038,
+      "step": 502
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.9473255813953487e-07,
+      "loss": 0.5984,
+      "step": 503
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.947209302325581e-07,
+      "loss": 0.5581,
+      "step": 504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.947093023255814e-07,
+      "loss": 0.5777,
+      "step": 505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9469767441860463e-07,
+      "loss": 0.6457,
+      "step": 506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.946860465116279e-07,
+      "loss": 0.6123,
+      "step": 507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9467441860465115e-07,
+      "loss": 0.6412,
+      "step": 508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.946627906976744e-07,
+      "loss": 0.6106,
+      "step": 509
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.946511627906977e-07,
+      "loss": 0.563,
+      "step": 510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.946395348837209e-07,
+      "loss": 0.6146,
+      "step": 511
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9462790697674414e-07,
+      "loss": 0.5774,
+      "step": 512
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9461627906976743e-07,
+      "loss": 0.5922,
+      "step": 513
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.946046511627907e-07,
+      "loss": 0.6216,
+      "step": 514
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9459302325581396e-07,
+      "loss": 0.5887,
+      "step": 515
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.945813953488372e-07,
+      "loss": 0.654,
+      "step": 516
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.945697674418604e-07,
+      "loss": 0.6232,
+      "step": 517
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9574468085106383,
+      "eval_loss": 0.6163627505302429,
+      "eval_runtime": 1.5652,
+      "eval_samples_per_second": 660.62,
+      "eval_steps_per_second": 10.861,
+      "step": 517
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.945581395348837e-07,
+      "loss": 0.6576,
+      "step": 518
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9454651162790695e-07,
+      "loss": 0.5613,
+      "step": 519
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.9453488372093024e-07,
+      "loss": 0.5871,
+      "step": 520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.9452325581395347e-07,
+      "loss": 0.5776,
+      "step": 521
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.9451162790697676e-07,
+      "loss": 0.5983,
+      "step": 522
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.945e-07,
+      "loss": 0.5871,
+      "step": 523
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.9448837209302323e-07,
+      "loss": 0.5968,
+      "step": 524
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.9447674418604647e-07,
+      "loss": 0.6094,
+      "step": 525
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.9446511627906976e-07,
+      "loss": 0.6325,
+      "step": 526
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.9445348837209304e-07,
+      "loss": 0.6417,
+      "step": 527
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.944418604651163e-07,
+      "loss": 0.6214,
+      "step": 528
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.944302325581395e-07,
+      "loss": 0.6458,
+      "step": 529
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.9441860465116275e-07,
+      "loss": 0.5449,
+      "step": 530
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.9440697674418604e-07,
+      "loss": 0.5547,
+      "step": 531
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.9439534883720927e-07,
+      "loss": 0.6293,
+      "step": 532
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.9438372093023256e-07,
+      "loss": 0.6299,
+      "step": 533
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.943720930232558e-07,
+      "loss": 0.5693,
+      "step": 534
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.943604651162791e-07,
+      "loss": 0.695,
+      "step": 535
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.943488372093023e-07,
+      "loss": 0.5622,
+      "step": 536
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.9433720930232555e-07,
+      "loss": 0.5622,
+      "step": 537
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.943255813953488e-07,
+      "loss": 0.5466,
+      "step": 538
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.943139534883721e-07,
+      "loss": 0.6232,
+      "step": 539
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.943023255813953e-07,
+      "loss": 0.5918,
+      "step": 540
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.942906976744186e-07,
+      "loss": 0.6131,
+      "step": 541
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.9427906976744184e-07,
+      "loss": 0.598,
+      "step": 542
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.942674418604651e-07,
+      "loss": 0.6634,
+      "step": 543
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.9425581395348836e-07,
+      "loss": 0.6058,
+      "step": 544
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.942441860465116e-07,
+      "loss": 0.5575,
+      "step": 545
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.942325581395349e-07,
+      "loss": 0.644,
+      "step": 546
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.942209302325581e-07,
+      "loss": 0.551,
+      "step": 547
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.942093023255814e-07,
+      "loss": 0.5918,
+      "step": 548
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.9419767441860464e-07,
+      "loss": 0.6013,
+      "step": 549
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.941860465116279e-07,
+      "loss": 0.5636,
+      "step": 550
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.9417441860465117e-07,
+      "loss": 0.5689,
+      "step": 551
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.941627906976744e-07,
+      "loss": 0.5948,
+      "step": 552
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.9415116279069764e-07,
+      "loss": 0.6319,
+      "step": 553
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.941395348837209e-07,
+      "loss": 0.6602,
+      "step": 554
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.941279069767442e-07,
+      "loss": 0.6247,
+      "step": 555
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.9411627906976745e-07,
+      "loss": 0.5374,
+      "step": 556
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.941046511627907e-07,
+      "loss": 0.548,
+      "step": 557
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.940930232558139e-07,
+      "loss": 0.6089,
+      "step": 558
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.940813953488372e-07,
+      "loss": 0.5668,
+      "step": 559
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.9406976744186044e-07,
+      "loss": 0.5988,
+      "step": 560
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.9405813953488373e-07,
+      "loss": 0.5442,
+      "step": 561
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.9404651162790697e-07,
+      "loss": 0.5482,
+      "step": 562
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.940348837209302e-07,
+      "loss": 0.5618,
+      "step": 563
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.940232558139535e-07,
+      "loss": 0.689,
+      "step": 564
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.940116279069767e-07,
+      "loss": 0.597,
+      "step": 565
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.9399999999999996e-07,
+      "loss": 0.5432,
+      "step": 566
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.9398837209302325e-07,
+      "loss": 0.5639,
+      "step": 567
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.9397674418604654e-07,
+      "loss": 0.5614,
+      "step": 568
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.9396511627906977e-07,
+      "loss": 0.614,
+      "step": 569
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.93953488372093e-07,
+      "loss": 0.5507,
+      "step": 570
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.9394186046511624e-07,
+      "loss": 0.6244,
+      "step": 571
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.9393023255813953e-07,
+      "loss": 0.6964,
+      "step": 572
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.9391860465116276e-07,
+      "loss": 0.6664,
+      "step": 573
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.9390697674418605e-07,
+      "loss": 0.5604,
+      "step": 574
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.938953488372093e-07,
+      "loss": 0.6084,
+      "step": 575
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.938837209302326e-07,
+      "loss": 0.5801,
+      "step": 576
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.938720930232558e-07,
+      "loss": 0.6176,
+      "step": 577
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.9386046511627905e-07,
+      "loss": 0.5815,
+      "step": 578
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.938488372093023e-07,
+      "loss": 0.5313,
+      "step": 579
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.9383720930232557e-07,
+      "loss": 0.6067,
+      "step": 580
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.938255813953488e-07,
+      "loss": 0.6189,
+      "step": 581
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.938139534883721e-07,
+      "loss": 0.6149,
+      "step": 582
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.9380232558139533e-07,
+      "loss": 0.5705,
+      "step": 583
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.9379069767441856e-07,
+      "loss": 0.6552,
+      "step": 584
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.9377906976744185e-07,
+      "loss": 0.6595,
+      "step": 585
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.937674418604651e-07,
+      "loss": 0.6399,
+      "step": 586
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.937558139534884e-07,
+      "loss": 0.5816,
+      "step": 587
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.937441860465116e-07,
+      "loss": 0.6723,
+      "step": 588
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.937325581395349e-07,
+      "loss": 0.5273,
+      "step": 589
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.9372093023255813e-07,
+      "loss": 0.6271,
+      "step": 590
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.9370930232558137e-07,
+      "loss": 0.6007,
+      "step": 591
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.936976744186046e-07,
+      "loss": 0.5913,
+      "step": 592
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.936860465116279e-07,
+      "loss": 0.5562,
+      "step": 593
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.9367441860465113e-07,
+      "loss": 0.6552,
+      "step": 594
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.936627906976744e-07,
+      "loss": 0.6267,
+      "step": 595
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.9365116279069765e-07,
+      "loss": 0.5799,
+      "step": 596
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.9363953488372094e-07,
+      "loss": 0.5704,
+      "step": 597
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.936279069767442e-07,
+      "loss": 0.5444,
+      "step": 598
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.936162790697674e-07,
+      "loss": 0.5952,
+      "step": 599
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.936046511627907e-07,
+      "loss": 0.4911,
+      "step": 600
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.9359302325581393e-07,
+      "loss": 0.5714,
+      "step": 601
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.935813953488372e-07,
+      "loss": 0.6159,
+      "step": 602
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.9356976744186046e-07,
+      "loss": 0.5976,
+      "step": 603
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.935581395348837e-07,
+      "loss": 0.5656,
+      "step": 604
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.93546511627907e-07,
+      "loss": 0.5657,
+      "step": 605
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.935348837209302e-07,
+      "loss": 0.5693,
+      "step": 606
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.9352325581395345e-07,
+      "loss": 0.5271,
+      "step": 607
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.9351162790697674e-07,
+      "loss": 0.6077,
+      "step": 608
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.9350000000000003e-07,
+      "loss": 0.5275,
+      "step": 609
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.9348837209302326e-07,
+      "loss": 0.6308,
+      "step": 610
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.934767441860465e-07,
+      "loss": 0.5985,
+      "step": 611
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.9346511627906973e-07,
+      "loss": 0.6206,
+      "step": 612
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.9345348837209297e-07,
+      "loss": 0.5999,
+      "step": 613
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.9344186046511626e-07,
+      "loss": 0.5322,
+      "step": 614
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.9343023255813954e-07,
+      "loss": 0.6012,
+      "step": 615
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.934186046511628e-07,
+      "loss": 0.5798,
+      "step": 616
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.93406976744186e-07,
+      "loss": 0.5413,
+      "step": 617
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.933953488372093e-07,
+      "loss": 0.5582,
+      "step": 618
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.9338372093023254e-07,
+      "loss": 0.605,
+      "step": 619
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.933720930232558e-07,
+      "loss": 0.6029,
+      "step": 620
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.9336046511627906e-07,
+      "loss": 0.6158,
+      "step": 621
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.9334883720930235e-07,
+      "loss": 0.5296,
+      "step": 622
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.933372093023256e-07,
+      "loss": 0.5618,
+      "step": 623
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.933255813953488e-07,
+      "loss": 0.5279,
+      "step": 624
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.9331395348837206e-07,
+      "loss": 0.5527,
+      "step": 625
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.9330232558139534e-07,
+      "loss": 0.5131,
+      "step": 626
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.932906976744186e-07,
+      "loss": 0.5057,
+      "step": 627
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.9327906976744187e-07,
+      "loss": 0.5982,
+      "step": 628
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.932674418604651e-07,
+      "loss": 0.5161,
+      "step": 629
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.932558139534884e-07,
+      "loss": 0.5718,
+      "step": 630
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.932441860465116e-07,
+      "loss": 0.5267,
+      "step": 631
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.9323255813953486e-07,
+      "loss": 0.5738,
+      "step": 632
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.932209302325581e-07,
+      "loss": 0.6487,
+      "step": 633
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.932093023255814e-07,
+      "loss": 0.5981,
+      "step": 634
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.931976744186046e-07,
+      "loss": 0.6277,
+      "step": 635
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.931860465116279e-07,
+      "loss": 0.5494,
+      "step": 636
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.9317441860465114e-07,
+      "loss": 0.5571,
+      "step": 637
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.931627906976744e-07,
+      "loss": 0.5324,
+      "step": 638
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.9315116279069767e-07,
+      "loss": 0.5185,
+      "step": 639
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.931395348837209e-07,
+      "loss": 0.5721,
+      "step": 640
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.931279069767442e-07,
+      "loss": 0.5033,
+      "step": 641
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.931162790697674e-07,
+      "loss": 0.6425,
+      "step": 642
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.931046511627907e-07,
+      "loss": 0.5531,
+      "step": 643
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.9309302325581395e-07,
+      "loss": 0.6051,
+      "step": 644
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.930813953488372e-07,
+      "loss": 0.5681,
+      "step": 645
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.930697674418604e-07,
+      "loss": 0.5173,
+      "step": 646
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.930581395348837e-07,
+      "loss": 0.5864,
+      "step": 647
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.9304651162790694e-07,
+      "loss": 0.5471,
+      "step": 648
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.9303488372093023e-07,
+      "loss": 0.5953,
+      "step": 649
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.9302325581395347e-07,
+      "loss": 0.6172,
+      "step": 650
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.9301162790697675e-07,
+      "loss": 0.5916,
+      "step": 651
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.93e-07,
+      "loss": 0.6281,
+      "step": 652
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.929883720930232e-07,
+      "loss": 0.5215,
+      "step": 653
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.929767441860465e-07,
+      "loss": 0.6908,
+      "step": 654
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.9296511627906975e-07,
+      "loss": 0.6121,
+      "step": 655
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.9295348837209304e-07,
+      "loss": 0.5545,
+      "step": 656
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.9294186046511627e-07,
+      "loss": 0.5302,
+      "step": 657
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.929302325581395e-07,
+      "loss": 0.6341,
+      "step": 658
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.929186046511628e-07,
+      "loss": 0.5779,
+      "step": 659
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.9290697674418603e-07,
+      "loss": 0.5296,
+      "step": 660
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.9289534883720927e-07,
+      "loss": 0.5928,
+      "step": 661
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.9288372093023255e-07,
+      "loss": 0.5832,
+      "step": 662
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.9287209302325584e-07,
+      "loss": 0.529,
+      "step": 663
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.928604651162791e-07,
+      "loss": 0.5207,
+      "step": 664
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.928488372093023e-07,
+      "loss": 0.5863,
+      "step": 665
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.9283720930232555e-07,
+      "loss": 0.6204,
+      "step": 666
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.928255813953488e-07,
+      "loss": 0.6076,
+      "step": 667
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.9281395348837207e-07,
+      "loss": 0.5861,
+      "step": 668
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.9280232558139536e-07,
+      "loss": 0.5457,
+      "step": 669
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.927906976744186e-07,
+      "loss": 0.538,
+      "step": 670
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.9277906976744183e-07,
+      "loss": 0.5263,
+      "step": 671
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.927674418604651e-07,
+      "loss": 0.5418,
+      "step": 672
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.9275581395348835e-07,
+      "loss": 0.5703,
+      "step": 673
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.927441860465116e-07,
+      "loss": 0.5277,
+      "step": 674
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.927325581395349e-07,
+      "loss": 0.5281,
+      "step": 675
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.9272093023255817e-07,
+      "loss": 0.5274,
+      "step": 676
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.927093023255814e-07,
+      "loss": 0.5868,
+      "step": 677
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.9269767441860464e-07,
+      "loss": 0.6086,
+      "step": 678
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.9268604651162787e-07,
+      "loss": 0.6795,
+      "step": 679
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9267441860465116e-07,
+      "loss": 0.572,
+      "step": 680
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.926627906976744e-07,
+      "loss": 0.5182,
+      "step": 681
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.926511627906977e-07,
+      "loss": 0.5419,
+      "step": 682
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.926395348837209e-07,
+      "loss": 0.5401,
+      "step": 683
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.926279069767442e-07,
+      "loss": 0.5562,
+      "step": 684
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9261627906976744e-07,
+      "loss": 0.5114,
+      "step": 685
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.926046511627907e-07,
+      "loss": 0.5328,
+      "step": 686
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.925930232558139e-07,
+      "loss": 0.5124,
+      "step": 687
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.925813953488372e-07,
+      "loss": 0.4886,
+      "step": 688
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.9256976744186044e-07,
+      "loss": 0.5113,
+      "step": 689
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.925581395348837e-07,
+      "loss": 0.5201,
+      "step": 690
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.9254651162790696e-07,
+      "loss": 0.4878,
+      "step": 691
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.925348837209302e-07,
+      "loss": 0.4964,
+      "step": 692
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.925232558139535e-07,
+      "loss": 0.507,
+      "step": 693
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.925116279069767e-07,
+      "loss": 0.59,
+      "step": 694
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.925e-07,
+      "loss": 0.5378,
+      "step": 695
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9248837209302324e-07,
+      "loss": 0.4742,
+      "step": 696
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9247674418604653e-07,
+      "loss": 0.5295,
+      "step": 697
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9246511627906976e-07,
+      "loss": 0.5363,
+      "step": 698
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.92453488372093e-07,
+      "loss": 0.4857,
+      "step": 699
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9244186046511623e-07,
+      "loss": 0.5253,
+      "step": 700
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.924302325581395e-07,
+      "loss": 0.4991,
+      "step": 701
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.9241860465116276e-07,
+      "loss": 0.4955,
+      "step": 702
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.9240697674418605e-07,
+      "loss": 0.5026,
+      "step": 703
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.923953488372093e-07,
+      "loss": 0.6005,
+      "step": 704
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.9238372093023257e-07,
+      "loss": 0.622,
+      "step": 705
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.923720930232558e-07,
+      "loss": 0.49,
+      "step": 706
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.9236046511627904e-07,
+      "loss": 0.5325,
+      "step": 707
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.923488372093023e-07,
+      "loss": 0.5486,
+      "step": 708
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.9233720930232556e-07,
+      "loss": 0.5222,
+      "step": 709
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.9232558139534885e-07,
+      "loss": 0.4931,
+      "step": 710
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.923139534883721e-07,
+      "loss": 0.5068,
+      "step": 711
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.923023255813953e-07,
+      "loss": 0.4784,
+      "step": 712
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.922906976744186e-07,
+      "loss": 0.4761,
+      "step": 713
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.9227906976744185e-07,
+      "loss": 0.5165,
+      "step": 714
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.922674418604651e-07,
+      "loss": 0.5167,
+      "step": 715
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.9225581395348837e-07,
+      "loss": 0.6105,
+      "step": 716
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.922441860465116e-07,
+      "loss": 0.5303,
+      "step": 717
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.922325581395349e-07,
+      "loss": 0.4702,
+      "step": 718
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.9222093023255813e-07,
+      "loss": 0.5752,
+      "step": 719
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.9220930232558136e-07,
+      "loss": 0.4993,
+      "step": 720
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.921976744186046e-07,
+      "loss": 0.5443,
+      "step": 721
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.921860465116279e-07,
+      "loss": 0.5505,
+      "step": 722
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.921744186046512e-07,
+      "loss": 0.5594,
+      "step": 723
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.921627906976744e-07,
+      "loss": 0.5442,
+      "step": 724
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.9215116279069764e-07,
+      "loss": 0.5466,
+      "step": 725
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.9213953488372093e-07,
+      "loss": 0.5624,
+      "step": 726
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9212790697674417e-07,
+      "loss": 0.5375,
+      "step": 727
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.921162790697674e-07,
+      "loss": 0.5458,
+      "step": 728
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.921046511627907e-07,
+      "loss": 0.5164,
+      "step": 729
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9209302325581393e-07,
+      "loss": 0.5113,
+      "step": 730
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.920813953488372e-07,
+      "loss": 0.6623,
+      "step": 731
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9206976744186045e-07,
+      "loss": 0.5364,
+      "step": 732
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.920581395348837e-07,
+      "loss": 0.487,
+      "step": 733
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.92046511627907e-07,
+      "loss": 0.5038,
+      "step": 734
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.920348837209302e-07,
+      "loss": 0.5281,
+      "step": 735
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.920232558139535e-07,
+      "loss": 0.5413,
+      "step": 736
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.9201162790697673e-07,
+      "loss": 0.5139,
+      "step": 737
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.92e-07,
+      "loss": 0.518,
+      "step": 738
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.9198837209302326e-07,
+      "loss": 0.5516,
+      "step": 739
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.919767441860465e-07,
+      "loss": 0.4968,
+      "step": 740
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.919651162790697e-07,
+      "loss": 0.5788,
+      "step": 741
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.91953488372093e-07,
+      "loss": 0.4995,
+      "step": 742
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.9194186046511625e-07,
+      "loss": 0.5179,
+      "step": 743
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.9193023255813954e-07,
+      "loss": 0.4851,
+      "step": 744
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.9191860465116277e-07,
+      "loss": 0.6331,
+      "step": 745
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.91906976744186e-07,
+      "loss": 0.4922,
+      "step": 746
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.918953488372093e-07,
+      "loss": 0.6002,
+      "step": 747
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.9188372093023253e-07,
+      "loss": 0.4861,
+      "step": 748
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.918720930232558e-07,
+      "loss": 0.5708,
+      "step": 749
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.9186046511627906e-07,
+      "loss": 0.5347,
+      "step": 750
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.9184883720930234e-07,
+      "loss": 0.4634,
+      "step": 751
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.918372093023256e-07,
+      "loss": 0.5173,
+      "step": 752
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.918255813953488e-07,
+      "loss": 0.5442,
+      "step": 753
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.9181395348837205e-07,
+      "loss": 0.5563,
+      "step": 754
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.9180232558139534e-07,
+      "loss": 0.4777,
+      "step": 755
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.9179069767441857e-07,
+      "loss": 0.5052,
+      "step": 756
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.9177906976744186e-07,
+      "loss": 0.4815,
+      "step": 757
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.917674418604651e-07,
+      "loss": 0.5317,
+      "step": 758
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.917558139534884e-07,
+      "loss": 0.5012,
+      "step": 759
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.917441860465116e-07,
+      "loss": 0.4815,
+      "step": 760
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.9173255813953485e-07,
+      "loss": 0.5188,
+      "step": 761
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.917209302325581e-07,
+      "loss": 0.5335,
+      "step": 762
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.917093023255814e-07,
+      "loss": 0.526,
+      "step": 763
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.9169767441860467e-07,
+      "loss": 0.4579,
+      "step": 764
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.916860465116279e-07,
+      "loss": 0.5202,
+      "step": 765
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.9167441860465114e-07,
+      "loss": 0.5124,
+      "step": 766
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.916627906976744e-07,
+      "loss": 0.4351,
+      "step": 767
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.9165116279069766e-07,
+      "loss": 0.4948,
+      "step": 768
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.916395348837209e-07,
+      "loss": 0.483,
+      "step": 769
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.916279069767442e-07,
+      "loss": 0.4783,
+      "step": 770
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.916162790697674e-07,
+      "loss": 0.4701,
+      "step": 771
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.916046511627907e-07,
+      "loss": 0.5515,
+      "step": 772
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.9159302325581394e-07,
+      "loss": 0.5016,
+      "step": 773
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.915813953488372e-07,
+      "loss": 0.5838,
+      "step": 774
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.915697674418604e-07,
+      "loss": 0.5683,
+      "step": 775
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.915581395348837e-07,
+      "loss": 0.4484,
+      "step": 776
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.91546511627907e-07,
+      "loss": 0.4879,
+      "step": 777
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.915348837209302e-07,
+      "loss": 0.5255,
+      "step": 778
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9152325581395346e-07,
+      "loss": 0.4549,
+      "step": 779
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.9151162790697675e-07,
+      "loss": 0.522,
+      "step": 780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.915e-07,
+      "loss": 0.4794,
+      "step": 781
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.914883720930232e-07,
+      "loss": 0.4947,
+      "step": 782
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.914767441860465e-07,
+      "loss": 0.6258,
+      "step": 783
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9146511627906974e-07,
+      "loss": 0.5678,
+      "step": 784
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9145348837209303e-07,
+      "loss": 0.4841,
+      "step": 785
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.9144186046511627e-07,
+      "loss": 0.5238,
+      "step": 786
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.914302325581395e-07,
+      "loss": 0.5649,
+      "step": 787
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.914186046511628e-07,
+      "loss": 0.4732,
+      "step": 788
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.91406976744186e-07,
+      "loss": 0.4971,
+      "step": 789
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.913953488372093e-07,
+      "loss": 0.4779,
+      "step": 790
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.9138372093023255e-07,
+      "loss": 0.452,
+      "step": 791
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.9137209302325584e-07,
+      "loss": 0.4796,
+      "step": 792
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.9136046511627907e-07,
+      "loss": 0.4698,
+      "step": 793
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.913488372093023e-07,
+      "loss": 0.4742,
+      "step": 794
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.9133720930232554e-07,
+      "loss": 0.446,
+      "step": 795
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.9132558139534883e-07,
+      "loss": 0.4716,
+      "step": 796
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.9131395348837206e-07,
+      "loss": 0.4454,
+      "step": 797
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.9130232558139535e-07,
+      "loss": 0.5143,
+      "step": 798
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.912906976744186e-07,
+      "loss": 0.5105,
+      "step": 799
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.912790697674418e-07,
+      "loss": 0.532,
+      "step": 800
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.912674418604651e-07,
+      "loss": 0.451,
+      "step": 801
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.9125581395348835e-07,
+      "loss": 0.4789,
+      "step": 802
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.9124418604651163e-07,
+      "loss": 0.5408,
+      "step": 803
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9123255813953487e-07,
+      "loss": 0.4928,
+      "step": 804
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9122093023255816e-07,
+      "loss": 0.449,
+      "step": 805
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.912093023255814e-07,
+      "loss": 0.4496,
+      "step": 806
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9119767441860463e-07,
+      "loss": 0.4774,
+      "step": 807
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9118604651162786e-07,
+      "loss": 0.4441,
+      "step": 808
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.9117441860465115e-07,
+      "loss": 0.4425,
+      "step": 809
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.911627906976744e-07,
+      "loss": 0.6037,
+      "step": 810
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.911511627906977e-07,
+      "loss": 0.4553,
+      "step": 811
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.911395348837209e-07,
+      "loss": 0.4142,
+      "step": 812
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.911279069767442e-07,
+      "loss": 0.4762,
+      "step": 813
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.9111627906976743e-07,
+      "loss": 0.5514,
+      "step": 814
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.9110465116279067e-07,
+      "loss": 0.5005,
+      "step": 815
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.910930232558139e-07,
+      "loss": 0.4571,
+      "step": 816
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.910813953488372e-07,
+      "loss": 0.5553,
+      "step": 817
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.910697674418605e-07,
+      "loss": 0.4224,
+      "step": 818
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.910581395348837e-07,
+      "loss": 0.5937,
+      "step": 819
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.9104651162790695e-07,
+      "loss": 0.471,
+      "step": 820
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.9103488372093024e-07,
+      "loss": 0.4352,
+      "step": 821
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.910232558139535e-07,
+      "loss": 0.4377,
+      "step": 822
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.910116279069767e-07,
+      "loss": 0.4213,
+      "step": 823
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.91e-07,
+      "loss": 0.4873,
+      "step": 824
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.9098837209302323e-07,
+      "loss": 0.468,
+      "step": 825
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.909767441860465e-07,
+      "loss": 0.4056,
+      "step": 826
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.9096511627906976e-07,
+      "loss": 0.4793,
+      "step": 827
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.90953488372093e-07,
+      "loss": 0.4395,
+      "step": 828
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.9094186046511623e-07,
+      "loss": 0.3891,
+      "step": 829
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.909302325581395e-07,
+      "loss": 0.3948,
+      "step": 830
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.909186046511628e-07,
+      "loss": 0.4829,
+      "step": 831
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.9090697674418604e-07,
+      "loss": 0.4399,
+      "step": 832
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.908953488372093e-07,
+      "loss": 0.4754,
+      "step": 833
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.9088372093023256e-07,
+      "loss": 0.5812,
+      "step": 834
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.908720930232558e-07,
+      "loss": 0.4475,
+      "step": 835
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.9086046511627903e-07,
+      "loss": 0.4351,
+      "step": 836
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.908488372093023e-07,
+      "loss": 0.4297,
+      "step": 837
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.9083720930232556e-07,
+      "loss": 0.4848,
+      "step": 838
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.9082558139534884e-07,
+      "loss": 0.6103,
+      "step": 839
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.908139534883721e-07,
+      "loss": 0.3617,
+      "step": 840
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.908023255813953e-07,
+      "loss": 0.4456,
+      "step": 841
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.907906976744186e-07,
+      "loss": 0.4274,
+      "step": 842
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.9077906976744184e-07,
+      "loss": 0.3998,
+      "step": 843
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.9076744186046513e-07,
+      "loss": 0.6279,
+      "step": 844
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.9075581395348836e-07,
+      "loss": 0.4642,
+      "step": 845
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.9074418604651165e-07,
+      "loss": 0.5132,
+      "step": 846
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.907325581395349e-07,
+      "loss": 0.4876,
+      "step": 847
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.907209302325581e-07,
+      "loss": 0.4772,
+      "step": 848
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.9070930232558136e-07,
+      "loss": 0.4307,
+      "step": 849
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.9069767441860464e-07,
+      "loss": 0.5181,
+      "step": 850
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.906860465116279e-07,
+      "loss": 0.4053,
+      "step": 851
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.9067441860465117e-07,
+      "loss": 0.3974,
+      "step": 852
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.906627906976744e-07,
+      "loss": 0.5176,
+      "step": 853
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.9065116279069764e-07,
+      "loss": 0.3936,
+      "step": 854
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.906395348837209e-07,
+      "loss": 0.4842,
+      "step": 855
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.9062790697674416e-07,
+      "loss": 0.4376,
+      "step": 856
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.906162790697674e-07,
+      "loss": 0.5369,
+      "step": 857
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.906046511627907e-07,
+      "loss": 0.4935,
+      "step": 858
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.9059302325581397e-07,
+      "loss": 0.4771,
+      "step": 859
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.905813953488372e-07,
+      "loss": 0.428,
+      "step": 860
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.9056976744186044e-07,
+      "loss": 0.4079,
+      "step": 861
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.905581395348837e-07,
+      "loss": 0.4482,
+      "step": 862
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.9054651162790697e-07,
+      "loss": 0.3944,
+      "step": 863
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.905348837209302e-07,
+      "loss": 0.4725,
+      "step": 864
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.905232558139535e-07,
+      "loss": 0.4737,
+      "step": 865
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.905116279069767e-07,
+      "loss": 0.498,
+      "step": 866
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.905e-07,
+      "loss": 0.4139,
+      "step": 867
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.9048837209302325e-07,
+      "loss": 0.3925,
+      "step": 868
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.904767441860465e-07,
+      "loss": 0.5261,
+      "step": 869
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.904651162790697e-07,
+      "loss": 0.4489,
+      "step": 870
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.90453488372093e-07,
+      "loss": 0.4101,
+      "step": 871
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.904418604651163e-07,
+      "loss": 0.4624,
+      "step": 872
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.9043023255813953e-07,
+      "loss": 0.3652,
+      "step": 873
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.9041860465116277e-07,
+      "loss": 0.4052,
+      "step": 874
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.9040697674418605e-07,
+      "loss": 0.4454,
+      "step": 875
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.903953488372093e-07,
+      "loss": 0.4846,
+      "step": 876
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.903837209302325e-07,
+      "loss": 0.4179,
+      "step": 877
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.903720930232558e-07,
+      "loss": 0.4287,
+      "step": 878
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.9036046511627905e-07,
+      "loss": 0.4032,
+      "step": 879
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.9034883720930234e-07,
+      "loss": 0.4262,
+      "step": 880
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.9033720930232557e-07,
+      "loss": 0.4322,
+      "step": 881
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.903255813953488e-07,
+      "loss": 0.3802,
+      "step": 882
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9031395348837204e-07,
+      "loss": 0.4777,
+      "step": 883
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9030232558139533e-07,
+      "loss": 0.4134,
+      "step": 884
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.902906976744186e-07,
+      "loss": 0.4529,
+      "step": 885
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9027906976744185e-07,
+      "loss": 0.4285,
+      "step": 886
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.902674418604651e-07,
+      "loss": 0.4936,
+      "step": 887
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.902558139534884e-07,
+      "loss": 0.3857,
+      "step": 888
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.902441860465116e-07,
+      "loss": 0.3933,
+      "step": 889
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9023255813953485e-07,
+      "loss": 0.3709,
+      "step": 890
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9022093023255814e-07,
+      "loss": 0.3423,
+      "step": 891
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.9020930232558137e-07,
+      "loss": 0.4439,
+      "step": 892
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.9019767441860466e-07,
+      "loss": 0.3202,
+      "step": 893
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.901860465116279e-07,
+      "loss": 0.3892,
+      "step": 894
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.9017441860465113e-07,
+      "loss": 0.3202,
+      "step": 895
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.901627906976744e-07,
+      "loss": 0.4431,
+      "step": 896
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.9015116279069765e-07,
+      "loss": 0.4615,
+      "step": 897
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.9013953488372094e-07,
+      "loss": 0.3832,
+      "step": 898
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.901279069767442e-07,
+      "loss": 0.4266,
+      "step": 899
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.9011627906976747e-07,
+      "loss": 0.3961,
+      "step": 900
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.901046511627907e-07,
+      "loss": 0.4078,
+      "step": 901
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.9009302325581394e-07,
+      "loss": 0.4475,
+      "step": 902
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.9008139534883717e-07,
+      "loss": 0.3859,
+      "step": 903
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.9006976744186046e-07,
+      "loss": 0.4794,
+      "step": 904
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.900581395348837e-07,
+      "loss": 0.3422,
+      "step": 905
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.90046511627907e-07,
+      "loss": 0.7503,
+      "step": 906
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.900348837209302e-07,
+      "loss": 0.4257,
+      "step": 907
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.9002325581395345e-07,
+      "loss": 0.3751,
+      "step": 908
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.9001162790697674e-07,
+      "loss": 0.3914,
+      "step": 909
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.9e-07,
+      "loss": 0.3599,
+      "step": 910
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.899883720930232e-07,
+      "loss": 0.3942,
+      "step": 911
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.899767441860465e-07,
+      "loss": 0.3448,
+      "step": 912
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.899651162790698e-07,
+      "loss": 0.4159,
+      "step": 913
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.89953488372093e-07,
+      "loss": 0.3586,
+      "step": 914
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.8994186046511626e-07,
+      "loss": 0.3025,
+      "step": 915
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.899302325581395e-07,
+      "loss": 0.3984,
+      "step": 916
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.899186046511628e-07,
+      "loss": 0.4752,
+      "step": 917
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.89906976744186e-07,
+      "loss": 0.403,
+      "step": 918
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.898953488372093e-07,
+      "loss": 0.3901,
+      "step": 919
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.8988372093023254e-07,
+      "loss": 0.4042,
+      "step": 920
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.8987209302325583e-07,
+      "loss": 0.372,
+      "step": 921
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.8986046511627906e-07,
+      "loss": 0.4162,
+      "step": 922
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.898488372093023e-07,
+      "loss": 0.4544,
+      "step": 923
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.8983720930232553e-07,
+      "loss": 0.3449,
+      "step": 924
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.898255813953488e-07,
+      "loss": 0.3002,
+      "step": 925
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.898139534883721e-07,
+      "loss": 0.3684,
+      "step": 926
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.8980232558139535e-07,
+      "loss": 0.3931,
+      "step": 927
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.897906976744186e-07,
+      "loss": 0.3526,
+      "step": 928
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.8977906976744187e-07,
+      "loss": 0.3629,
+      "step": 929
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.897674418604651e-07,
+      "loss": 0.3995,
+      "step": 930
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.8975581395348834e-07,
+      "loss": 0.3649,
+      "step": 931
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.8974418604651163e-07,
+      "loss": 0.3913,
+      "step": 932
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.8973255813953486e-07,
+      "loss": 0.4817,
+      "step": 933
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8972093023255815e-07,
+      "loss": 0.334,
+      "step": 934
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.897093023255814e-07,
+      "loss": 0.3602,
+      "step": 935
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.896976744186046e-07,
+      "loss": 0.4181,
+      "step": 936
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8968604651162786e-07,
+      "loss": 0.3659,
+      "step": 937
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8967441860465115e-07,
+      "loss": 0.3933,
+      "step": 938
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.8966279069767443e-07,
+      "loss": 0.3825,
+      "step": 939
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.8965116279069767e-07,
+      "loss": 0.342,
+      "step": 940
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.896395348837209e-07,
+      "loss": 0.2992,
+      "step": 941
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.896279069767442e-07,
+      "loss": 0.5959,
+      "step": 942
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.8961627906976743e-07,
+      "loss": 0.4869,
+      "step": 943
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.8960465116279066e-07,
+      "loss": 0.3007,
+      "step": 944
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.8959302325581395e-07,
+      "loss": 0.5336,
+      "step": 945
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.895813953488372e-07,
+      "loss": 0.3366,
+      "step": 946
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.895697674418605e-07,
+      "loss": 0.4294,
+      "step": 947
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.895581395348837e-07,
+      "loss": 0.3453,
+      "step": 948
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.8954651162790694e-07,
+      "loss": 0.3282,
+      "step": 949
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.8953488372093023e-07,
+      "loss": 0.4037,
+      "step": 950
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.8952325581395347e-07,
+      "loss": 0.3605,
+      "step": 951
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.895116279069767e-07,
+      "loss": 0.2871,
+      "step": 952
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.895e-07,
+      "loss": 0.3684,
+      "step": 953
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.894883720930233e-07,
+      "loss": 0.3436,
+      "step": 954
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.894767441860465e-07,
+      "loss": 0.473,
+      "step": 955
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.8946511627906975e-07,
+      "loss": 0.3886,
+      "step": 956
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.89453488372093e-07,
+      "loss": 0.3355,
+      "step": 957
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.894418604651163e-07,
+      "loss": 0.498,
+      "step": 958
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.894302325581395e-07,
+      "loss": 0.3406,
+      "step": 959
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.894186046511628e-07,
+      "loss": 0.2956,
+      "step": 960
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.8940697674418603e-07,
+      "loss": 0.3464,
+      "step": 961
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.8939534883720927e-07,
+      "loss": 0.417,
+      "step": 962
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.8938372093023256e-07,
+      "loss": 0.3474,
+      "step": 963
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.893720930232558e-07,
+      "loss": 0.3367,
+      "step": 964
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.89360465116279e-07,
+      "loss": 0.3819,
+      "step": 965
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.893488372093023e-07,
+      "loss": 0.2943,
+      "step": 966
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.893372093023256e-07,
+      "loss": 0.3596,
+      "step": 967
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.8932558139534884e-07,
+      "loss": 0.4479,
+      "step": 968
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.8931395348837207e-07,
+      "loss": 0.7429,
+      "step": 969
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.893023255813953e-07,
+      "loss": 0.2924,
+      "step": 970
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.892906976744186e-07,
+      "loss": 0.4029,
+      "step": 971
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.8927906976744183e-07,
+      "loss": 0.4229,
+      "step": 972
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.892674418604651e-07,
+      "loss": 0.3479,
+      "step": 973
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.8925581395348836e-07,
+      "loss": 0.4132,
+      "step": 974
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.8924418604651164e-07,
+      "loss": 0.3325,
+      "step": 975
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.892325581395349e-07,
+      "loss": 0.3156,
+      "step": 976
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.892209302325581e-07,
+      "loss": 0.4555,
+      "step": 977
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.8920930232558135e-07,
+      "loss": 0.2596,
+      "step": 978
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.8919767441860464e-07,
+      "loss": 0.664,
+      "step": 979
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.891860465116279e-07,
+      "loss": 0.2594,
+      "step": 980
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.8917441860465116e-07,
+      "loss": 0.3204,
+      "step": 981
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.891627906976744e-07,
+      "loss": 0.307,
+      "step": 982
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.8915116279069763e-07,
+      "loss": 0.3122,
+      "step": 983
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.891395348837209e-07,
+      "loss": 0.4457,
+      "step": 984
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.8912790697674415e-07,
+      "loss": 0.3252,
+      "step": 985
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.8911627906976744e-07,
+      "loss": 0.342,
+      "step": 986
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.891046511627907e-07,
+      "loss": 0.3187,
+      "step": 987
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.8909302325581397e-07,
+      "loss": 0.3245,
+      "step": 988
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.890813953488372e-07,
+      "loss": 0.3877,
+      "step": 989
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.8906976744186044e-07,
+      "loss": 0.3262,
+      "step": 990
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.8905813953488367e-07,
+      "loss": 0.3504,
+      "step": 991
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.8904651162790696e-07,
+      "loss": 0.3645,
+      "step": 992
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.8903488372093025e-07,
+      "loss": 0.2873,
+      "step": 993
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.890232558139535e-07,
+      "loss": 0.2946,
+      "step": 994
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.890116279069767e-07,
+      "loss": 0.3769,
+      "step": 995
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.89e-07,
+      "loss": 0.3396,
+      "step": 996
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.8898837209302324e-07,
+      "loss": 0.2808,
+      "step": 997
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.889767441860465e-07,
+      "loss": 0.227,
+      "step": 998
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.8896511627906977e-07,
+      "loss": 0.3525,
+      "step": 999
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.88953488372093e-07,
+      "loss": 0.2693,
+      "step": 1000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.889418604651163e-07,
+      "loss": 0.2869,
+      "step": 1001
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.889302325581395e-07,
+      "loss": 0.3227,
+      "step": 1002
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.8891860465116276e-07,
+      "loss": 0.2448,
+      "step": 1003
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.8890697674418605e-07,
+      "loss": 0.274,
+      "step": 1004
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.888953488372093e-07,
+      "loss": 0.3129,
+      "step": 1005
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.888837209302325e-07,
+      "loss": 0.3582,
+      "step": 1006
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.888720930232558e-07,
+      "loss": 0.3197,
+      "step": 1007
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.888604651162791e-07,
+      "loss": 0.3092,
+      "step": 1008
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8884883720930233e-07,
+      "loss": 0.2505,
+      "step": 1009
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.8883720930232557e-07,
+      "loss": 0.3274,
+      "step": 1010
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.888255813953488e-07,
+      "loss": 0.3679,
+      "step": 1011
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.888139534883721e-07,
+      "loss": 0.417,
+      "step": 1012
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.888023255813953e-07,
+      "loss": 0.3431,
+      "step": 1013
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.887906976744186e-07,
+      "loss": 0.2794,
+      "step": 1014
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.8877906976744185e-07,
+      "loss": 0.3019,
+      "step": 1015
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.887674418604651e-07,
+      "loss": 0.3228,
+      "step": 1016
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.8875581395348837e-07,
+      "loss": 0.7409,
+      "step": 1017
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.887441860465116e-07,
+      "loss": 0.3594,
+      "step": 1018
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.8873255813953484e-07,
+      "loss": 0.2873,
+      "step": 1019
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.8872093023255813e-07,
+      "loss": 0.2951,
+      "step": 1020
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.887093023255814e-07,
+      "loss": 0.3314,
+      "step": 1021
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.8869767441860465e-07,
+      "loss": 0.245,
+      "step": 1022
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.886860465116279e-07,
+      "loss": 0.3289,
+      "step": 1023
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.886744186046511e-07,
+      "loss": 0.2906,
+      "step": 1024
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.886627906976744e-07,
+      "loss": 0.3434,
+      "step": 1025
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.8865116279069765e-07,
+      "loss": 0.2385,
+      "step": 1026
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8863953488372093e-07,
+      "loss": 0.274,
+      "step": 1027
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8862790697674417e-07,
+      "loss": 0.2529,
+      "step": 1028
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8861627906976746e-07,
+      "loss": 0.2793,
+      "step": 1029
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.886046511627907e-07,
+      "loss": 0.3041,
+      "step": 1030
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.8859302325581393e-07,
+      "loss": 0.3607,
+      "step": 1031
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.8858139534883716e-07,
+      "loss": 0.6064,
+      "step": 1032
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.8856976744186045e-07,
+      "loss": 0.3664,
+      "step": 1033
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.8855813953488374e-07,
+      "loss": 0.2277,
+      "step": 1034
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.9564796905222437,
+      "eval_loss": 0.31594955921173096,
+      "eval_runtime": 1.5584,
+      "eval_samples_per_second": 663.488,
+      "eval_steps_per_second": 10.908,
+      "step": 1034
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.88546511627907e-07,
+      "loss": 0.2625,
+      "step": 1035
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.885348837209302e-07,
+      "loss": 0.2705,
+      "step": 1036
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.8852325581395345e-07,
+      "loss": 0.3023,
+      "step": 1037
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.8851162790697673e-07,
+      "loss": 0.3033,
+      "step": 1038
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.8849999999999997e-07,
+      "loss": 0.2748,
+      "step": 1039
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.8848837209302326e-07,
+      "loss": 0.2935,
+      "step": 1040
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.884767441860465e-07,
+      "loss": 0.2427,
+      "step": 1041
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.884651162790698e-07,
+      "loss": 0.4108,
+      "step": 1042
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.88453488372093e-07,
+      "loss": 0.2173,
+      "step": 1043
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.8844186046511625e-07,
+      "loss": 0.2407,
+      "step": 1044
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.884302325581395e-07,
+      "loss": 0.3424,
+      "step": 1045
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.884186046511628e-07,
+      "loss": 0.2349,
+      "step": 1046
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.8840697674418606e-07,
+      "loss": 0.2501,
+      "step": 1047
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.883953488372093e-07,
+      "loss": 0.2712,
+      "step": 1048
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.8838372093023253e-07,
+      "loss": 0.2234,
+      "step": 1049
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.883720930232558e-07,
+      "loss": 0.2392,
+      "step": 1050
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.8836046511627906e-07,
+      "loss": 0.4822,
+      "step": 1051
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.883488372093023e-07,
+      "loss": 0.2528,
+      "step": 1052
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.883372093023256e-07,
+      "loss": 0.2816,
+      "step": 1053
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.883255813953488e-07,
+      "loss": 0.4371,
+      "step": 1054
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.883139534883721e-07,
+      "loss": 0.2402,
+      "step": 1055
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.8830232558139534e-07,
+      "loss": 0.2918,
+      "step": 1056
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.882906976744186e-07,
+      "loss": 0.2612,
+      "step": 1057
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.8827906976744186e-07,
+      "loss": 0.2202,
+      "step": 1058
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.882674418604651e-07,
+      "loss": 0.3171,
+      "step": 1059
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.8825581395348833e-07,
+      "loss": 0.2922,
+      "step": 1060
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.882441860465116e-07,
+      "loss": 0.2647,
+      "step": 1061
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.882325581395349e-07,
+      "loss": 0.2325,
+      "step": 1062
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.8822093023255814e-07,
+      "loss": 0.2288,
+      "step": 1063
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.882093023255814e-07,
+      "loss": 0.3264,
+      "step": 1064
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.881976744186046e-07,
+      "loss": 0.5481,
+      "step": 1065
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.881860465116279e-07,
+      "loss": 0.2554,
+      "step": 1066
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.8817441860465114e-07,
+      "loss": 0.3074,
+      "step": 1067
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.8816279069767443e-07,
+      "loss": 0.2542,
+      "step": 1068
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.8815116279069766e-07,
+      "loss": 0.3013,
+      "step": 1069
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.881395348837209e-07,
+      "loss": 0.2738,
+      "step": 1070
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.881279069767442e-07,
+      "loss": 0.247,
+      "step": 1071
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.881162790697674e-07,
+      "loss": 0.5489,
+      "step": 1072
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.8810465116279066e-07,
+      "loss": 0.251,
+      "step": 1073
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.8809302325581394e-07,
+      "loss": 0.2399,
+      "step": 1074
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.8808139534883723e-07,
+      "loss": 0.203,
+      "step": 1075
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.8806976744186047e-07,
+      "loss": 0.2885,
+      "step": 1076
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.880581395348837e-07,
+      "loss": 0.339,
+      "step": 1077
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.8804651162790694e-07,
+      "loss": 0.2405,
+      "step": 1078
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.880348837209302e-07,
+      "loss": 0.3342,
+      "step": 1079
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.8802325581395346e-07,
+      "loss": 0.2189,
+      "step": 1080
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.8801162790697675e-07,
+      "loss": 0.2326,
+      "step": 1081
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.88e-07,
+      "loss": 0.2287,
+      "step": 1082
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.8798837209302327e-07,
+      "loss": 0.2213,
+      "step": 1083
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.879767441860465e-07,
+      "loss": 0.2156,
+      "step": 1084
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.8796511627906974e-07,
+      "loss": 0.3939,
+      "step": 1085
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.87953488372093e-07,
+      "loss": 0.3582,
+      "step": 1086
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.8794186046511627e-07,
+      "loss": 0.3826,
+      "step": 1087
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.8793023255813956e-07,
+      "loss": 0.2186,
+      "step": 1088
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.879186046511628e-07,
+      "loss": 0.3516,
+      "step": 1089
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.87906976744186e-07,
+      "loss": 0.2641,
+      "step": 1090
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.8789534883720926e-07,
+      "loss": 0.2205,
+      "step": 1091
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.8788372093023255e-07,
+      "loss": 0.2317,
+      "step": 1092
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.878720930232558e-07,
+      "loss": 0.4036,
+      "step": 1093
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.8786046511627907e-07,
+      "loss": 0.2378,
+      "step": 1094
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.878488372093023e-07,
+      "loss": 0.3519,
+      "step": 1095
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.878372093023256e-07,
+      "loss": 0.2214,
+      "step": 1096
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.8782558139534883e-07,
+      "loss": 0.2038,
+      "step": 1097
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.8781395348837207e-07,
+      "loss": 0.2019,
+      "step": 1098
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.878023255813953e-07,
+      "loss": 0.3173,
+      "step": 1099
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.877906976744186e-07,
+      "loss": 0.1846,
+      "step": 1100
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.877790697674418e-07,
+      "loss": 0.6026,
+      "step": 1101
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.877674418604651e-07,
+      "loss": 0.2181,
+      "step": 1102
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.8775581395348835e-07,
+      "loss": 0.2125,
+      "step": 1103
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.8774418604651164e-07,
+      "loss": 0.2131,
+      "step": 1104
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.8773255813953487e-07,
+      "loss": 0.2514,
+      "step": 1105
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.877209302325581e-07,
+      "loss": 0.226,
+      "step": 1106
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.877093023255814e-07,
+      "loss": 0.1957,
+      "step": 1107
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.8769767441860463e-07,
+      "loss": 0.2457,
+      "step": 1108
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.876860465116279e-07,
+      "loss": 0.3524,
+      "step": 1109
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.8767441860465115e-07,
+      "loss": 0.2022,
+      "step": 1110
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.876627906976744e-07,
+      "loss": 0.2764,
+      "step": 1111
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.876511627906977e-07,
+      "loss": 0.1791,
+      "step": 1112
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.876395348837209e-07,
+      "loss": 0.212,
+      "step": 1113
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.8762790697674415e-07,
+      "loss": 0.4745,
+      "step": 1114
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.8761627906976744e-07,
+      "loss": 0.2352,
+      "step": 1115
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.876046511627907e-07,
+      "loss": 0.2981,
+      "step": 1116
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.8759302325581396e-07,
+      "loss": 0.2123,
+      "step": 1117
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.875813953488372e-07,
+      "loss": 0.2299,
+      "step": 1118
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.8756976744186043e-07,
+      "loss": 0.219,
+      "step": 1119
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.875581395348837e-07,
+      "loss": 0.3785,
+      "step": 1120
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.8754651162790695e-07,
+      "loss": 0.4357,
+      "step": 1121
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.8753488372093024e-07,
+      "loss": 0.1732,
+      "step": 1122
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.875232558139535e-07,
+      "loss": 0.2673,
+      "step": 1123
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.875116279069767e-07,
+      "loss": 0.1909,
+      "step": 1124
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.875e-07,
+      "loss": 0.4429,
+      "step": 1125
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8748837209302324e-07,
+      "loss": 0.2098,
+      "step": 1126
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8747674418604647e-07,
+      "loss": 0.2329,
+      "step": 1127
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8746511627906976e-07,
+      "loss": 0.1664,
+      "step": 1128
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8745348837209305e-07,
+      "loss": 0.2604,
+      "step": 1129
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.874418604651163e-07,
+      "loss": 0.2304,
+      "step": 1130
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.874302325581395e-07,
+      "loss": 0.2215,
+      "step": 1131
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.8741860465116275e-07,
+      "loss": 0.1959,
+      "step": 1132
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.8740697674418604e-07,
+      "loss": 0.3371,
+      "step": 1133
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.873953488372093e-07,
+      "loss": 0.7185,
+      "step": 1134
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.8738372093023256e-07,
+      "loss": 0.2142,
+      "step": 1135
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.873720930232558e-07,
+      "loss": 0.2214,
+      "step": 1136
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.873604651162791e-07,
+      "loss": 0.2866,
+      "step": 1137
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.873488372093023e-07,
+      "loss": 0.1691,
+      "step": 1138
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.8733720930232556e-07,
+      "loss": 0.2031,
+      "step": 1139
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.873255813953488e-07,
+      "loss": 0.1949,
+      "step": 1140
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.873139534883721e-07,
+      "loss": 0.2313,
+      "step": 1141
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.8730232558139537e-07,
+      "loss": 0.2918,
+      "step": 1142
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.872906976744186e-07,
+      "loss": 0.1858,
+      "step": 1143
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.8727906976744184e-07,
+      "loss": 0.1717,
+      "step": 1144
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.872674418604651e-07,
+      "loss": 0.2066,
+      "step": 1145
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.8725581395348836e-07,
+      "loss": 0.1986,
+      "step": 1146
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.872441860465116e-07,
+      "loss": 0.6503,
+      "step": 1147
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.872325581395349e-07,
+      "loss": 0.2312,
+      "step": 1148
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.872209302325581e-07,
+      "loss": 0.2714,
+      "step": 1149
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.872093023255814e-07,
+      "loss": 0.2193,
+      "step": 1150
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.8719767441860465e-07,
+      "loss": 0.2759,
+      "step": 1151
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.871860465116279e-07,
+      "loss": 0.1848,
+      "step": 1152
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.871744186046511e-07,
+      "loss": 0.2483,
+      "step": 1153
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.871627906976744e-07,
+      "loss": 0.2046,
+      "step": 1154
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.8715116279069764e-07,
+      "loss": 0.2849,
+      "step": 1155
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.8713953488372093e-07,
+      "loss": 0.2844,
+      "step": 1156
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.8712790697674416e-07,
+      "loss": 0.3356,
+      "step": 1157
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.8711627906976745e-07,
+      "loss": 0.1657,
+      "step": 1158
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.871046511627907e-07,
+      "loss": 0.2084,
+      "step": 1159
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.870930232558139e-07,
+      "loss": 0.1688,
+      "step": 1160
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.870813953488372e-07,
+      "loss": 0.38,
+      "step": 1161
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8706976744186045e-07,
+      "loss": 0.193,
+      "step": 1162
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8705813953488373e-07,
+      "loss": 0.1736,
+      "step": 1163
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8704651162790697e-07,
+      "loss": 0.2433,
+      "step": 1164
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.870348837209302e-07,
+      "loss": 0.2126,
+      "step": 1165
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.870232558139535e-07,
+      "loss": 0.2802,
+      "step": 1166
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8701162790697673e-07,
+      "loss": 0.1684,
+      "step": 1167
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8699999999999996e-07,
+      "loss": 0.2296,
+      "step": 1168
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.8698837209302325e-07,
+      "loss": 0.1598,
+      "step": 1169
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.869767441860465e-07,
+      "loss": 0.2991,
+      "step": 1170
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.869651162790698e-07,
+      "loss": 0.2041,
+      "step": 1171
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.86953488372093e-07,
+      "loss": 0.2465,
+      "step": 1172
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8694186046511624e-07,
+      "loss": 0.1779,
+      "step": 1173
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8693023255813953e-07,
+      "loss": 0.1856,
+      "step": 1174
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8691860465116277e-07,
+      "loss": 0.4353,
+      "step": 1175
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.8690697674418606e-07,
+      "loss": 0.1548,
+      "step": 1176
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.868953488372093e-07,
+      "loss": 0.5035,
+      "step": 1177
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.8688372093023253e-07,
+      "loss": 0.1552,
+      "step": 1178
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.868720930232558e-07,
+      "loss": 0.2914,
+      "step": 1179
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.8686046511627905e-07,
+      "loss": 0.1598,
+      "step": 1180
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.868488372093023e-07,
+      "loss": 0.1832,
+      "step": 1181
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.868372093023256e-07,
+      "loss": 0.1991,
+      "step": 1182
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.8682558139534886e-07,
+      "loss": 0.2222,
+      "step": 1183
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.868139534883721e-07,
+      "loss": 0.1709,
+      "step": 1184
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.8680232558139533e-07,
+      "loss": 0.2596,
+      "step": 1185
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.8679069767441857e-07,
+      "loss": 0.2359,
+      "step": 1186
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.8677906976744186e-07,
+      "loss": 0.2897,
+      "step": 1187
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.867674418604651e-07,
+      "loss": 0.1792,
+      "step": 1188
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.867558139534884e-07,
+      "loss": 0.1976,
+      "step": 1189
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.867441860465116e-07,
+      "loss": 0.2613,
+      "step": 1190
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.867325581395349e-07,
+      "loss": 0.1873,
+      "step": 1191
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.8672093023255814e-07,
+      "loss": 0.1735,
+      "step": 1192
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.8670930232558137e-07,
+      "loss": 0.3398,
+      "step": 1193
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.866976744186046e-07,
+      "loss": 0.1679,
+      "step": 1194
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.866860465116279e-07,
+      "loss": 0.1684,
+      "step": 1195
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.8667441860465113e-07,
+      "loss": 0.1879,
+      "step": 1196
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.866627906976744e-07,
+      "loss": 0.4695,
+      "step": 1197
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.8665116279069766e-07,
+      "loss": 0.1625,
+      "step": 1198
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.866395348837209e-07,
+      "loss": 0.1766,
+      "step": 1199
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.866279069767442e-07,
+      "loss": 0.3011,
+      "step": 1200
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.866162790697674e-07,
+      "loss": 0.2357,
+      "step": 1201
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.866046511627907e-07,
+      "loss": 0.2026,
+      "step": 1202
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.8659302325581394e-07,
+      "loss": 0.3494,
+      "step": 1203
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.865813953488372e-07,
+      "loss": 0.2297,
+      "step": 1204
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.8656976744186046e-07,
+      "loss": 0.1599,
+      "step": 1205
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.865581395348837e-07,
+      "loss": 0.2013,
+      "step": 1206
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.8654651162790693e-07,
+      "loss": 0.1785,
+      "step": 1207
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.865348837209302e-07,
+      "loss": 0.3071,
+      "step": 1208
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.8652325581395345e-07,
+      "loss": 0.2369,
+      "step": 1209
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.8651162790697674e-07,
+      "loss": 0.2453,
+      "step": 1210
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.865e-07,
+      "loss": 0.1511,
+      "step": 1211
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.8648837209302327e-07,
+      "loss": 0.3308,
+      "step": 1212
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.864767441860465e-07,
+      "loss": 0.4335,
+      "step": 1213
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.8646511627906974e-07,
+      "loss": 0.5395,
+      "step": 1214
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.86453488372093e-07,
+      "loss": 0.1904,
+      "step": 1215
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.8644186046511626e-07,
+      "loss": 0.2184,
+      "step": 1216
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.8643023255813955e-07,
+      "loss": 0.1717,
+      "step": 1217
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.864186046511628e-07,
+      "loss": 0.2739,
+      "step": 1218
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.86406976744186e-07,
+      "loss": 0.1671,
+      "step": 1219
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.863953488372093e-07,
+      "loss": 0.742,
+      "step": 1220
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.8638372093023254e-07,
+      "loss": 0.1701,
+      "step": 1221
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.863720930232558e-07,
+      "loss": 0.1391,
+      "step": 1222
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.8636046511627907e-07,
+      "loss": 0.1753,
+      "step": 1223
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.863488372093023e-07,
+      "loss": 0.1244,
+      "step": 1224
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.863372093023256e-07,
+      "loss": 0.1415,
+      "step": 1225
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.863255813953488e-07,
+      "loss": 0.215,
+      "step": 1226
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.8631395348837206e-07,
+      "loss": 0.2297,
+      "step": 1227
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.863023255813953e-07,
+      "loss": 0.1726,
+      "step": 1228
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.862906976744186e-07,
+      "loss": 0.2018,
+      "step": 1229
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8627906976744187e-07,
+      "loss": 0.1594,
+      "step": 1230
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.862674418604651e-07,
+      "loss": 0.1941,
+      "step": 1231
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8625581395348834e-07,
+      "loss": 0.48,
+      "step": 1232
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.8624418604651163e-07,
+      "loss": 0.8055,
+      "step": 1233
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.8623255813953487e-07,
+      "loss": 0.2293,
+      "step": 1234
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.862209302325581e-07,
+      "loss": 0.6754,
+      "step": 1235
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.862093023255814e-07,
+      "loss": 0.2101,
+      "step": 1236
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.861976744186047e-07,
+      "loss": 0.1368,
+      "step": 1237
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.861860465116279e-07,
+      "loss": 0.2303,
+      "step": 1238
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8617441860465115e-07,
+      "loss": 0.1948,
+      "step": 1239
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.861627906976744e-07,
+      "loss": 0.1976,
+      "step": 1240
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.8615116279069767e-07,
+      "loss": 0.1664,
+      "step": 1241
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.861395348837209e-07,
+      "loss": 0.227,
+      "step": 1242
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.861279069767442e-07,
+      "loss": 0.1717,
+      "step": 1243
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.8611627906976743e-07,
+      "loss": 0.2918,
+      "step": 1244
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.861046511627907e-07,
+      "loss": 0.1731,
+      "step": 1245
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.8609302325581395e-07,
+      "loss": 0.2213,
+      "step": 1246
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.860813953488372e-07,
+      "loss": 0.1335,
+      "step": 1247
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.860697674418604e-07,
+      "loss": 0.2564,
+      "step": 1248
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.860581395348837e-07,
+      "loss": 0.4124,
+      "step": 1249
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.8604651162790695e-07,
+      "loss": 0.2069,
+      "step": 1250
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.8603488372093023e-07,
+      "loss": 0.209,
+      "step": 1251
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.8602325581395347e-07,
+      "loss": 0.1958,
+      "step": 1252
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.860116279069767e-07,
+      "loss": 0.2949,
+      "step": 1253
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.86e-07,
+      "loss": 0.2727,
+      "step": 1254
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.8598837209302323e-07,
+      "loss": 0.2383,
+      "step": 1255
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.859767441860465e-07,
+      "loss": 0.1301,
+      "step": 1256
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.8596511627906975e-07,
+      "loss": 0.1962,
+      "step": 1257
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.8595348837209304e-07,
+      "loss": 0.1685,
+      "step": 1258
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.859418604651163e-07,
+      "loss": 0.1302,
+      "step": 1259
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.859302325581395e-07,
+      "loss": 0.1369,
+      "step": 1260
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.8591860465116275e-07,
+      "loss": 0.1147,
+      "step": 1261
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.8590697674418603e-07,
+      "loss": 0.2745,
+      "step": 1262
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.8589534883720927e-07,
+      "loss": 0.1706,
+      "step": 1263
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.8588372093023256e-07,
+      "loss": 0.1515,
+      "step": 1264
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.858720930232558e-07,
+      "loss": 0.1554,
+      "step": 1265
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.858604651162791e-07,
+      "loss": 0.3902,
+      "step": 1266
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.858488372093023e-07,
+      "loss": 0.1256,
+      "step": 1267
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.8583720930232555e-07,
+      "loss": 0.1703,
+      "step": 1268
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.8582558139534884e-07,
+      "loss": 0.1416,
+      "step": 1269
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.858139534883721e-07,
+      "loss": 0.167,
+      "step": 1270
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.8580232558139536e-07,
+      "loss": 0.132,
+      "step": 1271
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.857906976744186e-07,
+      "loss": 0.1579,
+      "step": 1272
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.8577906976744183e-07,
+      "loss": 0.2326,
+      "step": 1273
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.857674418604651e-07,
+      "loss": 0.206,
+      "step": 1274
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.8575581395348836e-07,
+      "loss": 0.1489,
+      "step": 1275
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.857441860465116e-07,
+      "loss": 0.249,
+      "step": 1276
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.857325581395349e-07,
+      "loss": 0.1586,
+      "step": 1277
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.857209302325581e-07,
+      "loss": 0.4254,
+      "step": 1278
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.857093023255814e-07,
+      "loss": 0.1227,
+      "step": 1279
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.8569767441860464e-07,
+      "loss": 0.1423,
+      "step": 1280
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.856860465116279e-07,
+      "loss": 0.1569,
+      "step": 1281
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.856744186046511e-07,
+      "loss": 0.1327,
+      "step": 1282
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.856627906976744e-07,
+      "loss": 0.1254,
+      "step": 1283
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.856511627906977e-07,
+      "loss": 0.1153,
+      "step": 1284
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.856395348837209e-07,
+      "loss": 0.1709,
+      "step": 1285
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.8562790697674416e-07,
+      "loss": 0.1454,
+      "step": 1286
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.8561627906976744e-07,
+      "loss": 0.1706,
+      "step": 1287
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.856046511627907e-07,
+      "loss": 0.1709,
+      "step": 1288
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.855930232558139e-07,
+      "loss": 0.1521,
+      "step": 1289
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.855813953488372e-07,
+      "loss": 0.168,
+      "step": 1290
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.855697674418605e-07,
+      "loss": 0.1299,
+      "step": 1291
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.8555813953488373e-07,
+      "loss": 0.2862,
+      "step": 1292
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.8554651162790696e-07,
+      "loss": 0.2462,
+      "step": 1293
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.855348837209302e-07,
+      "loss": 0.1886,
+      "step": 1294
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.855232558139535e-07,
+      "loss": 0.1077,
+      "step": 1295
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.855116279069767e-07,
+      "loss": 0.1519,
+      "step": 1296
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.855e-07,
+      "loss": 0.0969,
+      "step": 1297
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.8548837209302324e-07,
+      "loss": 0.1674,
+      "step": 1298
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.8547674418604653e-07,
+      "loss": 0.1479,
+      "step": 1299
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.8546511627906977e-07,
+      "loss": 0.1619,
+      "step": 1300
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.85453488372093e-07,
+      "loss": 0.0968,
+      "step": 1301
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.8544186046511624e-07,
+      "loss": 0.3504,
+      "step": 1302
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.854302325581395e-07,
+      "loss": 0.1231,
+      "step": 1303
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.8541860465116276e-07,
+      "loss": 0.1171,
+      "step": 1304
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.8540697674418605e-07,
+      "loss": 0.164,
+      "step": 1305
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.853953488372093e-07,
+      "loss": 0.2448,
+      "step": 1306
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.853837209302325e-07,
+      "loss": 0.2495,
+      "step": 1307
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.853720930232558e-07,
+      "loss": 0.1222,
+      "step": 1308
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.8536046511627904e-07,
+      "loss": 0.2323,
+      "step": 1309
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.8534883720930233e-07,
+      "loss": 0.5159,
+      "step": 1310
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.8533720930232557e-07,
+      "loss": 0.1324,
+      "step": 1311
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.8532558139534886e-07,
+      "loss": 0.116,
+      "step": 1312
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.853139534883721e-07,
+      "loss": 0.1275,
+      "step": 1313
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.853023255813953e-07,
+      "loss": 0.1126,
+      "step": 1314
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.8529069767441856e-07,
+      "loss": 0.1221,
+      "step": 1315
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.8527906976744185e-07,
+      "loss": 0.1523,
+      "step": 1316
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.852674418604651e-07,
+      "loss": 0.2238,
+      "step": 1317
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.8525581395348837e-07,
+      "loss": 0.0995,
+      "step": 1318
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.852441860465116e-07,
+      "loss": 0.2371,
+      "step": 1319
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.852325581395349e-07,
+      "loss": 0.3613,
+      "step": 1320
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.8522093023255813e-07,
+      "loss": 0.1157,
+      "step": 1321
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.8520930232558137e-07,
+      "loss": 0.1463,
+      "step": 1322
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.851976744186046e-07,
+      "loss": 0.7299,
+      "step": 1323
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.851860465116279e-07,
+      "loss": 0.1943,
+      "step": 1324
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.851744186046512e-07,
+      "loss": 0.1032,
+      "step": 1325
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.851627906976744e-07,
+      "loss": 0.0899,
+      "step": 1326
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8515116279069765e-07,
+      "loss": 0.2248,
+      "step": 1327
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8513953488372094e-07,
+      "loss": 0.1329,
+      "step": 1328
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8512790697674417e-07,
+      "loss": 0.2021,
+      "step": 1329
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.851162790697674e-07,
+      "loss": 0.1261,
+      "step": 1330
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.851046511627907e-07,
+      "loss": 0.1311,
+      "step": 1331
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8509302325581393e-07,
+      "loss": 0.2329,
+      "step": 1332
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.850813953488372e-07,
+      "loss": 0.1227,
+      "step": 1333
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8506976744186045e-07,
+      "loss": 0.2196,
+      "step": 1334
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.850581395348837e-07,
+      "loss": 0.0958,
+      "step": 1335
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.850465116279069e-07,
+      "loss": 1.1332,
+      "step": 1336
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.850348837209302e-07,
+      "loss": 0.1523,
+      "step": 1337
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.850232558139535e-07,
+      "loss": 0.1145,
+      "step": 1338
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.8501162790697674e-07,
+      "loss": 0.1563,
+      "step": 1339
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.8499999999999997e-07,
+      "loss": 0.1414,
+      "step": 1340
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.8498837209302326e-07,
+      "loss": 0.1007,
+      "step": 1341
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.849767441860465e-07,
+      "loss": 0.0997,
+      "step": 1342
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.8496511627906973e-07,
+      "loss": 0.3156,
+      "step": 1343
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.84953488372093e-07,
+      "loss": 0.0967,
+      "step": 1344
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.8494186046511625e-07,
+      "loss": 0.1293,
+      "step": 1345
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.8493023255813954e-07,
+      "loss": 0.1965,
+      "step": 1346
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.849186046511628e-07,
+      "loss": 1.1678,
+      "step": 1347
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.84906976744186e-07,
+      "loss": 0.112,
+      "step": 1348
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.848953488372093e-07,
+      "loss": 0.1669,
+      "step": 1349
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.8488372093023254e-07,
+      "loss": 0.122,
+      "step": 1350
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.848720930232558e-07,
+      "loss": 0.3113,
+      "step": 1351
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.8486046511627906e-07,
+      "loss": 0.9757,
+      "step": 1352
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.8484883720930235e-07,
+      "loss": 0.1636,
+      "step": 1353
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.848372093023256e-07,
+      "loss": 0.1099,
+      "step": 1354
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.848255813953488e-07,
+      "loss": 0.1222,
+      "step": 1355
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.8481395348837205e-07,
+      "loss": 0.1195,
+      "step": 1356
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.8480232558139534e-07,
+      "loss": 0.1757,
+      "step": 1357
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.847906976744186e-07,
+      "loss": 0.2163,
+      "step": 1358
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.8477906976744186e-07,
+      "loss": 0.2622,
+      "step": 1359
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.847674418604651e-07,
+      "loss": 0.1092,
+      "step": 1360
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.8475581395348833e-07,
+      "loss": 0.0886,
+      "step": 1361
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.847441860465116e-07,
+      "loss": 0.1891,
+      "step": 1362
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.8473255813953486e-07,
+      "loss": 0.0861,
+      "step": 1363
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.8472093023255815e-07,
+      "loss": 0.1866,
+      "step": 1364
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.847093023255814e-07,
+      "loss": 0.1043,
+      "step": 1365
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.8469767441860467e-07,
+      "loss": 0.2509,
+      "step": 1366
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.846860465116279e-07,
+      "loss": 0.0915,
+      "step": 1367
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.8467441860465114e-07,
+      "loss": 0.153,
+      "step": 1368
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.846627906976744e-07,
+      "loss": 0.1883,
+      "step": 1369
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.8465116279069766e-07,
+      "loss": 0.2857,
+      "step": 1370
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.846395348837209e-07,
+      "loss": 0.1227,
+      "step": 1371
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.846279069767442e-07,
+      "loss": 0.164,
+      "step": 1372
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.846162790697674e-07,
+      "loss": 0.1382,
+      "step": 1373
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.846046511627907e-07,
+      "loss": 0.1065,
+      "step": 1374
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.8459302325581395e-07,
+      "loss": 0.2427,
+      "step": 1375
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.845813953488372e-07,
+      "loss": 0.1246,
+      "step": 1376
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.845697674418604e-07,
+      "loss": 0.4684,
+      "step": 1377
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.845581395348837e-07,
+      "loss": 0.1421,
+      "step": 1378
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.84546511627907e-07,
+      "loss": 0.0875,
+      "step": 1379
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.8453488372093023e-07,
+      "loss": 0.0973,
+      "step": 1380
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.8452325581395346e-07,
+      "loss": 0.1938,
+      "step": 1381
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.8451162790697675e-07,
+      "loss": 0.1197,
+      "step": 1382
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.845e-07,
+      "loss": 0.1037,
+      "step": 1383
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.844883720930232e-07,
+      "loss": 0.1299,
+      "step": 1384
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.844767441860465e-07,
+      "loss": 0.1571,
+      "step": 1385
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.8446511627906975e-07,
+      "loss": 0.0974,
+      "step": 1386
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.8445348837209303e-07,
+      "loss": 0.0898,
+      "step": 1387
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.8444186046511627e-07,
+      "loss": 0.1534,
+      "step": 1388
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.844302325581395e-07,
+      "loss": 0.0974,
+      "step": 1389
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.8441860465116274e-07,
+      "loss": 0.0894,
+      "step": 1390
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.8440697674418603e-07,
+      "loss": 0.1294,
+      "step": 1391
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.843953488372093e-07,
+      "loss": 0.2126,
+      "step": 1392
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.8438372093023255e-07,
+      "loss": 0.1466,
+      "step": 1393
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.843720930232558e-07,
+      "loss": 0.0981,
+      "step": 1394
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.843604651162791e-07,
+      "loss": 0.1093,
+      "step": 1395
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.843488372093023e-07,
+      "loss": 0.1969,
+      "step": 1396
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.8433720930232554e-07,
+      "loss": 0.0782,
+      "step": 1397
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.8432558139534883e-07,
+      "loss": 0.7296,
+      "step": 1398
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.8431395348837207e-07,
+      "loss": 0.1866,
+      "step": 1399
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.8430232558139536e-07,
+      "loss": 0.2155,
+      "step": 1400
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.842906976744186e-07,
+      "loss": 0.1495,
+      "step": 1401
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.8427906976744183e-07,
+      "loss": 0.0906,
+      "step": 1402
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.842674418604651e-07,
+      "loss": 0.1407,
+      "step": 1403
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.8425581395348835e-07,
+      "loss": 0.132,
+      "step": 1404
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.8424418604651164e-07,
+      "loss": 1.019,
+      "step": 1405
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.842325581395349e-07,
+      "loss": 0.2254,
+      "step": 1406
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.8422093023255816e-07,
+      "loss": 0.1789,
+      "step": 1407
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.842093023255814e-07,
+      "loss": 0.1009,
+      "step": 1408
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.8419767441860463e-07,
+      "loss": 0.1181,
+      "step": 1409
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.8418604651162787e-07,
+      "loss": 0.1093,
+      "step": 1410
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.8417441860465116e-07,
+      "loss": 0.0754,
+      "step": 1411
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.841627906976744e-07,
+      "loss": 0.1027,
+      "step": 1412
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.841511627906977e-07,
+      "loss": 0.1277,
+      "step": 1413
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.841395348837209e-07,
+      "loss": 0.1146,
+      "step": 1414
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.8412790697674415e-07,
+      "loss": 0.1129,
+      "step": 1415
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.8411627906976744e-07,
+      "loss": 0.0658,
+      "step": 1416
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.8410465116279067e-07,
+      "loss": 0.1389,
+      "step": 1417
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.8409302325581396e-07,
+      "loss": 0.0701,
+      "step": 1418
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.840813953488372e-07,
+      "loss": 0.1084,
+      "step": 1419
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.840697674418605e-07,
+      "loss": 0.0864,
+      "step": 1420
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.840581395348837e-07,
+      "loss": 0.0951,
+      "step": 1421
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.8404651162790696e-07,
+      "loss": 0.1011,
+      "step": 1422
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.840348837209302e-07,
+      "loss": 0.1081,
+      "step": 1423
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.840232558139535e-07,
+      "loss": 0.0992,
+      "step": 1424
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.840116279069767e-07,
+      "loss": 0.1042,
+      "step": 1425
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.84e-07,
+      "loss": 0.2077,
+      "step": 1426
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.8398837209302324e-07,
+      "loss": 0.1179,
+      "step": 1427
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.839767441860465e-07,
+      "loss": 0.103,
+      "step": 1428
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.8396511627906976e-07,
+      "loss": 0.0955,
+      "step": 1429
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.83953488372093e-07,
+      "loss": 0.1644,
+      "step": 1430
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8394186046511623e-07,
+      "loss": 0.1542,
+      "step": 1431
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.839302325581395e-07,
+      "loss": 0.1719,
+      "step": 1432
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.839186046511628e-07,
+      "loss": 0.0982,
+      "step": 1433
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.8390697674418604e-07,
+      "loss": 0.2573,
+      "step": 1434
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.838953488372093e-07,
+      "loss": 0.1889,
+      "step": 1435
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.838837209302325e-07,
+      "loss": 0.0925,
+      "step": 1436
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.838720930232558e-07,
+      "loss": 0.0732,
+      "step": 1437
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.8386046511627904e-07,
+      "loss": 0.1068,
+      "step": 1438
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.838488372093023e-07,
+      "loss": 0.2178,
+      "step": 1439
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.8383720930232556e-07,
+      "loss": 0.0968,
+      "step": 1440
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.8382558139534885e-07,
+      "loss": 0.2528,
+      "step": 1441
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.838139534883721e-07,
+      "loss": 0.1962,
+      "step": 1442
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.838023255813953e-07,
+      "loss": 0.1103,
+      "step": 1443
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.8379069767441855e-07,
+      "loss": 0.1661,
+      "step": 1444
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.8377906976744184e-07,
+      "loss": 0.258,
+      "step": 1445
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.8376744186046513e-07,
+      "loss": 0.3086,
+      "step": 1446
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.8375581395348837e-07,
+      "loss": 0.1329,
+      "step": 1447
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.837441860465116e-07,
+      "loss": 0.0935,
+      "step": 1448
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.837325581395349e-07,
+      "loss": 0.0916,
+      "step": 1449
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.837209302325581e-07,
+      "loss": 0.1479,
+      "step": 1450
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.8370930232558136e-07,
+      "loss": 0.2435,
+      "step": 1451
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.8369767441860465e-07,
+      "loss": 0.096,
+      "step": 1452
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.836860465116279e-07,
+      "loss": 0.0705,
+      "step": 1453
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.8367441860465117e-07,
+      "loss": 0.0982,
+      "step": 1454
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.836627906976744e-07,
+      "loss": 0.1265,
+      "step": 1455
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.8365116279069764e-07,
+      "loss": 0.1317,
+      "step": 1456
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.8363953488372093e-07,
+      "loss": 0.0865,
+      "step": 1457
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.8362790697674416e-07,
+      "loss": 0.0721,
+      "step": 1458
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.8361627906976745e-07,
+      "loss": 0.1926,
+      "step": 1459
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.836046511627907e-07,
+      "loss": 0.1617,
+      "step": 1460
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.83593023255814e-07,
+      "loss": 0.0972,
+      "step": 1461
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.835813953488372e-07,
+      "loss": 0.0963,
+      "step": 1462
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.8356976744186045e-07,
+      "loss": 0.2559,
+      "step": 1463
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.835581395348837e-07,
+      "loss": 0.0712,
+      "step": 1464
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.8354651162790697e-07,
+      "loss": 0.0903,
+      "step": 1465
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.835348837209302e-07,
+      "loss": 0.0701,
+      "step": 1466
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.835232558139535e-07,
+      "loss": 0.1172,
+      "step": 1467
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.8351162790697673e-07,
+      "loss": 0.0872,
+      "step": 1468
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.8349999999999996e-07,
+      "loss": 0.0773,
+      "step": 1469
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.8348837209302325e-07,
+      "loss": 0.0595,
+      "step": 1470
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.834767441860465e-07,
+      "loss": 0.7168,
+      "step": 1471
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.834651162790697e-07,
+      "loss": 0.2074,
+      "step": 1472
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.83453488372093e-07,
+      "loss": 0.0691,
+      "step": 1473
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.834418604651163e-07,
+      "loss": 0.1314,
+      "step": 1474
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.8343023255813953e-07,
+      "loss": 0.2209,
+      "step": 1475
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.8341860465116277e-07,
+      "loss": 0.0887,
+      "step": 1476
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.83406976744186e-07,
+      "loss": 0.1071,
+      "step": 1477
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.833953488372093e-07,
+      "loss": 0.0659,
+      "step": 1478
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.8338372093023253e-07,
+      "loss": 0.0865,
+      "step": 1479
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.833720930232558e-07,
+      "loss": 0.073,
+      "step": 1480
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.8336046511627905e-07,
+      "loss": 0.0979,
+      "step": 1481
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.8334883720930234e-07,
+      "loss": 0.095,
+      "step": 1482
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.833372093023256e-07,
+      "loss": 0.5357,
+      "step": 1483
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.833255813953488e-07,
+      "loss": 0.0848,
+      "step": 1484
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.8331395348837205e-07,
+      "loss": 0.0848,
+      "step": 1485
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.8330232558139533e-07,
+      "loss": 0.0823,
+      "step": 1486
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.832906976744186e-07,
+      "loss": 0.1177,
+      "step": 1487
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.8327906976744186e-07,
+      "loss": 0.3026,
+      "step": 1488
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.832674418604651e-07,
+      "loss": 0.2294,
+      "step": 1489
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.8325581395348833e-07,
+      "loss": 0.059,
+      "step": 1490
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.832441860465116e-07,
+      "loss": 0.0748,
+      "step": 1491
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.8323255813953485e-07,
+      "loss": 0.1101,
+      "step": 1492
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.8322093023255814e-07,
+      "loss": 0.0756,
+      "step": 1493
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.832093023255814e-07,
+      "loss": 0.0992,
+      "step": 1494
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.8319767441860466e-07,
+      "loss": 0.1462,
+      "step": 1495
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.831860465116279e-07,
+      "loss": 0.0944,
+      "step": 1496
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.8317441860465113e-07,
+      "loss": 0.0745,
+      "step": 1497
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.8316279069767437e-07,
+      "loss": 0.0666,
+      "step": 1498
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.8315116279069766e-07,
+      "loss": 0.0877,
+      "step": 1499
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.8313953488372095e-07,
+      "loss": 0.353,
+      "step": 1500
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.831279069767442e-07,
+      "loss": 0.1901,
+      "step": 1501
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.831162790697674e-07,
+      "loss": 0.0519,
+      "step": 1502
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.831046511627907e-07,
+      "loss": 0.0762,
+      "step": 1503
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.8309302325581394e-07,
+      "loss": 0.0838,
+      "step": 1504
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.830813953488372e-07,
+      "loss": 0.0896,
+      "step": 1505
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.8306976744186046e-07,
+      "loss": 0.136,
+      "step": 1506
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.830581395348837e-07,
+      "loss": 0.079,
+      "step": 1507
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.83046511627907e-07,
+      "loss": 0.0801,
+      "step": 1508
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.830348837209302e-07,
+      "loss": 0.0768,
+      "step": 1509
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.8302325581395346e-07,
+      "loss": 0.0647,
+      "step": 1510
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.8301162790697674e-07,
+      "loss": 0.1073,
+      "step": 1511
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.83e-07,
+      "loss": 0.166,
+      "step": 1512
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8298837209302327e-07,
+      "loss": 0.0904,
+      "step": 1513
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.829767441860465e-07,
+      "loss": 0.1107,
+      "step": 1514
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.829651162790698e-07,
+      "loss": 0.1759,
+      "step": 1515
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8295348837209303e-07,
+      "loss": 1.0694,
+      "step": 1516
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.8294186046511626e-07,
+      "loss": 0.1357,
+      "step": 1517
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.829302325581395e-07,
+      "loss": 0.0591,
+      "step": 1518
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.829186046511628e-07,
+      "loss": 1.1984,
+      "step": 1519
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.82906976744186e-07,
+      "loss": 0.0691,
+      "step": 1520
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.828953488372093e-07,
+      "loss": 0.1042,
+      "step": 1521
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.8288372093023254e-07,
+      "loss": 0.6405,
+      "step": 1522
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.828720930232558e-07,
+      "loss": 0.0597,
+      "step": 1523
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.8286046511627907e-07,
+      "loss": 0.0687,
+      "step": 1524
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.828488372093023e-07,
+      "loss": 0.0721,
+      "step": 1525
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.8283720930232554e-07,
+      "loss": 0.057,
+      "step": 1526
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.828255813953488e-07,
+      "loss": 0.051,
+      "step": 1527
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.828139534883721e-07,
+      "loss": 0.0739,
+      "step": 1528
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.8280232558139535e-07,
+      "loss": 0.2102,
+      "step": 1529
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.827906976744186e-07,
+      "loss": 0.0607,
+      "step": 1530
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.827790697674418e-07,
+      "loss": 0.0817,
+      "step": 1531
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.827674418604651e-07,
+      "loss": 0.056,
+      "step": 1532
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.8275581395348834e-07,
+      "loss": 0.0615,
+      "step": 1533
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.8274418604651163e-07,
+      "loss": 0.3677,
+      "step": 1534
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.8273255813953487e-07,
+      "loss": 0.1599,
+      "step": 1535
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.8272093023255815e-07,
+      "loss": 0.3276,
+      "step": 1536
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.827093023255814e-07,
+      "loss": 0.0805,
+      "step": 1537
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.826976744186046e-07,
+      "loss": 0.1398,
+      "step": 1538
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.8268604651162786e-07,
+      "loss": 0.2097,
+      "step": 1539
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.8267441860465115e-07,
+      "loss": 0.0611,
+      "step": 1540
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.8266279069767444e-07,
+      "loss": 0.1376,
+      "step": 1541
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.8265116279069767e-07,
+      "loss": 0.099,
+      "step": 1542
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.826395348837209e-07,
+      "loss": 0.2874,
+      "step": 1543
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8262790697674414e-07,
+      "loss": 0.1007,
+      "step": 1544
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8261627906976743e-07,
+      "loss": 0.0832,
+      "step": 1545
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8260465116279067e-07,
+      "loss": 0.0847,
+      "step": 1546
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.8259302325581395e-07,
+      "loss": 0.0629,
+      "step": 1547
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.825813953488372e-07,
+      "loss": 0.1113,
+      "step": 1548
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.825697674418605e-07,
+      "loss": 0.2742,
+      "step": 1549
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.825581395348837e-07,
+      "loss": 0.0442,
+      "step": 1550
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.8254651162790695e-07,
+      "loss": 0.1951,
+      "step": 1551
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9661508704061895,
+      "eval_loss": 0.16998709738254547,
+      "eval_runtime": 1.5409,
+      "eval_samples_per_second": 671.02,
+      "eval_steps_per_second": 11.032,
+      "step": 1551
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.825348837209302e-07,
+      "loss": 0.1114,
+      "step": 1552
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.8252325581395347e-07,
+      "loss": 0.1229,
+      "step": 1553
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.8251162790697676e-07,
+      "loss": 0.0771,
+      "step": 1554
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.825e-07,
+      "loss": 0.1257,
+      "step": 1555
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.8248837209302323e-07,
+      "loss": 0.2118,
+      "step": 1556
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.824767441860465e-07,
+      "loss": 0.1411,
+      "step": 1557
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.8246511627906975e-07,
+      "loss": 0.3719,
+      "step": 1558
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.82453488372093e-07,
+      "loss": 0.088,
+      "step": 1559
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.824418604651163e-07,
+      "loss": 0.1201,
+      "step": 1560
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.824302325581395e-07,
+      "loss": 0.0726,
+      "step": 1561
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.824186046511628e-07,
+      "loss": 0.0779,
+      "step": 1562
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.8240697674418604e-07,
+      "loss": 0.0562,
+      "step": 1563
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.8239534883720927e-07,
+      "loss": 0.1171,
+      "step": 1564
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.8238372093023256e-07,
+      "loss": 0.0727,
+      "step": 1565
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.823720930232558e-07,
+      "loss": 0.2369,
+      "step": 1566
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.823604651162791e-07,
+      "loss": 0.0855,
+      "step": 1567
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.823488372093023e-07,
+      "loss": 0.064,
+      "step": 1568
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.823372093023256e-07,
+      "loss": 0.0555,
+      "step": 1569
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.8232558139534884e-07,
+      "loss": 0.0994,
+      "step": 1570
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.823139534883721e-07,
+      "loss": 0.048,
+      "step": 1571
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.823023255813953e-07,
+      "loss": 0.4061,
+      "step": 1572
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.822906976744186e-07,
+      "loss": 0.606,
+      "step": 1573
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.8227906976744184e-07,
+      "loss": 0.3202,
+      "step": 1574
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.822674418604651e-07,
+      "loss": 0.0672,
+      "step": 1575
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.8225581395348836e-07,
+      "loss": 0.2215,
+      "step": 1576
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.822441860465116e-07,
+      "loss": 0.1213,
+      "step": 1577
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.822325581395349e-07,
+      "loss": 0.1666,
+      "step": 1578
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.822209302325581e-07,
+      "loss": 0.2207,
+      "step": 1579
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.8220930232558135e-07,
+      "loss": 0.1102,
+      "step": 1580
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.8219767441860464e-07,
+      "loss": 0.0727,
+      "step": 1581
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.8218604651162793e-07,
+      "loss": 0.0484,
+      "step": 1582
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.8217441860465116e-07,
+      "loss": 0.0664,
+      "step": 1583
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.821627906976744e-07,
+      "loss": 0.0599,
+      "step": 1584
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.8215116279069763e-07,
+      "loss": 0.0659,
+      "step": 1585
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.821395348837209e-07,
+      "loss": 0.0521,
+      "step": 1586
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.8212790697674416e-07,
+      "loss": 0.1342,
+      "step": 1587
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.8211627906976745e-07,
+      "loss": 0.1899,
+      "step": 1588
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.821046511627907e-07,
+      "loss": 0.0753,
+      "step": 1589
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.8209302325581397e-07,
+      "loss": 0.0568,
+      "step": 1590
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.820813953488372e-07,
+      "loss": 0.0649,
+      "step": 1591
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.8206976744186044e-07,
+      "loss": 0.0525,
+      "step": 1592
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.820581395348837e-07,
+      "loss": 0.0527,
+      "step": 1593
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.8204651162790696e-07,
+      "loss": 0.1082,
+      "step": 1594
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.8203488372093025e-07,
+      "loss": 0.0568,
+      "step": 1595
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.820232558139535e-07,
+      "loss": 0.082,
+      "step": 1596
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.820116279069767e-07,
+      "loss": 0.1074,
+      "step": 1597
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.8199999999999996e-07,
+      "loss": 0.0771,
+      "step": 1598
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.8198837209302325e-07,
+      "loss": 0.0531,
+      "step": 1599
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.819767441860465e-07,
+      "loss": 0.0554,
+      "step": 1600
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.8196511627906977e-07,
+      "loss": 0.2763,
+      "step": 1601
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.81953488372093e-07,
+      "loss": 0.0662,
+      "step": 1602
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.819418604651163e-07,
+      "loss": 0.1698,
+      "step": 1603
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.8193023255813953e-07,
+      "loss": 0.123,
+      "step": 1604
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.8191860465116276e-07,
+      "loss": 0.0626,
+      "step": 1605
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.81906976744186e-07,
+      "loss": 0.1109,
+      "step": 1606
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.818953488372093e-07,
+      "loss": 0.0573,
+      "step": 1607
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.818837209302326e-07,
+      "loss": 0.0448,
+      "step": 1608
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.818720930232558e-07,
+      "loss": 0.0815,
+      "step": 1609
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.8186046511627905e-07,
+      "loss": 0.0969,
+      "step": 1610
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.8184883720930233e-07,
+      "loss": 0.0723,
+      "step": 1611
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.8183720930232557e-07,
+      "loss": 0.0602,
+      "step": 1612
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.818255813953488e-07,
+      "loss": 0.0703,
+      "step": 1613
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.818139534883721e-07,
+      "loss": 0.1198,
+      "step": 1614
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.8180232558139533e-07,
+      "loss": 0.052,
+      "step": 1615
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.817906976744186e-07,
+      "loss": 0.0852,
+      "step": 1616
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.8177906976744185e-07,
+      "loss": 0.047,
+      "step": 1617
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.817674418604651e-07,
+      "loss": 0.0576,
+      "step": 1618
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.817558139534884e-07,
+      "loss": 0.0692,
+      "step": 1619
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.817441860465116e-07,
+      "loss": 0.0396,
+      "step": 1620
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8173255813953484e-07,
+      "loss": 0.1671,
+      "step": 1621
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8172093023255813e-07,
+      "loss": 0.2834,
+      "step": 1622
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8170930232558137e-07,
+      "loss": 0.071,
+      "step": 1623
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8169767441860466e-07,
+      "loss": 0.0403,
+      "step": 1624
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.816860465116279e-07,
+      "loss": 0.0874,
+      "step": 1625
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.8167441860465113e-07,
+      "loss": 0.0499,
+      "step": 1626
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.816627906976744e-07,
+      "loss": 0.0414,
+      "step": 1627
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.8165116279069765e-07,
+      "loss": 0.0624,
+      "step": 1628
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.8163953488372094e-07,
+      "loss": 0.053,
+      "step": 1629
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.816279069767442e-07,
+      "loss": 0.0481,
+      "step": 1630
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.816162790697674e-07,
+      "loss": 0.0551,
+      "step": 1631
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.816046511627907e-07,
+      "loss": 0.0488,
+      "step": 1632
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.8159302325581393e-07,
+      "loss": 0.0708,
+      "step": 1633
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.8158139534883717e-07,
+      "loss": 0.0567,
+      "step": 1634
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.8156976744186046e-07,
+      "loss": 0.0584,
+      "step": 1635
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.8155813953488374e-07,
+      "loss": 0.0795,
+      "step": 1636
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.81546511627907e-07,
+      "loss": 0.0602,
+      "step": 1637
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.815348837209302e-07,
+      "loss": 0.0669,
+      "step": 1638
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.8152325581395345e-07,
+      "loss": 0.1177,
+      "step": 1639
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.8151162790697674e-07,
+      "loss": 0.0433,
+      "step": 1640
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.8149999999999997e-07,
+      "loss": 0.0737,
+      "step": 1641
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.8148837209302326e-07,
+      "loss": 0.0904,
+      "step": 1642
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.814767441860465e-07,
+      "loss": 0.1449,
+      "step": 1643
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.814651162790698e-07,
+      "loss": 0.0585,
+      "step": 1644
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.81453488372093e-07,
+      "loss": 0.1839,
+      "step": 1645
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.8144186046511626e-07,
+      "loss": 0.0743,
+      "step": 1646
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.814302325581395e-07,
+      "loss": 0.1147,
+      "step": 1647
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.814186046511628e-07,
+      "loss": 0.1372,
+      "step": 1648
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.8140697674418607e-07,
+      "loss": 0.0515,
+      "step": 1649
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.813953488372093e-07,
+      "loss": 0.042,
+      "step": 1650
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.8138372093023254e-07,
+      "loss": 0.0452,
+      "step": 1651
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.8137209302325577e-07,
+      "loss": 0.0529,
+      "step": 1652
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.8136046511627906e-07,
+      "loss": 0.0734,
+      "step": 1653
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.813488372093023e-07,
+      "loss": 0.0734,
+      "step": 1654
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.813372093023256e-07,
+      "loss": 0.1391,
+      "step": 1655
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.813255813953488e-07,
+      "loss": 0.0874,
+      "step": 1656
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.813139534883721e-07,
+      "loss": 0.0895,
+      "step": 1657
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.8130232558139534e-07,
+      "loss": 0.0529,
+      "step": 1658
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.812906976744186e-07,
+      "loss": 0.0438,
+      "step": 1659
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.812790697674418e-07,
+      "loss": 0.0693,
+      "step": 1660
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.812674418604651e-07,
+      "loss": 0.0532,
+      "step": 1661
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.812558139534884e-07,
+      "loss": 0.0659,
+      "step": 1662
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.812441860465116e-07,
+      "loss": 0.3905,
+      "step": 1663
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.8123255813953486e-07,
+      "loss": 0.1372,
+      "step": 1664
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.8122093023255815e-07,
+      "loss": 0.1632,
+      "step": 1665
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.812093023255814e-07,
+      "loss": 0.1246,
+      "step": 1666
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.811976744186046e-07,
+      "loss": 0.0577,
+      "step": 1667
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.811860465116279e-07,
+      "loss": 0.0668,
+      "step": 1668
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.8117441860465114e-07,
+      "loss": 0.0792,
+      "step": 1669
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.8116279069767443e-07,
+      "loss": 0.2678,
+      "step": 1670
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.8115116279069767e-07,
+      "loss": 0.0498,
+      "step": 1671
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.811395348837209e-07,
+      "loss": 0.0546,
+      "step": 1672
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.811279069767442e-07,
+      "loss": 0.0447,
+      "step": 1673
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.811162790697674e-07,
+      "loss": 0.0607,
+      "step": 1674
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.8110465116279066e-07,
+      "loss": 0.0837,
+      "step": 1675
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.8109302325581395e-07,
+      "loss": 0.0842,
+      "step": 1676
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.810813953488372e-07,
+      "loss": 0.1814,
+      "step": 1677
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.8106976744186047e-07,
+      "loss": 0.0727,
+      "step": 1678
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.810581395348837e-07,
+      "loss": 0.1859,
+      "step": 1679
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.8104651162790694e-07,
+      "loss": 0.0683,
+      "step": 1680
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.8103488372093023e-07,
+      "loss": 0.0819,
+      "step": 1681
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.8102325581395346e-07,
+      "loss": 0.047,
+      "step": 1682
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.8101162790697675e-07,
+      "loss": 0.0791,
+      "step": 1683
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.81e-07,
+      "loss": 0.7716,
+      "step": 1684
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.809883720930232e-07,
+      "loss": 0.1348,
+      "step": 1685
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.809767441860465e-07,
+      "loss": 0.1164,
+      "step": 1686
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.8096511627906975e-07,
+      "loss": 0.0479,
+      "step": 1687
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.80953488372093e-07,
+      "loss": 0.0856,
+      "step": 1688
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.8094186046511627e-07,
+      "loss": 0.0912,
+      "step": 1689
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.8093023255813956e-07,
+      "loss": 0.0926,
+      "step": 1690
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.809186046511628e-07,
+      "loss": 0.0474,
+      "step": 1691
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.8090697674418603e-07,
+      "loss": 0.8106,
+      "step": 1692
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.8089534883720926e-07,
+      "loss": 0.1257,
+      "step": 1693
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.8088372093023255e-07,
+      "loss": 0.8841,
+      "step": 1694
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.808720930232558e-07,
+      "loss": 0.0502,
+      "step": 1695
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.808604651162791e-07,
+      "loss": 0.0325,
+      "step": 1696
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.808488372093023e-07,
+      "loss": 0.0332,
+      "step": 1697
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.808372093023256e-07,
+      "loss": 0.7336,
+      "step": 1698
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.8082558139534883e-07,
+      "loss": 0.0923,
+      "step": 1699
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.8081395348837207e-07,
+      "loss": 0.0376,
+      "step": 1700
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.808023255813953e-07,
+      "loss": 0.7566,
+      "step": 1701
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.807906976744186e-07,
+      "loss": 0.052,
+      "step": 1702
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.807790697674419e-07,
+      "loss": 1.0337,
+      "step": 1703
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.807674418604651e-07,
+      "loss": 0.1352,
+      "step": 1704
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.8075581395348835e-07,
+      "loss": 0.0532,
+      "step": 1705
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.807441860465116e-07,
+      "loss": 0.0384,
+      "step": 1706
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.807325581395349e-07,
+      "loss": 0.0395,
+      "step": 1707
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.807209302325581e-07,
+      "loss": 0.0422,
+      "step": 1708
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.807093023255814e-07,
+      "loss": 0.2758,
+      "step": 1709
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.8069767441860463e-07,
+      "loss": 0.0449,
+      "step": 1710
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.806860465116279e-07,
+      "loss": 0.0314,
+      "step": 1711
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.8067441860465116e-07,
+      "loss": 0.067,
+      "step": 1712
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.806627906976744e-07,
+      "loss": 0.0465,
+      "step": 1713
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.8065116279069763e-07,
+      "loss": 0.0659,
+      "step": 1714
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.806395348837209e-07,
+      "loss": 0.0594,
+      "step": 1715
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.8062790697674415e-07,
+      "loss": 0.0398,
+      "step": 1716
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.8061627906976744e-07,
+      "loss": 0.0362,
+      "step": 1717
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.806046511627907e-07,
+      "loss": 0.0322,
+      "step": 1718
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.8059302325581396e-07,
+      "loss": 0.0412,
+      "step": 1719
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.805813953488372e-07,
+      "loss": 0.0551,
+      "step": 1720
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.8056976744186043e-07,
+      "loss": 0.0547,
+      "step": 1721
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.805581395348837e-07,
+      "loss": 0.1123,
+      "step": 1722
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.8054651162790696e-07,
+      "loss": 0.0952,
+      "step": 1723
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.8053488372093025e-07,
+      "loss": 0.2207,
+      "step": 1724
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.805232558139535e-07,
+      "loss": 0.3771,
+      "step": 1725
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.805116279069767e-07,
+      "loss": 0.0334,
+      "step": 1726
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.805e-07,
+      "loss": 0.0674,
+      "step": 1727
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.8048837209302324e-07,
+      "loss": 0.0856,
+      "step": 1728
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.804767441860465e-07,
+      "loss": 0.0397,
+      "step": 1729
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.8046511627906976e-07,
+      "loss": 0.088,
+      "step": 1730
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.80453488372093e-07,
+      "loss": 0.1009,
+      "step": 1731
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.804418604651163e-07,
+      "loss": 0.0321,
+      "step": 1732
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.804302325581395e-07,
+      "loss": 0.1755,
+      "step": 1733
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.8041860465116276e-07,
+      "loss": 0.7748,
+      "step": 1734
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.8040697674418604e-07,
+      "loss": 0.1224,
+      "step": 1735
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.803953488372093e-07,
+      "loss": 0.1147,
+      "step": 1736
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.8038372093023257e-07,
+      "loss": 0.0397,
+      "step": 1737
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.803720930232558e-07,
+      "loss": 0.0555,
+      "step": 1738
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.8036046511627904e-07,
+      "loss": 0.2595,
+      "step": 1739
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.8034883720930233e-07,
+      "loss": 0.3811,
+      "step": 1740
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.8033720930232556e-07,
+      "loss": 0.0585,
+      "step": 1741
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.803255813953488e-07,
+      "loss": 0.0304,
+      "step": 1742
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.803139534883721e-07,
+      "loss": 0.0524,
+      "step": 1743
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.8030232558139537e-07,
+      "loss": 0.0382,
+      "step": 1744
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.802906976744186e-07,
+      "loss": 0.1777,
+      "step": 1745
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.8027906976744184e-07,
+      "loss": 0.0501,
+      "step": 1746
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.802674418604651e-07,
+      "loss": 0.0541,
+      "step": 1747
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.8025581395348837e-07,
+      "loss": 0.0517,
+      "step": 1748
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.802441860465116e-07,
+      "loss": 0.054,
+      "step": 1749
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.802325581395349e-07,
+      "loss": 0.0823,
+      "step": 1750
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.802209302325581e-07,
+      "loss": 0.6728,
+      "step": 1751
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.802093023255814e-07,
+      "loss": 0.055,
+      "step": 1752
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.8019767441860465e-07,
+      "loss": 0.0814,
+      "step": 1753
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.801860465116279e-07,
+      "loss": 0.0679,
+      "step": 1754
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.801744186046511e-07,
+      "loss": 0.0289,
+      "step": 1755
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.801627906976744e-07,
+      "loss": 0.0434,
+      "step": 1756
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.801511627906977e-07,
+      "loss": 0.0712,
+      "step": 1757
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.8013953488372093e-07,
+      "loss": 0.0348,
+      "step": 1758
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.8012790697674417e-07,
+      "loss": 0.0655,
+      "step": 1759
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.801162790697674e-07,
+      "loss": 0.188,
+      "step": 1760
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.801046511627907e-07,
+      "loss": 0.0428,
+      "step": 1761
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.800930232558139e-07,
+      "loss": 0.0518,
+      "step": 1762
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.800813953488372e-07,
+      "loss": 0.1169,
+      "step": 1763
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.8006976744186045e-07,
+      "loss": 0.0318,
+      "step": 1764
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.8005813953488374e-07,
+      "loss": 0.0447,
+      "step": 1765
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.8004651162790697e-07,
+      "loss": 0.0332,
+      "step": 1766
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.800348837209302e-07,
+      "loss": 0.0593,
+      "step": 1767
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.8002325581395344e-07,
+      "loss": 0.0361,
+      "step": 1768
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.8001162790697673e-07,
+      "loss": 0.0833,
+      "step": 1769
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.7999999999999997e-07,
+      "loss": 1.7339,
+      "step": 1770
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.7998837209302325e-07,
+      "loss": 0.0392,
+      "step": 1771
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.799767441860465e-07,
+      "loss": 0.0543,
+      "step": 1772
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.799651162790698e-07,
+      "loss": 0.0357,
+      "step": 1773
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.79953488372093e-07,
+      "loss": 0.1722,
+      "step": 1774
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.7994186046511625e-07,
+      "loss": 0.0882,
+      "step": 1775
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.7993023255813954e-07,
+      "loss": 0.0517,
+      "step": 1776
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.7991860465116277e-07,
+      "loss": 0.0506,
+      "step": 1777
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.7990697674418606e-07,
+      "loss": 1.0148,
+      "step": 1778
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.798953488372093e-07,
+      "loss": 0.0717,
+      "step": 1779
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.7988372093023253e-07,
+      "loss": 0.0539,
+      "step": 1780
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.798720930232558e-07,
+      "loss": 0.031,
+      "step": 1781
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.7986046511627905e-07,
+      "loss": 0.0502,
+      "step": 1782
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.798488372093023e-07,
+      "loss": 0.0371,
+      "step": 1783
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.798372093023256e-07,
+      "loss": 0.0454,
+      "step": 1784
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.798255813953488e-07,
+      "loss": 0.5584,
+      "step": 1785
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.798139534883721e-07,
+      "loss": 0.0565,
+      "step": 1786
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.7980232558139534e-07,
+      "loss": 0.0753,
+      "step": 1787
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.7979069767441857e-07,
+      "loss": 0.0719,
+      "step": 1788
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.7977906976744186e-07,
+      "loss": 0.0559,
+      "step": 1789
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.797674418604651e-07,
+      "loss": 0.0283,
+      "step": 1790
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.797558139534884e-07,
+      "loss": 0.0446,
+      "step": 1791
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.797441860465116e-07,
+      "loss": 0.0368,
+      "step": 1792
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.7973255813953485e-07,
+      "loss": 0.3666,
+      "step": 1793
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.7972093023255814e-07,
+      "loss": 0.3796,
+      "step": 1794
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.797093023255814e-07,
+      "loss": 0.0358,
+      "step": 1795
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.796976744186046e-07,
+      "loss": 0.0697,
+      "step": 1796
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.796860465116279e-07,
+      "loss": 0.0519,
+      "step": 1797
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.796744186046512e-07,
+      "loss": 0.2312,
+      "step": 1798
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.796627906976744e-07,
+      "loss": 0.0421,
+      "step": 1799
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.7965116279069766e-07,
+      "loss": 0.0448,
+      "step": 1800
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.796395348837209e-07,
+      "loss": 0.0578,
+      "step": 1801
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.796279069767442e-07,
+      "loss": 0.0885,
+      "step": 1802
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.796162790697674e-07,
+      "loss": 0.0397,
+      "step": 1803
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.796046511627907e-07,
+      "loss": 0.0744,
+      "step": 1804
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.7959302325581394e-07,
+      "loss": 0.0944,
+      "step": 1805
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.7958139534883723e-07,
+      "loss": 0.169,
+      "step": 1806
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.7956976744186046e-07,
+      "loss": 0.0847,
+      "step": 1807
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.795581395348837e-07,
+      "loss": 0.9377,
+      "step": 1808
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.7954651162790693e-07,
+      "loss": 0.0411,
+      "step": 1809
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.795348837209302e-07,
+      "loss": 0.1275,
+      "step": 1810
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.795232558139535e-07,
+      "loss": 0.0288,
+      "step": 1811
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.7951162790697675e-07,
+      "loss": 0.036,
+      "step": 1812
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.795e-07,
+      "loss": 0.0468,
+      "step": 1813
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.794883720930232e-07,
+      "loss": 0.0776,
+      "step": 1814
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.794767441860465e-07,
+      "loss": 0.0687,
+      "step": 1815
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.7946511627906974e-07,
+      "loss": 0.3023,
+      "step": 1816
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.7945348837209303e-07,
+      "loss": 0.0838,
+      "step": 1817
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.7944186046511626e-07,
+      "loss": 0.0812,
+      "step": 1818
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.7943023255813955e-07,
+      "loss": 0.0387,
+      "step": 1819
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.794186046511628e-07,
+      "loss": 0.0504,
+      "step": 1820
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.79406976744186e-07,
+      "loss": 0.1168,
+      "step": 1821
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.7939534883720926e-07,
+      "loss": 0.0311,
+      "step": 1822
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.7938372093023255e-07,
+      "loss": 0.0873,
+      "step": 1823
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.793720930232558e-07,
+      "loss": 0.0591,
+      "step": 1824
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.7936046511627907e-07,
+      "loss": 0.0293,
+      "step": 1825
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.793488372093023e-07,
+      "loss": 0.0262,
+      "step": 1826
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.793372093023256e-07,
+      "loss": 0.0324,
+      "step": 1827
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.7932558139534883e-07,
+      "loss": 0.0351,
+      "step": 1828
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.7931395348837206e-07,
+      "loss": 0.1129,
+      "step": 1829
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.7930232558139535e-07,
+      "loss": 0.393,
+      "step": 1830
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.792906976744186e-07,
+      "loss": 0.0409,
+      "step": 1831
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.792790697674419e-07,
+      "loss": 0.0473,
+      "step": 1832
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.792674418604651e-07,
+      "loss": 0.2692,
+      "step": 1833
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.7925581395348835e-07,
+      "loss": 0.0608,
+      "step": 1834
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.7924418604651163e-07,
+      "loss": 0.0324,
+      "step": 1835
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.7923255813953487e-07,
+      "loss": 0.0539,
+      "step": 1836
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.792209302325581e-07,
+      "loss": 0.0399,
+      "step": 1837
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.792093023255814e-07,
+      "loss": 0.3401,
+      "step": 1838
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.7919767441860463e-07,
+      "loss": 0.0709,
+      "step": 1839
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.791860465116279e-07,
+      "loss": 0.0323,
+      "step": 1840
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.7917441860465115e-07,
+      "loss": 0.1862,
+      "step": 1841
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.791627906976744e-07,
+      "loss": 0.0535,
+      "step": 1842
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.791511627906976e-07,
+      "loss": 0.0418,
+      "step": 1843
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.791395348837209e-07,
+      "loss": 0.0321,
+      "step": 1844
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.791279069767442e-07,
+      "loss": 0.0652,
+      "step": 1845
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.7911627906976743e-07,
+      "loss": 0.0407,
+      "step": 1846
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.7910465116279067e-07,
+      "loss": 0.1428,
+      "step": 1847
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.7909302325581396e-07,
+      "loss": 0.0352,
+      "step": 1848
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.790813953488372e-07,
+      "loss": 0.0424,
+      "step": 1849
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.7906976744186043e-07,
+      "loss": 0.0413,
+      "step": 1850
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.790581395348837e-07,
+      "loss": 0.0459,
+      "step": 1851
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.79046511627907e-07,
+      "loss": 0.1531,
+      "step": 1852
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.7903488372093024e-07,
+      "loss": 0.1072,
+      "step": 1853
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.7902325581395347e-07,
+      "loss": 0.0544,
+      "step": 1854
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.790116279069767e-07,
+      "loss": 0.1074,
+      "step": 1855
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.79e-07,
+      "loss": 0.0258,
+      "step": 1856
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.7898837209302323e-07,
+      "loss": 0.0478,
+      "step": 1857
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.789767441860465e-07,
+      "loss": 0.0398,
+      "step": 1858
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.7896511627906976e-07,
+      "loss": 0.0345,
+      "step": 1859
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.7895348837209304e-07,
+      "loss": 0.0268,
+      "step": 1860
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.789418604651163e-07,
+      "loss": 0.0274,
+      "step": 1861
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.789302325581395e-07,
+      "loss": 0.1242,
+      "step": 1862
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.7891860465116275e-07,
+      "loss": 0.1412,
+      "step": 1863
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.7890697674418604e-07,
+      "loss": 0.0736,
+      "step": 1864
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.7889534883720927e-07,
+      "loss": 0.0469,
+      "step": 1865
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.7888372093023256e-07,
+      "loss": 0.1012,
+      "step": 1866
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.788720930232558e-07,
+      "loss": 0.0273,
+      "step": 1867
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.7886046511627903e-07,
+      "loss": 0.0402,
+      "step": 1868
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.788488372093023e-07,
+      "loss": 0.0407,
+      "step": 1869
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.7883720930232555e-07,
+      "loss": 0.045,
+      "step": 1870
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.7882558139534884e-07,
+      "loss": 0.6259,
+      "step": 1871
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.788139534883721e-07,
+      "loss": 0.2232,
+      "step": 1872
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.7880232558139537e-07,
+      "loss": 0.0425,
+      "step": 1873
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.787906976744186e-07,
+      "loss": 0.0478,
+      "step": 1874
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.7877906976744184e-07,
+      "loss": 0.0408,
+      "step": 1875
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.7876744186046507e-07,
+      "loss": 0.0481,
+      "step": 1876
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.7875581395348836e-07,
+      "loss": 0.0266,
+      "step": 1877
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.787441860465116e-07,
+      "loss": 0.0274,
+      "step": 1878
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.787325581395349e-07,
+      "loss": 0.0689,
+      "step": 1879
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.787209302325581e-07,
+      "loss": 0.412,
+      "step": 1880
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.787093023255814e-07,
+      "loss": 0.0269,
+      "step": 1881
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.7869767441860464e-07,
+      "loss": 0.203,
+      "step": 1882
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.786860465116279e-07,
+      "loss": 0.0295,
+      "step": 1883
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.7867441860465117e-07,
+      "loss": 0.4241,
+      "step": 1884
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.786627906976744e-07,
+      "loss": 0.1521,
+      "step": 1885
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.786511627906977e-07,
+      "loss": 0.0226,
+      "step": 1886
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.786395348837209e-07,
+      "loss": 0.0413,
+      "step": 1887
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.7862790697674416e-07,
+      "loss": 0.1371,
+      "step": 1888
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.7861627906976745e-07,
+      "loss": 0.0445,
+      "step": 1889
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.786046511627907e-07,
+      "loss": 0.0288,
+      "step": 1890
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.785930232558139e-07,
+      "loss": 0.0293,
+      "step": 1891
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.785813953488372e-07,
+      "loss": 0.5699,
+      "step": 1892
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.7856976744186044e-07,
+      "loss": 0.3151,
+      "step": 1893
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.7855813953488373e-07,
+      "loss": 0.0519,
+      "step": 1894
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.7854651162790697e-07,
+      "loss": 0.073,
+      "step": 1895
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.785348837209302e-07,
+      "loss": 0.0323,
+      "step": 1896
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.7852325581395344e-07,
+      "loss": 0.0558,
+      "step": 1897
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.785116279069767e-07,
+      "loss": 0.0913,
+      "step": 1898
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.785e-07,
+      "loss": 0.0679,
+      "step": 1899
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.7848837209302325e-07,
+      "loss": 0.0215,
+      "step": 1900
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.784767441860465e-07,
+      "loss": 0.0346,
+      "step": 1901
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.7846511627906977e-07,
+      "loss": 0.0454,
+      "step": 1902
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.78453488372093e-07,
+      "loss": 0.0541,
+      "step": 1903
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.7844186046511624e-07,
+      "loss": 0.0365,
+      "step": 1904
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.7843023255813953e-07,
+      "loss": 0.0224,
+      "step": 1905
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 2.784186046511628e-07,
+      "loss": 0.039,
+      "step": 1906
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 2.7840697674418605e-07,
+      "loss": 0.059,
+      "step": 1907
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 2.783953488372093e-07,
+      "loss": 0.0653,
+      "step": 1908
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 2.783837209302325e-07,
+      "loss": 0.0349,
+      "step": 1909
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 2.783720930232558e-07,
+      "loss": 0.0428,
+      "step": 1910
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.7836046511627905e-07,
+      "loss": 0.023,
+      "step": 1911
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.7834883720930234e-07,
+      "loss": 0.0374,
+      "step": 1912
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.7833720930232557e-07,
+      "loss": 0.0418,
+      "step": 1913
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.7832558139534886e-07,
+      "loss": 0.1358,
+      "step": 1914
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 2.783139534883721e-07,
+      "loss": 0.3153,
+      "step": 1915
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.7830232558139533e-07,
+      "loss": 0.0656,
+      "step": 1916
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.7829069767441856e-07,
+      "loss": 0.0252,
+      "step": 1917
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.7827906976744185e-07,
+      "loss": 0.0247,
+      "step": 1918
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.782674418604651e-07,
+      "loss": 0.0259,
+      "step": 1919
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2.782558139534884e-07,
+      "loss": 0.035,
+      "step": 1920
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 2.782441860465116e-07,
+      "loss": 0.0369,
+      "step": 1921
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 2.7823255813953485e-07,
+      "loss": 0.1233,
+      "step": 1922
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 2.7822093023255813e-07,
+      "loss": 0.0683,
+      "step": 1923
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 2.7820930232558137e-07,
+      "loss": 0.0334,
+      "step": 1924
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 2.7819767441860466e-07,
+      "loss": 0.0444,
+      "step": 1925
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.781860465116279e-07,
+      "loss": 0.1396,
+      "step": 1926
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.781744186046512e-07,
+      "loss": 0.027,
+      "step": 1927
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.781627906976744e-07,
+      "loss": 0.0326,
+      "step": 1928
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.7815116279069765e-07,
+      "loss": 0.0488,
+      "step": 1929
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 2.781395348837209e-07,
+      "loss": 0.0771,
+      "step": 1930
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.781279069767442e-07,
+      "loss": 0.0195,
+      "step": 1931
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.781162790697674e-07,
+      "loss": 0.0475,
+      "step": 1932
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.781046511627907e-07,
+      "loss": 0.0548,
+      "step": 1933
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.7809302325581393e-07,
+      "loss": 0.0261,
+      "step": 1934
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.780813953488372e-07,
+      "loss": 0.0529,
+      "step": 1935
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 2.7806976744186046e-07,
+      "loss": 1.7739,
+      "step": 1936
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.780581395348837e-07,
+      "loss": 0.0472,
+      "step": 1937
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.78046511627907e-07,
+      "loss": 0.0373,
+      "step": 1938
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.780348837209302e-07,
+      "loss": 0.0427,
+      "step": 1939
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.780232558139535e-07,
+      "loss": 1.5795,
+      "step": 1940
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 2.7801162790697674e-07,
+      "loss": 0.0453,
+      "step": 1941
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2.78e-07,
+      "loss": 1.0947,
+      "step": 1942
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2.779883720930232e-07,
+      "loss": 0.0378,
+      "step": 1943
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2.779767441860465e-07,
+      "loss": 0.0287,
+      "step": 1944
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2.7796511627906973e-07,
+      "loss": 0.0357,
+      "step": 1945
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2.77953488372093e-07,
+      "loss": 0.0562,
+      "step": 1946
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.7794186046511626e-07,
+      "loss": 0.02,
+      "step": 1947
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.7793023255813954e-07,
+      "loss": 0.0476,
+      "step": 1948
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.779186046511628e-07,
+      "loss": 0.0373,
+      "step": 1949
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.77906976744186e-07,
+      "loss": 0.0598,
+      "step": 1950
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 2.7789534883720925e-07,
+      "loss": 0.0405,
+      "step": 1951
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 2.7788372093023254e-07,
+      "loss": 0.6594,
+      "step": 1952
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 2.7787209302325583e-07,
+      "loss": 0.0439,
+      "step": 1953
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 2.7786046511627906e-07,
+      "loss": 0.0585,
+      "step": 1954
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 2.778488372093023e-07,
+      "loss": 0.0222,
+      "step": 1955
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 2.778372093023256e-07,
+      "loss": 0.0278,
+      "step": 1956
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.778255813953488e-07,
+      "loss": 0.0331,
+      "step": 1957
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.7781395348837206e-07,
+      "loss": 0.079,
+      "step": 1958
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.7780232558139534e-07,
+      "loss": 0.0332,
+      "step": 1959
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.777906976744186e-07,
+      "loss": 0.6048,
+      "step": 1960
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.7777906976744187e-07,
+      "loss": 0.8192,
+      "step": 1961
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 2.777674418604651e-07,
+      "loss": 0.0323,
+      "step": 1962
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.7775581395348834e-07,
+      "loss": 0.0549,
+      "step": 1963
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.7774418604651163e-07,
+      "loss": 0.0316,
+      "step": 1964
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.7773255813953486e-07,
+      "loss": 0.777,
+      "step": 1965
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.7772093023255815e-07,
+      "loss": 0.052,
+      "step": 1966
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 2.777093023255814e-07,
+      "loss": 0.0329,
+      "step": 1967
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2.7769767441860467e-07,
+      "loss": 0.0297,
+      "step": 1968
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2.776860465116279e-07,
+      "loss": 0.0297,
+      "step": 1969
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2.7767441860465114e-07,
+      "loss": 0.0341,
+      "step": 1970
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2.776627906976744e-07,
+      "loss": 0.0257,
+      "step": 1971
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2.7765116279069767e-07,
+      "loss": 0.09,
+      "step": 1972
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 2.776395348837209e-07,
+      "loss": 0.0279,
+      "step": 1973
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 2.776279069767442e-07,
+      "loss": 0.0216,
+      "step": 1974
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 2.776162790697674e-07,
+      "loss": 0.0236,
+      "step": 1975
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 2.7760465116279066e-07,
+      "loss": 0.0373,
+      "step": 1976
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 2.7759302325581395e-07,
+      "loss": 0.0767,
+      "step": 1977
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.775813953488372e-07,
+      "loss": 0.0869,
+      "step": 1978
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.7756976744186047e-07,
+      "loss": 0.0304,
+      "step": 1979
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.775581395348837e-07,
+      "loss": 0.0359,
+      "step": 1980
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.77546511627907e-07,
+      "loss": 0.3321,
+      "step": 1981
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 2.7753488372093023e-07,
+      "loss": 1.2878,
+      "step": 1982
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.7752325581395347e-07,
+      "loss": 0.024,
+      "step": 1983
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.775116279069767e-07,
+      "loss": 0.7795,
+      "step": 1984
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.775e-07,
+      "loss": 0.0505,
+      "step": 1985
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.774883720930232e-07,
+      "loss": 0.0215,
+      "step": 1986
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.774767441860465e-07,
+      "loss": 0.0457,
+      "step": 1987
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.7746511627906975e-07,
+      "loss": 0.0275,
+      "step": 1988
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.7745348837209304e-07,
+      "loss": 0.037,
+      "step": 1989
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.7744186046511627e-07,
+      "loss": 0.0216,
+      "step": 1990
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.774302325581395e-07,
+      "loss": 0.0288,
+      "step": 1991
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.7741860465116274e-07,
+      "loss": 0.0224,
+      "step": 1992
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 2.7740697674418603e-07,
+      "loss": 0.0376,
+      "step": 1993
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.773953488372093e-07,
+      "loss": 0.0326,
+      "step": 1994
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.7738372093023255e-07,
+      "loss": 0.045,
+      "step": 1995
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.773720930232558e-07,
+      "loss": 0.0515,
+      "step": 1996
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.77360465116279e-07,
+      "loss": 0.0273,
+      "step": 1997
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.773488372093023e-07,
+      "loss": 0.0396,
+      "step": 1998
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.7733720930232555e-07,
+      "loss": 0.062,
+      "step": 1999
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.7732558139534884e-07,
+      "loss": 0.0924,
+      "step": 2000
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.7731395348837207e-07,
+      "loss": 0.0205,
+      "step": 2001
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.7730232558139536e-07,
+      "loss": 0.1755,
+      "step": 2002
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 2.772906976744186e-07,
+      "loss": 0.0293,
+      "step": 2003
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.7727906976744183e-07,
+      "loss": 0.0433,
+      "step": 2004
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.7726744186046507e-07,
+      "loss": 0.0485,
+      "step": 2005
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.7725581395348835e-07,
+      "loss": 0.0627,
+      "step": 2006
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.7724418604651164e-07,
+      "loss": 0.029,
+      "step": 2007
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.772325581395349e-07,
+      "loss": 0.063,
+      "step": 2008
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 2.772209302325581e-07,
+      "loss": 0.1002,
+      "step": 2009
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 2.772093023255814e-07,
+      "loss": 0.0231,
+      "step": 2010
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 2.7719767441860464e-07,
+      "loss": 0.0747,
+      "step": 2011
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 2.7718604651162787e-07,
+      "loss": 0.0326,
+      "step": 2012
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 2.7717441860465116e-07,
+      "loss": 0.067,
+      "step": 2013
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.771627906976744e-07,
+      "loss": 0.0297,
+      "step": 2014
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.771511627906977e-07,
+      "loss": 0.0246,
+      "step": 2015
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.771395348837209e-07,
+      "loss": 0.0226,
+      "step": 2016
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.7712790697674415e-07,
+      "loss": 0.047,
+      "step": 2017
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 2.7711627906976744e-07,
+      "loss": 0.032,
+      "step": 2018
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.771046511627907e-07,
+      "loss": 0.028,
+      "step": 2019
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.7709302325581396e-07,
+      "loss": 0.04,
+      "step": 2020
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.770813953488372e-07,
+      "loss": 0.2328,
+      "step": 2021
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.770697674418605e-07,
+      "loss": 0.0479,
+      "step": 2022
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.770581395348837e-07,
+      "loss": 0.0588,
+      "step": 2023
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2.7704651162790696e-07,
+      "loss": 0.0467,
+      "step": 2024
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 2.770348837209302e-07,
+      "loss": 0.0238,
+      "step": 2025
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 2.770232558139535e-07,
+      "loss": 0.1031,
+      "step": 2026
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 2.770116279069767e-07,
+      "loss": 0.0209,
+      "step": 2027
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 2.77e-07,
+      "loss": 0.1102,
+      "step": 2028
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 2.7698837209302324e-07,
+      "loss": 0.0771,
+      "step": 2029
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.769767441860465e-07,
+      "loss": 0.0267,
+      "step": 2030
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.7696511627906976e-07,
+      "loss": 0.0513,
+      "step": 2031
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.76953488372093e-07,
+      "loss": 0.0248,
+      "step": 2032
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.769418604651163e-07,
+      "loss": 0.0246,
+      "step": 2033
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 2.769302325581395e-07,
+      "loss": 0.0443,
+      "step": 2034
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 2.769186046511628e-07,
+      "loss": 0.0341,
+      "step": 2035
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 2.7690697674418605e-07,
+      "loss": 0.0343,
+      "step": 2036
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 2.768953488372093e-07,
+      "loss": 0.0272,
+      "step": 2037
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 2.768837209302325e-07,
+      "loss": 0.0816,
+      "step": 2038
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 2.768720930232558e-07,
+      "loss": 0.066,
+      "step": 2039
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.7686046511627904e-07,
+      "loss": 0.0402,
+      "step": 2040
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.7684883720930233e-07,
+      "loss": 0.0191,
+      "step": 2041
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.7683720930232556e-07,
+      "loss": 0.0242,
+      "step": 2042
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.7682558139534885e-07,
+      "loss": 0.0285,
+      "step": 2043
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 2.768139534883721e-07,
+      "loss": 0.0424,
+      "step": 2044
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.768023255813953e-07,
+      "loss": 0.026,
+      "step": 2045
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.7679069767441856e-07,
+      "loss": 1.8445,
+      "step": 2046
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.7677906976744185e-07,
+      "loss": 0.0346,
+      "step": 2047
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.7676744186046513e-07,
+      "loss": 0.0306,
+      "step": 2048
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 2.7675581395348837e-07,
+      "loss": 0.8233,
+      "step": 2049
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.767441860465116e-07,
+      "loss": 0.0539,
+      "step": 2050
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.7673255813953484e-07,
+      "loss": 0.031,
+      "step": 2051
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.7672093023255813e-07,
+      "loss": 0.3501,
+      "step": 2052
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.7670930232558136e-07,
+      "loss": 0.0289,
+      "step": 2053
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.7669767441860465e-07,
+      "loss": 1.8158,
+      "step": 2054
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2.766860465116279e-07,
+      "loss": 0.024,
+      "step": 2055
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 2.766744186046512e-07,
+      "loss": 0.0964,
+      "step": 2056
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 2.766627906976744e-07,
+      "loss": 0.0558,
+      "step": 2057
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 2.7665116279069765e-07,
+      "loss": 0.0271,
+      "step": 2058
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 2.766395348837209e-07,
+      "loss": 0.027,
+      "step": 2059
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 2.7662790697674417e-07,
+      "loss": 1.197,
+      "step": 2060
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.7661627906976746e-07,
+      "loss": 0.1298,
+      "step": 2061
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.766046511627907e-07,
+      "loss": 0.0215,
+      "step": 2062
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.7659302325581393e-07,
+      "loss": 0.0514,
+      "step": 2063
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.765813953488372e-07,
+      "loss": 1.5546,
+      "step": 2064
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 2.7656976744186045e-07,
+      "loss": 0.0228,
+      "step": 2065
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.765581395348837e-07,
+      "loss": 0.019,
+      "step": 2066
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.76546511627907e-07,
+      "loss": 0.0683,
+      "step": 2067
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.765348837209302e-07,
+      "loss": 0.0446,
+      "step": 2068
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.9787234042553191,
+      "eval_loss": 0.15925627946853638,
+      "eval_runtime": 1.5415,
+      "eval_samples_per_second": 670.765,
+      "eval_steps_per_second": 11.028,
+      "step": 2068
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.765232558139535e-07,
+      "loss": 0.0278,
+      "step": 2069
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 2.7651162790697673e-07,
+      "loss": 0.0375,
+      "step": 2070
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 2.7649999999999997e-07,
+      "loss": 0.0193,
+      "step": 2071
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 2.7648837209302326e-07,
+      "loss": 0.0373,
+      "step": 2072
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 2.764767441860465e-07,
+      "loss": 0.0349,
+      "step": 2073
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 2.764651162790698e-07,
+      "loss": 0.028,
+      "step": 2074
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 2.76453488372093e-07,
+      "loss": 0.0389,
+      "step": 2075
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2.7644186046511625e-07,
+      "loss": 1.6831,
+      "step": 2076
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2.7643023255813954e-07,
+      "loss": 0.0531,
+      "step": 2077
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2.7641860465116277e-07,
+      "loss": 0.0291,
+      "step": 2078
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2.76406976744186e-07,
+      "loss": 1.3613,
+      "step": 2079
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2.763953488372093e-07,
+      "loss": 0.0371,
+      "step": 2080
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.7638372093023253e-07,
+      "loss": 0.0468,
+      "step": 2081
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.763720930232558e-07,
+      "loss": 0.039,
+      "step": 2082
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.7636046511627906e-07,
+      "loss": 0.0383,
+      "step": 2083
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.763488372093023e-07,
+      "loss": 0.0231,
+      "step": 2084
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.763372093023256e-07,
+      "loss": 0.2867,
+      "step": 2085
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 2.763255813953488e-07,
+      "loss": 1.1142,
+      "step": 2086
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 2.7631395348837205e-07,
+      "loss": 0.0296,
+      "step": 2087
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 2.7630232558139534e-07,
+      "loss": 0.0219,
+      "step": 2088
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 2.762906976744186e-07,
+      "loss": 0.0399,
+      "step": 2089
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 2.7627906976744186e-07,
+      "loss": 0.0368,
+      "step": 2090
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 2.762674418604651e-07,
+      "loss": 0.0311,
+      "step": 2091
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 2.7625581395348833e-07,
+      "loss": 0.0255,
+      "step": 2092
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 2.762441860465116e-07,
+      "loss": 0.0271,
+      "step": 2093
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 2.7623255813953485e-07,
+      "loss": 0.5034,
+      "step": 2094
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 2.7622093023255814e-07,
+      "loss": 0.0297,
+      "step": 2095
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 2.762093023255814e-07,
+      "loss": 0.0961,
+      "step": 2096
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.7619767441860467e-07,
+      "loss": 0.02,
+      "step": 2097
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.761860465116279e-07,
+      "loss": 0.0316,
+      "step": 2098
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.7617441860465114e-07,
+      "loss": 0.0219,
+      "step": 2099
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.7616279069767437e-07,
+      "loss": 0.0273,
+      "step": 2100
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 2.7615116279069766e-07,
+      "loss": 0.0245,
+      "step": 2101
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.7613953488372095e-07,
+      "loss": 0.0364,
+      "step": 2102
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.761279069767442e-07,
+      "loss": 0.0466,
+      "step": 2103
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.761162790697674e-07,
+      "loss": 0.0347,
+      "step": 2104
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.7610465116279065e-07,
+      "loss": 0.0159,
+      "step": 2105
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2.7609302325581394e-07,
+      "loss": 0.0221,
+      "step": 2106
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 2.760813953488372e-07,
+      "loss": 0.0294,
+      "step": 2107
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 2.7606976744186047e-07,
+      "loss": 0.0193,
+      "step": 2108
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 2.760581395348837e-07,
+      "loss": 0.0203,
+      "step": 2109
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 2.76046511627907e-07,
+      "loss": 0.0796,
+      "step": 2110
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 2.760348837209302e-07,
+      "loss": 0.2204,
+      "step": 2111
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.7602325581395346e-07,
+      "loss": 0.0747,
+      "step": 2112
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.760116279069767e-07,
+      "loss": 0.0635,
+      "step": 2113
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.76e-07,
+      "loss": 0.0334,
+      "step": 2114
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.7598837209302327e-07,
+      "loss": 0.0492,
+      "step": 2115
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.759767441860465e-07,
+      "loss": 0.0537,
+      "step": 2116
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 2.7596511627906974e-07,
+      "loss": 0.0177,
+      "step": 2117
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.7595348837209303e-07,
+      "loss": 0.0782,
+      "step": 2118
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.7594186046511627e-07,
+      "loss": 0.0302,
+      "step": 2119
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.759302325581395e-07,
+      "loss": 0.0382,
+      "step": 2120
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.759186046511628e-07,
+      "loss": 0.072,
+      "step": 2121
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 2.75906976744186e-07,
+      "loss": 0.0309,
+      "step": 2122
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 2.758953488372093e-07,
+      "loss": 0.0178,
+      "step": 2123
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 2.7588372093023255e-07,
+      "loss": 0.1033,
+      "step": 2124
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 2.758720930232558e-07,
+      "loss": 0.683,
+      "step": 2125
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 2.7586046511627907e-07,
+      "loss": 0.0332,
+      "step": 2126
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 2.758488372093023e-07,
+      "loss": 0.032,
+      "step": 2127
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.758372093023256e-07,
+      "loss": 0.0209,
+      "step": 2128
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.7582558139534883e-07,
+      "loss": 0.0242,
+      "step": 2129
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.7581395348837206e-07,
+      "loss": 0.0264,
+      "step": 2130
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.7580232558139535e-07,
+      "loss": 0.5161,
+      "step": 2131
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2.757906976744186e-07,
+      "loss": 0.0245,
+      "step": 2132
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.757790697674418e-07,
+      "loss": 0.062,
+      "step": 2133
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.757674418604651e-07,
+      "loss": 0.028,
+      "step": 2134
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.7575581395348835e-07,
+      "loss": 0.0363,
+      "step": 2135
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.7574418604651164e-07,
+      "loss": 0.0222,
+      "step": 2136
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 2.7573255813953487e-07,
+      "loss": 0.0533,
+      "step": 2137
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 2.757209302325581e-07,
+      "loss": 0.0301,
+      "step": 2138
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 2.757093023255814e-07,
+      "loss": 0.0273,
+      "step": 2139
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 2.7569767441860463e-07,
+      "loss": 0.944,
+      "step": 2140
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 2.7568604651162786e-07,
+      "loss": 0.0983,
+      "step": 2141
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 2.7567441860465115e-07,
+      "loss": 0.0188,
+      "step": 2142
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.7566279069767444e-07,
+      "loss": 0.0238,
+      "step": 2143
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.756511627906977e-07,
+      "loss": 0.0168,
+      "step": 2144
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.756395348837209e-07,
+      "loss": 1.2951,
+      "step": 2145
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.7562790697674415e-07,
+      "loss": 0.2591,
+      "step": 2146
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.7561627906976743e-07,
+      "loss": 0.2169,
+      "step": 2147
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 2.7560465116279067e-07,
+      "loss": 0.0387,
+      "step": 2148
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.7559302325581396e-07,
+      "loss": 0.0499,
+      "step": 2149
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.755813953488372e-07,
+      "loss": 0.0304,
+      "step": 2150
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.755697674418605e-07,
+      "loss": 0.0436,
+      "step": 2151
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.755581395348837e-07,
+      "loss": 0.0399,
+      "step": 2152
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 2.7554651162790695e-07,
+      "loss": 0.169,
+      "step": 2153
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2.755348837209302e-07,
+      "loss": 0.0313,
+      "step": 2154
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2.755232558139535e-07,
+      "loss": 0.0179,
+      "step": 2155
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2.7551162790697676e-07,
+      "loss": 0.022,
+      "step": 2156
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2.755e-07,
+      "loss": 0.0833,
+      "step": 2157
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2.7548837209302323e-07,
+      "loss": 0.0309,
+      "step": 2158
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 2.7547674418604647e-07,
+      "loss": 0.0679,
+      "step": 2159
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 2.7546511627906976e-07,
+      "loss": 0.0168,
+      "step": 2160
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 2.75453488372093e-07,
+      "loss": 0.023,
+      "step": 2161
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 2.754418604651163e-07,
+      "loss": 0.026,
+      "step": 2162
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 2.754302325581395e-07,
+      "loss": 0.1068,
+      "step": 2163
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.754186046511628e-07,
+      "loss": 0.0459,
+      "step": 2164
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.7540697674418604e-07,
+      "loss": 0.0295,
+      "step": 2165
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.753953488372093e-07,
+      "loss": 0.0269,
+      "step": 2166
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.753837209302325e-07,
+      "loss": 0.0153,
+      "step": 2167
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 2.753720930232558e-07,
+      "loss": 0.0355,
+      "step": 2168
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.753604651162791e-07,
+      "loss": 0.0366,
+      "step": 2169
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.753488372093023e-07,
+      "loss": 0.0304,
+      "step": 2170
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.7533720930232556e-07,
+      "loss": 0.0242,
+      "step": 2171
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.7532558139534884e-07,
+      "loss": 0.0178,
+      "step": 2172
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 2.753139534883721e-07,
+      "loss": 0.0285,
+      "step": 2173
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.753023255813953e-07,
+      "loss": 0.0208,
+      "step": 2174
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.752906976744186e-07,
+      "loss": 0.0313,
+      "step": 2175
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.7527906976744184e-07,
+      "loss": 0.2212,
+      "step": 2176
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.7526744186046513e-07,
+      "loss": 0.064,
+      "step": 2177
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.7525581395348836e-07,
+      "loss": 0.0376,
+      "step": 2178
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 2.752441860465116e-07,
+      "loss": 0.0373,
+      "step": 2179
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 2.752325581395349e-07,
+      "loss": 0.0396,
+      "step": 2180
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 2.752209302325581e-07,
+      "loss": 0.0256,
+      "step": 2181
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 2.752093023255814e-07,
+      "loss": 0.0287,
+      "step": 2182
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 2.7519767441860464e-07,
+      "loss": 0.3774,
+      "step": 2183
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 2.751860465116279e-07,
+      "loss": 0.0381,
+      "step": 2184
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.7517441860465117e-07,
+      "loss": 0.0176,
+      "step": 2185
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.751627906976744e-07,
+      "loss": 0.0229,
+      "step": 2186
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.7515116279069764e-07,
+      "loss": 0.0196,
+      "step": 2187
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.751395348837209e-07,
+      "loss": 0.0203,
+      "step": 2188
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2.7512790697674416e-07,
+      "loss": 0.0223,
+      "step": 2189
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.7511627906976745e-07,
+      "loss": 0.0198,
+      "step": 2190
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.751046511627907e-07,
+      "loss": 0.026,
+      "step": 2191
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.750930232558139e-07,
+      "loss": 0.0226,
+      "step": 2192
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.750813953488372e-07,
+      "loss": 0.0409,
+      "step": 2193
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 2.7506976744186044e-07,
+      "loss": 0.0158,
+      "step": 2194
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 2.750581395348837e-07,
+      "loss": 1.6434,
+      "step": 2195
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 2.7504651162790697e-07,
+      "loss": 0.0197,
+      "step": 2196
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 2.7503488372093026e-07,
+      "loss": 0.0281,
+      "step": 2197
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 2.750232558139535e-07,
+      "loss": 0.0312,
+      "step": 2198
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 2.750116279069767e-07,
+      "loss": 0.1757,
+      "step": 2199
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.7499999999999996e-07,
+      "loss": 0.2827,
+      "step": 2200
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.7498837209302325e-07,
+      "loss": 0.0325,
+      "step": 2201
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.749767441860465e-07,
+      "loss": 0.0335,
+      "step": 2202
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.7496511627906977e-07,
+      "loss": 0.0492,
+      "step": 2203
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.74953488372093e-07,
+      "loss": 0.3276,
+      "step": 2204
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 2.749418604651163e-07,
+      "loss": 0.0656,
+      "step": 2205
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 2.7493023255813953e-07,
+      "loss": 0.0196,
+      "step": 2206
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 2.7491860465116277e-07,
+      "loss": 0.0273,
+      "step": 2207
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 2.74906976744186e-07,
+      "loss": 0.0158,
+      "step": 2208
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 2.748953488372093e-07,
+      "loss": 0.0189,
+      "step": 2209
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 2.748837209302326e-07,
+      "loss": 0.0778,
+      "step": 2210
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.748720930232558e-07,
+      "loss": 0.0402,
+      "step": 2211
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.7486046511627905e-07,
+      "loss": 0.0347,
+      "step": 2212
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.748488372093023e-07,
+      "loss": 0.0171,
+      "step": 2213
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.7483720930232557e-07,
+      "loss": 0.0974,
+      "step": 2214
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2.748255813953488e-07,
+      "loss": 0.0327,
+      "step": 2215
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.748139534883721e-07,
+      "loss": 0.0199,
+      "step": 2216
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.7480232558139533e-07,
+      "loss": 0.0284,
+      "step": 2217
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.747906976744186e-07,
+      "loss": 0.0181,
+      "step": 2218
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.7477906976744185e-07,
+      "loss": 0.0203,
+      "step": 2219
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 2.747674418604651e-07,
+      "loss": 0.0135,
+      "step": 2220
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 2.747558139534883e-07,
+      "loss": 0.0332,
+      "step": 2221
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 2.747441860465116e-07,
+      "loss": 0.1599,
+      "step": 2222
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 2.747325581395349e-07,
+      "loss": 0.0284,
+      "step": 2223
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 2.7472093023255814e-07,
+      "loss": 0.0271,
+      "step": 2224
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 2.7470930232558137e-07,
+      "loss": 0.0235,
+      "step": 2225
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 2.7469767441860466e-07,
+      "loss": 0.029,
+      "step": 2226
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 2.746860465116279e-07,
+      "loss": 0.0241,
+      "step": 2227
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 2.7467441860465113e-07,
+      "loss": 0.0304,
+      "step": 2228
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 2.746627906976744e-07,
+      "loss": 0.0523,
+      "step": 2229
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 2.7465116279069765e-07,
+      "loss": 0.0593,
+      "step": 2230
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.7463953488372094e-07,
+      "loss": 0.0156,
+      "step": 2231
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.746279069767442e-07,
+      "loss": 0.03,
+      "step": 2232
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.746162790697674e-07,
+      "loss": 0.0227,
+      "step": 2233
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.746046511627907e-07,
+      "loss": 0.0226,
+      "step": 2234
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.7459302325581394e-07,
+      "loss": 0.0242,
+      "step": 2235
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 2.7458139534883717e-07,
+      "loss": 0.0241,
+      "step": 2236
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.7456976744186046e-07,
+      "loss": 0.0242,
+      "step": 2237
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.745581395348837e-07,
+      "loss": 0.0171,
+      "step": 2238
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.74546511627907e-07,
+      "loss": 0.0297,
+      "step": 2239
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.745348837209302e-07,
+      "loss": 0.2056,
+      "step": 2240
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2.7452325581395345e-07,
+      "loss": 0.0385,
+      "step": 2241
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 2.7451162790697674e-07,
+      "loss": 0.0216,
+      "step": 2242
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 2.745e-07,
+      "loss": 0.0485,
+      "step": 2243
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 2.7448837209302326e-07,
+      "loss": 0.0184,
+      "step": 2244
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 2.744767441860465e-07,
+      "loss": 0.1559,
+      "step": 2245
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 2.7446511627906974e-07,
+      "loss": 0.1365,
+      "step": 2246
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 2.74453488372093e-07,
+      "loss": 0.0692,
+      "step": 2247
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 2.7444186046511626e-07,
+      "loss": 0.0307,
+      "step": 2248
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 2.744302325581395e-07,
+      "loss": 0.0254,
+      "step": 2249
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 2.744186046511628e-07,
+      "loss": 0.0427,
+      "step": 2250
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 2.7440697674418607e-07,
+      "loss": 0.2262,
+      "step": 2251
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.743953488372093e-07,
+      "loss": 0.017,
+      "step": 2252
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.7438372093023254e-07,
+      "loss": 0.0168,
+      "step": 2253
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.743720930232558e-07,
+      "loss": 0.0211,
+      "step": 2254
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.7436046511627906e-07,
+      "loss": 0.0151,
+      "step": 2255
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 2.743488372093023e-07,
+      "loss": 0.0174,
+      "step": 2256
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 2.743372093023256e-07,
+      "loss": 0.0263,
+      "step": 2257
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 2.743255813953488e-07,
+      "loss": 0.0231,
+      "step": 2258
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 2.743139534883721e-07,
+      "loss": 0.0671,
+      "step": 2259
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 2.7430232558139535e-07,
+      "loss": 0.0166,
+      "step": 2260
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 2.742906976744186e-07,
+      "loss": 0.1439,
+      "step": 2261
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.742790697674418e-07,
+      "loss": 1.4457,
+      "step": 2262
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.742674418604651e-07,
+      "loss": 0.0184,
+      "step": 2263
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.742558139534884e-07,
+      "loss": 0.0198,
+      "step": 2264
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.7424418604651163e-07,
+      "loss": 0.021,
+      "step": 2265
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.7423255813953486e-07,
+      "loss": 0.0244,
+      "step": 2266
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2.742209302325581e-07,
+      "loss": 0.021,
+      "step": 2267
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.742093023255814e-07,
+      "loss": 0.0314,
+      "step": 2268
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.741976744186046e-07,
+      "loss": 0.0297,
+      "step": 2269
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.741860465116279e-07,
+      "loss": 0.033,
+      "step": 2270
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.7417441860465115e-07,
+      "loss": 0.0316,
+      "step": 2271
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 2.7416279069767443e-07,
+      "loss": 0.0225,
+      "step": 2272
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.7415116279069767e-07,
+      "loss": 0.1539,
+      "step": 2273
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.741395348837209e-07,
+      "loss": 0.0198,
+      "step": 2274
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.7412790697674414e-07,
+      "loss": 0.0268,
+      "step": 2275
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.7411627906976743e-07,
+      "loss": 0.0272,
+      "step": 2276
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 2.741046511627907e-07,
+      "loss": 0.0458,
+      "step": 2277
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 2.7409302325581395e-07,
+      "loss": 0.0254,
+      "step": 2278
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 2.740813953488372e-07,
+      "loss": 0.029,
+      "step": 2279
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 2.740697674418605e-07,
+      "loss": 1.3041,
+      "step": 2280
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 2.740581395348837e-07,
+      "loss": 0.0152,
+      "step": 2281
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 2.7404651162790694e-07,
+      "loss": 0.0543,
+      "step": 2282
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.7403488372093023e-07,
+      "loss": 0.0186,
+      "step": 2283
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.7402325581395347e-07,
+      "loss": 1.3974,
+      "step": 2284
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.7401162790697676e-07,
+      "loss": 0.071,
+      "step": 2285
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.74e-07,
+      "loss": 0.0428,
+      "step": 2286
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 2.7398837209302323e-07,
+      "loss": 0.0309,
+      "step": 2287
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2.739767441860465e-07,
+      "loss": 0.0209,
+      "step": 2288
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2.7396511627906975e-07,
+      "loss": 0.0226,
+      "step": 2289
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2.73953488372093e-07,
+      "loss": 0.0223,
+      "step": 2290
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2.739418604651163e-07,
+      "loss": 0.0146,
+      "step": 2291
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2.739302325581395e-07,
+      "loss": 0.0355,
+      "step": 2292
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.739186046511628e-07,
+      "loss": 0.1437,
+      "step": 2293
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.7390697674418603e-07,
+      "loss": 0.0462,
+      "step": 2294
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.7389534883720927e-07,
+      "loss": 0.0176,
+      "step": 2295
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.7388372093023256e-07,
+      "loss": 0.0207,
+      "step": 2296
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.738720930232558e-07,
+      "loss": 0.0378,
+      "step": 2297
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 2.738604651162791e-07,
+      "loss": 0.1249,
+      "step": 2298
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.738488372093023e-07,
+      "loss": 0.0896,
+      "step": 2299
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.7383720930232555e-07,
+      "loss": 0.0187,
+      "step": 2300
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.7382558139534884e-07,
+      "loss": 0.1073,
+      "step": 2301
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.7381395348837207e-07,
+      "loss": 0.0199,
+      "step": 2302
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.738023255813953e-07,
+      "loss": 0.0305,
+      "step": 2303
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.737906976744186e-07,
+      "loss": 0.0325,
+      "step": 2304
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.737790697674419e-07,
+      "loss": 0.0144,
+      "step": 2305
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.737674418604651e-07,
+      "loss": 0.0209,
+      "step": 2306
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.7375581395348836e-07,
+      "loss": 0.0146,
+      "step": 2307
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.737441860465116e-07,
+      "loss": 1.2561,
+      "step": 2308
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.737325581395349e-07,
+      "loss": 0.0226,
+      "step": 2309
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.737209302325581e-07,
+      "loss": 0.0369,
+      "step": 2310
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.737093023255814e-07,
+      "loss": 0.0201,
+      "step": 2311
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.7369767441860464e-07,
+      "loss": 0.0164,
+      "step": 2312
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.736860465116279e-07,
+      "loss": 1.4944,
+      "step": 2313
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.7367441860465116e-07,
+      "loss": 0.0234,
+      "step": 2314
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.736627906976744e-07,
+      "loss": 0.0241,
+      "step": 2315
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.7365116279069763e-07,
+      "loss": 0.0347,
+      "step": 2316
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.736395348837209e-07,
+      "loss": 0.0212,
+      "step": 2317
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.736279069767442e-07,
+      "loss": 0.0348,
+      "step": 2318
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.7361627906976744e-07,
+      "loss": 0.0225,
+      "step": 2319
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.736046511627907e-07,
+      "loss": 0.0204,
+      "step": 2320
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.735930232558139e-07,
+      "loss": 0.0631,
+      "step": 2321
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.735813953488372e-07,
+      "loss": 0.029,
+      "step": 2322
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.7356976744186044e-07,
+      "loss": 0.0154,
+      "step": 2323
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.735581395348837e-07,
+      "loss": 0.0314,
+      "step": 2324
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.7354651162790696e-07,
+      "loss": 0.51,
+      "step": 2325
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.7353488372093025e-07,
+      "loss": 0.0593,
+      "step": 2326
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.735232558139535e-07,
+      "loss": 0.2245,
+      "step": 2327
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.735116279069767e-07,
+      "loss": 0.0502,
+      "step": 2328
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.7349999999999995e-07,
+      "loss": 0.0213,
+      "step": 2329
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.7348837209302324e-07,
+      "loss": 0.0167,
+      "step": 2330
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.734767441860465e-07,
+      "loss": 0.0187,
+      "step": 2331
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.7346511627906977e-07,
+      "loss": 0.0193,
+      "step": 2332
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.73453488372093e-07,
+      "loss": 0.1345,
+      "step": 2333
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.734418604651163e-07,
+      "loss": 0.0356,
+      "step": 2334
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.734302325581395e-07,
+      "loss": 0.0162,
+      "step": 2335
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.7341860465116276e-07,
+      "loss": 0.0248,
+      "step": 2336
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.7340697674418605e-07,
+      "loss": 1.377,
+      "step": 2337
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.733953488372093e-07,
+      "loss": 0.0549,
+      "step": 2338
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.7338372093023257e-07,
+      "loss": 0.016,
+      "step": 2339
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.733720930232558e-07,
+      "loss": 0.0271,
+      "step": 2340
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.7336046511627904e-07,
+      "loss": 0.0581,
+      "step": 2341
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.7334883720930233e-07,
+      "loss": 0.0248,
+      "step": 2342
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.7333720930232557e-07,
+      "loss": 0.0381,
+      "step": 2343
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.733255813953488e-07,
+      "loss": 0.093,
+      "step": 2344
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.733139534883721e-07,
+      "loss": 0.0268,
+      "step": 2345
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.733023255813953e-07,
+      "loss": 0.0219,
+      "step": 2346
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.732906976744186e-07,
+      "loss": 0.024,
+      "step": 2347
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.7327906976744185e-07,
+      "loss": 0.0195,
+      "step": 2348
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.732674418604651e-07,
+      "loss": 0.0253,
+      "step": 2349
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.7325581395348837e-07,
+      "loss": 0.0223,
+      "step": 2350
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.732441860465116e-07,
+      "loss": 0.1778,
+      "step": 2351
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.732325581395349e-07,
+      "loss": 0.0303,
+      "step": 2352
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.7322093023255813e-07,
+      "loss": 0.0193,
+      "step": 2353
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.7320930232558136e-07,
+      "loss": 0.0237,
+      "step": 2354
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.7319767441860465e-07,
+      "loss": 0.0224,
+      "step": 2355
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.731860465116279e-07,
+      "loss": 0.0159,
+      "step": 2356
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.731744186046511e-07,
+      "loss": 1.3187,
+      "step": 2357
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.731627906976744e-07,
+      "loss": 0.047,
+      "step": 2358
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.731511627906977e-07,
+      "loss": 0.0137,
+      "step": 2359
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 2.7313953488372093e-07,
+      "loss": 0.013,
+      "step": 2360
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 2.7312790697674417e-07,
+      "loss": 0.3259,
+      "step": 2361
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 2.731162790697674e-07,
+      "loss": 0.0128,
+      "step": 2362
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 2.731046511627907e-07,
+      "loss": 0.0136,
+      "step": 2363
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 2.7309302325581393e-07,
+      "loss": 0.0227,
+      "step": 2364
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 2.730813953488372e-07,
+      "loss": 0.0214,
+      "step": 2365
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.7306976744186045e-07,
+      "loss": 0.1823,
+      "step": 2366
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.7305813953488374e-07,
+      "loss": 0.0527,
+      "step": 2367
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.73046511627907e-07,
+      "loss": 0.0227,
+      "step": 2368
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.730348837209302e-07,
+      "loss": 0.0395,
+      "step": 2369
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 2.7302325581395345e-07,
+      "loss": 0.0147,
+      "step": 2370
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.7301162790697673e-07,
+      "loss": 0.0137,
+      "step": 2371
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.73e-07,
+      "loss": 0.367,
+      "step": 2372
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.7298837209302326e-07,
+      "loss": 0.0557,
+      "step": 2373
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.729767441860465e-07,
+      "loss": 0.0253,
+      "step": 2374
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2.7296511627906973e-07,
+      "loss": 0.0153,
+      "step": 2375
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 2.72953488372093e-07,
+      "loss": 0.0881,
+      "step": 2376
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 2.7294186046511625e-07,
+      "loss": 0.0166,
+      "step": 2377
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 2.7293023255813954e-07,
+      "loss": 0.0191,
+      "step": 2378
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 2.729186046511628e-07,
+      "loss": 0.1268,
+      "step": 2379
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 2.7290697674418606e-07,
+      "loss": 0.0181,
+      "step": 2380
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 2.728953488372093e-07,
+      "loss": 1.7324,
+      "step": 2381
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 2.7288372093023253e-07,
+      "loss": 0.0156,
+      "step": 2382
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 2.7287209302325577e-07,
+      "loss": 0.013,
+      "step": 2383
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 2.7286046511627906e-07,
+      "loss": 0.543,
+      "step": 2384
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 2.728488372093023e-07,
+      "loss": 0.0214,
+      "step": 2385
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.728372093023256e-07,
+      "loss": 0.0204,
+      "step": 2386
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.728255813953488e-07,
+      "loss": 0.0175,
+      "step": 2387
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.728139534883721e-07,
+      "loss": 0.0126,
+      "step": 2388
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.7280232558139534e-07,
+      "loss": 0.0248,
+      "step": 2389
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.727906976744186e-07,
+      "loss": 0.0203,
+      "step": 2390
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 2.7277906976744186e-07,
+      "loss": 0.0445,
+      "step": 2391
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 2.727674418604651e-07,
+      "loss": 0.178,
+      "step": 2392
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 2.727558139534884e-07,
+      "loss": 0.0212,
+      "step": 2393
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 2.727441860465116e-07,
+      "loss": 0.0206,
+      "step": 2394
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 2.7273255813953486e-07,
+      "loss": 0.0165,
+      "step": 2395
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 2.727209302325581e-07,
+      "loss": 0.0328,
+      "step": 2396
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.727093023255814e-07,
+      "loss": 0.0568,
+      "step": 2397
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.726976744186046e-07,
+      "loss": 0.0248,
+      "step": 2398
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.726860465116279e-07,
+      "loss": 0.02,
+      "step": 2399
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.7267441860465114e-07,
+      "loss": 0.0138,
+      "step": 2400
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2.7266279069767443e-07,
+      "loss": 0.0312,
+      "step": 2401
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.7265116279069766e-07,
+      "loss": 0.0227,
+      "step": 2402
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.726395348837209e-07,
+      "loss": 0.0371,
+      "step": 2403
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.726279069767442e-07,
+      "loss": 0.0241,
+      "step": 2404
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.726162790697674e-07,
+      "loss": 0.0243,
+      "step": 2405
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 2.726046511627907e-07,
+      "loss": 0.0404,
+      "step": 2406
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 2.7259302325581394e-07,
+      "loss": 0.0188,
+      "step": 2407
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 2.725813953488372e-07,
+      "loss": 0.0478,
+      "step": 2408
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 2.7256976744186047e-07,
+      "loss": 0.0292,
+      "step": 2409
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 2.725581395348837e-07,
+      "loss": 0.0157,
+      "step": 2410
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 2.7254651162790694e-07,
+      "loss": 0.0208,
+      "step": 2411
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 2.725348837209302e-07,
+      "loss": 0.0433,
+      "step": 2412
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 2.725232558139535e-07,
+      "loss": 0.0302,
+      "step": 2413
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 2.7251162790697675e-07,
+      "loss": 0.0178,
+      "step": 2414
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 2.725e-07,
+      "loss": 0.0507,
+      "step": 2415
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 2.724883720930232e-07,
+      "loss": 0.0113,
+      "step": 2416
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.724767441860465e-07,
+      "loss": 0.0193,
+      "step": 2417
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.7246511627906974e-07,
+      "loss": 0.0357,
+      "step": 2418
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.7245348837209303e-07,
+      "loss": 0.015,
+      "step": 2419
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.7244186046511627e-07,
+      "loss": 0.0126,
+      "step": 2420
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.7243023255813956e-07,
+      "loss": 0.0152,
+      "step": 2421
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 2.724186046511628e-07,
+      "loss": 0.0138,
+      "step": 2422
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.72406976744186e-07,
+      "loss": 0.0224,
+      "step": 2423
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.7239534883720926e-07,
+      "loss": 0.0756,
+      "step": 2424
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.7238372093023255e-07,
+      "loss": 0.0237,
+      "step": 2425
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.7237209302325584e-07,
+      "loss": 0.0159,
+      "step": 2426
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2.7236046511627907e-07,
+      "loss": 0.0114,
+      "step": 2427
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 2.723488372093023e-07,
+      "loss": 0.0227,
+      "step": 2428
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 2.7233720930232554e-07,
+      "loss": 2.0084,
+      "step": 2429
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 2.7232558139534883e-07,
+      "loss": 0.0278,
+      "step": 2430
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 2.7231395348837207e-07,
+      "loss": 0.0191,
+      "step": 2431
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 2.7230232558139535e-07,
+      "loss": 0.0254,
+      "step": 2432
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.722906976744186e-07,
+      "loss": 0.0211,
+      "step": 2433
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.722790697674419e-07,
+      "loss": 0.023,
+      "step": 2434
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.722674418604651e-07,
+      "loss": 0.0242,
+      "step": 2435
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.7225581395348835e-07,
+      "loss": 0.0155,
+      "step": 2436
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 2.722441860465116e-07,
+      "loss": 0.0922,
+      "step": 2437
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.7223255813953487e-07,
+      "loss": 0.3055,
+      "step": 2438
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.722209302325581e-07,
+      "loss": 0.0265,
+      "step": 2439
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.722093023255814e-07,
+      "loss": 0.0873,
+      "step": 2440
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.7219767441860463e-07,
+      "loss": 0.0336,
+      "step": 2441
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 2.721860465116279e-07,
+      "loss": 1.5624,
+      "step": 2442
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.7217441860465115e-07,
+      "loss": 0.0411,
+      "step": 2443
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.721627906976744e-07,
+      "loss": 1.0365,
+      "step": 2444
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.721511627906977e-07,
+      "loss": 0.0128,
+      "step": 2445
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.721395348837209e-07,
+      "loss": 0.03,
+      "step": 2446
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 2.721279069767442e-07,
+      "loss": 0.0235,
+      "step": 2447
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.7211627906976744e-07,
+      "loss": 0.0266,
+      "step": 2448
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.7210465116279067e-07,
+      "loss": 0.0237,
+      "step": 2449
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.720930232558139e-07,
+      "loss": 0.0128,
+      "step": 2450
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.720813953488372e-07,
+      "loss": 0.011,
+      "step": 2451
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.7206976744186043e-07,
+      "loss": 0.0282,
+      "step": 2452
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.720581395348837e-07,
+      "loss": 0.0253,
+      "step": 2453
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.7204651162790695e-07,
+      "loss": 0.0131,
+      "step": 2454
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.7203488372093024e-07,
+      "loss": 0.0188,
+      "step": 2455
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.720232558139535e-07,
+      "loss": 0.033,
+      "step": 2456
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.720116279069767e-07,
+      "loss": 0.0241,
+      "step": 2457
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 2.7199999999999995e-07,
+      "loss": 0.0178,
+      "step": 2458
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.7198837209302324e-07,
+      "loss": 2.0906,
+      "step": 2459
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.719767441860465e-07,
+      "loss": 0.0426,
+      "step": 2460
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.7196511627906976e-07,
+      "loss": 0.0116,
+      "step": 2461
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.71953488372093e-07,
+      "loss": 0.0494,
+      "step": 2462
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 2.719418604651163e-07,
+      "loss": 0.0166,
+      "step": 2463
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.719302325581395e-07,
+      "loss": 0.0185,
+      "step": 2464
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.7191860465116275e-07,
+      "loss": 0.0259,
+      "step": 2465
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.7190697674418604e-07,
+      "loss": 0.0121,
+      "step": 2466
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.7189534883720933e-07,
+      "loss": 0.0303,
+      "step": 2467
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 2.7188372093023256e-07,
+      "loss": 0.0109,
+      "step": 2468
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.718720930232558e-07,
+      "loss": 0.0184,
+      "step": 2469
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.7186046511627904e-07,
+      "loss": 0.0215,
+      "step": 2470
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.718488372093023e-07,
+      "loss": 0.0145,
+      "step": 2471
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.7183720930232556e-07,
+      "loss": 0.0221,
+      "step": 2472
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 2.7182558139534885e-07,
+      "loss": 0.0208,
+      "step": 2473
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.718139534883721e-07,
+      "loss": 0.056,
+      "step": 2474
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.7180232558139537e-07,
+      "loss": 0.0122,
+      "step": 2475
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.717906976744186e-07,
+      "loss": 0.0256,
+      "step": 2476
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.7177906976744184e-07,
+      "loss": 0.0294,
+      "step": 2477
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.717674418604651e-07,
+      "loss": 0.0171,
+      "step": 2478
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2.7175581395348836e-07,
+      "loss": 0.0264,
+      "step": 2479
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.717441860465116e-07,
+      "loss": 0.0901,
+      "step": 2480
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.717325581395349e-07,
+      "loss": 0.0123,
+      "step": 2481
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.717209302325581e-07,
+      "loss": 0.2025,
+      "step": 2482
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.7170930232558136e-07,
+      "loss": 0.0542,
+      "step": 2483
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 2.7169767441860465e-07,
+      "loss": 0.0127,
+      "step": 2484
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.716860465116279e-07,
+      "loss": 1.7941,
+      "step": 2485
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.7167441860465117e-07,
+      "loss": 0.0144,
+      "step": 2486
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.716627906976744e-07,
+      "loss": 0.0185,
+      "step": 2487
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.716511627906977e-07,
+      "loss": 1.2798,
+      "step": 2488
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.7163953488372093e-07,
+      "loss": 0.0195,
+      "step": 2489
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.7162790697674416e-07,
+      "loss": 0.0199,
+      "step": 2490
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.716162790697674e-07,
+      "loss": 0.0162,
+      "step": 2491
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.716046511627907e-07,
+      "loss": 0.0141,
+      "step": 2492
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.715930232558139e-07,
+      "loss": 0.0182,
+      "step": 2493
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.715813953488372e-07,
+      "loss": 1.4994,
+      "step": 2494
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.7156976744186045e-07,
+      "loss": 0.0306,
+      "step": 2495
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.7155813953488373e-07,
+      "loss": 0.0288,
+      "step": 2496
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.7154651162790697e-07,
+      "loss": 0.0141,
+      "step": 2497
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.715348837209302e-07,
+      "loss": 0.0152,
+      "step": 2498
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.715232558139535e-07,
+      "loss": 0.0186,
+      "step": 2499
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.7151162790697673e-07,
+      "loss": 0.0163,
+      "step": 2500
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.715e-07,
+      "loss": 0.0237,
+      "step": 2501
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.7148837209302325e-07,
+      "loss": 0.0166,
+      "step": 2502
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.714767441860465e-07,
+      "loss": 0.015,
+      "step": 2503
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.714651162790697e-07,
+      "loss": 0.0252,
+      "step": 2504
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.71453488372093e-07,
+      "loss": 0.0139,
+      "step": 2505
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.7144186046511624e-07,
+      "loss": 0.0124,
+      "step": 2506
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.7143023255813953e-07,
+      "loss": 0.0183,
+      "step": 2507
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.7141860465116277e-07,
+      "loss": 0.8494,
+      "step": 2508
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.7140697674418606e-07,
+      "loss": 0.0131,
+      "step": 2509
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.713953488372093e-07,
+      "loss": 0.9585,
+      "step": 2510
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.7138372093023253e-07,
+      "loss": 0.1459,
+      "step": 2511
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.7137209302325576e-07,
+      "loss": 0.9829,
+      "step": 2512
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.7136046511627905e-07,
+      "loss": 0.0219,
+      "step": 2513
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.7134883720930234e-07,
+      "loss": 0.8278,
+      "step": 2514
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.713372093023256e-07,
+      "loss": 0.0226,
+      "step": 2515
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.713255813953488e-07,
+      "loss": 1.3814,
+      "step": 2516
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.713139534883721e-07,
+      "loss": 0.0187,
+      "step": 2517
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.7130232558139533e-07,
+      "loss": 0.0161,
+      "step": 2518
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.7129069767441857e-07,
+      "loss": 0.0135,
+      "step": 2519
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.7127906976744186e-07,
+      "loss": 0.0481,
+      "step": 2520
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.7126744186046514e-07,
+      "loss": 0.0208,
+      "step": 2521
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.712558139534884e-07,
+      "loss": 0.0475,
+      "step": 2522
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.712441860465116e-07,
+      "loss": 0.0174,
+      "step": 2523
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.7123255813953485e-07,
+      "loss": 0.1246,
+      "step": 2524
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 2.7122093023255814e-07,
+      "loss": 0.0192,
+      "step": 2525
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2.7120930232558137e-07,
+      "loss": 0.0131,
+      "step": 2526
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2.7119767441860466e-07,
+      "loss": 0.0137,
+      "step": 2527
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2.711860465116279e-07,
+      "loss": 0.0314,
+      "step": 2528
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2.7117441860465113e-07,
+      "loss": 0.0226,
+      "step": 2529
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2.711627906976744e-07,
+      "loss": 0.0196,
+      "step": 2530
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.7115116279069766e-07,
+      "loss": 0.016,
+      "step": 2531
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.711395348837209e-07,
+      "loss": 0.0156,
+      "step": 2532
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.711279069767442e-07,
+      "loss": 0.014,
+      "step": 2533
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.711162790697674e-07,
+      "loss": 0.0184,
+      "step": 2534
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 2.711046511627907e-07,
+      "loss": 0.0212,
+      "step": 2535
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.7109302325581394e-07,
+      "loss": 0.0108,
+      "step": 2536
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.7108139534883717e-07,
+      "loss": 0.0161,
+      "step": 2537
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.7106976744186046e-07,
+      "loss": 0.0379,
+      "step": 2538
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.710581395348837e-07,
+      "loss": 0.0175,
+      "step": 2539
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.71046511627907e-07,
+      "loss": 0.0242,
+      "step": 2540
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 2.710348837209302e-07,
+      "loss": 0.1527,
+      "step": 2541
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.710232558139535e-07,
+      "loss": 0.0169,
+      "step": 2542
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.7101162790697674e-07,
+      "loss": 0.406,
+      "step": 2543
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.71e-07,
+      "loss": 0.0175,
+      "step": 2544
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.709883720930232e-07,
+      "loss": 0.0171,
+      "step": 2545
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 2.709767441860465e-07,
+      "loss": 0.0116,
+      "step": 2546
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 2.7096511627906974e-07,
+      "loss": 0.017,
+      "step": 2547
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 2.70953488372093e-07,
+      "loss": 0.0151,
+      "step": 2548
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 2.7094186046511626e-07,
+      "loss": 0.1982,
+      "step": 2549
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 2.7093023255813955e-07,
+      "loss": 0.1421,
+      "step": 2550
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 2.709186046511628e-07,
+      "loss": 0.018,
+      "step": 2551
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2.70906976744186e-07,
+      "loss": 1.1791,
+      "step": 2552
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2.708953488372093e-07,
+      "loss": 0.0202,
+      "step": 2553
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2.7088372093023254e-07,
+      "loss": 0.0181,
+      "step": 2554
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2.7087209302325583e-07,
+      "loss": 0.0385,
+      "step": 2555
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2.7086046511627907e-07,
+      "loss": 0.0405,
+      "step": 2556
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.708488372093023e-07,
+      "loss": 0.0231,
+      "step": 2557
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.7083720930232554e-07,
+      "loss": 0.0249,
+      "step": 2558
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.708255813953488e-07,
+      "loss": 0.0182,
+      "step": 2559
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.7081395348837206e-07,
+      "loss": 0.0263,
+      "step": 2560
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.7080232558139535e-07,
+      "loss": 0.0111,
+      "step": 2561
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.707906976744186e-07,
+      "loss": 0.019,
+      "step": 2562
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.7077906976744187e-07,
+      "loss": 0.0427,
+      "step": 2563
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.707674418604651e-07,
+      "loss": 0.036,
+      "step": 2564
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.7075581395348834e-07,
+      "loss": 1.3689,
+      "step": 2565
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.707441860465116e-07,
+      "loss": 0.0149,
+      "step": 2566
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.7073255813953487e-07,
+      "loss": 0.0151,
+      "step": 2567
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.7072093023255815e-07,
+      "loss": 0.0247,
+      "step": 2568
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.707093023255814e-07,
+      "loss": 0.0264,
+      "step": 2569
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.706976744186046e-07,
+      "loss": 0.0195,
+      "step": 2570
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.706860465116279e-07,
+      "loss": 0.017,
+      "step": 2571
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 2.7067441860465115e-07,
+      "loss": 0.0209,
+      "step": 2572
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.706627906976744e-07,
+      "loss": 0.0175,
+      "step": 2573
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.7065116279069767e-07,
+      "loss": 0.0186,
+      "step": 2574
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.706395348837209e-07,
+      "loss": 0.0185,
+      "step": 2575
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.706279069767442e-07,
+      "loss": 0.0113,
+      "step": 2576
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.7061627906976743e-07,
+      "loss": 0.0121,
+      "step": 2577
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.7060465116279066e-07,
+      "loss": 0.0165,
+      "step": 2578
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.7059302325581395e-07,
+      "loss": 0.0186,
+      "step": 2579
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.705813953488372e-07,
+      "loss": 0.0164,
+      "step": 2580
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.705697674418605e-07,
+      "loss": 0.0126,
+      "step": 2581
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.705581395348837e-07,
+      "loss": 0.0148,
+      "step": 2582
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2.7054651162790695e-07,
+      "loss": 0.0152,
+      "step": 2583
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2.7053488372093023e-07,
+      "loss": 0.0124,
+      "step": 2584
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2.7052325581395347e-07,
+      "loss": 0.0175,
+      "step": 2585
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.9825918762088974,
+      "eval_loss": 0.15864484012126923,
+      "eval_runtime": 1.5463,
+      "eval_samples_per_second": 668.713,
+      "eval_steps_per_second": 10.994,
+      "step": 2585
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2.705116279069767e-07,
+      "loss": 0.49,
+      "step": 2586
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2.705e-07,
+      "loss": 0.0564,
+      "step": 2587
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 2.7048837209302323e-07,
+      "loss": 0.0164,
+      "step": 2588
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 2.704767441860465e-07,
+      "loss": 0.0101,
+      "step": 2589
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 2.7046511627906975e-07,
+      "loss": 0.0132,
+      "step": 2590
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 2.70453488372093e-07,
+      "loss": 0.0127,
+      "step": 2591
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 2.704418604651163e-07,
+      "loss": 0.0245,
+      "step": 2592
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.704302325581395e-07,
+      "loss": 0.3043,
+      "step": 2593
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.704186046511628e-07,
+      "loss": 0.0169,
+      "step": 2594
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.7040697674418603e-07,
+      "loss": 0.1051,
+      "step": 2595
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.703953488372093e-07,
+      "loss": 0.0182,
+      "step": 2596
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 2.7038372093023256e-07,
+      "loss": 0.0268,
+      "step": 2597
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.703720930232558e-07,
+      "loss": 0.014,
+      "step": 2598
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.7036046511627903e-07,
+      "loss": 0.0172,
+      "step": 2599
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.703488372093023e-07,
+      "loss": 1.4403,
+      "step": 2600
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.7033720930232555e-07,
+      "loss": 0.0386,
+      "step": 2601
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.7032558139534884e-07,
+      "loss": 0.0259,
+      "step": 2602
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 2.703139534883721e-07,
+      "loss": 0.0163,
+      "step": 2603
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 2.7030232558139536e-07,
+      "loss": 0.0365,
+      "step": 2604
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 2.702906976744186e-07,
+      "loss": 0.0185,
+      "step": 2605
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 2.7027906976744183e-07,
+      "loss": 0.0222,
+      "step": 2606
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 2.7026744186046507e-07,
+      "loss": 0.0125,
+      "step": 2607
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 2.7025581395348836e-07,
+      "loss": 0.0158,
+      "step": 2608
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.7024418604651165e-07,
+      "loss": 0.0219,
+      "step": 2609
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.702325581395349e-07,
+      "loss": 0.019,
+      "step": 2610
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.702209302325581e-07,
+      "loss": 0.0161,
+      "step": 2611
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.7020930232558135e-07,
+      "loss": 0.0677,
+      "step": 2612
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2.7019767441860464e-07,
+      "loss": 0.0166,
+      "step": 2613
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 2.701860465116279e-07,
+      "loss": 0.0105,
+      "step": 2614
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 2.7017441860465116e-07,
+      "loss": 0.0118,
+      "step": 2615
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 2.701627906976744e-07,
+      "loss": 0.0124,
+      "step": 2616
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 2.701511627906977e-07,
+      "loss": 0.0109,
+      "step": 2617
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 2.701395348837209e-07,
+      "loss": 0.0181,
+      "step": 2618
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 2.7012790697674416e-07,
+      "loss": 0.0669,
+      "step": 2619
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 2.701162790697674e-07,
+      "loss": 0.0138,
+      "step": 2620
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 2.701046511627907e-07,
+      "loss": 0.0757,
+      "step": 2621
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 2.7009302325581397e-07,
+      "loss": 2.0965,
+      "step": 2622
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 2.700813953488372e-07,
+      "loss": 0.0127,
+      "step": 2623
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.7006976744186044e-07,
+      "loss": 0.0118,
+      "step": 2624
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.7005813953488373e-07,
+      "loss": 0.0702,
+      "step": 2625
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.7004651162790696e-07,
+      "loss": 0.0223,
+      "step": 2626
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.700348837209302e-07,
+      "loss": 0.0224,
+      "step": 2627
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 2.700232558139535e-07,
+      "loss": 0.0144,
+      "step": 2628
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.700116279069767e-07,
+      "loss": 0.0181,
+      "step": 2629
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.7e-07,
+      "loss": 0.0121,
+      "step": 2630
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.6998837209302324e-07,
+      "loss": 0.0176,
+      "step": 2631
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.699767441860465e-07,
+      "loss": 0.0155,
+      "step": 2632
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.6996511627906977e-07,
+      "loss": 0.027,
+      "step": 2633
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 2.69953488372093e-07,
+      "loss": 0.012,
+      "step": 2634
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.699418604651163e-07,
+      "loss": 0.0281,
+      "step": 2635
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.699302325581395e-07,
+      "loss": 0.0233,
+      "step": 2636
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.6991860465116276e-07,
+      "loss": 0.0139,
+      "step": 2637
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.6990697674418605e-07,
+      "loss": 0.0263,
+      "step": 2638
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2.698953488372093e-07,
+      "loss": 2.0611,
+      "step": 2639
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.698837209302325e-07,
+      "loss": 0.0184,
+      "step": 2640
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.698720930232558e-07,
+      "loss": 0.0143,
+      "step": 2641
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.6986046511627904e-07,
+      "loss": 0.0215,
+      "step": 2642
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.6984883720930233e-07,
+      "loss": 0.0187,
+      "step": 2643
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 2.6983720930232557e-07,
+      "loss": 0.0101,
+      "step": 2644
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 2.698255813953488e-07,
+      "loss": 2.1253,
+      "step": 2645
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 2.698139534883721e-07,
+      "loss": 0.0372,
+      "step": 2646
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 2.698023255813953e-07,
+      "loss": 0.0195,
+      "step": 2647
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 2.697906976744186e-07,
+      "loss": 0.2742,
+      "step": 2648
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 2.6977906976744185e-07,
+      "loss": 0.0145,
+      "step": 2649
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 2.6976744186046514e-07,
+      "loss": 0.0115,
+      "step": 2650
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 2.6975581395348837e-07,
+      "loss": 0.0099,
+      "step": 2651
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 2.697441860465116e-07,
+      "loss": 0.1078,
+      "step": 2652
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 2.6973255813953484e-07,
+      "loss": 0.0106,
+      "step": 2653
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 2.6972093023255813e-07,
+      "loss": 0.0423,
+      "step": 2654
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 2.6970930232558137e-07,
+      "loss": 0.017,
+      "step": 2655
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 2.6969767441860465e-07,
+      "loss": 0.0119,
+      "step": 2656
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 2.696860465116279e-07,
+      "loss": 0.0354,
+      "step": 2657
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 2.696744186046512e-07,
+      "loss": 0.0172,
+      "step": 2658
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 2.696627906976744e-07,
+      "loss": 0.0184,
+      "step": 2659
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.6965116279069765e-07,
+      "loss": 1.2188,
+      "step": 2660
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.696395348837209e-07,
+      "loss": 0.0218,
+      "step": 2661
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.6962790697674417e-07,
+      "loss": 0.082,
+      "step": 2662
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.6961627906976746e-07,
+      "loss": 0.4465,
+      "step": 2663
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.696046511627907e-07,
+      "loss": 0.0131,
+      "step": 2664
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2.6959302325581393e-07,
+      "loss": 0.1284,
+      "step": 2665
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.6958139534883717e-07,
+      "loss": 0.0229,
+      "step": 2666
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.6956976744186045e-07,
+      "loss": 0.0125,
+      "step": 2667
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.695581395348837e-07,
+      "loss": 0.2017,
+      "step": 2668
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.69546511627907e-07,
+      "loss": 0.0117,
+      "step": 2669
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 2.695348837209302e-07,
+      "loss": 0.0126,
+      "step": 2670
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 2.695232558139535e-07,
+      "loss": 0.0162,
+      "step": 2671
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 2.6951162790697674e-07,
+      "loss": 0.0183,
+      "step": 2672
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 2.6949999999999997e-07,
+      "loss": 0.0317,
+      "step": 2673
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 2.694883720930232e-07,
+      "loss": 0.0175,
+      "step": 2674
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 2.694767441860465e-07,
+      "loss": 1.5317,
+      "step": 2675
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.694651162790698e-07,
+      "loss": 0.0224,
+      "step": 2676
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.69453488372093e-07,
+      "loss": 0.016,
+      "step": 2677
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.6944186046511625e-07,
+      "loss": 0.0216,
+      "step": 2678
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.6943023255813954e-07,
+      "loss": 0.0213,
+      "step": 2679
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 2.694186046511628e-07,
+      "loss": 0.0157,
+      "step": 2680
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 2.69406976744186e-07,
+      "loss": 0.3592,
+      "step": 2681
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 2.693953488372093e-07,
+      "loss": 0.0224,
+      "step": 2682
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 2.6938372093023254e-07,
+      "loss": 0.012,
+      "step": 2683
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 2.693720930232558e-07,
+      "loss": 0.0101,
+      "step": 2684
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 2.6936046511627906e-07,
+      "loss": 0.0165,
+      "step": 2685
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 2.693488372093023e-07,
+      "loss": 0.0173,
+      "step": 2686
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 2.693372093023256e-07,
+      "loss": 0.0169,
+      "step": 2687
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 2.693255813953488e-07,
+      "loss": 0.0137,
+      "step": 2688
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 2.693139534883721e-07,
+      "loss": 1.5222,
+      "step": 2689
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 2.6930232558139534e-07,
+      "loss": 0.0179,
+      "step": 2690
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.692906976744186e-07,
+      "loss": 0.0112,
+      "step": 2691
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.6927906976744186e-07,
+      "loss": 0.4889,
+      "step": 2692
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.692674418604651e-07,
+      "loss": 0.0252,
+      "step": 2693
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.6925581395348833e-07,
+      "loss": 0.0172,
+      "step": 2694
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.692441860465116e-07,
+      "loss": 0.0131,
+      "step": 2695
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2.6923255813953486e-07,
+      "loss": 0.0172,
+      "step": 2696
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 2.6922093023255815e-07,
+      "loss": 0.0088,
+      "step": 2697
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 2.692093023255814e-07,
+      "loss": 0.019,
+      "step": 2698
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 2.691976744186046e-07,
+      "loss": 0.0139,
+      "step": 2699
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 2.691860465116279e-07,
+      "loss": 0.0169,
+      "step": 2700
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 2.6917441860465114e-07,
+      "loss": 0.0194,
+      "step": 2701
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.6916279069767443e-07,
+      "loss": 0.0198,
+      "step": 2702
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.6915116279069766e-07,
+      "loss": 0.0134,
+      "step": 2703
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.6913953488372095e-07,
+      "loss": 1.6487,
+      "step": 2704
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.691279069767442e-07,
+      "loss": 0.0143,
+      "step": 2705
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 2.691162790697674e-07,
+      "loss": 0.0168,
+      "step": 2706
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 2.6910465116279066e-07,
+      "loss": 0.0439,
+      "step": 2707
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 2.6909302325581395e-07,
+      "loss": 0.0213,
+      "step": 2708
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 2.690813953488372e-07,
+      "loss": 0.0279,
+      "step": 2709
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 2.6906976744186047e-07,
+      "loss": 1.1442,
+      "step": 2710
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 2.690581395348837e-07,
+      "loss": 0.0138,
+      "step": 2711
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.69046511627907e-07,
+      "loss": 0.0153,
+      "step": 2712
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.6903488372093023e-07,
+      "loss": 0.0221,
+      "step": 2713
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.6902325581395346e-07,
+      "loss": 0.0234,
+      "step": 2714
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.690116279069767e-07,
+      "loss": 0.0151,
+      "step": 2715
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 2.69e-07,
+      "loss": 0.014,
+      "step": 2716
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.689883720930233e-07,
+      "loss": 1.3365,
+      "step": 2717
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.689767441860465e-07,
+      "loss": 0.0111,
+      "step": 2718
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.6896511627906975e-07,
+      "loss": 0.023,
+      "step": 2719
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.68953488372093e-07,
+      "loss": 0.0105,
+      "step": 2720
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.6894186046511627e-07,
+      "loss": 0.0227,
+      "step": 2721
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.689302325581395e-07,
+      "loss": 0.0225,
+      "step": 2722
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 2.689186046511628e-07,
+      "loss": 0.0135,
+      "step": 2723
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 2.6890697674418603e-07,
+      "loss": 0.0113,
+      "step": 2724
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 2.688953488372093e-07,
+      "loss": 0.0179,
+      "step": 2725
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 2.6888372093023255e-07,
+      "loss": 0.0103,
+      "step": 2726
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 2.688720930232558e-07,
+      "loss": 0.0119,
+      "step": 2727
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.68860465116279e-07,
+      "loss": 0.0125,
+      "step": 2728
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.688488372093023e-07,
+      "loss": 0.011,
+      "step": 2729
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.688372093023256e-07,
+      "loss": 1.8853,
+      "step": 2730
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.6882558139534883e-07,
+      "loss": 0.0121,
+      "step": 2731
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 2.6881395348837207e-07,
+      "loss": 2.1897,
+      "step": 2732
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.6880232558139536e-07,
+      "loss": 0.0145,
+      "step": 2733
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.687906976744186e-07,
+      "loss": 1.0864,
+      "step": 2734
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.6877906976744183e-07,
+      "loss": 0.0137,
+      "step": 2735
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.687674418604651e-07,
+      "loss": 0.0194,
+      "step": 2736
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 2.6875581395348835e-07,
+      "loss": 0.595,
+      "step": 2737
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 2.6874418604651164e-07,
+      "loss": 0.0076,
+      "step": 2738
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 2.687325581395349e-07,
+      "loss": 0.0186,
+      "step": 2739
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 2.687209302325581e-07,
+      "loss": 0.0227,
+      "step": 2740
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 2.687093023255814e-07,
+      "loss": 0.0147,
+      "step": 2741
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 2.6869767441860463e-07,
+      "loss": 0.1279,
+      "step": 2742
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.686860465116279e-07,
+      "loss": 0.0107,
+      "step": 2743
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.6867441860465116e-07,
+      "loss": 0.0179,
+      "step": 2744
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.686627906976744e-07,
+      "loss": 0.0648,
+      "step": 2745
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.686511627906977e-07,
+      "loss": 0.01,
+      "step": 2746
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2.686395348837209e-07,
+      "loss": 0.0091,
+      "step": 2747
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.6862790697674415e-07,
+      "loss": 0.0138,
+      "step": 2748
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.6861627906976744e-07,
+      "loss": 0.0297,
+      "step": 2749
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.6860465116279067e-07,
+      "loss": 0.0128,
+      "step": 2750
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.6859302325581396e-07,
+      "loss": 0.0123,
+      "step": 2751
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.685813953488372e-07,
+      "loss": 0.0132,
+      "step": 2752
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 2.6856976744186043e-07,
+      "loss": 0.0168,
+      "step": 2753
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 2.685581395348837e-07,
+      "loss": 0.0106,
+      "step": 2754
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 2.6854651162790696e-07,
+      "loss": 0.0098,
+      "step": 2755
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 2.685348837209302e-07,
+      "loss": 0.0119,
+      "step": 2756
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 2.685232558139535e-07,
+      "loss": 1.3748,
+      "step": 2757
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 2.6851162790697677e-07,
+      "loss": 0.0208,
+      "step": 2758
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.685e-07,
+      "loss": 0.0091,
+      "step": 2759
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.6848837209302324e-07,
+      "loss": 0.0159,
+      "step": 2760
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.6847674418604647e-07,
+      "loss": 0.0128,
+      "step": 2761
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.6846511627906976e-07,
+      "loss": 0.0142,
+      "step": 2762
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 2.68453488372093e-07,
+      "loss": 0.2679,
+      "step": 2763
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.684418604651163e-07,
+      "loss": 0.0181,
+      "step": 2764
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.684302325581395e-07,
+      "loss": 0.0165,
+      "step": 2765
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.684186046511628e-07,
+      "loss": 0.0178,
+      "step": 2766
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.6840697674418604e-07,
+      "loss": 0.0104,
+      "step": 2767
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 2.683953488372093e-07,
+      "loss": 0.011,
+      "step": 2768
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2.683837209302325e-07,
+      "loss": 0.0142,
+      "step": 2769
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2.683720930232558e-07,
+      "loss": 0.0628,
+      "step": 2770
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2.683604651162791e-07,
+      "loss": 0.0364,
+      "step": 2771
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2.683488372093023e-07,
+      "loss": 0.0166,
+      "step": 2772
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2.6833720930232556e-07,
+      "loss": 0.0172,
+      "step": 2773
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 2.683255813953488e-07,
+      "loss": 0.0187,
+      "step": 2774
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 2.683139534883721e-07,
+      "loss": 0.0106,
+      "step": 2775
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 2.683023255813953e-07,
+      "loss": 0.0281,
+      "step": 2776
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 2.682906976744186e-07,
+      "loss": 0.0125,
+      "step": 2777
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 2.6827906976744184e-07,
+      "loss": 0.0221,
+      "step": 2778
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.6826744186046513e-07,
+      "loss": 0.0129,
+      "step": 2779
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.6825581395348837e-07,
+      "loss": 0.0143,
+      "step": 2780
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.682441860465116e-07,
+      "loss": 0.0228,
+      "step": 2781
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.6823255813953484e-07,
+      "loss": 0.0123,
+      "step": 2782
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.682209302325581e-07,
+      "loss": 0.1218,
+      "step": 2783
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 2.682093023255814e-07,
+      "loss": 0.347,
+      "step": 2784
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 2.6819767441860465e-07,
+      "loss": 0.0138,
+      "step": 2785
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 2.681860465116279e-07,
+      "loss": 0.0166,
+      "step": 2786
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 2.6817441860465117e-07,
+      "loss": 0.017,
+      "step": 2787
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 2.681627906976744e-07,
+      "loss": 0.0442,
+      "step": 2788
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 2.6815116279069764e-07,
+      "loss": 0.0139,
+      "step": 2789
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 2.6813953488372093e-07,
+      "loss": 0.0142,
+      "step": 2790
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 2.6812790697674417e-07,
+      "loss": 0.0218,
+      "step": 2791
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 2.6811627906976745e-07,
+      "loss": 0.0264,
+      "step": 2792
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 2.681046511627907e-07,
+      "loss": 0.0097,
+      "step": 2793
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 2.680930232558139e-07,
+      "loss": 0.0138,
+      "step": 2794
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.680813953488372e-07,
+      "loss": 0.0318,
+      "step": 2795
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.6806976744186045e-07,
+      "loss": 0.0241,
+      "step": 2796
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.6805813953488374e-07,
+      "loss": 0.0135,
+      "step": 2797
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.6804651162790697e-07,
+      "loss": 0.1199,
+      "step": 2798
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2.680348837209302e-07,
+      "loss": 0.0132,
+      "step": 2799
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.680232558139535e-07,
+      "loss": 0.015,
+      "step": 2800
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.6801162790697673e-07,
+      "loss": 0.0269,
+      "step": 2801
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.6799999999999996e-07,
+      "loss": 0.0289,
+      "step": 2802
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.6798837209302325e-07,
+      "loss": 0.0144,
+      "step": 2803
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 2.679767441860465e-07,
+      "loss": 0.3609,
+      "step": 2804
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 2.679651162790698e-07,
+      "loss": 0.0106,
+      "step": 2805
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 2.67953488372093e-07,
+      "loss": 0.2812,
+      "step": 2806
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 2.6794186046511625e-07,
+      "loss": 0.0162,
+      "step": 2807
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 2.6793023255813953e-07,
+      "loss": 0.0137,
+      "step": 2808
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 2.6791860465116277e-07,
+      "loss": 0.0113,
+      "step": 2809
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.67906976744186e-07,
+      "loss": 0.0158,
+      "step": 2810
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.678953488372093e-07,
+      "loss": 0.0127,
+      "step": 2811
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.678837209302326e-07,
+      "loss": 0.012,
+      "step": 2812
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.678720930232558e-07,
+      "loss": 0.0091,
+      "step": 2813
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.6786046511627905e-07,
+      "loss": 0.0117,
+      "step": 2814
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 2.678488372093023e-07,
+      "loss": 0.0106,
+      "step": 2815
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 2.678372093023256e-07,
+      "loss": 0.0088,
+      "step": 2816
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 2.678255813953488e-07,
+      "loss": 0.0302,
+      "step": 2817
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 2.678139534883721e-07,
+      "loss": 0.0211,
+      "step": 2818
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 2.6780232558139533e-07,
+      "loss": 0.0097,
+      "step": 2819
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 2.677906976744186e-07,
+      "loss": 0.5264,
+      "step": 2820
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2.6777906976744186e-07,
+      "loss": 0.114,
+      "step": 2821
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2.677674418604651e-07,
+      "loss": 0.0176,
+      "step": 2822
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2.6775581395348833e-07,
+      "loss": 0.0249,
+      "step": 2823
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2.677441860465116e-07,
+      "loss": 0.0108,
+      "step": 2824
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2.677325581395349e-07,
+      "loss": 0.0292,
+      "step": 2825
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 2.6772093023255814e-07,
+      "loss": 0.9361,
+      "step": 2826
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 2.677093023255814e-07,
+      "loss": 0.0103,
+      "step": 2827
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 2.676976744186046e-07,
+      "loss": 0.0084,
+      "step": 2828
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 2.676860465116279e-07,
+      "loss": 0.0275,
+      "step": 2829
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 2.6767441860465113e-07,
+      "loss": 0.0184,
+      "step": 2830
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.676627906976744e-07,
+      "loss": 0.0143,
+      "step": 2831
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.6765116279069766e-07,
+      "loss": 0.0142,
+      "step": 2832
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.6763953488372095e-07,
+      "loss": 0.4817,
+      "step": 2833
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.676279069767442e-07,
+      "loss": 0.0084,
+      "step": 2834
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 2.676162790697674e-07,
+      "loss": 0.0107,
+      "step": 2835
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 2.6760465116279065e-07,
+      "loss": 0.0159,
+      "step": 2836
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 2.6759302325581394e-07,
+      "loss": 0.0149,
+      "step": 2837
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 2.6758139534883723e-07,
+      "loss": 0.0119,
+      "step": 2838
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 2.6756976744186046e-07,
+      "loss": 0.0151,
+      "step": 2839
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 2.675581395348837e-07,
+      "loss": 0.0127,
+      "step": 2840
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.67546511627907e-07,
+      "loss": 0.0129,
+      "step": 2841
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.675348837209302e-07,
+      "loss": 0.0162,
+      "step": 2842
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.6752325581395346e-07,
+      "loss": 0.0114,
+      "step": 2843
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.6751162790697674e-07,
+      "loss": 0.0393,
+      "step": 2844
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.675e-07,
+      "loss": 0.008,
+      "step": 2845
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 2.6748837209302327e-07,
+      "loss": 0.0177,
+      "step": 2846
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.674767441860465e-07,
+      "loss": 0.0147,
+      "step": 2847
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.6746511627906974e-07,
+      "loss": 0.0071,
+      "step": 2848
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.67453488372093e-07,
+      "loss": 0.037,
+      "step": 2849
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.6744186046511626e-07,
+      "loss": 0.4378,
+      "step": 2850
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2.674302325581395e-07,
+      "loss": 0.0109,
+      "step": 2851
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 2.674186046511628e-07,
+      "loss": 0.0087,
+      "step": 2852
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 2.67406976744186e-07,
+      "loss": 0.0468,
+      "step": 2853
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 2.673953488372093e-07,
+      "loss": 0.0112,
+      "step": 2854
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 2.6738372093023254e-07,
+      "loss": 0.0127,
+      "step": 2855
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 2.673720930232558e-07,
+      "loss": 0.5412,
+      "step": 2856
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 2.6736046511627907e-07,
+      "loss": 0.0128,
+      "step": 2857
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 2.673488372093023e-07,
+      "loss": 0.0195,
+      "step": 2858
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 2.673372093023256e-07,
+      "loss": 0.136,
+      "step": 2859
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 2.673255813953488e-07,
+      "loss": 0.0082,
+      "step": 2860
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 2.6731395348837206e-07,
+      "loss": 0.0123,
+      "step": 2861
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.6730232558139535e-07,
+      "loss": 0.6694,
+      "step": 2862
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.672906976744186e-07,
+      "loss": 0.0159,
+      "step": 2863
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.672790697674418e-07,
+      "loss": 0.0571,
+      "step": 2864
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.672674418604651e-07,
+      "loss": 0.218,
+      "step": 2865
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 2.672558139534884e-07,
+      "loss": 0.0129,
+      "step": 2866
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.6724418604651163e-07,
+      "loss": 0.0307,
+      "step": 2867
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.6723255813953487e-07,
+      "loss": 2.2489,
+      "step": 2868
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.672209302325581e-07,
+      "loss": 0.0142,
+      "step": 2869
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.672093023255814e-07,
+      "loss": 0.0142,
+      "step": 2870
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 2.671976744186046e-07,
+      "loss": 0.0129,
+      "step": 2871
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.671860465116279e-07,
+      "loss": 0.0912,
+      "step": 2872
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.6717441860465115e-07,
+      "loss": 0.011,
+      "step": 2873
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.6716279069767444e-07,
+      "loss": 0.0142,
+      "step": 2874
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.6715116279069767e-07,
+      "loss": 0.0131,
+      "step": 2875
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.671395348837209e-07,
+      "loss": 0.0436,
+      "step": 2876
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2.6712790697674414e-07,
+      "loss": 0.0116,
+      "step": 2877
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.6711627906976743e-07,
+      "loss": 0.0146,
+      "step": 2878
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.671046511627907e-07,
+      "loss": 0.0081,
+      "step": 2879
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.6709302325581395e-07,
+      "loss": 0.0084,
+      "step": 2880
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.670813953488372e-07,
+      "loss": 0.0091,
+      "step": 2881
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 2.670697674418604e-07,
+      "loss": 0.0243,
+      "step": 2882
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 2.670581395348837e-07,
+      "loss": 0.0168,
+      "step": 2883
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 2.6704651162790695e-07,
+      "loss": 0.0479,
+      "step": 2884
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 2.6703488372093024e-07,
+      "loss": 1.3792,
+      "step": 2885
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 2.6702325581395347e-07,
+      "loss": 0.011,
+      "step": 2886
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 2.6701162790697676e-07,
+      "loss": 0.0177,
+      "step": 2887
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 2.67e-07,
+      "loss": 0.0114,
+      "step": 2888
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 2.6698837209302323e-07,
+      "loss": 0.0165,
+      "step": 2889
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 2.6697674418604647e-07,
+      "loss": 0.0252,
+      "step": 2890
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 2.6696511627906975e-07,
+      "loss": 0.0078,
+      "step": 2891
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 2.6695348837209304e-07,
+      "loss": 0.0144,
+      "step": 2892
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 2.669418604651163e-07,
+      "loss": 0.0163,
+      "step": 2893
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 2.669302325581395e-07,
+      "loss": 0.0585,
+      "step": 2894
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 2.669186046511628e-07,
+      "loss": 0.0156,
+      "step": 2895
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 2.6690697674418604e-07,
+      "loss": 0.0101,
+      "step": 2896
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 2.6689534883720927e-07,
+      "loss": 0.0096,
+      "step": 2897
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.6688372093023256e-07,
+      "loss": 0.0138,
+      "step": 2898
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.668720930232558e-07,
+      "loss": 0.0099,
+      "step": 2899
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.668604651162791e-07,
+      "loss": 1.5995,
+      "step": 2900
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.668488372093023e-07,
+      "loss": 0.0108,
+      "step": 2901
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 2.6683720930232555e-07,
+      "loss": 0.0124,
+      "step": 2902
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.668255813953488e-07,
+      "loss": 0.0133,
+      "step": 2903
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.668139534883721e-07,
+      "loss": 0.0259,
+      "step": 2904
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.668023255813953e-07,
+      "loss": 0.0143,
+      "step": 2905
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.667906976744186e-07,
+      "loss": 0.0156,
+      "step": 2906
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.6677906976744184e-07,
+      "loss": 0.0129,
+      "step": 2907
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2.667674418604651e-07,
+      "loss": 0.1777,
+      "step": 2908
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 2.6675581395348836e-07,
+      "loss": 0.0232,
+      "step": 2909
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 2.667441860465116e-07,
+      "loss": 0.2719,
+      "step": 2910
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 2.667325581395349e-07,
+      "loss": 0.0136,
+      "step": 2911
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 2.667209302325581e-07,
+      "loss": 0.0132,
+      "step": 2912
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 2.667093023255814e-07,
+      "loss": 0.0156,
+      "step": 2913
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.6669767441860464e-07,
+      "loss": 0.0107,
+      "step": 2914
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.666860465116279e-07,
+      "loss": 0.0186,
+      "step": 2915
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.6667441860465116e-07,
+      "loss": 0.0157,
+      "step": 2916
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.666627906976744e-07,
+      "loss": 0.0223,
+      "step": 2917
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 2.6665116279069763e-07,
+      "loss": 0.0112,
+      "step": 2918
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 2.666395348837209e-07,
+      "loss": 0.0126,
+      "step": 2919
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 2.666279069767442e-07,
+      "loss": 0.0103,
+      "step": 2920
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 2.6661627906976745e-07,
+      "loss": 0.0276,
+      "step": 2921
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 2.666046511627907e-07,
+      "loss": 0.0975,
+      "step": 2922
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 2.665930232558139e-07,
+      "loss": 0.0109,
+      "step": 2923
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 2.665813953488372e-07,
+      "loss": 0.0101,
+      "step": 2924
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 2.6656976744186044e-07,
+      "loss": 0.5298,
+      "step": 2925
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 2.6655813953488373e-07,
+      "loss": 0.0092,
+      "step": 2926
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 2.6654651162790696e-07,
+      "loss": 0.0095,
+      "step": 2927
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 2.6653488372093025e-07,
+      "loss": 0.0081,
+      "step": 2928
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.665232558139535e-07,
+      "loss": 0.0167,
+      "step": 2929
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.665116279069767e-07,
+      "loss": 0.0153,
+      "step": 2930
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.6649999999999996e-07,
+      "loss": 0.0153,
+      "step": 2931
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.6648837209302325e-07,
+      "loss": 0.019,
+      "step": 2932
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2.6647674418604653e-07,
+      "loss": 0.0152,
+      "step": 2933
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.6646511627906977e-07,
+      "loss": 0.042,
+      "step": 2934
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.66453488372093e-07,
+      "loss": 0.0102,
+      "step": 2935
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.6644186046511624e-07,
+      "loss": 0.0254,
+      "step": 2936
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.6643023255813953e-07,
+      "loss": 0.0203,
+      "step": 2937
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.6641860465116276e-07,
+      "loss": 0.0368,
+      "step": 2938
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 2.6640697674418605e-07,
+      "loss": 0.0088,
+      "step": 2939
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 2.663953488372093e-07,
+      "loss": 0.0092,
+      "step": 2940
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 2.663837209302326e-07,
+      "loss": 1.5703,
+      "step": 2941
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 2.663720930232558e-07,
+      "loss": 0.0086,
+      "step": 2942
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 2.6636046511627905e-07,
+      "loss": 0.0077,
+      "step": 2943
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 2.663488372093023e-07,
+      "loss": 0.0151,
+      "step": 2944
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 2.6633720930232557e-07,
+      "loss": 0.1626,
+      "step": 2945
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 2.6632558139534886e-07,
+      "loss": 0.0078,
+      "step": 2946
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 2.663139534883721e-07,
+      "loss": 0.0153,
+      "step": 2947
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 2.6630232558139533e-07,
+      "loss": 0.1178,
+      "step": 2948
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 2.662906976744186e-07,
+      "loss": 0.0097,
+      "step": 2949
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.6627906976744185e-07,
+      "loss": 0.0121,
+      "step": 2950
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.662674418604651e-07,
+      "loss": 0.0115,
+      "step": 2951
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.662558139534884e-07,
+      "loss": 0.0108,
+      "step": 2952
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.662441860465116e-07,
+      "loss": 0.0191,
+      "step": 2953
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 2.662325581395349e-07,
+      "loss": 2.3836,
+      "step": 2954
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2.6622093023255813e-07,
+      "loss": 0.0454,
+      "step": 2955
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2.6620930232558137e-07,
+      "loss": 0.0118,
+      "step": 2956
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2.661976744186046e-07,
+      "loss": 0.0106,
+      "step": 2957
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2.661860465116279e-07,
+      "loss": 0.0107,
+      "step": 2958
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2.6617441860465113e-07,
+      "loss": 0.0082,
+      "step": 2959
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 2.661627906976744e-07,
+      "loss": 0.013,
+      "step": 2960
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 2.6615116279069765e-07,
+      "loss": 0.0137,
+      "step": 2961
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 2.6613953488372094e-07,
+      "loss": 0.0452,
+      "step": 2962
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 2.661279069767442e-07,
+      "loss": 0.0423,
+      "step": 2963
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 2.661162790697674e-07,
+      "loss": 0.0193,
+      "step": 2964
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.661046511627907e-07,
+      "loss": 0.0083,
+      "step": 2965
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.6609302325581393e-07,
+      "loss": 0.0094,
+      "step": 2966
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.660813953488372e-07,
+      "loss": 0.0119,
+      "step": 2967
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.6606976744186046e-07,
+      "loss": 0.0315,
+      "step": 2968
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.660581395348837e-07,
+      "loss": 0.0089,
+      "step": 2969
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 2.66046511627907e-07,
+      "loss": 0.0159,
+      "step": 2970
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 2.660348837209302e-07,
+      "loss": 0.0165,
+      "step": 2971
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 2.6602325581395345e-07,
+      "loss": 0.015,
+      "step": 2972
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 2.6601162790697674e-07,
+      "loss": 0.0111,
+      "step": 2973
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 2.66e-07,
+      "loss": 0.0169,
+      "step": 2974
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 2.6598837209302326e-07,
+      "loss": 0.0077,
+      "step": 2975
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 2.659767441860465e-07,
+      "loss": 0.0166,
+      "step": 2976
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 2.6596511627906973e-07,
+      "loss": 2.2451,
+      "step": 2977
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 2.65953488372093e-07,
+      "loss": 0.0093,
+      "step": 2978
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 2.6594186046511626e-07,
+      "loss": 0.0917,
+      "step": 2979
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 2.6593023255813954e-07,
+      "loss": 0.2622,
+      "step": 2980
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.659186046511628e-07,
+      "loss": 0.0421,
+      "step": 2981
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.65906976744186e-07,
+      "loss": 0.0125,
+      "step": 2982
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.658953488372093e-07,
+      "loss": 0.0875,
+      "step": 2983
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.6588372093023254e-07,
+      "loss": 0.0106,
+      "step": 2984
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2.6587209302325577e-07,
+      "loss": 0.0086,
+      "step": 2985
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.6586046511627906e-07,
+      "loss": 0.0203,
+      "step": 2986
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.6584883720930235e-07,
+      "loss": 0.0168,
+      "step": 2987
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.658372093023256e-07,
+      "loss": 0.0101,
+      "step": 2988
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.658255813953488e-07,
+      "loss": 0.0116,
+      "step": 2989
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 2.6581395348837205e-07,
+      "loss": 0.0081,
+      "step": 2990
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.6580232558139534e-07,
+      "loss": 0.0082,
+      "step": 2991
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.657906976744186e-07,
+      "loss": 0.0192,
+      "step": 2992
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.6577906976744187e-07,
+      "loss": 0.0098,
+      "step": 2993
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.657674418604651e-07,
+      "loss": 0.0114,
+      "step": 2994
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.657558139534884e-07,
+      "loss": 0.0169,
+      "step": 2995
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 2.657441860465116e-07,
+      "loss": 1.4897,
+      "step": 2996
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.6573255813953486e-07,
+      "loss": 0.0077,
+      "step": 2997
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.657209302325581e-07,
+      "loss": 0.0078,
+      "step": 2998
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.657093023255814e-07,
+      "loss": 0.025,
+      "step": 2999
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.656976744186046e-07,
+      "loss": 0.0137,
+      "step": 3000
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 2.656860465116279e-07,
+      "loss": 0.0125,
+      "step": 3001
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 2.6567441860465114e-07,
+      "loss": 0.0134,
+      "step": 3002
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 2.6566279069767443e-07,
+      "loss": 0.0147,
+      "step": 3003
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 2.6565116279069767e-07,
+      "loss": 0.1354,
+      "step": 3004
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 2.656395348837209e-07,
+      "loss": 0.0103,
+      "step": 3005
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 2.656279069767442e-07,
+      "loss": 0.011,
+      "step": 3006
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2.656162790697674e-07,
+      "loss": 0.0116,
+      "step": 3007
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2.656046511627907e-07,
+      "loss": 0.0109,
+      "step": 3008
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2.6559302325581395e-07,
+      "loss": 0.01,
+      "step": 3009
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2.655813953488372e-07,
+      "loss": 0.0119,
+      "step": 3010
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2.655697674418604e-07,
+      "loss": 0.0097,
+      "step": 3011
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.655581395348837e-07,
+      "loss": 0.0084,
+      "step": 3012
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.6554651162790694e-07,
+      "loss": 0.0149,
+      "step": 3013
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.6553488372093023e-07,
+      "loss": 0.0135,
+      "step": 3014
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.6552325581395347e-07,
+      "loss": 0.0098,
+      "step": 3015
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 2.6551162790697675e-07,
+      "loss": 0.0137,
+      "step": 3016
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.655e-07,
+      "loss": 0.0114,
+      "step": 3017
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.654883720930232e-07,
+      "loss": 0.0092,
+      "step": 3018
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.654767441860465e-07,
+      "loss": 0.0173,
+      "step": 3019
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.6546511627906975e-07,
+      "loss": 0.0087,
+      "step": 3020
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 2.6545348837209304e-07,
+      "loss": 0.0149,
+      "step": 3021
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.6544186046511627e-07,
+      "loss": 0.0132,
+      "step": 3022
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.654302325581395e-07,
+      "loss": 0.0107,
+      "step": 3023
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.654186046511628e-07,
+      "loss": 0.0125,
+      "step": 3024
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.6540697674418603e-07,
+      "loss": 0.009,
+      "step": 3025
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.6539534883720926e-07,
+      "loss": 0.0082,
+      "step": 3026
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 2.6538372093023255e-07,
+      "loss": 0.0189,
+      "step": 3027
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 2.6537209302325584e-07,
+      "loss": 0.0489,
+      "step": 3028
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 2.653604651162791e-07,
+      "loss": 0.0109,
+      "step": 3029
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 2.653488372093023e-07,
+      "loss": 0.0126,
+      "step": 3030
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 2.6533720930232555e-07,
+      "loss": 0.0123,
+      "step": 3031
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 2.6532558139534883e-07,
+      "loss": 0.4152,
+      "step": 3032
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.6531395348837207e-07,
+      "loss": 0.0107,
+      "step": 3033
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.6530232558139536e-07,
+      "loss": 0.0154,
+      "step": 3034
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.652906976744186e-07,
+      "loss": 0.0254,
+      "step": 3035
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.6527906976744183e-07,
+      "loss": 0.0176,
+      "step": 3036
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2.652674418604651e-07,
+      "loss": 0.016,
+      "step": 3037
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 2.6525581395348835e-07,
+      "loss": 0.0105,
+      "step": 3038
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 2.652441860465116e-07,
+      "loss": 0.0331,
+      "step": 3039
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 2.652325581395349e-07,
+      "loss": 0.0134,
+      "step": 3040
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 2.6522093023255816e-07,
+      "loss": 0.289,
+      "step": 3041
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 2.652093023255814e-07,
+      "loss": 0.0145,
+      "step": 3042
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 2.6519767441860463e-07,
+      "loss": 0.1514,
+      "step": 3043
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 2.6518604651162787e-07,
+      "loss": 0.0132,
+      "step": 3044
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 2.6517441860465116e-07,
+      "loss": 0.0137,
+      "step": 3045
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 2.651627906976744e-07,
+      "loss": 0.0117,
+      "step": 3046
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 2.651511627906977e-07,
+      "loss": 0.0134,
+      "step": 3047
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.651395348837209e-07,
+      "loss": 0.0069,
+      "step": 3048
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.651279069767442e-07,
+      "loss": 0.0083,
+      "step": 3049
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.6511627906976744e-07,
+      "loss": 0.0064,
+      "step": 3050
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.651046511627907e-07,
+      "loss": 0.0135,
+      "step": 3051
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 2.650930232558139e-07,
+      "loss": 0.0159,
+      "step": 3052
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.650813953488372e-07,
+      "loss": 0.0102,
+      "step": 3053
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.6506976744186043e-07,
+      "loss": 0.0106,
+      "step": 3054
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.650581395348837e-07,
+      "loss": 0.0144,
+      "step": 3055
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.6504651162790696e-07,
+      "loss": 0.0125,
+      "step": 3056
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.6503488372093025e-07,
+      "loss": 0.0202,
+      "step": 3057
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 2.650232558139535e-07,
+      "loss": 0.435,
+      "step": 3058
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2.650116279069767e-07,
+      "loss": 0.0282,
+      "step": 3059
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2.65e-07,
+      "loss": 0.0084,
+      "step": 3060
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2.6498837209302324e-07,
+      "loss": 0.0477,
+      "step": 3061
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2.6497674418604653e-07,
+      "loss": 0.0165,
+      "step": 3062
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2.6496511627906976e-07,
+      "loss": 1.7256,
+      "step": 3063
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.64953488372093e-07,
+      "loss": 0.0102,
+      "step": 3064
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.6494186046511623e-07,
+      "loss": 0.0383,
+      "step": 3065
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.649302325581395e-07,
+      "loss": 0.0077,
+      "step": 3066
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.6491860465116276e-07,
+      "loss": 0.007,
+      "step": 3067
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 2.6490697674418604e-07,
+      "loss": 0.0105,
+      "step": 3068
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.648953488372093e-07,
+      "loss": 0.0087,
+      "step": 3069
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.6488372093023257e-07,
+      "loss": 0.0075,
+      "step": 3070
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.648720930232558e-07,
+      "loss": 0.0244,
+      "step": 3071
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.6486046511627904e-07,
+      "loss": 0.0079,
+      "step": 3072
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 2.6484883720930233e-07,
+      "loss": 0.0081,
+      "step": 3073
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.6483720930232556e-07,
+      "loss": 0.0119,
+      "step": 3074
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.6482558139534885e-07,
+      "loss": 0.0145,
+      "step": 3075
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.648139534883721e-07,
+      "loss": 0.1029,
+      "step": 3076
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.648023255813953e-07,
+      "loss": 0.0123,
+      "step": 3077
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 2.647906976744186e-07,
+      "loss": 0.052,
+      "step": 3078
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 2.6477906976744184e-07,
+      "loss": 0.0105,
+      "step": 3079
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 2.647674418604651e-07,
+      "loss": 0.0092,
+      "step": 3080
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 2.6475581395348837e-07,
+      "loss": 0.0119,
+      "step": 3081
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 2.6474418604651166e-07,
+      "loss": 0.0145,
+      "step": 3082
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 2.647325581395349e-07,
+      "loss": 0.0131,
+      "step": 3083
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.647209302325581e-07,
+      "loss": 0.011,
+      "step": 3084
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.6470930232558136e-07,
+      "loss": 0.0195,
+      "step": 3085
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.6469767441860465e-07,
+      "loss": 0.0094,
+      "step": 3086
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.646860465116279e-07,
+      "loss": 0.0161,
+      "step": 3087
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.6467441860465117e-07,
+      "loss": 0.0142,
+      "step": 3088
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 2.646627906976744e-07,
+      "loss": 0.0061,
+      "step": 3089
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2.6465116279069764e-07,
+      "loss": 0.0071,
+      "step": 3090
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2.6463953488372093e-07,
+      "loss": 0.2305,
+      "step": 3091
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2.6462790697674417e-07,
+      "loss": 0.0073,
+      "step": 3092
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2.646162790697674e-07,
+      "loss": 0.0092,
+      "step": 3093
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2.646046511627907e-07,
+      "loss": 0.016,
+      "step": 3094
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.645930232558139e-07,
+      "loss": 0.0115,
+      "step": 3095
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.645813953488372e-07,
+      "loss": 0.0095,
+      "step": 3096
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.6456976744186045e-07,
+      "loss": 0.0227,
+      "step": 3097
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.645581395348837e-07,
+      "loss": 0.0196,
+      "step": 3098
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 2.6454651162790697e-07,
+      "loss": 0.0091,
+      "step": 3099
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 2.645348837209302e-07,
+      "loss": 0.011,
+      "step": 3100
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 2.645232558139535e-07,
+      "loss": 0.0102,
+      "step": 3101
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 2.6451162790697673e-07,
+      "loss": 0.0102,
+      "step": 3102
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.9816247582205029,
+      "eval_loss": 0.1632627248764038,
+      "eval_runtime": 1.5691,
+      "eval_samples_per_second": 658.968,
+      "eval_steps_per_second": 10.834,
+      "step": 3102
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 2.645e-07,
+      "loss": 0.0129,
+      "step": 3103
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 2.6448837209302325e-07,
+      "loss": 0.0107,
+      "step": 3104
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 2.644767441860465e-07,
+      "loss": 0.0069,
+      "step": 3105
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 2.644651162790697e-07,
+      "loss": 0.0105,
+      "step": 3106
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 2.64453488372093e-07,
+      "loss": 0.0102,
+      "step": 3107
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 2.6444186046511625e-07,
+      "loss": 0.0158,
+      "step": 3108
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 2.6443023255813954e-07,
+      "loss": 0.0101,
+      "step": 3109
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 2.6441860465116277e-07,
+      "loss": 0.0093,
+      "step": 3110
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 2.6440697674418606e-07,
+      "loss": 0.0122,
+      "step": 3111
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 2.643953488372093e-07,
+      "loss": 0.0076,
+      "step": 3112
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 2.6438372093023253e-07,
+      "loss": 0.0095,
+      "step": 3113
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 2.643720930232558e-07,
+      "loss": 0.0094,
+      "step": 3114
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2.6436046511627905e-07,
+      "loss": 0.0137,
+      "step": 3115
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2.6434883720930234e-07,
+      "loss": 0.0074,
+      "step": 3116
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2.643372093023256e-07,
+      "loss": 0.166,
+      "step": 3117
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2.643255813953488e-07,
+      "loss": 0.022,
+      "step": 3118
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2.6431395348837205e-07,
+      "loss": 0.0145,
+      "step": 3119
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2.6430232558139534e-07,
+      "loss": 0.0356,
+      "step": 3120
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 2.6429069767441857e-07,
+      "loss": 0.0153,
+      "step": 3121
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 2.6427906976744186e-07,
+      "loss": 0.0141,
+      "step": 3122
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 2.642674418604651e-07,
+      "loss": 0.0108,
+      "step": 3123
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 2.642558139534884e-07,
+      "loss": 0.0157,
+      "step": 3124
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 2.642441860465116e-07,
+      "loss": 0.0128,
+      "step": 3125
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 2.6423255813953485e-07,
+      "loss": 0.0126,
+      "step": 3126
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 2.642209302325581e-07,
+      "loss": 0.0113,
+      "step": 3127
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 2.642093023255814e-07,
+      "loss": 0.0104,
+      "step": 3128
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 2.6419767441860466e-07,
+      "loss": 0.009,
+      "step": 3129
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 2.641860465116279e-07,
+      "loss": 0.0134,
+      "step": 3130
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 2.6417441860465114e-07,
+      "loss": 0.0131,
+      "step": 3131
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 2.641627906976744e-07,
+      "loss": 0.0086,
+      "step": 3132
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 2.6415116279069766e-07,
+      "loss": 0.0073,
+      "step": 3133
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 2.641395348837209e-07,
+      "loss": 0.0134,
+      "step": 3134
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 2.641279069767442e-07,
+      "loss": 0.0195,
+      "step": 3135
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 2.6411627906976747e-07,
+      "loss": 0.0172,
+      "step": 3136
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 2.641046511627907e-07,
+      "loss": 0.049,
+      "step": 3137
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 2.6409302325581394e-07,
+      "loss": 0.0173,
+      "step": 3138
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 2.640813953488372e-07,
+      "loss": 0.0079,
+      "step": 3139
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 2.6406976744186046e-07,
+      "loss": 0.0127,
+      "step": 3140
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 2.640581395348837e-07,
+      "loss": 0.0073,
+      "step": 3141
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 2.64046511627907e-07,
+      "loss": 0.0112,
+      "step": 3142
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 2.640348837209302e-07,
+      "loss": 0.0132,
+      "step": 3143
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 2.6402325581395346e-07,
+      "loss": 0.0071,
+      "step": 3144
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 2.6401162790697675e-07,
+      "loss": 0.0115,
+      "step": 3145
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 2.64e-07,
+      "loss": 0.0103,
+      "step": 3146
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 2.639883720930232e-07,
+      "loss": 0.0078,
+      "step": 3147
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 2.639767441860465e-07,
+      "loss": 0.0089,
+      "step": 3148
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 2.6396511627906974e-07,
+      "loss": 0.0158,
+      "step": 3149
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 2.6395348837209303e-07,
+      "loss": 0.0109,
+      "step": 3150
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 2.6394186046511626e-07,
+      "loss": 0.0121,
+      "step": 3151
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 2.639302325581395e-07,
+      "loss": 0.0074,
+      "step": 3152
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 2.639186046511628e-07,
+      "loss": 0.0125,
+      "step": 3153
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 2.63906976744186e-07,
+      "loss": 0.0099,
+      "step": 3154
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 2.638953488372093e-07,
+      "loss": 0.01,
+      "step": 3155
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 2.6388372093023255e-07,
+      "loss": 0.4903,
+      "step": 3156
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 2.6387209302325583e-07,
+      "loss": 0.0394,
+      "step": 3157
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 2.6386046511627907e-07,
+      "loss": 0.1996,
+      "step": 3158
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 2.638488372093023e-07,
+      "loss": 0.0135,
+      "step": 3159
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 2.6383720930232554e-07,
+      "loss": 0.0089,
+      "step": 3160
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 2.6382558139534883e-07,
+      "loss": 0.0139,
+      "step": 3161
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 2.6381395348837206e-07,
+      "loss": 0.0073,
+      "step": 3162
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 2.6380232558139535e-07,
+      "loss": 0.0119,
+      "step": 3163
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 2.637906976744186e-07,
+      "loss": 0.0119,
+      "step": 3164
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 2.637790697674419e-07,
+      "loss": 0.6772,
+      "step": 3165
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 2.637674418604651e-07,
+      "loss": 0.0075,
+      "step": 3166
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 2.6375581395348835e-07,
+      "loss": 0.0078,
+      "step": 3167
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 2.6374418604651163e-07,
+      "loss": 0.0098,
+      "step": 3168
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 2.6373255813953487e-07,
+      "loss": 0.0151,
+      "step": 3169
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 2.6372093023255816e-07,
+      "loss": 0.0092,
+      "step": 3170
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 2.637093023255814e-07,
+      "loss": 0.0153,
+      "step": 3171
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 2.6369767441860463e-07,
+      "loss": 0.0093,
+      "step": 3172
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 2.6368604651162786e-07,
+      "loss": 0.0235,
+      "step": 3173
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 2.6367441860465115e-07,
+      "loss": 0.0116,
+      "step": 3174
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 2.636627906976744e-07,
+      "loss": 0.0081,
+      "step": 3175
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 2.636511627906977e-07,
+      "loss": 0.0085,
+      "step": 3176
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 2.636395348837209e-07,
+      "loss": 1.4302,
+      "step": 3177
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 2.636279069767442e-07,
+      "loss": 0.008,
+      "step": 3178
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 2.6361627906976743e-07,
+      "loss": 0.0081,
+      "step": 3179
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 2.6360465116279067e-07,
+      "loss": 0.0116,
+      "step": 3180
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 2.635930232558139e-07,
+      "loss": 0.0106,
+      "step": 3181
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 2.635813953488372e-07,
+      "loss": 0.103,
+      "step": 3182
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 2.635697674418605e-07,
+      "loss": 0.0089,
+      "step": 3183
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 2.635581395348837e-07,
+      "loss": 0.4876,
+      "step": 3184
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 2.6354651162790695e-07,
+      "loss": 0.0103,
+      "step": 3185
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 2.6353488372093024e-07,
+      "loss": 0.0085,
+      "step": 3186
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 2.635232558139535e-07,
+      "loss": 0.0086,
+      "step": 3187
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 2.635116279069767e-07,
+      "loss": 0.0115,
+      "step": 3188
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 2.635e-07,
+      "loss": 2.2244,
+      "step": 3189
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 2.634883720930233e-07,
+      "loss": 0.4162,
+      "step": 3190
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 2.634767441860465e-07,
+      "loss": 0.2591,
+      "step": 3191
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 2.6346511627906976e-07,
+      "loss": 0.0109,
+      "step": 3192
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 2.63453488372093e-07,
+      "loss": 0.0107,
+      "step": 3193
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 2.634418604651163e-07,
+      "loss": 0.0223,
+      "step": 3194
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 2.634302325581395e-07,
+      "loss": 0.1203,
+      "step": 3195
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 2.634186046511628e-07,
+      "loss": 0.0117,
+      "step": 3196
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 2.6340697674418604e-07,
+      "loss": 0.0083,
+      "step": 3197
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 2.6339534883720927e-07,
+      "loss": 0.0051,
+      "step": 3198
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 2.6338372093023256e-07,
+      "loss": 0.0115,
+      "step": 3199
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 2.633720930232558e-07,
+      "loss": 0.0074,
+      "step": 3200
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 2.6336046511627903e-07,
+      "loss": 0.0165,
+      "step": 3201
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 2.633488372093023e-07,
+      "loss": 1.6314,
+      "step": 3202
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 2.6333720930232556e-07,
+      "loss": 0.0113,
+      "step": 3203
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 2.6332558139534884e-07,
+      "loss": 0.0111,
+      "step": 3204
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 2.633139534883721e-07,
+      "loss": 0.0123,
+      "step": 3205
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 2.633023255813953e-07,
+      "loss": 0.013,
+      "step": 3206
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 2.632906976744186e-07,
+      "loss": 0.0101,
+      "step": 3207
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 2.6327906976744184e-07,
+      "loss": 0.0075,
+      "step": 3208
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 2.632674418604651e-07,
+      "loss": 1.3921,
+      "step": 3209
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 2.6325581395348836e-07,
+      "loss": 0.0089,
+      "step": 3210
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 2.6324418604651165e-07,
+      "loss": 1.4893,
+      "step": 3211
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 2.632325581395349e-07,
+      "loss": 0.0086,
+      "step": 3212
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 2.632209302325581e-07,
+      "loss": 0.0155,
+      "step": 3213
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 2.6320930232558135e-07,
+      "loss": 0.008,
+      "step": 3214
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 2.6319767441860464e-07,
+      "loss": 0.0099,
+      "step": 3215
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 2.631860465116279e-07,
+      "loss": 0.0076,
+      "step": 3216
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 2.6317441860465117e-07,
+      "loss": 0.0088,
+      "step": 3217
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 2.631627906976744e-07,
+      "loss": 0.0157,
+      "step": 3218
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 2.631511627906977e-07,
+      "loss": 0.0457,
+      "step": 3219
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 2.631395348837209e-07,
+      "loss": 0.0076,
+      "step": 3220
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 2.6312790697674416e-07,
+      "loss": 0.0144,
+      "step": 3221
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 2.631162790697674e-07,
+      "loss": 0.0116,
+      "step": 3222
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 2.631046511627907e-07,
+      "loss": 0.0179,
+      "step": 3223
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 2.6309302325581397e-07,
+      "loss": 0.0072,
+      "step": 3224
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 2.630813953488372e-07,
+      "loss": 0.009,
+      "step": 3225
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 2.6306976744186044e-07,
+      "loss": 1.0556,
+      "step": 3226
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 2.630581395348837e-07,
+      "loss": 0.0095,
+      "step": 3227
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 2.6304651162790697e-07,
+      "loss": 0.0084,
+      "step": 3228
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 2.630348837209302e-07,
+      "loss": 0.0097,
+      "step": 3229
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 2.630232558139535e-07,
+      "loss": 0.0076,
+      "step": 3230
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 2.630116279069767e-07,
+      "loss": 2.1927,
+      "step": 3231
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 2.63e-07,
+      "loss": 0.0078,
+      "step": 3232
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 2.6298837209302325e-07,
+      "loss": 0.0066,
+      "step": 3233
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 2.629767441860465e-07,
+      "loss": 0.0154,
+      "step": 3234
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 2.629651162790697e-07,
+      "loss": 0.0104,
+      "step": 3235
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 2.62953488372093e-07,
+      "loss": 0.0105,
+      "step": 3236
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 2.629418604651163e-07,
+      "loss": 0.0262,
+      "step": 3237
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 2.6293023255813953e-07,
+      "loss": 0.3005,
+      "step": 3238
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 2.6291860465116277e-07,
+      "loss": 0.0125,
+      "step": 3239
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 2.6290697674418605e-07,
+      "loss": 0.0092,
+      "step": 3240
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 2.628953488372093e-07,
+      "loss": 0.012,
+      "step": 3241
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 2.628837209302325e-07,
+      "loss": 0.0492,
+      "step": 3242
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 2.628720930232558e-07,
+      "loss": 0.0079,
+      "step": 3243
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 2.6286046511627905e-07,
+      "loss": 0.0153,
+      "step": 3244
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 2.6284883720930234e-07,
+      "loss": 0.0068,
+      "step": 3245
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 2.6283720930232557e-07,
+      "loss": 0.0126,
+      "step": 3246
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 2.628255813953488e-07,
+      "loss": 0.0061,
+      "step": 3247
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 2.628139534883721e-07,
+      "loss": 0.0125,
+      "step": 3248
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 2.6280232558139533e-07,
+      "loss": 0.0071,
+      "step": 3249
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 2.627906976744186e-07,
+      "loss": 0.0141,
+      "step": 3250
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 2.6277906976744185e-07,
+      "loss": 0.0112,
+      "step": 3251
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 2.627674418604651e-07,
+      "loss": 0.0071,
+      "step": 3252
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 2.627558139534884e-07,
+      "loss": 0.0087,
+      "step": 3253
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 2.627441860465116e-07,
+      "loss": 0.009,
+      "step": 3254
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 2.6273255813953485e-07,
+      "loss": 0.0081,
+      "step": 3255
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 2.6272093023255813e-07,
+      "loss": 1.0843,
+      "step": 3256
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 2.6270930232558137e-07,
+      "loss": 0.8701,
+      "step": 3257
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 2.6269767441860466e-07,
+      "loss": 0.2604,
+      "step": 3258
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 2.626860465116279e-07,
+      "loss": 0.0092,
+      "step": 3259
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 2.6267441860465113e-07,
+      "loss": 0.01,
+      "step": 3260
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 2.626627906976744e-07,
+      "loss": 0.1562,
+      "step": 3261
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 2.6265116279069765e-07,
+      "loss": 0.0886,
+      "step": 3262
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 2.6263953488372094e-07,
+      "loss": 0.0136,
+      "step": 3263
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 2.626279069767442e-07,
+      "loss": 0.0057,
+      "step": 3264
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 2.6261627906976746e-07,
+      "loss": 0.009,
+      "step": 3265
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 2.626046511627907e-07,
+      "loss": 0.0151,
+      "step": 3266
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 2.6259302325581393e-07,
+      "loss": 0.9944,
+      "step": 3267
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 2.6258139534883717e-07,
+      "loss": 0.0084,
+      "step": 3268
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 2.6256976744186046e-07,
+      "loss": 0.0071,
+      "step": 3269
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 2.625581395348837e-07,
+      "loss": 0.0118,
+      "step": 3270
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 2.62546511627907e-07,
+      "loss": 0.0138,
+      "step": 3271
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 2.625348837209302e-07,
+      "loss": 1.2628,
+      "step": 3272
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 2.625232558139535e-07,
+      "loss": 0.0056,
+      "step": 3273
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 2.6251162790697674e-07,
+      "loss": 0.01,
+      "step": 3274
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 2.625e-07,
+      "loss": 0.0102,
+      "step": 3275
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 2.624883720930232e-07,
+      "loss": 0.0497,
+      "step": 3276
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 2.624767441860465e-07,
+      "loss": 0.0116,
+      "step": 3277
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 2.624651162790698e-07,
+      "loss": 0.0054,
+      "step": 3278
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 2.62453488372093e-07,
+      "loss": 0.0098,
+      "step": 3279
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 2.6244186046511626e-07,
+      "loss": 0.0137,
+      "step": 3280
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 2.624302325581395e-07,
+      "loss": 0.931,
+      "step": 3281
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 2.624186046511628e-07,
+      "loss": 0.0091,
+      "step": 3282
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 2.62406976744186e-07,
+      "loss": 0.0112,
+      "step": 3283
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 2.623953488372093e-07,
+      "loss": 0.7381,
+      "step": 3284
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 2.6238372093023254e-07,
+      "loss": 0.0114,
+      "step": 3285
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 2.6237209302325583e-07,
+      "loss": 0.0093,
+      "step": 3286
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 2.6236046511627906e-07,
+      "loss": 0.0099,
+      "step": 3287
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 2.623488372093023e-07,
+      "loss": 0.0089,
+      "step": 3288
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 2.6233720930232553e-07,
+      "loss": 0.0124,
+      "step": 3289
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 2.623255813953488e-07,
+      "loss": 2.3695,
+      "step": 3290
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 2.623139534883721e-07,
+      "loss": 0.0107,
+      "step": 3291
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 2.6230232558139534e-07,
+      "loss": 0.0229,
+      "step": 3292
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 2.622906976744186e-07,
+      "loss": 0.3523,
+      "step": 3293
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 2.6227906976744187e-07,
+      "loss": 0.007,
+      "step": 3294
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 2.622674418604651e-07,
+      "loss": 0.0269,
+      "step": 3295
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 2.6225581395348834e-07,
+      "loss": 0.007,
+      "step": 3296
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 2.6224418604651163e-07,
+      "loss": 0.0225,
+      "step": 3297
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 2.6223255813953486e-07,
+      "loss": 0.0126,
+      "step": 3298
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 2.6222093023255815e-07,
+      "loss": 0.0066,
+      "step": 3299
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 2.622093023255814e-07,
+      "loss": 0.0071,
+      "step": 3300
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 2.621976744186046e-07,
+      "loss": 0.0084,
+      "step": 3301
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 2.6218604651162786e-07,
+      "loss": 0.0173,
+      "step": 3302
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 2.6217441860465114e-07,
+      "loss": 0.0093,
+      "step": 3303
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 2.6216279069767443e-07,
+      "loss": 0.0068,
+      "step": 3304
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 2.6215116279069767e-07,
+      "loss": 0.011,
+      "step": 3305
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 2.621395348837209e-07,
+      "loss": 0.0128,
+      "step": 3306
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 2.621279069767442e-07,
+      "loss": 0.0115,
+      "step": 3307
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 2.621162790697674e-07,
+      "loss": 0.0092,
+      "step": 3308
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 2.6210465116279066e-07,
+      "loss": 0.0113,
+      "step": 3309
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 2.6209302325581395e-07,
+      "loss": 0.0061,
+      "step": 3310
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 2.620813953488372e-07,
+      "loss": 0.0063,
+      "step": 3311
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 2.6206976744186047e-07,
+      "loss": 0.4264,
+      "step": 3312
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 2.620581395348837e-07,
+      "loss": 0.0084,
+      "step": 3313
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 2.6204651162790694e-07,
+      "loss": 0.0105,
+      "step": 3314
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 2.6203488372093023e-07,
+      "loss": 0.0144,
+      "step": 3315
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 2.6202325581395347e-07,
+      "loss": 0.0131,
+      "step": 3316
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 2.6201162790697676e-07,
+      "loss": 0.0106,
+      "step": 3317
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 2.62e-07,
+      "loss": 0.0073,
+      "step": 3318
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 2.619883720930233e-07,
+      "loss": 0.0158,
+      "step": 3319
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 2.619767441860465e-07,
+      "loss": 0.0062,
+      "step": 3320
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 2.6196511627906975e-07,
+      "loss": 0.1297,
+      "step": 3321
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 2.61953488372093e-07,
+      "loss": 0.008,
+      "step": 3322
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 2.6194186046511627e-07,
+      "loss": 0.0123,
+      "step": 3323
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 2.619302325581395e-07,
+      "loss": 0.0098,
+      "step": 3324
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 2.619186046511628e-07,
+      "loss": 0.0104,
+      "step": 3325
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 2.6190697674418603e-07,
+      "loss": 1.2269,
+      "step": 3326
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2.618953488372093e-07,
+      "loss": 0.0114,
+      "step": 3327
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2.6188372093023255e-07,
+      "loss": 0.0103,
+      "step": 3328
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2.618720930232558e-07,
+      "loss": 0.0145,
+      "step": 3329
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2.61860465116279e-07,
+      "loss": 0.0066,
+      "step": 3330
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2.618488372093023e-07,
+      "loss": 0.0072,
+      "step": 3331
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2.618372093023256e-07,
+      "loss": 0.0076,
+      "step": 3332
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 2.6182558139534884e-07,
+      "loss": 0.1302,
+      "step": 3333
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 2.6181395348837207e-07,
+      "loss": 0.0095,
+      "step": 3334
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 2.618023255813953e-07,
+      "loss": 0.0071,
+      "step": 3335
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 2.617906976744186e-07,
+      "loss": 0.0677,
+      "step": 3336
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 2.6177906976744183e-07,
+      "loss": 0.0061,
+      "step": 3337
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 2.617674418604651e-07,
+      "loss": 0.0075,
+      "step": 3338
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 2.6175581395348835e-07,
+      "loss": 0.0123,
+      "step": 3339
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 2.6174418604651164e-07,
+      "loss": 0.0097,
+      "step": 3340
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 2.617325581395349e-07,
+      "loss": 1.0497,
+      "step": 3341
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 2.617209302325581e-07,
+      "loss": 0.0084,
+      "step": 3342
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 2.6170930232558135e-07,
+      "loss": 0.0139,
+      "step": 3343
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 2.6169767441860464e-07,
+      "loss": 0.6558,
+      "step": 3344
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 2.616860465116279e-07,
+      "loss": 0.0093,
+      "step": 3345
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 2.6167441860465116e-07,
+      "loss": 0.0081,
+      "step": 3346
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 2.616627906976744e-07,
+      "loss": 0.0072,
+      "step": 3347
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 2.616511627906977e-07,
+      "loss": 0.0065,
+      "step": 3348
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 2.616395348837209e-07,
+      "loss": 0.0122,
+      "step": 3349
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 2.6162790697674415e-07,
+      "loss": 0.0075,
+      "step": 3350
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 2.6161627906976744e-07,
+      "loss": 0.0112,
+      "step": 3351
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 2.616046511627907e-07,
+      "loss": 0.0057,
+      "step": 3352
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 2.6159302325581396e-07,
+      "loss": 0.0099,
+      "step": 3353
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 2.615813953488372e-07,
+      "loss": 0.0655,
+      "step": 3354
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 2.6156976744186044e-07,
+      "loss": 1.9129,
+      "step": 3355
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 2.6155813953488367e-07,
+      "loss": 0.0155,
+      "step": 3356
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 2.6154651162790696e-07,
+      "loss": 0.0105,
+      "step": 3357
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 2.6153488372093025e-07,
+      "loss": 0.009,
+      "step": 3358
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 2.615232558139535e-07,
+      "loss": 0.0111,
+      "step": 3359
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 2.615116279069767e-07,
+      "loss": 0.0095,
+      "step": 3360
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 2.615e-07,
+      "loss": 0.0113,
+      "step": 3361
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 2.6148837209302324e-07,
+      "loss": 0.0111,
+      "step": 3362
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 2.614767441860465e-07,
+      "loss": 0.0101,
+      "step": 3363
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 2.6146511627906976e-07,
+      "loss": 0.0083,
+      "step": 3364
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 2.61453488372093e-07,
+      "loss": 0.1693,
+      "step": 3365
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 2.614418604651163e-07,
+      "loss": 0.009,
+      "step": 3366
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 2.614302325581395e-07,
+      "loss": 0.0087,
+      "step": 3367
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 2.6141860465116276e-07,
+      "loss": 0.0072,
+      "step": 3368
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 2.6140697674418605e-07,
+      "loss": 0.0135,
+      "step": 3369
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 2.613953488372093e-07,
+      "loss": 0.011,
+      "step": 3370
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 2.613837209302325e-07,
+      "loss": 0.0174,
+      "step": 3371
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 2.613720930232558e-07,
+      "loss": 0.0334,
+      "step": 3372
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 2.613604651162791e-07,
+      "loss": 0.0306,
+      "step": 3373
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 2.6134883720930233e-07,
+      "loss": 0.0117,
+      "step": 3374
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 2.6133720930232556e-07,
+      "loss": 0.0321,
+      "step": 3375
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 2.613255813953488e-07,
+      "loss": 0.0118,
+      "step": 3376
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 2.613139534883721e-07,
+      "loss": 0.0064,
+      "step": 3377
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 2.613023255813953e-07,
+      "loss": 1.0823,
+      "step": 3378
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 2.612906976744186e-07,
+      "loss": 0.0117,
+      "step": 3379
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 2.6127906976744185e-07,
+      "loss": 0.0052,
+      "step": 3380
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 2.6126744186046513e-07,
+      "loss": 0.01,
+      "step": 3381
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 2.6125581395348837e-07,
+      "loss": 0.0115,
+      "step": 3382
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 2.612441860465116e-07,
+      "loss": 0.0096,
+      "step": 3383
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 2.6123255813953484e-07,
+      "loss": 0.0512,
+      "step": 3384
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 2.6122093023255813e-07,
+      "loss": 0.0114,
+      "step": 3385
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 2.612093023255814e-07,
+      "loss": 0.0089,
+      "step": 3386
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 2.6119767441860465e-07,
+      "loss": 1.717,
+      "step": 3387
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 2.611860465116279e-07,
+      "loss": 0.0108,
+      "step": 3388
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 2.611744186046511e-07,
+      "loss": 0.0103,
+      "step": 3389
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 2.611627906976744e-07,
+      "loss": 0.0106,
+      "step": 3390
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 2.6115116279069765e-07,
+      "loss": 0.0082,
+      "step": 3391
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 2.6113953488372093e-07,
+      "loss": 0.009,
+      "step": 3392
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 2.6112790697674417e-07,
+      "loss": 0.0083,
+      "step": 3393
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 2.6111627906976746e-07,
+      "loss": 0.0141,
+      "step": 3394
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 2.611046511627907e-07,
+      "loss": 0.0074,
+      "step": 3395
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 2.6109302325581393e-07,
+      "loss": 0.0073,
+      "step": 3396
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 2.6108139534883716e-07,
+      "loss": 0.0081,
+      "step": 3397
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 2.6106976744186045e-07,
+      "loss": 0.0093,
+      "step": 3398
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 2.6105813953488374e-07,
+      "loss": 0.0056,
+      "step": 3399
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 2.61046511627907e-07,
+      "loss": 0.1351,
+      "step": 3400
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 2.610348837209302e-07,
+      "loss": 0.0054,
+      "step": 3401
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 2.610232558139535e-07,
+      "loss": 0.1127,
+      "step": 3402
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 2.6101162790697673e-07,
+      "loss": 0.0114,
+      "step": 3403
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 2.6099999999999997e-07,
+      "loss": 0.0062,
+      "step": 3404
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 2.6098837209302326e-07,
+      "loss": 0.0114,
+      "step": 3405
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 2.609767441860465e-07,
+      "loss": 0.1093,
+      "step": 3406
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 2.609651162790698e-07,
+      "loss": 0.0089,
+      "step": 3407
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 2.60953488372093e-07,
+      "loss": 0.0069,
+      "step": 3408
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 2.6094186046511625e-07,
+      "loss": 0.0084,
+      "step": 3409
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 2.609302325581395e-07,
+      "loss": 0.0068,
+      "step": 3410
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 2.609186046511628e-07,
+      "loss": 0.0064,
+      "step": 3411
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 2.6090697674418606e-07,
+      "loss": 0.009,
+      "step": 3412
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 2.608953488372093e-07,
+      "loss": 0.0076,
+      "step": 3413
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 2.6088372093023253e-07,
+      "loss": 0.0104,
+      "step": 3414
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 2.608720930232558e-07,
+      "loss": 0.2506,
+      "step": 3415
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 2.6086046511627906e-07,
+      "loss": 0.0173,
+      "step": 3416
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 2.608488372093023e-07,
+      "loss": 0.0143,
+      "step": 3417
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 2.608372093023256e-07,
+      "loss": 0.0127,
+      "step": 3418
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 2.608255813953488e-07,
+      "loss": 0.0114,
+      "step": 3419
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 2.608139534883721e-07,
+      "loss": 0.0106,
+      "step": 3420
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 2.6080232558139534e-07,
+      "loss": 0.0097,
+      "step": 3421
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 2.6079069767441857e-07,
+      "loss": 0.0173,
+      "step": 3422
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 2.6077906976744186e-07,
+      "loss": 0.011,
+      "step": 3423
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 2.607674418604651e-07,
+      "loss": 0.0062,
+      "step": 3424
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 2.6075581395348833e-07,
+      "loss": 0.0093,
+      "step": 3425
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 2.607441860465116e-07,
+      "loss": 0.0097,
+      "step": 3426
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 2.607325581395349e-07,
+      "loss": 0.0562,
+      "step": 3427
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 2.6072093023255814e-07,
+      "loss": 0.0056,
+      "step": 3428
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 2.607093023255814e-07,
+      "loss": 0.0086,
+      "step": 3429
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 2.606976744186046e-07,
+      "loss": 0.009,
+      "step": 3430
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 2.606860465116279e-07,
+      "loss": 0.0089,
+      "step": 3431
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 2.6067441860465114e-07,
+      "loss": 0.0099,
+      "step": 3432
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 2.606627906976744e-07,
+      "loss": 0.0164,
+      "step": 3433
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 2.6065116279069766e-07,
+      "loss": 0.0115,
+      "step": 3434
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 2.606395348837209e-07,
+      "loss": 0.006,
+      "step": 3435
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 2.606279069767442e-07,
+      "loss": 0.0101,
+      "step": 3436
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 2.606162790697674e-07,
+      "loss": 0.0143,
+      "step": 3437
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 2.6060465116279065e-07,
+      "loss": 0.0222,
+      "step": 3438
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 2.6059302325581394e-07,
+      "loss": 0.007,
+      "step": 3439
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 2.6058139534883723e-07,
+      "loss": 0.0095,
+      "step": 3440
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 2.6056976744186047e-07,
+      "loss": 2.4307,
+      "step": 3441
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 2.605581395348837e-07,
+      "loss": 0.0086,
+      "step": 3442
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 2.6054651162790694e-07,
+      "loss": 0.0121,
+      "step": 3443
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 2.605348837209302e-07,
+      "loss": 0.0098,
+      "step": 3444
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 2.6052325581395346e-07,
+      "loss": 0.0117,
+      "step": 3445
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 2.6051162790697675e-07,
+      "loss": 0.0088,
+      "step": 3446
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 2.605e-07,
+      "loss": 0.0079,
+      "step": 3447
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 2.6048837209302327e-07,
+      "loss": 0.0087,
+      "step": 3448
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 2.604767441860465e-07,
+      "loss": 0.0446,
+      "step": 3449
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 2.6046511627906974e-07,
+      "loss": 0.007,
+      "step": 3450
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 2.60453488372093e-07,
+      "loss": 0.0258,
+      "step": 3451
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 2.6044186046511627e-07,
+      "loss": 2.1163,
+      "step": 3452
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 2.6043023255813955e-07,
+      "loss": 0.0129,
+      "step": 3453
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 2.604186046511628e-07,
+      "loss": 0.0077,
+      "step": 3454
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 2.60406976744186e-07,
+      "loss": 0.0067,
+      "step": 3455
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 2.603953488372093e-07,
+      "loss": 0.1997,
+      "step": 3456
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 2.6038372093023255e-07,
+      "loss": 0.0067,
+      "step": 3457
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 2.603720930232558e-07,
+      "loss": 0.0069,
+      "step": 3458
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 2.6036046511627907e-07,
+      "loss": 0.0079,
+      "step": 3459
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 2.603488372093023e-07,
+      "loss": 0.0107,
+      "step": 3460
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 2.603372093023256e-07,
+      "loss": 1.4513,
+      "step": 3461
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 2.6032558139534883e-07,
+      "loss": 0.0088,
+      "step": 3462
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 2.6031395348837206e-07,
+      "loss": 0.0073,
+      "step": 3463
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 2.603023255813953e-07,
+      "loss": 0.0089,
+      "step": 3464
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 2.602906976744186e-07,
+      "loss": 0.01,
+      "step": 3465
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 2.602790697674418e-07,
+      "loss": 0.0082,
+      "step": 3466
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 2.602674418604651e-07,
+      "loss": 0.0074,
+      "step": 3467
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 2.6025581395348835e-07,
+      "loss": 0.0055,
+      "step": 3468
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 2.6024418604651164e-07,
+      "loss": 0.0088,
+      "step": 3469
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 2.6023255813953487e-07,
+      "loss": 0.011,
+      "step": 3470
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 2.602209302325581e-07,
+      "loss": 0.0155,
+      "step": 3471
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 2.602093023255814e-07,
+      "loss": 0.012,
+      "step": 3472
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 2.6019767441860463e-07,
+      "loss": 0.0052,
+      "step": 3473
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 2.601860465116279e-07,
+      "loss": 0.007,
+      "step": 3474
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 2.6017441860465115e-07,
+      "loss": 0.0196,
+      "step": 3475
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 2.601627906976744e-07,
+      "loss": 0.0065,
+      "step": 3476
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 2.601511627906977e-07,
+      "loss": 0.0081,
+      "step": 3477
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 2.601395348837209e-07,
+      "loss": 0.0081,
+      "step": 3478
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 2.6012790697674415e-07,
+      "loss": 0.0073,
+      "step": 3479
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 2.6011627906976743e-07,
+      "loss": 0.0059,
+      "step": 3480
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 2.601046511627907e-07,
+      "loss": 0.0126,
+      "step": 3481
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 2.6009302325581396e-07,
+      "loss": 0.0093,
+      "step": 3482
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 2.600813953488372e-07,
+      "loss": 0.0773,
+      "step": 3483
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 2.6006976744186043e-07,
+      "loss": 0.0088,
+      "step": 3484
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 2.600581395348837e-07,
+      "loss": 0.0071,
+      "step": 3485
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 2.6004651162790695e-07,
+      "loss": 0.0086,
+      "step": 3486
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 2.6003488372093024e-07,
+      "loss": 0.0052,
+      "step": 3487
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 2.600232558139535e-07,
+      "loss": 0.008,
+      "step": 3488
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 2.600116279069767e-07,
+      "loss": 0.0076,
+      "step": 3489
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 2.6e-07,
+      "loss": 0.0108,
+      "step": 3490
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 2.5998837209302323e-07,
+      "loss": 0.0197,
+      "step": 3491
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 2.5997674418604647e-07,
+      "loss": 0.0208,
+      "step": 3492
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 2.5996511627906976e-07,
+      "loss": 0.0063,
+      "step": 3493
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 2.5995348837209305e-07,
+      "loss": 0.0061,
+      "step": 3494
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 2.599418604651163e-07,
+      "loss": 0.0083,
+      "step": 3495
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 2.599302325581395e-07,
+      "loss": 0.0088,
+      "step": 3496
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 2.5991860465116275e-07,
+      "loss": 0.0094,
+      "step": 3497
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 2.5990697674418604e-07,
+      "loss": 0.0075,
+      "step": 3498
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 2.598953488372093e-07,
+      "loss": 0.009,
+      "step": 3499
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 2.5988372093023256e-07,
+      "loss": 0.0084,
+      "step": 3500
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 2.598720930232558e-07,
+      "loss": 0.0109,
+      "step": 3501
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 2.598604651162791e-07,
+      "loss": 0.0052,
+      "step": 3502
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 2.598488372093023e-07,
+      "loss": 0.0069,
+      "step": 3503
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 2.5983720930232556e-07,
+      "loss": 0.0094,
+      "step": 3504
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 2.598255813953488e-07,
+      "loss": 0.0788,
+      "step": 3505
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 2.598139534883721e-07,
+      "loss": 0.3591,
+      "step": 3506
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 2.5980232558139537e-07,
+      "loss": 0.0098,
+      "step": 3507
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 2.597906976744186e-07,
+      "loss": 0.0082,
+      "step": 3508
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 2.5977906976744184e-07,
+      "loss": 0.0095,
+      "step": 3509
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 2.5976744186046513e-07,
+      "loss": 0.0057,
+      "step": 3510
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 2.5975581395348836e-07,
+      "loss": 0.0061,
+      "step": 3511
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 2.597441860465116e-07,
+      "loss": 0.0067,
+      "step": 3512
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 2.597325581395349e-07,
+      "loss": 0.0086,
+      "step": 3513
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 2.597209302325581e-07,
+      "loss": 0.0097,
+      "step": 3514
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 2.597093023255814e-07,
+      "loss": 0.0052,
+      "step": 3515
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 2.5969767441860464e-07,
+      "loss": 0.0074,
+      "step": 3516
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 2.596860465116279e-07,
+      "loss": 0.024,
+      "step": 3517
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 2.596744186046511e-07,
+      "loss": 0.009,
+      "step": 3518
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.596627906976744e-07,
+      "loss": 0.0066,
+      "step": 3519
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.5965116279069764e-07,
+      "loss": 1.7836,
+      "step": 3520
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.5963953488372093e-07,
+      "loss": 0.0074,
+      "step": 3521
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.5962790697674416e-07,
+      "loss": 0.011,
+      "step": 3522
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.5961627906976745e-07,
+      "loss": 0.0323,
+      "step": 3523
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.596046511627907e-07,
+      "loss": 0.0051,
+      "step": 3524
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.595930232558139e-07,
+      "loss": 0.0956,
+      "step": 3525
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.595813953488372e-07,
+      "loss": 0.0228,
+      "step": 3526
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.5956976744186044e-07,
+      "loss": 0.0118,
+      "step": 3527
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.5955813953488373e-07,
+      "loss": 0.4404,
+      "step": 3528
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.5954651162790697e-07,
+      "loss": 0.0152,
+      "step": 3529
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.595348837209302e-07,
+      "loss": 0.008,
+      "step": 3530
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.595232558139535e-07,
+      "loss": 0.0089,
+      "step": 3531
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.595116279069767e-07,
+      "loss": 0.0076,
+      "step": 3532
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.5949999999999996e-07,
+      "loss": 0.0082,
+      "step": 3533
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 2.5948837209302325e-07,
+      "loss": 0.0077,
+      "step": 3534
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 2.5947674418604654e-07,
+      "loss": 0.0074,
+      "step": 3535
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 2.5946511627906977e-07,
+      "loss": 0.0063,
+      "step": 3536
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 2.59453488372093e-07,
+      "loss": 0.0058,
+      "step": 3537
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 2.5944186046511624e-07,
+      "loss": 0.0318,
+      "step": 3538
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.5943023255813953e-07,
+      "loss": 1.7343,
+      "step": 3539
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.5941860465116277e-07,
+      "loss": 0.0093,
+      "step": 3540
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.5940697674418605e-07,
+      "loss": 0.0084,
+      "step": 3541
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.593953488372093e-07,
+      "loss": 0.0141,
+      "step": 3542
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.593837209302325e-07,
+      "loss": 0.0078,
+      "step": 3543
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.593720930232558e-07,
+      "loss": 0.007,
+      "step": 3544
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.5936046511627905e-07,
+      "loss": 0.0127,
+      "step": 3545
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.593488372093023e-07,
+      "loss": 0.0645,
+      "step": 3546
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.5933720930232557e-07,
+      "loss": 0.0066,
+      "step": 3547
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.5932558139534886e-07,
+      "loss": 0.0084,
+      "step": 3548
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.593139534883721e-07,
+      "loss": 0.0071,
+      "step": 3549
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 2.5930232558139533e-07,
+      "loss": 0.0091,
+      "step": 3550
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 2.5929069767441857e-07,
+      "loss": 0.163,
+      "step": 3551
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 2.5927906976744185e-07,
+      "loss": 0.0077,
+      "step": 3552
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 2.592674418604651e-07,
+      "loss": 0.0064,
+      "step": 3553
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 2.592558139534884e-07,
+      "loss": 0.0117,
+      "step": 3554
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.592441860465116e-07,
+      "loss": 0.0098,
+      "step": 3555
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.592325581395349e-07,
+      "loss": 0.2203,
+      "step": 3556
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.5922093023255814e-07,
+      "loss": 0.0085,
+      "step": 3557
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.5920930232558137e-07,
+      "loss": 0.0073,
+      "step": 3558
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.591976744186046e-07,
+      "loss": 0.0074,
+      "step": 3559
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.591860465116279e-07,
+      "loss": 1.6644,
+      "step": 3560
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.591744186046512e-07,
+      "loss": 0.0099,
+      "step": 3561
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.591627906976744e-07,
+      "loss": 0.0053,
+      "step": 3562
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.5915116279069765e-07,
+      "loss": 0.0112,
+      "step": 3563
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.5913953488372094e-07,
+      "loss": 1.2571,
+      "step": 3564
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2.591279069767442e-07,
+      "loss": 0.0069,
+      "step": 3565
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2.591162790697674e-07,
+      "loss": 0.0066,
+      "step": 3566
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2.591046511627907e-07,
+      "loss": 0.0083,
+      "step": 3567
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2.5909302325581394e-07,
+      "loss": 0.3045,
+      "step": 3568
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2.590813953488372e-07,
+      "loss": 0.0052,
+      "step": 3569
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.5906976744186046e-07,
+      "loss": 0.0078,
+      "step": 3570
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.590581395348837e-07,
+      "loss": 0.0063,
+      "step": 3571
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.5904651162790693e-07,
+      "loss": 0.0059,
+      "step": 3572
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.590348837209302e-07,
+      "loss": 0.0116,
+      "step": 3573
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.5902325581395345e-07,
+      "loss": 0.0246,
+      "step": 3574
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.5901162790697674e-07,
+      "loss": 0.1121,
+      "step": 3575
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.59e-07,
+      "loss": 0.027,
+      "step": 3576
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.5898837209302326e-07,
+      "loss": 0.0096,
+      "step": 3577
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.589767441860465e-07,
+      "loss": 0.0118,
+      "step": 3578
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.5896511627906974e-07,
+      "loss": 0.0122,
+      "step": 3579
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.58953488372093e-07,
+      "loss": 0.006,
+      "step": 3580
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.5894186046511626e-07,
+      "loss": 0.0056,
+      "step": 3581
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.5893023255813955e-07,
+      "loss": 0.0049,
+      "step": 3582
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.589186046511628e-07,
+      "loss": 0.0075,
+      "step": 3583
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.58906976744186e-07,
+      "loss": 0.0079,
+      "step": 3584
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.588953488372093e-07,
+      "loss": 0.0094,
+      "step": 3585
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 2.5888372093023254e-07,
+      "loss": 0.005,
+      "step": 3586
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 2.588720930232558e-07,
+      "loss": 0.0154,
+      "step": 3587
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 2.5886046511627906e-07,
+      "loss": 0.0136,
+      "step": 3588
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 2.5884883720930235e-07,
+      "loss": 0.0057,
+      "step": 3589
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 2.588372093023256e-07,
+      "loss": 0.0104,
+      "step": 3590
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.588255813953488e-07,
+      "loss": 0.0069,
+      "step": 3591
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.5881395348837206e-07,
+      "loss": 0.0074,
+      "step": 3592
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.5880232558139535e-07,
+      "loss": 0.006,
+      "step": 3593
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.587906976744186e-07,
+      "loss": 0.0088,
+      "step": 3594
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.5877906976744187e-07,
+      "loss": 0.0733,
+      "step": 3595
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.587674418604651e-07,
+      "loss": 0.006,
+      "step": 3596
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.5875581395348834e-07,
+      "loss": 0.0104,
+      "step": 3597
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.5874418604651163e-07,
+      "loss": 0.0099,
+      "step": 3598
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.5873255813953486e-07,
+      "loss": 1.8108,
+      "step": 3599
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.587209302325581e-07,
+      "loss": 0.0077,
+      "step": 3600
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.587093023255814e-07,
+      "loss": 0.005,
+      "step": 3601
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.586976744186047e-07,
+      "loss": 0.0456,
+      "step": 3602
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.586860465116279e-07,
+      "loss": 1.0069,
+      "step": 3603
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.5867441860465115e-07,
+      "loss": 0.0073,
+      "step": 3604
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.586627906976744e-07,
+      "loss": 0.018,
+      "step": 3605
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.5865116279069767e-07,
+      "loss": 0.0055,
+      "step": 3606
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.586395348837209e-07,
+      "loss": 0.0058,
+      "step": 3607
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.586279069767442e-07,
+      "loss": 0.0064,
+      "step": 3608
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.5861627906976743e-07,
+      "loss": 0.0182,
+      "step": 3609
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.586046511627907e-07,
+      "loss": 0.0083,
+      "step": 3610
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.5859302325581395e-07,
+      "loss": 0.0067,
+      "step": 3611
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.585813953488372e-07,
+      "loss": 0.0091,
+      "step": 3612
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.585697674418604e-07,
+      "loss": 0.4142,
+      "step": 3613
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.585581395348837e-07,
+      "loss": 0.0718,
+      "step": 3614
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.5854651162790695e-07,
+      "loss": 0.0612,
+      "step": 3615
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.5853488372093023e-07,
+      "loss": 0.0063,
+      "step": 3616
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.5852325581395347e-07,
+      "loss": 0.1073,
+      "step": 3617
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.5851162790697676e-07,
+      "loss": 0.0065,
+      "step": 3618
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.585e-07,
+      "loss": 0.0047,
+      "step": 3619
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.9845261121856866,
+      "eval_loss": 0.1568414270877838,
+      "eval_runtime": 1.5577,
+      "eval_samples_per_second": 663.804,
+      "eval_steps_per_second": 10.914,
+      "step": 3619
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.5848837209302323e-07,
+      "loss": 0.0053,
+      "step": 3620
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.584767441860465e-07,
+      "loss": 0.0055,
+      "step": 3621
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.5846511627906975e-07,
+      "loss": 0.2307,
+      "step": 3622
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.5845348837209304e-07,
+      "loss": 0.0058,
+      "step": 3623
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.584418604651163e-07,
+      "loss": 0.0084,
+      "step": 3624
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.584302325581395e-07,
+      "loss": 0.01,
+      "step": 3625
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.5841860465116274e-07,
+      "loss": 1.9355,
+      "step": 3626
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.5840697674418603e-07,
+      "loss": 0.0137,
+      "step": 3627
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.5839534883720927e-07,
+      "loss": 0.0112,
+      "step": 3628
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.5838372093023256e-07,
+      "loss": 0.2974,
+      "step": 3629
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.583720930232558e-07,
+      "loss": 0.0194,
+      "step": 3630
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.583604651162791e-07,
+      "loss": 0.0166,
+      "step": 3631
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.583488372093023e-07,
+      "loss": 0.0097,
+      "step": 3632
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.5833720930232555e-07,
+      "loss": 0.0101,
+      "step": 3633
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.5832558139534884e-07,
+      "loss": 0.0648,
+      "step": 3634
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.5831395348837207e-07,
+      "loss": 0.0103,
+      "step": 3635
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.5830232558139536e-07,
+      "loss": 0.0055,
+      "step": 3636
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.582906976744186e-07,
+      "loss": 0.006,
+      "step": 3637
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.5827906976744183e-07,
+      "loss": 0.0083,
+      "step": 3638
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.582674418604651e-07,
+      "loss": 0.0465,
+      "step": 3639
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.5825581395348836e-07,
+      "loss": 0.0085,
+      "step": 3640
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.582441860465116e-07,
+      "loss": 0.005,
+      "step": 3641
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.582325581395349e-07,
+      "loss": 0.0057,
+      "step": 3642
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.5822093023255817e-07,
+      "loss": 0.7326,
+      "step": 3643
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.582093023255814e-07,
+      "loss": 0.0074,
+      "step": 3644
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.5819767441860464e-07,
+      "loss": 0.2848,
+      "step": 3645
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.5818604651162787e-07,
+      "loss": 0.0057,
+      "step": 3646
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.5817441860465116e-07,
+      "loss": 0.0074,
+      "step": 3647
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.581627906976744e-07,
+      "loss": 0.0049,
+      "step": 3648
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.581511627906977e-07,
+      "loss": 0.0104,
+      "step": 3649
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.581395348837209e-07,
+      "loss": 0.0082,
+      "step": 3650
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.5812790697674416e-07,
+      "loss": 0.0066,
+      "step": 3651
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.5811627906976744e-07,
+      "loss": 0.0153,
+      "step": 3652
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.581046511627907e-07,
+      "loss": 0.0077,
+      "step": 3653
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.580930232558139e-07,
+      "loss": 1.5496,
+      "step": 3654
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.580813953488372e-07,
+      "loss": 0.0082,
+      "step": 3655
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.580697674418605e-07,
+      "loss": 0.0285,
+      "step": 3656
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.580581395348837e-07,
+      "loss": 0.0192,
+      "step": 3657
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.5804651162790696e-07,
+      "loss": 0.1164,
+      "step": 3658
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.580348837209302e-07,
+      "loss": 0.0292,
+      "step": 3659
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.580232558139535e-07,
+      "loss": 0.0053,
+      "step": 3660
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.580116279069767e-07,
+      "loss": 0.0057,
+      "step": 3661
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.58e-07,
+      "loss": 0.0079,
+      "step": 3662
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.5798837209302324e-07,
+      "loss": 0.0083,
+      "step": 3663
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.5797674418604653e-07,
+      "loss": 0.0047,
+      "step": 3664
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.5796511627906977e-07,
+      "loss": 2.6049,
+      "step": 3665
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.57953488372093e-07,
+      "loss": 0.1748,
+      "step": 3666
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.5794186046511624e-07,
+      "loss": 0.0065,
+      "step": 3667
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.579302325581395e-07,
+      "loss": 0.0043,
+      "step": 3668
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.5791860465116276e-07,
+      "loss": 0.0099,
+      "step": 3669
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.5790697674418605e-07,
+      "loss": 0.0154,
+      "step": 3670
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.578953488372093e-07,
+      "loss": 0.0044,
+      "step": 3671
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.5788372093023257e-07,
+      "loss": 0.006,
+      "step": 3672
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.578720930232558e-07,
+      "loss": 0.0051,
+      "step": 3673
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.5786046511627904e-07,
+      "loss": 0.0056,
+      "step": 3674
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.5784883720930233e-07,
+      "loss": 0.0097,
+      "step": 3675
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.5783720930232557e-07,
+      "loss": 1.7248,
+      "step": 3676
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.5782558139534885e-07,
+      "loss": 0.0082,
+      "step": 3677
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.578139534883721e-07,
+      "loss": 0.0065,
+      "step": 3678
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.578023255813953e-07,
+      "loss": 0.0057,
+      "step": 3679
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.5779069767441856e-07,
+      "loss": 0.0098,
+      "step": 3680
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.5777906976744185e-07,
+      "loss": 0.005,
+      "step": 3681
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.577674418604651e-07,
+      "loss": 0.0068,
+      "step": 3682
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.5775581395348837e-07,
+      "loss": 0.025,
+      "step": 3683
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.577441860465116e-07,
+      "loss": 0.0073,
+      "step": 3684
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.577325581395349e-07,
+      "loss": 0.0042,
+      "step": 3685
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.5772093023255813e-07,
+      "loss": 0.0086,
+      "step": 3686
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.5770930232558136e-07,
+      "loss": 0.6779,
+      "step": 3687
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.5769767441860465e-07,
+      "loss": 0.0059,
+      "step": 3688
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.576860465116279e-07,
+      "loss": 0.0325,
+      "step": 3689
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.576744186046512e-07,
+      "loss": 0.0072,
+      "step": 3690
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.576627906976744e-07,
+      "loss": 0.0055,
+      "step": 3691
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.5765116279069765e-07,
+      "loss": 0.0192,
+      "step": 3692
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.5763953488372094e-07,
+      "loss": 0.0072,
+      "step": 3693
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.5762790697674417e-07,
+      "loss": 0.0075,
+      "step": 3694
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.576162790697674e-07,
+      "loss": 0.0071,
+      "step": 3695
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.576046511627907e-07,
+      "loss": 0.0103,
+      "step": 3696
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.57593023255814e-07,
+      "loss": 0.0062,
+      "step": 3697
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.575813953488372e-07,
+      "loss": 0.2178,
+      "step": 3698
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.5756976744186045e-07,
+      "loss": 0.0055,
+      "step": 3699
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.575581395348837e-07,
+      "loss": 0.0092,
+      "step": 3700
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.57546511627907e-07,
+      "loss": 2.1337,
+      "step": 3701
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.575348837209302e-07,
+      "loss": 0.0049,
+      "step": 3702
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.575232558139535e-07,
+      "loss": 0.0064,
+      "step": 3703
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.5751162790697673e-07,
+      "loss": 0.0088,
+      "step": 3704
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.5749999999999997e-07,
+      "loss": 0.0138,
+      "step": 3705
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.5748837209302326e-07,
+      "loss": 0.0205,
+      "step": 3706
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.574767441860465e-07,
+      "loss": 0.0084,
+      "step": 3707
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.5746511627906973e-07,
+      "loss": 0.0082,
+      "step": 3708
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.57453488372093e-07,
+      "loss": 0.0068,
+      "step": 3709
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.574418604651163e-07,
+      "loss": 0.0097,
+      "step": 3710
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.5743023255813954e-07,
+      "loss": 0.0106,
+      "step": 3711
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.574186046511628e-07,
+      "loss": 0.0059,
+      "step": 3712
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.57406976744186e-07,
+      "loss": 0.0109,
+      "step": 3713
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.573953488372093e-07,
+      "loss": 0.012,
+      "step": 3714
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.5738372093023253e-07,
+      "loss": 0.0054,
+      "step": 3715
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.573720930232558e-07,
+      "loss": 0.0168,
+      "step": 3716
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.5736046511627906e-07,
+      "loss": 0.0093,
+      "step": 3717
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.5734883720930235e-07,
+      "loss": 0.0101,
+      "step": 3718
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.573372093023256e-07,
+      "loss": 0.0089,
+      "step": 3719
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.573255813953488e-07,
+      "loss": 0.0094,
+      "step": 3720
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.5731395348837205e-07,
+      "loss": 0.0058,
+      "step": 3721
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.5730232558139534e-07,
+      "loss": 0.0106,
+      "step": 3722
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.572906976744186e-07,
+      "loss": 0.0067,
+      "step": 3723
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.5727906976744186e-07,
+      "loss": 0.0072,
+      "step": 3724
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.572674418604651e-07,
+      "loss": 0.007,
+      "step": 3725
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.572558139534884e-07,
+      "loss": 0.0061,
+      "step": 3726
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.572441860465116e-07,
+      "loss": 0.3156,
+      "step": 3727
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.5723255813953486e-07,
+      "loss": 0.0062,
+      "step": 3728
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.5722093023255815e-07,
+      "loss": 0.0096,
+      "step": 3729
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.572093023255814e-07,
+      "loss": 0.0109,
+      "step": 3730
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.5719767441860467e-07,
+      "loss": 0.01,
+      "step": 3731
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.571860465116279e-07,
+      "loss": 0.0087,
+      "step": 3732
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.5717441860465114e-07,
+      "loss": 0.0085,
+      "step": 3733
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.571627906976744e-07,
+      "loss": 0.012,
+      "step": 3734
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.5715116279069766e-07,
+      "loss": 0.0052,
+      "step": 3735
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.571395348837209e-07,
+      "loss": 0.0085,
+      "step": 3736
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.571279069767442e-07,
+      "loss": 0.0079,
+      "step": 3737
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.571162790697674e-07,
+      "loss": 0.016,
+      "step": 3738
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.571046511627907e-07,
+      "loss": 0.0054,
+      "step": 3739
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.5709302325581394e-07,
+      "loss": 0.0155,
+      "step": 3740
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.570813953488372e-07,
+      "loss": 0.0062,
+      "step": 3741
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.570697674418604e-07,
+      "loss": 0.0065,
+      "step": 3742
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.570581395348837e-07,
+      "loss": 0.0103,
+      "step": 3743
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.57046511627907e-07,
+      "loss": 0.005,
+      "step": 3744
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.5703488372093023e-07,
+      "loss": 0.0094,
+      "step": 3745
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.5702325581395346e-07,
+      "loss": 0.0086,
+      "step": 3746
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.5701162790697675e-07,
+      "loss": 0.0098,
+      "step": 3747
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.57e-07,
+      "loss": 0.0054,
+      "step": 3748
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.569883720930232e-07,
+      "loss": 0.0106,
+      "step": 3749
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.569767441860465e-07,
+      "loss": 0.9596,
+      "step": 3750
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.569651162790698e-07,
+      "loss": 0.0135,
+      "step": 3751
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.5695348837209303e-07,
+      "loss": 0.0057,
+      "step": 3752
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.5694186046511627e-07,
+      "loss": 0.2127,
+      "step": 3753
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.569302325581395e-07,
+      "loss": 0.0061,
+      "step": 3754
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.5691860465116274e-07,
+      "loss": 0.0126,
+      "step": 3755
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.56906976744186e-07,
+      "loss": 0.005,
+      "step": 3756
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.568953488372093e-07,
+      "loss": 0.0099,
+      "step": 3757
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.5688372093023255e-07,
+      "loss": 0.6434,
+      "step": 3758
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.568720930232558e-07,
+      "loss": 0.0522,
+      "step": 3759
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.5686046511627907e-07,
+      "loss": 0.0048,
+      "step": 3760
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.568488372093023e-07,
+      "loss": 0.0084,
+      "step": 3761
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.5683720930232554e-07,
+      "loss": 0.0057,
+      "step": 3762
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.5682558139534883e-07,
+      "loss": 0.1028,
+      "step": 3763
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.5681395348837207e-07,
+      "loss": 0.2225,
+      "step": 3764
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.5680232558139535e-07,
+      "loss": 0.016,
+      "step": 3765
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.567906976744186e-07,
+      "loss": 0.0077,
+      "step": 3766
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.567790697674418e-07,
+      "loss": 0.0062,
+      "step": 3767
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.567674418604651e-07,
+      "loss": 0.0088,
+      "step": 3768
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.5675581395348835e-07,
+      "loss": 0.0096,
+      "step": 3769
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.5674418604651164e-07,
+      "loss": 0.0075,
+      "step": 3770
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.5673255813953487e-07,
+      "loss": 0.0397,
+      "step": 3771
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.5672093023255816e-07,
+      "loss": 0.2203,
+      "step": 3772
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.567093023255814e-07,
+      "loss": 0.0073,
+      "step": 3773
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.5669767441860463e-07,
+      "loss": 0.0108,
+      "step": 3774
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.5668604651162787e-07,
+      "loss": 0.0102,
+      "step": 3775
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.5667441860465115e-07,
+      "loss": 0.01,
+      "step": 3776
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.566627906976744e-07,
+      "loss": 0.0072,
+      "step": 3777
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.566511627906977e-07,
+      "loss": 0.0058,
+      "step": 3778
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.566395348837209e-07,
+      "loss": 0.006,
+      "step": 3779
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.566279069767442e-07,
+      "loss": 0.0088,
+      "step": 3780
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.5661627906976744e-07,
+      "loss": 0.0071,
+      "step": 3781
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.5660465116279067e-07,
+      "loss": 0.0071,
+      "step": 3782
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.5659302325581396e-07,
+      "loss": 0.0052,
+      "step": 3783
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.565813953488372e-07,
+      "loss": 0.0047,
+      "step": 3784
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.565697674418605e-07,
+      "loss": 0.0154,
+      "step": 3785
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.565581395348837e-07,
+      "loss": 0.0053,
+      "step": 3786
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.5654651162790695e-07,
+      "loss": 0.0067,
+      "step": 3787
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.565348837209302e-07,
+      "loss": 0.009,
+      "step": 3788
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.565232558139535e-07,
+      "loss": 0.0069,
+      "step": 3789
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.565116279069767e-07,
+      "loss": 0.0143,
+      "step": 3790
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.565e-07,
+      "loss": 0.0116,
+      "step": 3791
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.5648837209302324e-07,
+      "loss": 0.0093,
+      "step": 3792
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.564767441860465e-07,
+      "loss": 0.3523,
+      "step": 3793
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.5646511627906976e-07,
+      "loss": 0.0072,
+      "step": 3794
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.56453488372093e-07,
+      "loss": 0.0207,
+      "step": 3795
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.5644186046511623e-07,
+      "loss": 0.3851,
+      "step": 3796
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.564302325581395e-07,
+      "loss": 0.0088,
+      "step": 3797
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.564186046511628e-07,
+      "loss": 0.0047,
+      "step": 3798
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.5640697674418604e-07,
+      "loss": 0.0072,
+      "step": 3799
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.563953488372093e-07,
+      "loss": 0.0073,
+      "step": 3800
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.5638372093023256e-07,
+      "loss": 0.0044,
+      "step": 3801
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.563720930232558e-07,
+      "loss": 0.0072,
+      "step": 3802
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.5636046511627904e-07,
+      "loss": 0.0087,
+      "step": 3803
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.563488372093023e-07,
+      "loss": 0.4283,
+      "step": 3804
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.563372093023256e-07,
+      "loss": 0.0098,
+      "step": 3805
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.5632558139534885e-07,
+      "loss": 0.0052,
+      "step": 3806
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.563139534883721e-07,
+      "loss": 0.0041,
+      "step": 3807
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.563023255813953e-07,
+      "loss": 0.0089,
+      "step": 3808
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.5629069767441855e-07,
+      "loss": 0.0473,
+      "step": 3809
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.5627906976744184e-07,
+      "loss": 0.0054,
+      "step": 3810
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.5626744186046513e-07,
+      "loss": 0.0071,
+      "step": 3811
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.5625581395348836e-07,
+      "loss": 0.0072,
+      "step": 3812
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.562441860465116e-07,
+      "loss": 0.0051,
+      "step": 3813
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.562325581395349e-07,
+      "loss": 0.0071,
+      "step": 3814
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.562209302325581e-07,
+      "loss": 0.0082,
+      "step": 3815
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.5620930232558136e-07,
+      "loss": 0.0051,
+      "step": 3816
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.5619767441860465e-07,
+      "loss": 0.0067,
+      "step": 3817
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.561860465116279e-07,
+      "loss": 0.7292,
+      "step": 3818
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.5617441860465117e-07,
+      "loss": 0.0073,
+      "step": 3819
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.561627906976744e-07,
+      "loss": 0.0061,
+      "step": 3820
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.5615116279069764e-07,
+      "loss": 0.0044,
+      "step": 3821
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.5613953488372093e-07,
+      "loss": 0.9845,
+      "step": 3822
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.5612790697674416e-07,
+      "loss": 0.0094,
+      "step": 3823
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.5611627906976745e-07,
+      "loss": 0.0096,
+      "step": 3824
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.561046511627907e-07,
+      "loss": 0.006,
+      "step": 3825
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.56093023255814e-07,
+      "loss": 0.0061,
+      "step": 3826
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.560813953488372e-07,
+      "loss": 0.0185,
+      "step": 3827
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.5606976744186045e-07,
+      "loss": 0.0061,
+      "step": 3828
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.560581395348837e-07,
+      "loss": 0.0065,
+      "step": 3829
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5604651162790697e-07,
+      "loss": 0.0091,
+      "step": 3830
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.560348837209302e-07,
+      "loss": 0.0124,
+      "step": 3831
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.560232558139535e-07,
+      "loss": 0.1312,
+      "step": 3832
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5601162790697673e-07,
+      "loss": 0.0089,
+      "step": 3833
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.56e-07,
+      "loss": 0.0094,
+      "step": 3834
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5598837209302325e-07,
+      "loss": 0.0066,
+      "step": 3835
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.559767441860465e-07,
+      "loss": 0.0119,
+      "step": 3836
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.559651162790698e-07,
+      "loss": 0.0055,
+      "step": 3837
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.55953488372093e-07,
+      "loss": 0.0058,
+      "step": 3838
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.559418604651163e-07,
+      "loss": 0.0071,
+      "step": 3839
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5593023255813953e-07,
+      "loss": 0.0054,
+      "step": 3840
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5591860465116277e-07,
+      "loss": 0.0042,
+      "step": 3841
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.55906976744186e-07,
+      "loss": 0.0053,
+      "step": 3842
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.558953488372093e-07,
+      "loss": 0.0068,
+      "step": 3843
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5588372093023253e-07,
+      "loss": 0.0059,
+      "step": 3844
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.558720930232558e-07,
+      "loss": 0.0046,
+      "step": 3845
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5586046511627905e-07,
+      "loss": 0.0156,
+      "step": 3846
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5584883720930234e-07,
+      "loss": 0.0073,
+      "step": 3847
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.558372093023256e-07,
+      "loss": 0.006,
+      "step": 3848
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.558255813953488e-07,
+      "loss": 0.01,
+      "step": 3849
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5581395348837204e-07,
+      "loss": 0.0084,
+      "step": 3850
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5580232558139533e-07,
+      "loss": 0.01,
+      "step": 3851
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.557906976744186e-07,
+      "loss": 0.0058,
+      "step": 3852
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5577906976744186e-07,
+      "loss": 0.0065,
+      "step": 3853
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.557674418604651e-07,
+      "loss": 0.0084,
+      "step": 3854
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.557558139534884e-07,
+      "loss": 0.0054,
+      "step": 3855
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.557441860465116e-07,
+      "loss": 0.0046,
+      "step": 3856
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5573255813953485e-07,
+      "loss": 0.0122,
+      "step": 3857
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5572093023255814e-07,
+      "loss": 0.0065,
+      "step": 3858
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5570930232558137e-07,
+      "loss": 0.005,
+      "step": 3859
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5569767441860466e-07,
+      "loss": 0.3367,
+      "step": 3860
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.556860465116279e-07,
+      "loss": 0.0186,
+      "step": 3861
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5567441860465113e-07,
+      "loss": 0.0108,
+      "step": 3862
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5566279069767437e-07,
+      "loss": 0.0118,
+      "step": 3863
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5565116279069766e-07,
+      "loss": 0.0037,
+      "step": 3864
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5563953488372094e-07,
+      "loss": 0.0058,
+      "step": 3865
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.556279069767442e-07,
+      "loss": 0.0128,
+      "step": 3866
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.556162790697674e-07,
+      "loss": 0.0079,
+      "step": 3867
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.556046511627907e-07,
+      "loss": 1.724,
+      "step": 3868
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5559302325581394e-07,
+      "loss": 0.0067,
+      "step": 3869
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5558139534883717e-07,
+      "loss": 0.0079,
+      "step": 3870
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5556976744186046e-07,
+      "loss": 0.0062,
+      "step": 3871
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.555581395348837e-07,
+      "loss": 0.0049,
+      "step": 3872
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.55546511627907e-07,
+      "loss": 0.0173,
+      "step": 3873
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.555348837209302e-07,
+      "loss": 0.128,
+      "step": 3874
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5552325581395345e-07,
+      "loss": 0.0084,
+      "step": 3875
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5551162790697674e-07,
+      "loss": 0.1767,
+      "step": 3876
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.555e-07,
+      "loss": 0.0179,
+      "step": 3877
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5548837209302327e-07,
+      "loss": 0.0277,
+      "step": 3878
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.554767441860465e-07,
+      "loss": 0.5359,
+      "step": 3879
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.554651162790698e-07,
+      "loss": 0.0072,
+      "step": 3880
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.55453488372093e-07,
+      "loss": 0.0149,
+      "step": 3881
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.5544186046511626e-07,
+      "loss": 0.0072,
+      "step": 3882
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.554302325581395e-07,
+      "loss": 0.0133,
+      "step": 3883
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.554186046511628e-07,
+      "loss": 0.0092,
+      "step": 3884
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.55406976744186e-07,
+      "loss": 0.0056,
+      "step": 3885
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.553953488372093e-07,
+      "loss": 0.0081,
+      "step": 3886
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.5538372093023254e-07,
+      "loss": 0.0861,
+      "step": 3887
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.5537209302325583e-07,
+      "loss": 0.022,
+      "step": 3888
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.5536046511627907e-07,
+      "loss": 0.0046,
+      "step": 3889
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.553488372093023e-07,
+      "loss": 0.0072,
+      "step": 3890
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.5533720930232554e-07,
+      "loss": 0.0147,
+      "step": 3891
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.553255813953488e-07,
+      "loss": 0.0048,
+      "step": 3892
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.553139534883721e-07,
+      "loss": 0.0077,
+      "step": 3893
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.5530232558139535e-07,
+      "loss": 0.0097,
+      "step": 3894
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.552906976744186e-07,
+      "loss": 0.02,
+      "step": 3895
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.552790697674418e-07,
+      "loss": 0.0078,
+      "step": 3896
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.552674418604651e-07,
+      "loss": 0.0075,
+      "step": 3897
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.5525581395348834e-07,
+      "loss": 0.0101,
+      "step": 3898
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.5524418604651163e-07,
+      "loss": 0.0192,
+      "step": 3899
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.5523255813953487e-07,
+      "loss": 0.0093,
+      "step": 3900
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.5522093023255815e-07,
+      "loss": 0.0055,
+      "step": 3901
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.552093023255814e-07,
+      "loss": 0.0183,
+      "step": 3902
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.551976744186046e-07,
+      "loss": 0.006,
+      "step": 3903
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.5518604651162786e-07,
+      "loss": 0.0074,
+      "step": 3904
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.5517441860465115e-07,
+      "loss": 0.009,
+      "step": 3905
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.5516279069767444e-07,
+      "loss": 0.0066,
+      "step": 3906
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.5515116279069767e-07,
+      "loss": 0.1216,
+      "step": 3907
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.551395348837209e-07,
+      "loss": 0.5603,
+      "step": 3908
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.551279069767442e-07,
+      "loss": 0.0668,
+      "step": 3909
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.5511627906976743e-07,
+      "loss": 0.0057,
+      "step": 3910
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.5510465116279066e-07,
+      "loss": 0.0043,
+      "step": 3911
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.5509302325581395e-07,
+      "loss": 0.0075,
+      "step": 3912
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.550813953488372e-07,
+      "loss": 0.0068,
+      "step": 3913
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.550697674418605e-07,
+      "loss": 0.0079,
+      "step": 3914
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.550581395348837e-07,
+      "loss": 0.0089,
+      "step": 3915
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.5504651162790695e-07,
+      "loss": 0.0057,
+      "step": 3916
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.550348837209302e-07,
+      "loss": 0.0067,
+      "step": 3917
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.5502325581395347e-07,
+      "loss": 0.0084,
+      "step": 3918
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.5501162790697676e-07,
+      "loss": 0.039,
+      "step": 3919
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.55e-07,
+      "loss": 0.0119,
+      "step": 3920
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.5498837209302323e-07,
+      "loss": 0.0073,
+      "step": 3921
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.549767441860465e-07,
+      "loss": 0.0054,
+      "step": 3922
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.5496511627906975e-07,
+      "loss": 0.006,
+      "step": 3923
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.54953488372093e-07,
+      "loss": 0.0086,
+      "step": 3924
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.549418604651163e-07,
+      "loss": 0.0047,
+      "step": 3925
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.549302325581395e-07,
+      "loss": 0.0053,
+      "step": 3926
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.549186046511628e-07,
+      "loss": 0.008,
+      "step": 3927
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.5490697674418603e-07,
+      "loss": 0.0075,
+      "step": 3928
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.5489534883720927e-07,
+      "loss": 0.0056,
+      "step": 3929
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.5488372093023256e-07,
+      "loss": 0.0074,
+      "step": 3930
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.548720930232558e-07,
+      "loss": 0.0051,
+      "step": 3931
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.548604651162791e-07,
+      "loss": 0.0071,
+      "step": 3932
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.548488372093023e-07,
+      "loss": 0.3007,
+      "step": 3933
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.548372093023256e-07,
+      "loss": 0.0137,
+      "step": 3934
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.5482558139534884e-07,
+      "loss": 0.0053,
+      "step": 3935
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.548139534883721e-07,
+      "loss": 0.0222,
+      "step": 3936
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.548023255813953e-07,
+      "loss": 0.0279,
+      "step": 3937
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.547906976744186e-07,
+      "loss": 0.0069,
+      "step": 3938
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.5477906976744183e-07,
+      "loss": 0.0059,
+      "step": 3939
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.547674418604651e-07,
+      "loss": 0.0044,
+      "step": 3940
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.5475581395348836e-07,
+      "loss": 0.0101,
+      "step": 3941
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.547441860465116e-07,
+      "loss": 0.0068,
+      "step": 3942
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.547325581395349e-07,
+      "loss": 0.0061,
+      "step": 3943
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.547209302325581e-07,
+      "loss": 0.0046,
+      "step": 3944
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.5470930232558135e-07,
+      "loss": 0.0047,
+      "step": 3945
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.5469767441860464e-07,
+      "loss": 0.0058,
+      "step": 3946
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.5468604651162793e-07,
+      "loss": 0.0598,
+      "step": 3947
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.5467441860465116e-07,
+      "loss": 0.0087,
+      "step": 3948
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.546627906976744e-07,
+      "loss": 0.0078,
+      "step": 3949
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.5465116279069763e-07,
+      "loss": 0.0056,
+      "step": 3950
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.546395348837209e-07,
+      "loss": 0.0067,
+      "step": 3951
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.5462790697674416e-07,
+      "loss": 0.0052,
+      "step": 3952
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.5461627906976744e-07,
+      "loss": 0.005,
+      "step": 3953
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.546046511627907e-07,
+      "loss": 0.0052,
+      "step": 3954
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.5459302325581397e-07,
+      "loss": 0.0087,
+      "step": 3955
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.545813953488372e-07,
+      "loss": 0.0052,
+      "step": 3956
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.5456976744186044e-07,
+      "loss": 0.0064,
+      "step": 3957
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.545581395348837e-07,
+      "loss": 0.0057,
+      "step": 3958
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.5454651162790696e-07,
+      "loss": 0.0068,
+      "step": 3959
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.5453488372093025e-07,
+      "loss": 0.0093,
+      "step": 3960
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.545232558139535e-07,
+      "loss": 1.2162,
+      "step": 3961
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.545116279069767e-07,
+      "loss": 0.0044,
+      "step": 3962
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.545e-07,
+      "loss": 0.0083,
+      "step": 3963
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.5448837209302324e-07,
+      "loss": 0.0486,
+      "step": 3964
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.544767441860465e-07,
+      "loss": 0.0056,
+      "step": 3965
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.5446511627906977e-07,
+      "loss": 0.0048,
+      "step": 3966
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.54453488372093e-07,
+      "loss": 0.0039,
+      "step": 3967
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.544418604651163e-07,
+      "loss": 0.0075,
+      "step": 3968
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.544302325581395e-07,
+      "loss": 0.0072,
+      "step": 3969
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.5441860465116276e-07,
+      "loss": 0.0043,
+      "step": 3970
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.54406976744186e-07,
+      "loss": 0.0046,
+      "step": 3971
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.543953488372093e-07,
+      "loss": 0.889,
+      "step": 3972
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.5438372093023257e-07,
+      "loss": 0.0065,
+      "step": 3973
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.543720930232558e-07,
+      "loss": 0.0053,
+      "step": 3974
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.5436046511627904e-07,
+      "loss": 0.007,
+      "step": 3975
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.5434883720930233e-07,
+      "loss": 0.005,
+      "step": 3976
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.5433720930232557e-07,
+      "loss": 0.0071,
+      "step": 3977
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.543255813953488e-07,
+      "loss": 0.0187,
+      "step": 3978
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.543139534883721e-07,
+      "loss": 0.0089,
+      "step": 3979
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.543023255813953e-07,
+      "loss": 0.0045,
+      "step": 3980
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.542906976744186e-07,
+      "loss": 0.0059,
+      "step": 3981
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.5427906976744185e-07,
+      "loss": 0.0241,
+      "step": 3982
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.542674418604651e-07,
+      "loss": 0.0069,
+      "step": 3983
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.5425581395348837e-07,
+      "loss": 0.0068,
+      "step": 3984
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.542441860465116e-07,
+      "loss": 0.0081,
+      "step": 3985
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.5423255813953484e-07,
+      "loss": 0.0119,
+      "step": 3986
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.5422093023255813e-07,
+      "loss": 0.0073,
+      "step": 3987
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.542093023255814e-07,
+      "loss": 0.0067,
+      "step": 3988
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.5419767441860465e-07,
+      "loss": 0.0226,
+      "step": 3989
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.541860465116279e-07,
+      "loss": 0.0937,
+      "step": 3990
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.541744186046511e-07,
+      "loss": 0.0057,
+      "step": 3991
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.541627906976744e-07,
+      "loss": 0.0042,
+      "step": 3992
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.5415116279069765e-07,
+      "loss": 0.0091,
+      "step": 3993
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.5413953488372094e-07,
+      "loss": 0.0057,
+      "step": 3994
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.5412790697674417e-07,
+      "loss": 0.0096,
+      "step": 3995
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.541162790697674e-07,
+      "loss": 0.0062,
+      "step": 3996
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.541046511627907e-07,
+      "loss": 0.005,
+      "step": 3997
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.5409302325581393e-07,
+      "loss": 0.0054,
+      "step": 3998
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.5408139534883717e-07,
+      "loss": 0.007,
+      "step": 3999
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.5406976744186045e-07,
+      "loss": 0.0138,
+      "step": 4000
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.5405813953488374e-07,
+      "loss": 0.0072,
+      "step": 4001
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.54046511627907e-07,
+      "loss": 0.0093,
+      "step": 4002
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.540348837209302e-07,
+      "loss": 0.0056,
+      "step": 4003
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.5402325581395345e-07,
+      "loss": 0.4625,
+      "step": 4004
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.5401162790697674e-07,
+      "loss": 0.0314,
+      "step": 4005
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.5399999999999997e-07,
+      "loss": 1.8738,
+      "step": 4006
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.5398837209302326e-07,
+      "loss": 1.8039,
+      "step": 4007
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.539767441860465e-07,
+      "loss": 0.3208,
+      "step": 4008
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.539651162790698e-07,
+      "loss": 0.0467,
+      "step": 4009
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.53953488372093e-07,
+      "loss": 0.0052,
+      "step": 4010
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.5394186046511625e-07,
+      "loss": 0.0127,
+      "step": 4011
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.539302325581395e-07,
+      "loss": 0.004,
+      "step": 4012
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.539186046511628e-07,
+      "loss": 0.0045,
+      "step": 4013
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.5390697674418607e-07,
+      "loss": 0.0078,
+      "step": 4014
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.538953488372093e-07,
+      "loss": 0.0104,
+      "step": 4015
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.5388372093023254e-07,
+      "loss": 0.0041,
+      "step": 4016
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.538720930232558e-07,
+      "loss": 0.2388,
+      "step": 4017
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.5386046511627906e-07,
+      "loss": 0.0049,
+      "step": 4018
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.538488372093023e-07,
+      "loss": 0.0068,
+      "step": 4019
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.538372093023256e-07,
+      "loss": 0.0039,
+      "step": 4020
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.538255813953488e-07,
+      "loss": 0.0068,
+      "step": 4021
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.538139534883721e-07,
+      "loss": 0.007,
+      "step": 4022
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.5380232558139534e-07,
+      "loss": 0.0054,
+      "step": 4023
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.537906976744186e-07,
+      "loss": 0.0062,
+      "step": 4024
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.537790697674418e-07,
+      "loss": 0.0048,
+      "step": 4025
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.537674418604651e-07,
+      "loss": 1.1412,
+      "step": 4026
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.537558139534884e-07,
+      "loss": 0.0046,
+      "step": 4027
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.537441860465116e-07,
+      "loss": 0.0067,
+      "step": 4028
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.5373255813953486e-07,
+      "loss": 0.0048,
+      "step": 4029
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.5372093023255815e-07,
+      "loss": 0.0082,
+      "step": 4030
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.537093023255814e-07,
+      "loss": 0.0048,
+      "step": 4031
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.536976744186046e-07,
+      "loss": 1.2435,
+      "step": 4032
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.536860465116279e-07,
+      "loss": 1.3195,
+      "step": 4033
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.5367441860465114e-07,
+      "loss": 0.0049,
+      "step": 4034
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.5366279069767443e-07,
+      "loss": 0.0084,
+      "step": 4035
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.5365116279069766e-07,
+      "loss": 0.0915,
+      "step": 4036
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.536395348837209e-07,
+      "loss": 0.0065,
+      "step": 4037
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.536279069767442e-07,
+      "loss": 0.0072,
+      "step": 4038
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.536162790697674e-07,
+      "loss": 0.0082,
+      "step": 4039
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.5360465116279066e-07,
+      "loss": 0.0069,
+      "step": 4040
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.5359302325581395e-07,
+      "loss": 0.0038,
+      "step": 4041
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.5358139534883723e-07,
+      "loss": 0.0046,
+      "step": 4042
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.5356976744186047e-07,
+      "loss": 0.0043,
+      "step": 4043
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.535581395348837e-07,
+      "loss": 0.0052,
+      "step": 4044
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.5354651162790694e-07,
+      "loss": 0.0059,
+      "step": 4045
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.5353488372093023e-07,
+      "loss": 0.0051,
+      "step": 4046
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.5352325581395346e-07,
+      "loss": 0.0044,
+      "step": 4047
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.5351162790697675e-07,
+      "loss": 0.0075,
+      "step": 4048
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.535e-07,
+      "loss": 0.0051,
+      "step": 4049
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.534883720930232e-07,
+      "loss": 0.0074,
+      "step": 4050
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.534767441860465e-07,
+      "loss": 0.0055,
+      "step": 4051
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.5346511627906975e-07,
+      "loss": 0.0084,
+      "step": 4052
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.53453488372093e-07,
+      "loss": 0.0037,
+      "step": 4053
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.5344186046511627e-07,
+      "loss": 0.0064,
+      "step": 4054
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.5343023255813956e-07,
+      "loss": 0.0069,
+      "step": 4055
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.534186046511628e-07,
+      "loss": 0.0199,
+      "step": 4056
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.5340697674418603e-07,
+      "loss": 0.0042,
+      "step": 4057
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.5339534883720926e-07,
+      "loss": 0.0061,
+      "step": 4058
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.5338372093023255e-07,
+      "loss": 0.0043,
+      "step": 4059
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.533720930232558e-07,
+      "loss": 0.0052,
+      "step": 4060
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.533604651162791e-07,
+      "loss": 0.0073,
+      "step": 4061
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.533488372093023e-07,
+      "loss": 0.0091,
+      "step": 4062
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.533372093023256e-07,
+      "loss": 0.0183,
+      "step": 4063
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.5332558139534883e-07,
+      "loss": 0.0057,
+      "step": 4064
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.5331395348837207e-07,
+      "loss": 0.0043,
+      "step": 4065
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.533023255813953e-07,
+      "loss": 0.0041,
+      "step": 4066
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.532906976744186e-07,
+      "loss": 2.2398,
+      "step": 4067
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.532790697674419e-07,
+      "loss": 0.0078,
+      "step": 4068
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.532674418604651e-07,
+      "loss": 0.0062,
+      "step": 4069
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.5325581395348835e-07,
+      "loss": 0.0067,
+      "step": 4070
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.5324418604651164e-07,
+      "loss": 0.0068,
+      "step": 4071
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.532325581395349e-07,
+      "loss": 0.0056,
+      "step": 4072
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.532209302325581e-07,
+      "loss": 0.0037,
+      "step": 4073
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.532093023255814e-07,
+      "loss": 0.1416,
+      "step": 4074
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.5319767441860463e-07,
+      "loss": 1.6484,
+      "step": 4075
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.531860465116279e-07,
+      "loss": 0.1085,
+      "step": 4076
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.5317441860465116e-07,
+      "loss": 0.015,
+      "step": 4077
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.531627906976744e-07,
+      "loss": 0.0045,
+      "step": 4078
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.5315116279069763e-07,
+      "loss": 0.0057,
+      "step": 4079
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.531395348837209e-07,
+      "loss": 0.0712,
+      "step": 4080
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.531279069767442e-07,
+      "loss": 0.4356,
+      "step": 4081
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.5311627906976744e-07,
+      "loss": 0.0078,
+      "step": 4082
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.5310465116279067e-07,
+      "loss": 0.0069,
+      "step": 4083
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.5309302325581396e-07,
+      "loss": 0.006,
+      "step": 4084
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.530813953488372e-07,
+      "loss": 0.0036,
+      "step": 4085
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.5306976744186043e-07,
+      "loss": 0.0062,
+      "step": 4086
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.530581395348837e-07,
+      "loss": 0.005,
+      "step": 4087
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.5304651162790696e-07,
+      "loss": 0.0047,
+      "step": 4088
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.5303488372093024e-07,
+      "loss": 0.0161,
+      "step": 4089
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.530232558139535e-07,
+      "loss": 0.0056,
+      "step": 4090
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.530116279069767e-07,
+      "loss": 0.0048,
+      "step": 4091
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.53e-07,
+      "loss": 0.0057,
+      "step": 4092
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.5298837209302324e-07,
+      "loss": 0.0056,
+      "step": 4093
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.5297674418604647e-07,
+      "loss": 0.0039,
+      "step": 4094
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.5296511627906976e-07,
+      "loss": 1.9191,
+      "step": 4095
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.5295348837209305e-07,
+      "loss": 0.0075,
+      "step": 4096
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.529418604651163e-07,
+      "loss": 0.0089,
+      "step": 4097
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.529302325581395e-07,
+      "loss": 0.0049,
+      "step": 4098
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.5291860465116275e-07,
+      "loss": 0.0056,
+      "step": 4099
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.5290697674418604e-07,
+      "loss": 0.0064,
+      "step": 4100
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.528953488372093e-07,
+      "loss": 1.1254,
+      "step": 4101
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.5288372093023257e-07,
+      "loss": 0.0062,
+      "step": 4102
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.528720930232558e-07,
+      "loss": 0.0075,
+      "step": 4103
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.5286046511627904e-07,
+      "loss": 0.0075,
+      "step": 4104
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.528488372093023e-07,
+      "loss": 0.0059,
+      "step": 4105
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.5283720930232556e-07,
+      "loss": 0.031,
+      "step": 4106
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.528255813953488e-07,
+      "loss": 0.005,
+      "step": 4107
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.528139534883721e-07,
+      "loss": 0.4581,
+      "step": 4108
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.5280232558139537e-07,
+      "loss": 0.086,
+      "step": 4109
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.527906976744186e-07,
+      "loss": 0.0037,
+      "step": 4110
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.5277906976744184e-07,
+      "loss": 0.005,
+      "step": 4111
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.527674418604651e-07,
+      "loss": 0.0055,
+      "step": 4112
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.5275581395348837e-07,
+      "loss": 0.0084,
+      "step": 4113
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.527441860465116e-07,
+      "loss": 0.0055,
+      "step": 4114
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.527325581395349e-07,
+      "loss": 0.0047,
+      "step": 4115
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.527209302325581e-07,
+      "loss": 0.0054,
+      "step": 4116
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.527093023255814e-07,
+      "loss": 0.0051,
+      "step": 4117
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.5269767441860465e-07,
+      "loss": 0.1431,
+      "step": 4118
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.526860465116279e-07,
+      "loss": 0.0034,
+      "step": 4119
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.526744186046511e-07,
+      "loss": 0.0068,
+      "step": 4120
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.526627906976744e-07,
+      "loss": 0.0637,
+      "step": 4121
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.526511627906977e-07,
+      "loss": 1.4925,
+      "step": 4122
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.5263953488372093e-07,
+      "loss": 1.8599,
+      "step": 4123
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.5262790697674417e-07,
+      "loss": 0.0037,
+      "step": 4124
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.5261627906976745e-07,
+      "loss": 0.0038,
+      "step": 4125
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.526046511627907e-07,
+      "loss": 0.0064,
+      "step": 4126
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.525930232558139e-07,
+      "loss": 0.0044,
+      "step": 4127
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.525813953488372e-07,
+      "loss": 0.0068,
+      "step": 4128
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.5256976744186045e-07,
+      "loss": 0.0064,
+      "step": 4129
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.5255813953488374e-07,
+      "loss": 0.0053,
+      "step": 4130
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.5254651162790697e-07,
+      "loss": 0.0067,
+      "step": 4131
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.525348837209302e-07,
+      "loss": 0.0049,
+      "step": 4132
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.5252325581395344e-07,
+      "loss": 0.0069,
+      "step": 4133
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.5251162790697673e-07,
+      "loss": 0.0787,
+      "step": 4134
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.5249999999999996e-07,
+      "loss": 0.0051,
+      "step": 4135
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.5248837209302325e-07,
+      "loss": 0.0037,
+      "step": 4136
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.9835589941972921,
+      "eval_loss": 0.15391960740089417,
+      "eval_runtime": 1.5634,
+      "eval_samples_per_second": 661.384,
+      "eval_steps_per_second": 10.874,
+      "step": 4136
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.524767441860465e-07,
+      "loss": 0.0073,
+      "step": 4137
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.524651162790698e-07,
+      "loss": 0.0045,
+      "step": 4138
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.52453488372093e-07,
+      "loss": 1.1539,
+      "step": 4139
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.5244186046511625e-07,
+      "loss": 2.2052,
+      "step": 4140
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.5243023255813954e-07,
+      "loss": 0.0058,
+      "step": 4141
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.5241860465116277e-07,
+      "loss": 0.0045,
+      "step": 4142
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.5240697674418606e-07,
+      "loss": 1.1221,
+      "step": 4143
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.523953488372093e-07,
+      "loss": 2.6093,
+      "step": 4144
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.5238372093023253e-07,
+      "loss": 2.6265,
+      "step": 4145
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.523720930232558e-07,
+      "loss": 0.8886,
+      "step": 4146
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.5236046511627905e-07,
+      "loss": 0.0144,
+      "step": 4147
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.523488372093023e-07,
+      "loss": 0.0072,
+      "step": 4148
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.523372093023256e-07,
+      "loss": 0.0059,
+      "step": 4149
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.5232558139534886e-07,
+      "loss": 0.0039,
+      "step": 4150
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.523139534883721e-07,
+      "loss": 0.0043,
+      "step": 4151
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.5230232558139533e-07,
+      "loss": 0.0049,
+      "step": 4152
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.5229069767441857e-07,
+      "loss": 0.0069,
+      "step": 4153
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.5227906976744186e-07,
+      "loss": 0.0192,
+      "step": 4154
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.522674418604651e-07,
+      "loss": 0.0073,
+      "step": 4155
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.522558139534884e-07,
+      "loss": 0.0054,
+      "step": 4156
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.522441860465116e-07,
+      "loss": 0.006,
+      "step": 4157
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.5223255813953485e-07,
+      "loss": 0.0048,
+      "step": 4158
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.5222093023255814e-07,
+      "loss": 0.007,
+      "step": 4159
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.522093023255814e-07,
+      "loss": 0.0182,
+      "step": 4160
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.521976744186046e-07,
+      "loss": 0.0073,
+      "step": 4161
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.521860465116279e-07,
+      "loss": 0.0097,
+      "step": 4162
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.521744186046512e-07,
+      "loss": 0.9031,
+      "step": 4163
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.521627906976744e-07,
+      "loss": 0.0043,
+      "step": 4164
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.5215116279069766e-07,
+      "loss": 0.0116,
+      "step": 4165
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.521395348837209e-07,
+      "loss": 0.0044,
+      "step": 4166
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.521279069767442e-07,
+      "loss": 0.0182,
+      "step": 4167
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.521162790697674e-07,
+      "loss": 0.0067,
+      "step": 4168
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.521046511627907e-07,
+      "loss": 0.0059,
+      "step": 4169
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.5209302325581394e-07,
+      "loss": 0.0048,
+      "step": 4170
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.5208139534883723e-07,
+      "loss": 0.0073,
+      "step": 4171
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.5206976744186046e-07,
+      "loss": 0.0046,
+      "step": 4172
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.520581395348837e-07,
+      "loss": 0.0062,
+      "step": 4173
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.5204651162790693e-07,
+      "loss": 0.0066,
+      "step": 4174
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.520348837209302e-07,
+      "loss": 0.0037,
+      "step": 4175
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.520232558139535e-07,
+      "loss": 0.0086,
+      "step": 4176
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.5201162790697674e-07,
+      "loss": 0.0118,
+      "step": 4177
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.52e-07,
+      "loss": 0.0069,
+      "step": 4178
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.5198837209302327e-07,
+      "loss": 0.0052,
+      "step": 4179
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.519767441860465e-07,
+      "loss": 0.0656,
+      "step": 4180
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.5196511627906974e-07,
+      "loss": 0.046,
+      "step": 4181
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.5195348837209303e-07,
+      "loss": 0.0065,
+      "step": 4182
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.5194186046511626e-07,
+      "loss": 0.0074,
+      "step": 4183
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.5193023255813955e-07,
+      "loss": 0.0052,
+      "step": 4184
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.519186046511628e-07,
+      "loss": 0.008,
+      "step": 4185
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.51906976744186e-07,
+      "loss": 0.0048,
+      "step": 4186
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.5189534883720926e-07,
+      "loss": 0.0058,
+      "step": 4187
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.5188372093023254e-07,
+      "loss": 0.006,
+      "step": 4188
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.518720930232558e-07,
+      "loss": 0.0061,
+      "step": 4189
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.5186046511627907e-07,
+      "loss": 0.0061,
+      "step": 4190
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.518488372093023e-07,
+      "loss": 0.0115,
+      "step": 4191
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.518372093023256e-07,
+      "loss": 0.0053,
+      "step": 4192
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.518255813953488e-07,
+      "loss": 0.0054,
+      "step": 4193
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.5181395348837206e-07,
+      "loss": 0.0042,
+      "step": 4194
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.5180232558139535e-07,
+      "loss": 0.0112,
+      "step": 4195
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.517906976744186e-07,
+      "loss": 0.2612,
+      "step": 4196
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.5177906976744187e-07,
+      "loss": 0.0058,
+      "step": 4197
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.517674418604651e-07,
+      "loss": 0.0052,
+      "step": 4198
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.5175581395348834e-07,
+      "loss": 0.0054,
+      "step": 4199
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.5174418604651163e-07,
+      "loss": 0.0322,
+      "step": 4200
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.5173255813953487e-07,
+      "loss": 0.004,
+      "step": 4201
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.517209302325581e-07,
+      "loss": 0.0106,
+      "step": 4202
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.517093023255814e-07,
+      "loss": 0.0036,
+      "step": 4203
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.516976744186047e-07,
+      "loss": 0.0056,
+      "step": 4204
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.516860465116279e-07,
+      "loss": 0.0071,
+      "step": 4205
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.5167441860465115e-07,
+      "loss": 0.0052,
+      "step": 4206
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.516627906976744e-07,
+      "loss": 0.0099,
+      "step": 4207
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.5165116279069767e-07,
+      "loss": 0.0057,
+      "step": 4208
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.516395348837209e-07,
+      "loss": 0.0048,
+      "step": 4209
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.516279069767442e-07,
+      "loss": 0.0066,
+      "step": 4210
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.5161627906976743e-07,
+      "loss": 0.0039,
+      "step": 4211
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.5160465116279067e-07,
+      "loss": 0.0056,
+      "step": 4212
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.5159302325581395e-07,
+      "loss": 0.0058,
+      "step": 4213
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.515813953488372e-07,
+      "loss": 0.0078,
+      "step": 4214
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.515697674418604e-07,
+      "loss": 1.6933,
+      "step": 4215
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.515581395348837e-07,
+      "loss": 0.0064,
+      "step": 4216
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.51546511627907e-07,
+      "loss": 0.0069,
+      "step": 4217
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.5153488372093024e-07,
+      "loss": 0.0063,
+      "step": 4218
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.5152325581395347e-07,
+      "loss": 0.0075,
+      "step": 4219
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.515116279069767e-07,
+      "loss": 0.0727,
+      "step": 4220
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.515e-07,
+      "loss": 0.0069,
+      "step": 4221
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.5148837209302323e-07,
+      "loss": 0.4957,
+      "step": 4222
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.514767441860465e-07,
+      "loss": 0.006,
+      "step": 4223
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.5146511627906975e-07,
+      "loss": 0.2381,
+      "step": 4224
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.5145348837209304e-07,
+      "loss": 0.0065,
+      "step": 4225
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.514418604651163e-07,
+      "loss": 1.5487,
+      "step": 4226
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.514302325581395e-07,
+      "loss": 0.0053,
+      "step": 4227
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.5141860465116275e-07,
+      "loss": 0.007,
+      "step": 4228
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.5140697674418604e-07,
+      "loss": 0.0054,
+      "step": 4229
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.5139534883720927e-07,
+      "loss": 0.0036,
+      "step": 4230
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.5138372093023256e-07,
+      "loss": 0.2022,
+      "step": 4231
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.513720930232558e-07,
+      "loss": 2.9218,
+      "step": 4232
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.513604651162791e-07,
+      "loss": 0.006,
+      "step": 4233
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.513488372093023e-07,
+      "loss": 0.0057,
+      "step": 4234
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.5133720930232555e-07,
+      "loss": 0.0049,
+      "step": 4235
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.5132558139534884e-07,
+      "loss": 0.0074,
+      "step": 4236
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.513139534883721e-07,
+      "loss": 0.0089,
+      "step": 4237
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.5130232558139537e-07,
+      "loss": 0.0285,
+      "step": 4238
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.512906976744186e-07,
+      "loss": 0.0088,
+      "step": 4239
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.5127906976744184e-07,
+      "loss": 0.0097,
+      "step": 4240
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.5126744186046507e-07,
+      "loss": 0.0038,
+      "step": 4241
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.5125581395348836e-07,
+      "loss": 0.0035,
+      "step": 4242
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.512441860465116e-07,
+      "loss": 0.0071,
+      "step": 4243
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.512325581395349e-07,
+      "loss": 2.3995,
+      "step": 4244
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.512209302325581e-07,
+      "loss": 0.0045,
+      "step": 4245
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.512093023255814e-07,
+      "loss": 0.0067,
+      "step": 4246
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.5119767441860464e-07,
+      "loss": 0.005,
+      "step": 4247
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.511860465116279e-07,
+      "loss": 0.0064,
+      "step": 4248
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.5117441860465116e-07,
+      "loss": 0.0073,
+      "step": 4249
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.511627906976744e-07,
+      "loss": 0.0047,
+      "step": 4250
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.511511627906977e-07,
+      "loss": 1.3522,
+      "step": 4251
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.511395348837209e-07,
+      "loss": 0.8111,
+      "step": 4252
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.5112790697674416e-07,
+      "loss": 0.0041,
+      "step": 4253
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.5111627906976745e-07,
+      "loss": 0.3966,
+      "step": 4254
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.511046511627907e-07,
+      "loss": 0.0084,
+      "step": 4255
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.510930232558139e-07,
+      "loss": 0.0043,
+      "step": 4256
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.510813953488372e-07,
+      "loss": 0.0055,
+      "step": 4257
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.510697674418605e-07,
+      "loss": 0.005,
+      "step": 4258
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.5105813953488373e-07,
+      "loss": 0.0042,
+      "step": 4259
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.5104651162790696e-07,
+      "loss": 0.006,
+      "step": 4260
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.510348837209302e-07,
+      "loss": 0.0054,
+      "step": 4261
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.5102325581395343e-07,
+      "loss": 0.008,
+      "step": 4262
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.510116279069767e-07,
+      "loss": 0.0089,
+      "step": 4263
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.51e-07,
+      "loss": 0.0068,
+      "step": 4264
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.5098837209302325e-07,
+      "loss": 0.0094,
+      "step": 4265
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.509767441860465e-07,
+      "loss": 1.0731,
+      "step": 4266
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.5096511627906977e-07,
+      "loss": 0.0076,
+      "step": 4267
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.50953488372093e-07,
+      "loss": 0.0041,
+      "step": 4268
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.5094186046511624e-07,
+      "loss": 0.0041,
+      "step": 4269
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.5093023255813953e-07,
+      "loss": 0.0066,
+      "step": 4270
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.509186046511628e-07,
+      "loss": 0.0083,
+      "step": 4271
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.5090697674418605e-07,
+      "loss": 0.0356,
+      "step": 4272
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.508953488372093e-07,
+      "loss": 0.0052,
+      "step": 4273
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.508837209302325e-07,
+      "loss": 0.006,
+      "step": 4274
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.508720930232558e-07,
+      "loss": 0.0096,
+      "step": 4275
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.5086046511627905e-07,
+      "loss": 0.0092,
+      "step": 4276
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.5084883720930233e-07,
+      "loss": 0.0053,
+      "step": 4277
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.5083720930232557e-07,
+      "loss": 0.006,
+      "step": 4278
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.5082558139534886e-07,
+      "loss": 0.0067,
+      "step": 4279
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.508139534883721e-07,
+      "loss": 0.0053,
+      "step": 4280
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.5080232558139533e-07,
+      "loss": 0.0042,
+      "step": 4281
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.5079069767441856e-07,
+      "loss": 0.0134,
+      "step": 4282
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.5077906976744185e-07,
+      "loss": 0.0066,
+      "step": 4283
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.507674418604651e-07,
+      "loss": 0.0074,
+      "step": 4284
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.507558139534884e-07,
+      "loss": 0.0052,
+      "step": 4285
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.507441860465116e-07,
+      "loss": 0.0037,
+      "step": 4286
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.507325581395349e-07,
+      "loss": 0.0055,
+      "step": 4287
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.5072093023255813e-07,
+      "loss": 0.0299,
+      "step": 4288
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.5070930232558137e-07,
+      "loss": 0.0448,
+      "step": 4289
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.5069767441860466e-07,
+      "loss": 0.006,
+      "step": 4290
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.506860465116279e-07,
+      "loss": 0.0063,
+      "step": 4291
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.506744186046512e-07,
+      "loss": 0.1598,
+      "step": 4292
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.506627906976744e-07,
+      "loss": 0.0043,
+      "step": 4293
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.5065116279069765e-07,
+      "loss": 0.0045,
+      "step": 4294
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.506395348837209e-07,
+      "loss": 0.0064,
+      "step": 4295
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.506279069767442e-07,
+      "loss": 0.0069,
+      "step": 4296
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.506162790697674e-07,
+      "loss": 0.0054,
+      "step": 4297
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.506046511627907e-07,
+      "loss": 0.007,
+      "step": 4298
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.5059302325581393e-07,
+      "loss": 0.0032,
+      "step": 4299
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.505813953488372e-07,
+      "loss": 0.0039,
+      "step": 4300
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.5056976744186046e-07,
+      "loss": 0.0075,
+      "step": 4301
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.505581395348837e-07,
+      "loss": 0.0073,
+      "step": 4302
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.50546511627907e-07,
+      "loss": 0.0043,
+      "step": 4303
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.505348837209302e-07,
+      "loss": 0.0038,
+      "step": 4304
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.505232558139535e-07,
+      "loss": 0.0077,
+      "step": 4305
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.5051162790697674e-07,
+      "loss": 0.0062,
+      "step": 4306
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.5049999999999997e-07,
+      "loss": 0.0063,
+      "step": 4307
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.5048837209302326e-07,
+      "loss": 0.012,
+      "step": 4308
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.504767441860465e-07,
+      "loss": 0.0052,
+      "step": 4309
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.5046511627906973e-07,
+      "loss": 0.006,
+      "step": 4310
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.50453488372093e-07,
+      "loss": 0.0048,
+      "step": 4311
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.504418604651163e-07,
+      "loss": 0.0055,
+      "step": 4312
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.5043023255813954e-07,
+      "loss": 0.0057,
+      "step": 4313
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.504186046511628e-07,
+      "loss": 0.0073,
+      "step": 4314
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.50406976744186e-07,
+      "loss": 0.0054,
+      "step": 4315
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.5039534883720925e-07,
+      "loss": 0.0062,
+      "step": 4316
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.5038372093023254e-07,
+      "loss": 0.0224,
+      "step": 4317
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.503720930232558e-07,
+      "loss": 0.0072,
+      "step": 4318
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.5036046511627906e-07,
+      "loss": 0.0064,
+      "step": 4319
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.503488372093023e-07,
+      "loss": 0.0116,
+      "step": 4320
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.503372093023256e-07,
+      "loss": 0.0052,
+      "step": 4321
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.503255813953488e-07,
+      "loss": 0.0048,
+      "step": 4322
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.5031395348837205e-07,
+      "loss": 0.3763,
+      "step": 4323
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.5030232558139534e-07,
+      "loss": 0.0061,
+      "step": 4324
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.5029069767441863e-07,
+      "loss": 0.0036,
+      "step": 4325
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.5027906976744187e-07,
+      "loss": 0.4681,
+      "step": 4326
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.502674418604651e-07,
+      "loss": 0.0053,
+      "step": 4327
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.5025581395348834e-07,
+      "loss": 0.007,
+      "step": 4328
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.502441860465116e-07,
+      "loss": 0.0045,
+      "step": 4329
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.5023255813953486e-07,
+      "loss": 0.0053,
+      "step": 4330
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.5022093023255815e-07,
+      "loss": 0.4277,
+      "step": 4331
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.502093023255814e-07,
+      "loss": 0.0047,
+      "step": 4332
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.5019767441860467e-07,
+      "loss": 0.005,
+      "step": 4333
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.501860465116279e-07,
+      "loss": 0.006,
+      "step": 4334
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.5017441860465114e-07,
+      "loss": 0.0037,
+      "step": 4335
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.501627906976744e-07,
+      "loss": 0.0031,
+      "step": 4336
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.5015116279069767e-07,
+      "loss": 0.0038,
+      "step": 4337
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.501395348837209e-07,
+      "loss": 0.8897,
+      "step": 4338
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.501279069767442e-07,
+      "loss": 0.0064,
+      "step": 4339
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.501162790697674e-07,
+      "loss": 0.0056,
+      "step": 4340
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.501046511627907e-07,
+      "loss": 0.0068,
+      "step": 4341
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.5009302325581395e-07,
+      "loss": 0.0044,
+      "step": 4342
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.500813953488372e-07,
+      "loss": 0.0062,
+      "step": 4343
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.5006976744186047e-07,
+      "loss": 0.0072,
+      "step": 4344
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.500581395348837e-07,
+      "loss": 0.0051,
+      "step": 4345
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.50046511627907e-07,
+      "loss": 0.0071,
+      "step": 4346
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.5003488372093023e-07,
+      "loss": 1.7116,
+      "step": 4347
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.5002325581395347e-07,
+      "loss": 0.0061,
+      "step": 4348
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.500116279069767e-07,
+      "loss": 0.5006,
+      "step": 4349
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.5e-07,
+      "loss": 0.0057,
+      "step": 4350
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.499883720930232e-07,
+      "loss": 0.0907,
+      "step": 4351
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.499767441860465e-07,
+      "loss": 0.0033,
+      "step": 4352
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.4996511627906975e-07,
+      "loss": 0.0061,
+      "step": 4353
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.4995348837209304e-07,
+      "loss": 0.0397,
+      "step": 4354
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.4994186046511627e-07,
+      "loss": 0.0048,
+      "step": 4355
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.499302325581395e-07,
+      "loss": 0.004,
+      "step": 4356
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.4991860465116274e-07,
+      "loss": 0.0051,
+      "step": 4357
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.4990697674418603e-07,
+      "loss": 0.0053,
+      "step": 4358
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.498953488372093e-07,
+      "loss": 0.0058,
+      "step": 4359
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.4988372093023255e-07,
+      "loss": 0.0101,
+      "step": 4360
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.498720930232558e-07,
+      "loss": 0.05,
+      "step": 4361
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.498604651162791e-07,
+      "loss": 0.005,
+      "step": 4362
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.498488372093023e-07,
+      "loss": 0.004,
+      "step": 4363
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.4983720930232555e-07,
+      "loss": 0.0123,
+      "step": 4364
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.4982558139534883e-07,
+      "loss": 0.006,
+      "step": 4365
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.498139534883721e-07,
+      "loss": 0.0067,
+      "step": 4366
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.4980232558139536e-07,
+      "loss": 0.3039,
+      "step": 4367
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.497906976744186e-07,
+      "loss": 0.0086,
+      "step": 4368
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.4977906976744183e-07,
+      "loss": 0.0049,
+      "step": 4369
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.4976744186046506e-07,
+      "loss": 0.0037,
+      "step": 4370
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.4975581395348835e-07,
+      "loss": 0.0042,
+      "step": 4371
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.4974418604651164e-07,
+      "loss": 0.0039,
+      "step": 4372
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.497325581395349e-07,
+      "loss": 0.0075,
+      "step": 4373
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.497209302325581e-07,
+      "loss": 0.0034,
+      "step": 4374
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.497093023255814e-07,
+      "loss": 0.0045,
+      "step": 4375
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.4969767441860463e-07,
+      "loss": 0.0049,
+      "step": 4376
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.4968604651162787e-07,
+      "loss": 0.0037,
+      "step": 4377
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.4967441860465116e-07,
+      "loss": 0.0037,
+      "step": 4378
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.496627906976744e-07,
+      "loss": 0.0119,
+      "step": 4379
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.496511627906977e-07,
+      "loss": 0.0106,
+      "step": 4380
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.496395348837209e-07,
+      "loss": 0.02,
+      "step": 4381
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.4962790697674415e-07,
+      "loss": 0.0064,
+      "step": 4382
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.4961627906976744e-07,
+      "loss": 0.0041,
+      "step": 4383
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.496046511627907e-07,
+      "loss": 0.0037,
+      "step": 4384
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.4959302325581396e-07,
+      "loss": 0.01,
+      "step": 4385
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.495813953488372e-07,
+      "loss": 0.0036,
+      "step": 4386
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.495697674418605e-07,
+      "loss": 0.0131,
+      "step": 4387
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.495581395348837e-07,
+      "loss": 0.006,
+      "step": 4388
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.4954651162790696e-07,
+      "loss": 0.0037,
+      "step": 4389
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.495348837209302e-07,
+      "loss": 0.0041,
+      "step": 4390
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.495232558139535e-07,
+      "loss": 0.0058,
+      "step": 4391
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.495116279069767e-07,
+      "loss": 0.0047,
+      "step": 4392
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.495e-07,
+      "loss": 0.5697,
+      "step": 4393
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.4948837209302324e-07,
+      "loss": 0.0059,
+      "step": 4394
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.494767441860465e-07,
+      "loss": 0.0289,
+      "step": 4395
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.4946511627906976e-07,
+      "loss": 0.0045,
+      "step": 4396
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.49453488372093e-07,
+      "loss": 0.0043,
+      "step": 4397
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.494418604651163e-07,
+      "loss": 0.0061,
+      "step": 4398
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.494302325581395e-07,
+      "loss": 0.0063,
+      "step": 4399
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.494186046511628e-07,
+      "loss": 0.0055,
+      "step": 4400
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.4940697674418604e-07,
+      "loss": 0.0059,
+      "step": 4401
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.493953488372093e-07,
+      "loss": 0.009,
+      "step": 4402
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.493837209302325e-07,
+      "loss": 0.0048,
+      "step": 4403
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.493720930232558e-07,
+      "loss": 0.0152,
+      "step": 4404
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.4936046511627904e-07,
+      "loss": 0.0076,
+      "step": 4405
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.4934883720930233e-07,
+      "loss": 0.0046,
+      "step": 4406
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.4933720930232556e-07,
+      "loss": 0.0063,
+      "step": 4407
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.4932558139534885e-07,
+      "loss": 0.0049,
+      "step": 4408
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.493139534883721e-07,
+      "loss": 0.0034,
+      "step": 4409
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.493023255813953e-07,
+      "loss": 0.0253,
+      "step": 4410
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.4929069767441856e-07,
+      "loss": 0.0037,
+      "step": 4411
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.4927906976744184e-07,
+      "loss": 0.0064,
+      "step": 4412
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.4926744186046513e-07,
+      "loss": 0.0056,
+      "step": 4413
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.4925581395348837e-07,
+      "loss": 0.0029,
+      "step": 4414
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.492441860465116e-07,
+      "loss": 0.0043,
+      "step": 4415
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.492325581395349e-07,
+      "loss": 0.005,
+      "step": 4416
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.492209302325581e-07,
+      "loss": 0.0039,
+      "step": 4417
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.4920930232558136e-07,
+      "loss": 0.0041,
+      "step": 4418
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.4919767441860465e-07,
+      "loss": 0.0039,
+      "step": 4419
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.4918604651162794e-07,
+      "loss": 0.0038,
+      "step": 4420
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.4917441860465117e-07,
+      "loss": 0.0236,
+      "step": 4421
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.491627906976744e-07,
+      "loss": 0.0033,
+      "step": 4422
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.4915116279069764e-07,
+      "loss": 0.0036,
+      "step": 4423
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.491395348837209e-07,
+      "loss": 0.032,
+      "step": 4424
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.4912790697674417e-07,
+      "loss": 0.0048,
+      "step": 4425
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.4911627906976746e-07,
+      "loss": 0.0054,
+      "step": 4426
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.491046511627907e-07,
+      "loss": 0.4838,
+      "step": 4427
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.490930232558139e-07,
+      "loss": 0.006,
+      "step": 4428
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.490813953488372e-07,
+      "loss": 0.0056,
+      "step": 4429
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.4906976744186045e-07,
+      "loss": 0.0046,
+      "step": 4430
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.490581395348837e-07,
+      "loss": 0.004,
+      "step": 4431
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.4904651162790697e-07,
+      "loss": 0.0038,
+      "step": 4432
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.490348837209302e-07,
+      "loss": 0.2851,
+      "step": 4433
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.490232558139535e-07,
+      "loss": 0.0086,
+      "step": 4434
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.4901162790697673e-07,
+      "loss": 0.4247,
+      "step": 4435
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.4899999999999997e-07,
+      "loss": 0.0066,
+      "step": 4436
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.4898837209302325e-07,
+      "loss": 0.0044,
+      "step": 4437
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.489767441860465e-07,
+      "loss": 0.007,
+      "step": 4438
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.489651162790698e-07,
+      "loss": 0.7517,
+      "step": 4439
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.48953488372093e-07,
+      "loss": 0.0034,
+      "step": 4440
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.489418604651163e-07,
+      "loss": 0.004,
+      "step": 4441
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.4893023255813954e-07,
+      "loss": 0.0039,
+      "step": 4442
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.4891860465116277e-07,
+      "loss": 0.0062,
+      "step": 4443
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.48906976744186e-07,
+      "loss": 0.0051,
+      "step": 4444
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.488953488372093e-07,
+      "loss": 0.0045,
+      "step": 4445
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.4888372093023253e-07,
+      "loss": 2.6601,
+      "step": 4446
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.488720930232558e-07,
+      "loss": 0.0336,
+      "step": 4447
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.4886046511627905e-07,
+      "loss": 0.7918,
+      "step": 4448
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.488488372093023e-07,
+      "loss": 0.0055,
+      "step": 4449
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.488372093023256e-07,
+      "loss": 0.009,
+      "step": 4450
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.488255813953488e-07,
+      "loss": 0.0039,
+      "step": 4451
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.488139534883721e-07,
+      "loss": 0.0057,
+      "step": 4452
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.4880232558139534e-07,
+      "loss": 0.6753,
+      "step": 4453
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.487906976744186e-07,
+      "loss": 0.0034,
+      "step": 4454
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.4877906976744186e-07,
+      "loss": 0.0036,
+      "step": 4455
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.487674418604651e-07,
+      "loss": 0.0046,
+      "step": 4456
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.4875581395348833e-07,
+      "loss": 0.0039,
+      "step": 4457
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.487441860465116e-07,
+      "loss": 0.0076,
+      "step": 4458
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.4873255813953485e-07,
+      "loss": 0.003,
+      "step": 4459
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.4872093023255814e-07,
+      "loss": 0.0048,
+      "step": 4460
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.487093023255814e-07,
+      "loss": 0.0063,
+      "step": 4461
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.4869767441860467e-07,
+      "loss": 0.0069,
+      "step": 4462
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.486860465116279e-07,
+      "loss": 0.006,
+      "step": 4463
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.4867441860465114e-07,
+      "loss": 0.0036,
+      "step": 4464
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.4866279069767437e-07,
+      "loss": 0.0092,
+      "step": 4465
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.4865116279069766e-07,
+      "loss": 0.0039,
+      "step": 4466
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.4863953488372095e-07,
+      "loss": 0.0141,
+      "step": 4467
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.486279069767442e-07,
+      "loss": 0.0043,
+      "step": 4468
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.486162790697674e-07,
+      "loss": 0.0051,
+      "step": 4469
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.486046511627907e-07,
+      "loss": 0.003,
+      "step": 4470
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.4859302325581394e-07,
+      "loss": 0.0048,
+      "step": 4471
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.485813953488372e-07,
+      "loss": 0.0043,
+      "step": 4472
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.4856976744186046e-07,
+      "loss": 0.0043,
+      "step": 4473
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.485581395348837e-07,
+      "loss": 0.0052,
+      "step": 4474
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.48546511627907e-07,
+      "loss": 0.005,
+      "step": 4475
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.485348837209302e-07,
+      "loss": 0.0038,
+      "step": 4476
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.4852325581395346e-07,
+      "loss": 0.0099,
+      "step": 4477
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.485116279069767e-07,
+      "loss": 0.005,
+      "step": 4478
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.485e-07,
+      "loss": 0.0062,
+      "step": 4479
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.4848837209302327e-07,
+      "loss": 0.0073,
+      "step": 4480
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.484767441860465e-07,
+      "loss": 0.0055,
+      "step": 4481
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.4846511627906974e-07,
+      "loss": 0.0061,
+      "step": 4482
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.4845348837209303e-07,
+      "loss": 0.0053,
+      "step": 4483
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.4844186046511626e-07,
+      "loss": 0.0067,
+      "step": 4484
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.484302325581395e-07,
+      "loss": 0.0047,
+      "step": 4485
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.484186046511628e-07,
+      "loss": 0.0043,
+      "step": 4486
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.48406976744186e-07,
+      "loss": 0.0076,
+      "step": 4487
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.483953488372093e-07,
+      "loss": 0.0038,
+      "step": 4488
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.4838372093023255e-07,
+      "loss": 0.153,
+      "step": 4489
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.483720930232558e-07,
+      "loss": 0.0039,
+      "step": 4490
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.4836046511627907e-07,
+      "loss": 0.0059,
+      "step": 4491
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.483488372093023e-07,
+      "loss": 0.0153,
+      "step": 4492
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.483372093023256e-07,
+      "loss": 0.0036,
+      "step": 4493
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.4832558139534883e-07,
+      "loss": 0.0348,
+      "step": 4494
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.483139534883721e-07,
+      "loss": 0.0042,
+      "step": 4495
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.4830232558139535e-07,
+      "loss": 0.0058,
+      "step": 4496
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.482906976744186e-07,
+      "loss": 0.0038,
+      "step": 4497
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.482790697674418e-07,
+      "loss": 0.0052,
+      "step": 4498
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.482674418604651e-07,
+      "loss": 0.0044,
+      "step": 4499
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.4825581395348835e-07,
+      "loss": 0.0051,
+      "step": 4500
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.4824418604651163e-07,
+      "loss": 0.0031,
+      "step": 4501
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.4823255813953487e-07,
+      "loss": 0.0034,
+      "step": 4502
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.482209302325581e-07,
+      "loss": 0.1335,
+      "step": 4503
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.482093023255814e-07,
+      "loss": 0.0055,
+      "step": 4504
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.4819767441860463e-07,
+      "loss": 0.0033,
+      "step": 4505
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.4818604651162786e-07,
+      "loss": 0.0104,
+      "step": 4506
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.4817441860465115e-07,
+      "loss": 0.0075,
+      "step": 4507
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.4816279069767444e-07,
+      "loss": 0.2046,
+      "step": 4508
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.481511627906977e-07,
+      "loss": 0.0058,
+      "step": 4509
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.481395348837209e-07,
+      "loss": 0.0077,
+      "step": 4510
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.4812790697674414e-07,
+      "loss": 0.1148,
+      "step": 4511
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.4811627906976743e-07,
+      "loss": 0.0046,
+      "step": 4512
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.4810465116279067e-07,
+      "loss": 0.0034,
+      "step": 4513
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.4809302325581396e-07,
+      "loss": 0.0038,
+      "step": 4514
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.480813953488372e-07,
+      "loss": 0.0302,
+      "step": 4515
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.480697674418605e-07,
+      "loss": 0.0056,
+      "step": 4516
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.480581395348837e-07,
+      "loss": 0.0045,
+      "step": 4517
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.4804651162790695e-07,
+      "loss": 0.033,
+      "step": 4518
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.480348837209302e-07,
+      "loss": 0.004,
+      "step": 4519
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.480232558139535e-07,
+      "loss": 0.0064,
+      "step": 4520
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.4801162790697676e-07,
+      "loss": 0.0049,
+      "step": 4521
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.48e-07,
+      "loss": 0.0052,
+      "step": 4522
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.4798837209302323e-07,
+      "loss": 0.0093,
+      "step": 4523
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.479767441860465e-07,
+      "loss": 0.0171,
+      "step": 4524
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.4796511627906976e-07,
+      "loss": 0.0061,
+      "step": 4525
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.47953488372093e-07,
+      "loss": 1.7499,
+      "step": 4526
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.479418604651163e-07,
+      "loss": 0.0331,
+      "step": 4527
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.479302325581395e-07,
+      "loss": 0.0038,
+      "step": 4528
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.479186046511628e-07,
+      "loss": 0.0067,
+      "step": 4529
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.4790697674418604e-07,
+      "loss": 0.0055,
+      "step": 4530
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.4789534883720927e-07,
+      "loss": 0.005,
+      "step": 4531
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.478837209302325e-07,
+      "loss": 0.0066,
+      "step": 4532
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.478720930232558e-07,
+      "loss": 0.0041,
+      "step": 4533
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.478604651162791e-07,
+      "loss": 0.0059,
+      "step": 4534
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.478488372093023e-07,
+      "loss": 0.0054,
+      "step": 4535
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.4783720930232556e-07,
+      "loss": 0.0062,
+      "step": 4536
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.4782558139534884e-07,
+      "loss": 0.0061,
+      "step": 4537
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.478139534883721e-07,
+      "loss": 0.0042,
+      "step": 4538
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.478023255813953e-07,
+      "loss": 0.0051,
+      "step": 4539
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.477906976744186e-07,
+      "loss": 0.0131,
+      "step": 4540
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.4777906976744184e-07,
+      "loss": 0.0048,
+      "step": 4541
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.477674418604651e-07,
+      "loss": 0.0085,
+      "step": 4542
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.4775581395348836e-07,
+      "loss": 0.0051,
+      "step": 4543
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.477441860465116e-07,
+      "loss": 0.0051,
+      "step": 4544
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.477325581395349e-07,
+      "loss": 0.0085,
+      "step": 4545
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.477209302325581e-07,
+      "loss": 0.0035,
+      "step": 4546
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.477093023255814e-07,
+      "loss": 0.0035,
+      "step": 4547
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.4769767441860464e-07,
+      "loss": 0.0047,
+      "step": 4548
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.4768604651162793e-07,
+      "loss": 0.0037,
+      "step": 4549
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.4767441860465117e-07,
+      "loss": 0.0063,
+      "step": 4550
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.476627906976744e-07,
+      "loss": 0.0031,
+      "step": 4551
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.4765116279069764e-07,
+      "loss": 0.0063,
+      "step": 4552
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.476395348837209e-07,
+      "loss": 0.0036,
+      "step": 4553
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.4762790697674416e-07,
+      "loss": 0.0449,
+      "step": 4554
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.4761627906976745e-07,
+      "loss": 0.0039,
+      "step": 4555
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.476046511627907e-07,
+      "loss": 0.0062,
+      "step": 4556
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.475930232558139e-07,
+      "loss": 0.0046,
+      "step": 4557
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.475813953488372e-07,
+      "loss": 0.1329,
+      "step": 4558
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.4756976744186044e-07,
+      "loss": 0.0033,
+      "step": 4559
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.475581395348837e-07,
+      "loss": 2.7138,
+      "step": 4560
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.4754651162790697e-07,
+      "loss": 0.007,
+      "step": 4561
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.4753488372093025e-07,
+      "loss": 0.0049,
+      "step": 4562
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.475232558139535e-07,
+      "loss": 0.0046,
+      "step": 4563
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.475116279069767e-07,
+      "loss": 0.0039,
+      "step": 4564
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.4749999999999996e-07,
+      "loss": 0.005,
+      "step": 4565
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.4748837209302325e-07,
+      "loss": 0.004,
+      "step": 4566
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.474767441860465e-07,
+      "loss": 0.0047,
+      "step": 4567
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.4746511627906977e-07,
+      "loss": 0.0735,
+      "step": 4568
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.47453488372093e-07,
+      "loss": 0.0051,
+      "step": 4569
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.474418604651163e-07,
+      "loss": 0.0049,
+      "step": 4570
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.4743023255813953e-07,
+      "loss": 0.0076,
+      "step": 4571
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.4741860465116277e-07,
+      "loss": 0.0035,
+      "step": 4572
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.47406976744186e-07,
+      "loss": 0.004,
+      "step": 4573
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.473953488372093e-07,
+      "loss": 0.0045,
+      "step": 4574
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.473837209302326e-07,
+      "loss": 0.0086,
+      "step": 4575
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.473720930232558e-07,
+      "loss": 0.0051,
+      "step": 4576
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.4736046511627905e-07,
+      "loss": 0.0128,
+      "step": 4577
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.4734883720930234e-07,
+      "loss": 0.005,
+      "step": 4578
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.4733720930232557e-07,
+      "loss": 0.0063,
+      "step": 4579
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.473255813953488e-07,
+      "loss": 0.0032,
+      "step": 4580
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.473139534883721e-07,
+      "loss": 0.0042,
+      "step": 4581
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.4730232558139533e-07,
+      "loss": 0.0048,
+      "step": 4582
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.472906976744186e-07,
+      "loss": 0.0029,
+      "step": 4583
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.4727906976744185e-07,
+      "loss": 0.0052,
+      "step": 4584
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.472674418604651e-07,
+      "loss": 0.0039,
+      "step": 4585
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.472558139534883e-07,
+      "loss": 0.0071,
+      "step": 4586
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.472441860465116e-07,
+      "loss": 0.2976,
+      "step": 4587
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.472325581395349e-07,
+      "loss": 0.0072,
+      "step": 4588
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.4722093023255813e-07,
+      "loss": 0.0037,
+      "step": 4589
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.4720930232558137e-07,
+      "loss": 0.0042,
+      "step": 4590
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.4719767441860466e-07,
+      "loss": 0.0044,
+      "step": 4591
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.471860465116279e-07,
+      "loss": 0.0045,
+      "step": 4592
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.4717441860465113e-07,
+      "loss": 0.0068,
+      "step": 4593
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.471627906976744e-07,
+      "loss": 0.005,
+      "step": 4594
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.4715116279069765e-07,
+      "loss": 0.031,
+      "step": 4595
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.4713953488372094e-07,
+      "loss": 0.0077,
+      "step": 4596
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.471279069767442e-07,
+      "loss": 0.0037,
+      "step": 4597
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.471162790697674e-07,
+      "loss": 0.0034,
+      "step": 4598
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.471046511627907e-07,
+      "loss": 0.0048,
+      "step": 4599
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.4709302325581393e-07,
+      "loss": 1.823,
+      "step": 4600
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.470813953488372e-07,
+      "loss": 0.0034,
+      "step": 4601
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.4706976744186046e-07,
+      "loss": 0.1922,
+      "step": 4602
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.4705813953488375e-07,
+      "loss": 0.0056,
+      "step": 4603
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.47046511627907e-07,
+      "loss": 0.0069,
+      "step": 4604
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.470348837209302e-07,
+      "loss": 0.0034,
+      "step": 4605
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.4702325581395345e-07,
+      "loss": 0.0033,
+      "step": 4606
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.4701162790697674e-07,
+      "loss": 0.0114,
+      "step": 4607
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.47e-07,
+      "loss": 0.0065,
+      "step": 4608
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.4698837209302326e-07,
+      "loss": 0.0059,
+      "step": 4609
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.469767441860465e-07,
+      "loss": 0.0032,
+      "step": 4610
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.4696511627906973e-07,
+      "loss": 0.0033,
+      "step": 4611
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.46953488372093e-07,
+      "loss": 0.0039,
+      "step": 4612
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.4694186046511626e-07,
+      "loss": 0.0051,
+      "step": 4613
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.469302325581395e-07,
+      "loss": 0.0049,
+      "step": 4614
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.469186046511628e-07,
+      "loss": 0.0039,
+      "step": 4615
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.4690697674418607e-07,
+      "loss": 0.0034,
+      "step": 4616
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.468953488372093e-07,
+      "loss": 0.0091,
+      "step": 4617
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.4688372093023254e-07,
+      "loss": 0.0065,
+      "step": 4618
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.468720930232558e-07,
+      "loss": 0.0986,
+      "step": 4619
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.4686046511627906e-07,
+      "loss": 0.0035,
+      "step": 4620
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.468488372093023e-07,
+      "loss": 0.0043,
+      "step": 4621
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.468372093023256e-07,
+      "loss": 0.014,
+      "step": 4622
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.468255813953488e-07,
+      "loss": 0.0057,
+      "step": 4623
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.468139534883721e-07,
+      "loss": 0.0044,
+      "step": 4624
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.4680232558139534e-07,
+      "loss": 0.007,
+      "step": 4625
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.467906976744186e-07,
+      "loss": 0.0054,
+      "step": 4626
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.467790697674418e-07,
+      "loss": 0.0033,
+      "step": 4627
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.467674418604651e-07,
+      "loss": 0.0046,
+      "step": 4628
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.467558139534884e-07,
+      "loss": 0.0042,
+      "step": 4629
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.4674418604651163e-07,
+      "loss": 0.0038,
+      "step": 4630
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.4673255813953486e-07,
+      "loss": 0.0278,
+      "step": 4631
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.4672093023255815e-07,
+      "loss": 0.0031,
+      "step": 4632
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.467093023255814e-07,
+      "loss": 0.0042,
+      "step": 4633
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.466976744186046e-07,
+      "loss": 0.007,
+      "step": 4634
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.466860465116279e-07,
+      "loss": 0.0051,
+      "step": 4635
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.4667441860465114e-07,
+      "loss": 0.3037,
+      "step": 4636
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.4666279069767443e-07,
+      "loss": 0.0064,
+      "step": 4637
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.4665116279069767e-07,
+      "loss": 0.0057,
+      "step": 4638
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.466395348837209e-07,
+      "loss": 0.0033,
+      "step": 4639
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.4662790697674414e-07,
+      "loss": 0.1261,
+      "step": 4640
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.466162790697674e-07,
+      "loss": 0.0043,
+      "step": 4641
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.466046511627907e-07,
+      "loss": 0.003,
+      "step": 4642
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.4659302325581395e-07,
+      "loss": 0.0047,
+      "step": 4643
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.465813953488372e-07,
+      "loss": 0.0049,
+      "step": 4644
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.4656976744186047e-07,
+      "loss": 0.0047,
+      "step": 4645
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.465581395348837e-07,
+      "loss": 0.3606,
+      "step": 4646
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.4654651162790694e-07,
+      "loss": 0.0063,
+      "step": 4647
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.4653488372093023e-07,
+      "loss": 0.003,
+      "step": 4648
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.4652325581395347e-07,
+      "loss": 0.0043,
+      "step": 4649
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.4651162790697676e-07,
+      "loss": 1.062,
+      "step": 4650
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.465e-07,
+      "loss": 0.0054,
+      "step": 4651
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.464883720930232e-07,
+      "loss": 0.0042,
+      "step": 4652
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.464767441860465e-07,
+      "loss": 0.0049,
+      "step": 4653
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.9835589941972921,
+      "eval_loss": 0.1566050946712494,
+      "eval_runtime": 1.5639,
+      "eval_samples_per_second": 661.175,
+      "eval_steps_per_second": 10.87,
+      "step": 4653
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.4646511627906975e-07,
+      "loss": 0.0049,
+      "step": 4654
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.46453488372093e-07,
+      "loss": 0.0039,
+      "step": 4655
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.4644186046511627e-07,
+      "loss": 0.003,
+      "step": 4656
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.4643023255813956e-07,
+      "loss": 0.0032,
+      "step": 4657
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.464186046511628e-07,
+      "loss": 0.0041,
+      "step": 4658
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.4640697674418603e-07,
+      "loss": 0.003,
+      "step": 4659
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.4639534883720927e-07,
+      "loss": 0.0056,
+      "step": 4660
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.4638372093023255e-07,
+      "loss": 0.0036,
+      "step": 4661
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.463720930232558e-07,
+      "loss": 0.0075,
+      "step": 4662
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.463604651162791e-07,
+      "loss": 0.0106,
+      "step": 4663
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.463488372093023e-07,
+      "loss": 0.0054,
+      "step": 4664
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.4633720930232555e-07,
+      "loss": 0.0052,
+      "step": 4665
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.4632558139534884e-07,
+      "loss": 0.0052,
+      "step": 4666
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.4631395348837207e-07,
+      "loss": 0.0086,
+      "step": 4667
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.463023255813953e-07,
+      "loss": 0.0036,
+      "step": 4668
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.462906976744186e-07,
+      "loss": 0.0055,
+      "step": 4669
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.462790697674419e-07,
+      "loss": 0.0048,
+      "step": 4670
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.462674418604651e-07,
+      "loss": 0.0067,
+      "step": 4671
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.4625581395348835e-07,
+      "loss": 0.0207,
+      "step": 4672
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.462441860465116e-07,
+      "loss": 0.0047,
+      "step": 4673
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.462325581395349e-07,
+      "loss": 0.0033,
+      "step": 4674
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.462209302325581e-07,
+      "loss": 0.0105,
+      "step": 4675
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.462093023255814e-07,
+      "loss": 0.0032,
+      "step": 4676
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.4619767441860464e-07,
+      "loss": 0.0045,
+      "step": 4677
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.461860465116279e-07,
+      "loss": 0.0069,
+      "step": 4678
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.4617441860465116e-07,
+      "loss": 0.1421,
+      "step": 4679
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.461627906976744e-07,
+      "loss": 0.0034,
+      "step": 4680
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.4615116279069763e-07,
+      "loss": 0.0045,
+      "step": 4681
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.461395348837209e-07,
+      "loss": 0.54,
+      "step": 4682
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.461279069767442e-07,
+      "loss": 0.0051,
+      "step": 4683
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.4611627906976744e-07,
+      "loss": 0.5346,
+      "step": 4684
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.461046511627907e-07,
+      "loss": 0.0039,
+      "step": 4685
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.4609302325581397e-07,
+      "loss": 0.0056,
+      "step": 4686
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.460813953488372e-07,
+      "loss": 0.0046,
+      "step": 4687
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.4606976744186044e-07,
+      "loss": 0.0049,
+      "step": 4688
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.460581395348837e-07,
+      "loss": 0.0358,
+      "step": 4689
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.4604651162790696e-07,
+      "loss": 0.8691,
+      "step": 4690
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.4603488372093025e-07,
+      "loss": 0.0038,
+      "step": 4691
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.460232558139535e-07,
+      "loss": 1.3072,
+      "step": 4692
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.460116279069767e-07,
+      "loss": 0.006,
+      "step": 4693
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.4599999999999995e-07,
+      "loss": 0.0054,
+      "step": 4694
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.4598837209302324e-07,
+      "loss": 0.0036,
+      "step": 4695
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.4597674418604653e-07,
+      "loss": 0.0107,
+      "step": 4696
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.4596511627906976e-07,
+      "loss": 0.0055,
+      "step": 4697
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.45953488372093e-07,
+      "loss": 0.0034,
+      "step": 4698
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.459418604651163e-07,
+      "loss": 0.0309,
+      "step": 4699
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.459302325581395e-07,
+      "loss": 0.0039,
+      "step": 4700
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.4591860465116276e-07,
+      "loss": 0.0043,
+      "step": 4701
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.4590697674418605e-07,
+      "loss": 0.0076,
+      "step": 4702
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.458953488372093e-07,
+      "loss": 0.0047,
+      "step": 4703
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.4588372093023257e-07,
+      "loss": 0.0038,
+      "step": 4704
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.458720930232558e-07,
+      "loss": 0.0064,
+      "step": 4705
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.4586046511627904e-07,
+      "loss": 0.0045,
+      "step": 4706
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.4584883720930233e-07,
+      "loss": 0.0037,
+      "step": 4707
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.4583720930232556e-07,
+      "loss": 0.0046,
+      "step": 4708
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.458255813953488e-07,
+      "loss": 0.0047,
+      "step": 4709
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.458139534883721e-07,
+      "loss": 0.0045,
+      "step": 4710
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.458023255813954e-07,
+      "loss": 0.0046,
+      "step": 4711
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.457906976744186e-07,
+      "loss": 0.0032,
+      "step": 4712
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.4577906976744185e-07,
+      "loss": 0.0053,
+      "step": 4713
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.457674418604651e-07,
+      "loss": 0.0042,
+      "step": 4714
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.4575581395348837e-07,
+      "loss": 0.0035,
+      "step": 4715
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.457441860465116e-07,
+      "loss": 0.3577,
+      "step": 4716
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.457325581395349e-07,
+      "loss": 0.0038,
+      "step": 4717
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.4572093023255813e-07,
+      "loss": 0.0028,
+      "step": 4718
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.4570930232558136e-07,
+      "loss": 0.003,
+      "step": 4719
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.4569767441860465e-07,
+      "loss": 0.031,
+      "step": 4720
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.456860465116279e-07,
+      "loss": 0.0039,
+      "step": 4721
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.456744186046511e-07,
+      "loss": 0.0038,
+      "step": 4722
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.456627906976744e-07,
+      "loss": 0.0073,
+      "step": 4723
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.456511627906977e-07,
+      "loss": 0.0028,
+      "step": 4724
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.4563953488372093e-07,
+      "loss": 0.0055,
+      "step": 4725
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.4562790697674417e-07,
+      "loss": 0.0036,
+      "step": 4726
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.456162790697674e-07,
+      "loss": 0.0053,
+      "step": 4727
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.456046511627907e-07,
+      "loss": 0.004,
+      "step": 4728
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.4559302325581393e-07,
+      "loss": 0.0061,
+      "step": 4729
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.455813953488372e-07,
+      "loss": 0.004,
+      "step": 4730
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.4556976744186045e-07,
+      "loss": 0.0095,
+      "step": 4731
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.4555813953488374e-07,
+      "loss": 0.008,
+      "step": 4732
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.45546511627907e-07,
+      "loss": 0.6151,
+      "step": 4733
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.455348837209302e-07,
+      "loss": 0.0039,
+      "step": 4734
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.4552325581395344e-07,
+      "loss": 0.0043,
+      "step": 4735
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.4551162790697673e-07,
+      "loss": 0.0785,
+      "step": 4736
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.455e-07,
+      "loss": 0.003,
+      "step": 4737
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.4548837209302326e-07,
+      "loss": 0.0038,
+      "step": 4738
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.454767441860465e-07,
+      "loss": 0.0048,
+      "step": 4739
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.454651162790698e-07,
+      "loss": 0.0053,
+      "step": 4740
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.45453488372093e-07,
+      "loss": 0.0123,
+      "step": 4741
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4544186046511625e-07,
+      "loss": 0.0065,
+      "step": 4742
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4543023255813954e-07,
+      "loss": 0.0024,
+      "step": 4743
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.454186046511628e-07,
+      "loss": 0.0108,
+      "step": 4744
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4540697674418606e-07,
+      "loss": 0.0038,
+      "step": 4745
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.453953488372093e-07,
+      "loss": 0.0094,
+      "step": 4746
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4538372093023253e-07,
+      "loss": 0.0039,
+      "step": 4747
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4537209302325577e-07,
+      "loss": 0.0863,
+      "step": 4748
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4536046511627906e-07,
+      "loss": 0.0035,
+      "step": 4749
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.453488372093023e-07,
+      "loss": 0.0045,
+      "step": 4750
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.453372093023256e-07,
+      "loss": 0.006,
+      "step": 4751
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.453255813953488e-07,
+      "loss": 0.0036,
+      "step": 4752
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.453139534883721e-07,
+      "loss": 0.0049,
+      "step": 4753
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4530232558139534e-07,
+      "loss": 0.0052,
+      "step": 4754
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4529069767441857e-07,
+      "loss": 0.056,
+      "step": 4755
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4527906976744186e-07,
+      "loss": 0.0033,
+      "step": 4756
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.452674418604651e-07,
+      "loss": 0.0035,
+      "step": 4757
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.452558139534884e-07,
+      "loss": 0.0066,
+      "step": 4758
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.452441860465116e-07,
+      "loss": 0.0064,
+      "step": 4759
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.4523255813953486e-07,
+      "loss": 0.0087,
+      "step": 4760
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.4522093023255814e-07,
+      "loss": 0.0049,
+      "step": 4761
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.452093023255814e-07,
+      "loss": 0.0028,
+      "step": 4762
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.451976744186046e-07,
+      "loss": 0.0062,
+      "step": 4763
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.451860465116279e-07,
+      "loss": 0.0039,
+      "step": 4764
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.451744186046512e-07,
+      "loss": 0.0046,
+      "step": 4765
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.451627906976744e-07,
+      "loss": 0.0059,
+      "step": 4766
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.4515116279069766e-07,
+      "loss": 0.0054,
+      "step": 4767
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.451395348837209e-07,
+      "loss": 0.0032,
+      "step": 4768
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.451279069767442e-07,
+      "loss": 0.2202,
+      "step": 4769
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.451162790697674e-07,
+      "loss": 0.0034,
+      "step": 4770
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.451046511627907e-07,
+      "loss": 0.0032,
+      "step": 4771
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.4509302325581394e-07,
+      "loss": 0.6435,
+      "step": 4772
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.450813953488372e-07,
+      "loss": 0.0165,
+      "step": 4773
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.4506976744186047e-07,
+      "loss": 0.0044,
+      "step": 4774
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.450581395348837e-07,
+      "loss": 0.0028,
+      "step": 4775
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.4504651162790694e-07,
+      "loss": 1.4704,
+      "step": 4776
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.450348837209302e-07,
+      "loss": 0.0055,
+      "step": 4777
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.450232558139535e-07,
+      "loss": 0.0056,
+      "step": 4778
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.4501162790697675e-07,
+      "loss": 0.0047,
+      "step": 4779
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.45e-07,
+      "loss": 0.0039,
+      "step": 4780
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.449883720930232e-07,
+      "loss": 0.0031,
+      "step": 4781
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.449767441860465e-07,
+      "loss": 0.003,
+      "step": 4782
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.4496511627906974e-07,
+      "loss": 0.0055,
+      "step": 4783
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.4495348837209303e-07,
+      "loss": 0.0031,
+      "step": 4784
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.4494186046511627e-07,
+      "loss": 0.0034,
+      "step": 4785
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.4493023255813955e-07,
+      "loss": 0.005,
+      "step": 4786
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.449186046511628e-07,
+      "loss": 0.0042,
+      "step": 4787
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.44906976744186e-07,
+      "loss": 0.0034,
+      "step": 4788
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.4489534883720926e-07,
+      "loss": 2.5341,
+      "step": 4789
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.4488372093023255e-07,
+      "loss": 0.0028,
+      "step": 4790
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.4487209302325584e-07,
+      "loss": 0.0033,
+      "step": 4791
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.4486046511627907e-07,
+      "loss": 0.0073,
+      "step": 4792
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.448488372093023e-07,
+      "loss": 1.2173,
+      "step": 4793
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.448372093023256e-07,
+      "loss": 0.0039,
+      "step": 4794
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.4482558139534883e-07,
+      "loss": 0.0062,
+      "step": 4795
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.4481395348837207e-07,
+      "loss": 0.2184,
+      "step": 4796
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.4480232558139535e-07,
+      "loss": 0.0062,
+      "step": 4797
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.447906976744186e-07,
+      "loss": 0.0073,
+      "step": 4798
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.447790697674419e-07,
+      "loss": 0.0035,
+      "step": 4799
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.447674418604651e-07,
+      "loss": 0.0036,
+      "step": 4800
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.4475581395348835e-07,
+      "loss": 0.0048,
+      "step": 4801
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.447441860465116e-07,
+      "loss": 0.0049,
+      "step": 4802
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.4473255813953487e-07,
+      "loss": 0.0032,
+      "step": 4803
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.447209302325581e-07,
+      "loss": 0.0047,
+      "step": 4804
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.447093023255814e-07,
+      "loss": 0.004,
+      "step": 4805
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.4469767441860463e-07,
+      "loss": 0.0035,
+      "step": 4806
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.446860465116279e-07,
+      "loss": 0.0031,
+      "step": 4807
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.4467441860465115e-07,
+      "loss": 0.0036,
+      "step": 4808
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.446627906976744e-07,
+      "loss": 0.004,
+      "step": 4809
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.446511627906977e-07,
+      "loss": 0.0049,
+      "step": 4810
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.446395348837209e-07,
+      "loss": 0.0029,
+      "step": 4811
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.446279069767442e-07,
+      "loss": 0.04,
+      "step": 4812
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.4461627906976743e-07,
+      "loss": 0.1328,
+      "step": 4813
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.4460465116279067e-07,
+      "loss": 0.0155,
+      "step": 4814
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.4459302325581396e-07,
+      "loss": 0.0037,
+      "step": 4815
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.445813953488372e-07,
+      "loss": 0.0046,
+      "step": 4816
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.4456976744186043e-07,
+      "loss": 0.0046,
+      "step": 4817
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.445581395348837e-07,
+      "loss": 0.0027,
+      "step": 4818
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.44546511627907e-07,
+      "loss": 0.0038,
+      "step": 4819
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.4453488372093024e-07,
+      "loss": 0.0064,
+      "step": 4820
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.445232558139535e-07,
+      "loss": 0.0031,
+      "step": 4821
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.445116279069767e-07,
+      "loss": 0.004,
+      "step": 4822
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.445e-07,
+      "loss": 0.0046,
+      "step": 4823
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.4448837209302323e-07,
+      "loss": 0.0044,
+      "step": 4824
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.444767441860465e-07,
+      "loss": 0.0147,
+      "step": 4825
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.4446511627906976e-07,
+      "loss": 0.0036,
+      "step": 4826
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.44453488372093e-07,
+      "loss": 0.0421,
+      "step": 4827
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.444418604651163e-07,
+      "loss": 0.0027,
+      "step": 4828
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.444302325581395e-07,
+      "loss": 0.0046,
+      "step": 4829
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.4441860465116275e-07,
+      "loss": 0.0036,
+      "step": 4830
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.4440697674418604e-07,
+      "loss": 0.0068,
+      "step": 4831
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.4439534883720933e-07,
+      "loss": 0.0035,
+      "step": 4832
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.4438372093023256e-07,
+      "loss": 0.0026,
+      "step": 4833
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.443720930232558e-07,
+      "loss": 0.0063,
+      "step": 4834
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.4436046511627903e-07,
+      "loss": 0.003,
+      "step": 4835
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.443488372093023e-07,
+      "loss": 0.004,
+      "step": 4836
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.4433720930232556e-07,
+      "loss": 0.004,
+      "step": 4837
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.4432558139534885e-07,
+      "loss": 0.0053,
+      "step": 4838
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.443139534883721e-07,
+      "loss": 0.0037,
+      "step": 4839
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.4430232558139537e-07,
+      "loss": 0.0023,
+      "step": 4840
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.442906976744186e-07,
+      "loss": 2.302,
+      "step": 4841
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.4427906976744184e-07,
+      "loss": 0.0045,
+      "step": 4842
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.442674418604651e-07,
+      "loss": 0.0036,
+      "step": 4843
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.4425581395348836e-07,
+      "loss": 0.0048,
+      "step": 4844
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.4424418604651165e-07,
+      "loss": 1.7434,
+      "step": 4845
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.442325581395349e-07,
+      "loss": 0.0051,
+      "step": 4846
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.442209302325581e-07,
+      "loss": 0.0055,
+      "step": 4847
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.4420930232558136e-07,
+      "loss": 0.2531,
+      "step": 4848
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.4419767441860464e-07,
+      "loss": 0.0034,
+      "step": 4849
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.441860465116279e-07,
+      "loss": 0.0034,
+      "step": 4850
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.4417441860465117e-07,
+      "loss": 0.0031,
+      "step": 4851
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.441627906976744e-07,
+      "loss": 0.0035,
+      "step": 4852
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.441511627906977e-07,
+      "loss": 0.005,
+      "step": 4853
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.4413953488372093e-07,
+      "loss": 0.0036,
+      "step": 4854
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.4412790697674416e-07,
+      "loss": 0.004,
+      "step": 4855
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.441162790697674e-07,
+      "loss": 0.0054,
+      "step": 4856
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.441046511627907e-07,
+      "loss": 0.0039,
+      "step": 4857
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.440930232558139e-07,
+      "loss": 0.0074,
+      "step": 4858
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.440813953488372e-07,
+      "loss": 0.0386,
+      "step": 4859
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.4406976744186044e-07,
+      "loss": 0.0085,
+      "step": 4860
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.4405813953488373e-07,
+      "loss": 0.0202,
+      "step": 4861
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.4404651162790697e-07,
+      "loss": 0.0055,
+      "step": 4862
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.440348837209302e-07,
+      "loss": 0.1755,
+      "step": 4863
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.440232558139535e-07,
+      "loss": 0.003,
+      "step": 4864
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.440116279069767e-07,
+      "loss": 0.0053,
+      "step": 4865
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.44e-07,
+      "loss": 0.0044,
+      "step": 4866
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.4398837209302325e-07,
+      "loss": 0.003,
+      "step": 4867
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.439767441860465e-07,
+      "loss": 0.0032,
+      "step": 4868
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.4396511627906977e-07,
+      "loss": 0.0038,
+      "step": 4869
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.43953488372093e-07,
+      "loss": 0.0045,
+      "step": 4870
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.4394186046511624e-07,
+      "loss": 0.0054,
+      "step": 4871
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.4393023255813953e-07,
+      "loss": 0.0069,
+      "step": 4872
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.439186046511628e-07,
+      "loss": 0.0039,
+      "step": 4873
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.4390697674418606e-07,
+      "loss": 0.0463,
+      "step": 4874
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.438953488372093e-07,
+      "loss": 0.0048,
+      "step": 4875
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.438837209302325e-07,
+      "loss": 0.0032,
+      "step": 4876
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.4387209302325576e-07,
+      "loss": 0.0048,
+      "step": 4877
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.4386046511627905e-07,
+      "loss": 0.0029,
+      "step": 4878
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.4384883720930234e-07,
+      "loss": 0.005,
+      "step": 4879
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.4383720930232557e-07,
+      "loss": 0.0033,
+      "step": 4880
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.438255813953488e-07,
+      "loss": 0.0053,
+      "step": 4881
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.438139534883721e-07,
+      "loss": 0.011,
+      "step": 4882
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.4380232558139533e-07,
+      "loss": 0.0052,
+      "step": 4883
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.4379069767441857e-07,
+      "loss": 0.0034,
+      "step": 4884
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.4377906976744185e-07,
+      "loss": 0.0042,
+      "step": 4885
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.4376744186046514e-07,
+      "loss": 1.2706,
+      "step": 4886
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.437558139534884e-07,
+      "loss": 0.0054,
+      "step": 4887
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.437441860465116e-07,
+      "loss": 0.0032,
+      "step": 4888
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.4373255813953485e-07,
+      "loss": 0.006,
+      "step": 4889
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.4372093023255814e-07,
+      "loss": 0.004,
+      "step": 4890
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.4370930232558137e-07,
+      "loss": 0.0057,
+      "step": 4891
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.4369767441860466e-07,
+      "loss": 0.429,
+      "step": 4892
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.436860465116279e-07,
+      "loss": 0.0055,
+      "step": 4893
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.436744186046512e-07,
+      "loss": 0.0045,
+      "step": 4894
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.436627906976744e-07,
+      "loss": 0.225,
+      "step": 4895
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.4365116279069765e-07,
+      "loss": 0.005,
+      "step": 4896
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.436395348837209e-07,
+      "loss": 0.0048,
+      "step": 4897
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.436279069767442e-07,
+      "loss": 0.0052,
+      "step": 4898
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.436162790697674e-07,
+      "loss": 0.003,
+      "step": 4899
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.436046511627907e-07,
+      "loss": 0.0032,
+      "step": 4900
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.4359302325581394e-07,
+      "loss": 0.005,
+      "step": 4901
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.4358139534883717e-07,
+      "loss": 0.0027,
+      "step": 4902
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.4356976744186046e-07,
+      "loss": 0.0039,
+      "step": 4903
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.435581395348837e-07,
+      "loss": 0.0052,
+      "step": 4904
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.43546511627907e-07,
+      "loss": 0.0042,
+      "step": 4905
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.435348837209302e-07,
+      "loss": 0.0034,
+      "step": 4906
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.435232558139535e-07,
+      "loss": 0.004,
+      "step": 4907
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.4351162790697674e-07,
+      "loss": 0.0046,
+      "step": 4908
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.435e-07,
+      "loss": 0.0037,
+      "step": 4909
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.434883720930232e-07,
+      "loss": 0.0038,
+      "step": 4910
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.434767441860465e-07,
+      "loss": 0.0048,
+      "step": 4911
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.4346511627906974e-07,
+      "loss": 0.0048,
+      "step": 4912
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.43453488372093e-07,
+      "loss": 0.006,
+      "step": 4913
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.4344186046511626e-07,
+      "loss": 0.0064,
+      "step": 4914
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.4343023255813955e-07,
+      "loss": 0.0048,
+      "step": 4915
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.434186046511628e-07,
+      "loss": 0.0054,
+      "step": 4916
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.43406976744186e-07,
+      "loss": 0.0072,
+      "step": 4917
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.433953488372093e-07,
+      "loss": 0.005,
+      "step": 4918
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.4338372093023254e-07,
+      "loss": 0.003,
+      "step": 4919
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.4337209302325583e-07,
+      "loss": 0.0033,
+      "step": 4920
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.4336046511627906e-07,
+      "loss": 0.0034,
+      "step": 4921
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.433488372093023e-07,
+      "loss": 0.0068,
+      "step": 4922
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.433372093023256e-07,
+      "loss": 0.0094,
+      "step": 4923
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.433255813953488e-07,
+      "loss": 0.0062,
+      "step": 4924
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.4331395348837206e-07,
+      "loss": 0.0042,
+      "step": 4925
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.4330232558139535e-07,
+      "loss": 0.0285,
+      "step": 4926
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.4329069767441863e-07,
+      "loss": 0.0037,
+      "step": 4927
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.4327906976744187e-07,
+      "loss": 0.3383,
+      "step": 4928
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.432674418604651e-07,
+      "loss": 0.0032,
+      "step": 4929
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.4325581395348834e-07,
+      "loss": 0.0049,
+      "step": 4930
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.432441860465116e-07,
+      "loss": 0.005,
+      "step": 4931
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.4323255813953486e-07,
+      "loss": 0.0039,
+      "step": 4932
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.4322093023255815e-07,
+      "loss": 0.0027,
+      "step": 4933
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.432093023255814e-07,
+      "loss": 0.0031,
+      "step": 4934
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.431976744186046e-07,
+      "loss": 0.0038,
+      "step": 4935
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.431860465116279e-07,
+      "loss": 0.0046,
+      "step": 4936
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.4317441860465115e-07,
+      "loss": 0.0055,
+      "step": 4937
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.431627906976744e-07,
+      "loss": 0.0032,
+      "step": 4938
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.4315116279069767e-07,
+      "loss": 0.0028,
+      "step": 4939
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.4313953488372096e-07,
+      "loss": 0.0026,
+      "step": 4940
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.431279069767442e-07,
+      "loss": 2.5136,
+      "step": 4941
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.4311627906976743e-07,
+      "loss": 2.569,
+      "step": 4942
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.4310465116279066e-07,
+      "loss": 2.8453,
+      "step": 4943
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.4309302325581395e-07,
+      "loss": 0.0534,
+      "step": 4944
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.430813953488372e-07,
+      "loss": 0.0034,
+      "step": 4945
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.430697674418605e-07,
+      "loss": 0.0151,
+      "step": 4946
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.430581395348837e-07,
+      "loss": 0.0026,
+      "step": 4947
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.43046511627907e-07,
+      "loss": 0.0049,
+      "step": 4948
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.4303488372093023e-07,
+      "loss": 0.0082,
+      "step": 4949
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.4302325581395347e-07,
+      "loss": 0.0045,
+      "step": 4950
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.430116279069767e-07,
+      "loss": 0.003,
+      "step": 4951
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.43e-07,
+      "loss": 0.0025,
+      "step": 4952
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.4298837209302323e-07,
+      "loss": 0.0049,
+      "step": 4953
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.429767441860465e-07,
+      "loss": 0.0058,
+      "step": 4954
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.4296511627906975e-07,
+      "loss": 0.0031,
+      "step": 4955
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.42953488372093e-07,
+      "loss": 0.0031,
+      "step": 4956
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.429418604651163e-07,
+      "loss": 0.0047,
+      "step": 4957
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.429302325581395e-07,
+      "loss": 2.2122,
+      "step": 4958
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.429186046511628e-07,
+      "loss": 0.0277,
+      "step": 4959
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.4290697674418603e-07,
+      "loss": 0.0046,
+      "step": 4960
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.428953488372093e-07,
+      "loss": 0.0036,
+      "step": 4961
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.4288372093023256e-07,
+      "loss": 0.0038,
+      "step": 4962
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.428720930232558e-07,
+      "loss": 0.4208,
+      "step": 4963
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.4286046511627903e-07,
+      "loss": 0.0043,
+      "step": 4964
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.428488372093023e-07,
+      "loss": 0.0027,
+      "step": 4965
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.4283720930232555e-07,
+      "loss": 0.0052,
+      "step": 4966
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.4282558139534884e-07,
+      "loss": 0.0074,
+      "step": 4967
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.428139534883721e-07,
+      "loss": 0.0092,
+      "step": 4968
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.4280232558139536e-07,
+      "loss": 0.0108,
+      "step": 4969
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.427906976744186e-07,
+      "loss": 0.0045,
+      "step": 4970
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.4277906976744183e-07,
+      "loss": 0.0025,
+      "step": 4971
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.427674418604651e-07,
+      "loss": 0.0527,
+      "step": 4972
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.4275581395348836e-07,
+      "loss": 0.0029,
+      "step": 4973
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.4274418604651164e-07,
+      "loss": 0.011,
+      "step": 4974
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.427325581395349e-07,
+      "loss": 2.0643,
+      "step": 4975
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.427209302325581e-07,
+      "loss": 0.0066,
+      "step": 4976
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.427093023255814e-07,
+      "loss": 0.0037,
+      "step": 4977
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.4269767441860464e-07,
+      "loss": 0.0031,
+      "step": 4978
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.4268604651162787e-07,
+      "loss": 2.2181,
+      "step": 4979
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.4267441860465116e-07,
+      "loss": 0.0044,
+      "step": 4980
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.426627906976744e-07,
+      "loss": 0.0072,
+      "step": 4981
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.426511627906977e-07,
+      "loss": 0.1186,
+      "step": 4982
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.426395348837209e-07,
+      "loss": 0.0057,
+      "step": 4983
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.4262790697674416e-07,
+      "loss": 2.4359,
+      "step": 4984
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.426162790697674e-07,
+      "loss": 0.0037,
+      "step": 4985
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.426046511627907e-07,
+      "loss": 0.0063,
+      "step": 4986
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.4259302325581397e-07,
+      "loss": 0.7309,
+      "step": 4987
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.425813953488372e-07,
+      "loss": 1.9805,
+      "step": 4988
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.4256976744186044e-07,
+      "loss": 0.0046,
+      "step": 4989
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.425581395348837e-07,
+      "loss": 0.0061,
+      "step": 4990
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.4254651162790696e-07,
+      "loss": 0.0031,
+      "step": 4991
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.425348837209302e-07,
+      "loss": 0.0023,
+      "step": 4992
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.425232558139535e-07,
+      "loss": 0.0088,
+      "step": 4993
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.425116279069767e-07,
+      "loss": 0.0057,
+      "step": 4994
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.425e-07,
+      "loss": 2.7442,
+      "step": 4995
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.4248837209302324e-07,
+      "loss": 0.0027,
+      "step": 4996
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.424767441860465e-07,
+      "loss": 0.0046,
+      "step": 4997
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.4246511627906977e-07,
+      "loss": 0.0056,
+      "step": 4998
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.42453488372093e-07,
+      "loss": 0.1517,
+      "step": 4999
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.424418604651163e-07,
+      "loss": 0.0038,
+      "step": 5000
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.424302325581395e-07,
+      "loss": 0.0031,
+      "step": 5001
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.424186046511628e-07,
+      "loss": 0.003,
+      "step": 5002
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.4240697674418605e-07,
+      "loss": 0.005,
+      "step": 5003
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.423953488372093e-07,
+      "loss": 0.0034,
+      "step": 5004
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.423837209302325e-07,
+      "loss": 0.0058,
+      "step": 5005
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.423720930232558e-07,
+      "loss": 0.0032,
+      "step": 5006
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.4236046511627904e-07,
+      "loss": 0.0041,
+      "step": 5007
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.4234883720930233e-07,
+      "loss": 0.0087,
+      "step": 5008
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.4233720930232557e-07,
+      "loss": 0.0047,
+      "step": 5009
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.423255813953488e-07,
+      "loss": 0.0039,
+      "step": 5010
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.423139534883721e-07,
+      "loss": 0.1213,
+      "step": 5011
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.423023255813953e-07,
+      "loss": 0.0039,
+      "step": 5012
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.422906976744186e-07,
+      "loss": 0.0753,
+      "step": 5013
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.4227906976744185e-07,
+      "loss": 0.003,
+      "step": 5014
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.4226744186046514e-07,
+      "loss": 0.0029,
+      "step": 5015
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.4225581395348837e-07,
+      "loss": 0.0028,
+      "step": 5016
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.422441860465116e-07,
+      "loss": 0.0155,
+      "step": 5017
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.4223255813953484e-07,
+      "loss": 0.0031,
+      "step": 5018
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.4222093023255813e-07,
+      "loss": 0.0033,
+      "step": 5019
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.4220930232558137e-07,
+      "loss": 0.0031,
+      "step": 5020
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.4219767441860465e-07,
+      "loss": 0.0043,
+      "step": 5021
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.421860465116279e-07,
+      "loss": 0.0036,
+      "step": 5022
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.421744186046512e-07,
+      "loss": 0.0051,
+      "step": 5023
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.421627906976744e-07,
+      "loss": 0.0044,
+      "step": 5024
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.4215116279069765e-07,
+      "loss": 0.1866,
+      "step": 5025
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.421395348837209e-07,
+      "loss": 0.1872,
+      "step": 5026
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.4212790697674417e-07,
+      "loss": 0.0142,
+      "step": 5027
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.4211627906976746e-07,
+      "loss": 0.0106,
+      "step": 5028
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.421046511627907e-07,
+      "loss": 0.2159,
+      "step": 5029
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.4209302325581393e-07,
+      "loss": 0.0046,
+      "step": 5030
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.420813953488372e-07,
+      "loss": 0.0039,
+      "step": 5031
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.4206976744186045e-07,
+      "loss": 0.0032,
+      "step": 5032
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.420581395348837e-07,
+      "loss": 0.0039,
+      "step": 5033
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.42046511627907e-07,
+      "loss": 0.0031,
+      "step": 5034
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.420348837209302e-07,
+      "loss": 0.0056,
+      "step": 5035
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.420232558139535e-07,
+      "loss": 0.0023,
+      "step": 5036
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.4201162790697673e-07,
+      "loss": 0.0043,
+      "step": 5037
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.4199999999999997e-07,
+      "loss": 0.0035,
+      "step": 5038
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.419883720930232e-07,
+      "loss": 0.0049,
+      "step": 5039
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.419767441860465e-07,
+      "loss": 0.0026,
+      "step": 5040
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.419651162790698e-07,
+      "loss": 0.0056,
+      "step": 5041
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.41953488372093e-07,
+      "loss": 0.0041,
+      "step": 5042
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.4194186046511625e-07,
+      "loss": 0.0042,
+      "step": 5043
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.4193023255813954e-07,
+      "loss": 1.3967,
+      "step": 5044
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.419186046511628e-07,
+      "loss": 0.0087,
+      "step": 5045
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.41906976744186e-07,
+      "loss": 0.0028,
+      "step": 5046
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.418953488372093e-07,
+      "loss": 0.004,
+      "step": 5047
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.4188372093023253e-07,
+      "loss": 0.0027,
+      "step": 5048
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.418720930232558e-07,
+      "loss": 0.0055,
+      "step": 5049
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.4186046511627906e-07,
+      "loss": 0.2557,
+      "step": 5050
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.418488372093023e-07,
+      "loss": 0.0037,
+      "step": 5051
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.418372093023256e-07,
+      "loss": 0.0028,
+      "step": 5052
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.418255813953488e-07,
+      "loss": 0.0036,
+      "step": 5053
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.418139534883721e-07,
+      "loss": 0.0076,
+      "step": 5054
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.4180232558139534e-07,
+      "loss": 0.0035,
+      "step": 5055
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.4179069767441863e-07,
+      "loss": 0.0036,
+      "step": 5056
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.4177906976744186e-07,
+      "loss": 0.0063,
+      "step": 5057
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.417674418604651e-07,
+      "loss": 0.0045,
+      "step": 5058
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.4175581395348833e-07,
+      "loss": 0.0092,
+      "step": 5059
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.417441860465116e-07,
+      "loss": 0.0041,
+      "step": 5060
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.4173255813953486e-07,
+      "loss": 0.0035,
+      "step": 5061
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.4172093023255815e-07,
+      "loss": 0.0032,
+      "step": 5062
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.417093023255814e-07,
+      "loss": 0.2615,
+      "step": 5063
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.416976744186046e-07,
+      "loss": 0.0062,
+      "step": 5064
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.416860465116279e-07,
+      "loss": 0.0032,
+      "step": 5065
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.4167441860465114e-07,
+      "loss": 0.0051,
+      "step": 5066
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.4166279069767443e-07,
+      "loss": 0.0055,
+      "step": 5067
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.4165116279069766e-07,
+      "loss": 0.0042,
+      "step": 5068
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.4163953488372095e-07,
+      "loss": 0.0044,
+      "step": 5069
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.416279069767442e-07,
+      "loss": 0.0037,
+      "step": 5070
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.416162790697674e-07,
+      "loss": 0.8114,
+      "step": 5071
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.4160465116279066e-07,
+      "loss": 0.0029,
+      "step": 5072
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.4159302325581394e-07,
+      "loss": 0.0046,
+      "step": 5073
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.415813953488372e-07,
+      "loss": 0.0043,
+      "step": 5074
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.4156976744186047e-07,
+      "loss": 0.0024,
+      "step": 5075
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.415581395348837e-07,
+      "loss": 0.0037,
+      "step": 5076
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.41546511627907e-07,
+      "loss": 0.0033,
+      "step": 5077
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.4153488372093023e-07,
+      "loss": 0.0036,
+      "step": 5078
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.4152325581395346e-07,
+      "loss": 0.0053,
+      "step": 5079
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.415116279069767e-07,
+      "loss": 0.0025,
+      "step": 5080
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.415e-07,
+      "loss": 0.0033,
+      "step": 5081
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.414883720930233e-07,
+      "loss": 0.0027,
+      "step": 5082
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.414767441860465e-07,
+      "loss": 0.0203,
+      "step": 5083
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.4146511627906974e-07,
+      "loss": 0.0028,
+      "step": 5084
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.4145348837209303e-07,
+      "loss": 0.0054,
+      "step": 5085
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.4144186046511627e-07,
+      "loss": 0.0028,
+      "step": 5086
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.414302325581395e-07,
+      "loss": 0.0054,
+      "step": 5087
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.414186046511628e-07,
+      "loss": 0.0048,
+      "step": 5088
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.41406976744186e-07,
+      "loss": 0.0118,
+      "step": 5089
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.413953488372093e-07,
+      "loss": 0.0043,
+      "step": 5090
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.4138372093023255e-07,
+      "loss": 0.0027,
+      "step": 5091
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.413720930232558e-07,
+      "loss": 0.0045,
+      "step": 5092
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.41360465116279e-07,
+      "loss": 0.0053,
+      "step": 5093
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.413488372093023e-07,
+      "loss": 0.0292,
+      "step": 5094
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.413372093023256e-07,
+      "loss": 0.0029,
+      "step": 5095
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.4132558139534883e-07,
+      "loss": 0.0043,
+      "step": 5096
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.4131395348837207e-07,
+      "loss": 0.0048,
+      "step": 5097
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.4130232558139536e-07,
+      "loss": 0.0064,
+      "step": 5098
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.412906976744186e-07,
+      "loss": 0.0039,
+      "step": 5099
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.412790697674418e-07,
+      "loss": 0.0035,
+      "step": 5100
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.412674418604651e-07,
+      "loss": 2.5138,
+      "step": 5101
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.4125581395348835e-07,
+      "loss": 0.0102,
+      "step": 5102
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.4124418604651164e-07,
+      "loss": 0.0044,
+      "step": 5103
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.4123255813953487e-07,
+      "loss": 0.0049,
+      "step": 5104
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.412209302325581e-07,
+      "loss": 0.0048,
+      "step": 5105
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.412093023255814e-07,
+      "loss": 0.0041,
+      "step": 5106
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.4119767441860463e-07,
+      "loss": 0.0035,
+      "step": 5107
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.411860465116279e-07,
+      "loss": 0.0034,
+      "step": 5108
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.4117441860465115e-07,
+      "loss": 0.0051,
+      "step": 5109
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.4116279069767444e-07,
+      "loss": 0.0031,
+      "step": 5110
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.411511627906977e-07,
+      "loss": 0.0022,
+      "step": 5111
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.411395348837209e-07,
+      "loss": 0.0109,
+      "step": 5112
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.4112790697674415e-07,
+      "loss": 0.0034,
+      "step": 5113
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.4111627906976744e-07,
+      "loss": 0.0032,
+      "step": 5114
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.4110465116279067e-07,
+      "loss": 0.004,
+      "step": 5115
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.4109302325581396e-07,
+      "loss": 0.6167,
+      "step": 5116
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.410813953488372e-07,
+      "loss": 0.0055,
+      "step": 5117
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.4106976744186043e-07,
+      "loss": 0.0031,
+      "step": 5118
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.410581395348837e-07,
+      "loss": 0.004,
+      "step": 5119
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.4104651162790695e-07,
+      "loss": 0.1254,
+      "step": 5120
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.410348837209302e-07,
+      "loss": 0.0046,
+      "step": 5121
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.410232558139535e-07,
+      "loss": 0.0037,
+      "step": 5122
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.4101162790697677e-07,
+      "loss": 0.0037,
+      "step": 5123
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.41e-07,
+      "loss": 0.0048,
+      "step": 5124
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.4098837209302324e-07,
+      "loss": 0.0034,
+      "step": 5125
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.4097674418604647e-07,
+      "loss": 0.0033,
+      "step": 5126
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4096511627906976e-07,
+      "loss": 0.0032,
+      "step": 5127
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.40953488372093e-07,
+      "loss": 0.0033,
+      "step": 5128
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.409418604651163e-07,
+      "loss": 0.0031,
+      "step": 5129
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.409302325581395e-07,
+      "loss": 0.4823,
+      "step": 5130
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.409186046511628e-07,
+      "loss": 0.021,
+      "step": 5131
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.4090697674418604e-07,
+      "loss": 0.0025,
+      "step": 5132
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.408953488372093e-07,
+      "loss": 0.0046,
+      "step": 5133
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.408837209302325e-07,
+      "loss": 0.0056,
+      "step": 5134
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.408720930232558e-07,
+      "loss": 0.0033,
+      "step": 5135
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.408604651162791e-07,
+      "loss": 0.0023,
+      "step": 5136
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.408488372093023e-07,
+      "loss": 0.0051,
+      "step": 5137
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.4083720930232556e-07,
+      "loss": 0.0037,
+      "step": 5138
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.4082558139534885e-07,
+      "loss": 0.0765,
+      "step": 5139
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.408139534883721e-07,
+      "loss": 0.0049,
+      "step": 5140
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.408023255813953e-07,
+      "loss": 0.0046,
+      "step": 5141
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.407906976744186e-07,
+      "loss": 0.0051,
+      "step": 5142
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.4077906976744184e-07,
+      "loss": 1.9411,
+      "step": 5143
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.4076744186046513e-07,
+      "loss": 0.0047,
+      "step": 5144
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.4075581395348836e-07,
+      "loss": 0.0038,
+      "step": 5145
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.407441860465116e-07,
+      "loss": 0.0044,
+      "step": 5146
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.4073255813953483e-07,
+      "loss": 0.0032,
+      "step": 5147
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.407209302325581e-07,
+      "loss": 0.005,
+      "step": 5148
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.407093023255814e-07,
+      "loss": 0.0062,
+      "step": 5149
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.4069767441860465e-07,
+      "loss": 0.0042,
+      "step": 5150
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.406860465116279e-07,
+      "loss": 0.0045,
+      "step": 5151
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.4067441860465117e-07,
+      "loss": 0.0035,
+      "step": 5152
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.406627906976744e-07,
+      "loss": 0.0029,
+      "step": 5153
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.4065116279069764e-07,
+      "loss": 0.0024,
+      "step": 5154
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.4063953488372093e-07,
+      "loss": 0.0031,
+      "step": 5155
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.4062790697674416e-07,
+      "loss": 0.0235,
+      "step": 5156
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.4061627906976745e-07,
+      "loss": 0.1675,
+      "step": 5157
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.406046511627907e-07,
+      "loss": 0.0038,
+      "step": 5158
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.405930232558139e-07,
+      "loss": 0.0037,
+      "step": 5159
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.405813953488372e-07,
+      "loss": 0.003,
+      "step": 5160
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.4056976744186045e-07,
+      "loss": 0.0078,
+      "step": 5161
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.4055813953488373e-07,
+      "loss": 0.0669,
+      "step": 5162
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.4054651162790697e-07,
+      "loss": 0.0034,
+      "step": 5163
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.4053488372093026e-07,
+      "loss": 0.003,
+      "step": 5164
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.405232558139535e-07,
+      "loss": 0.0045,
+      "step": 5165
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.4051162790697673e-07,
+      "loss": 0.0027,
+      "step": 5166
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.4049999999999996e-07,
+      "loss": 0.2213,
+      "step": 5167
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.4048837209302325e-07,
+      "loss": 0.005,
+      "step": 5168
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.404767441860465e-07,
+      "loss": 0.0032,
+      "step": 5169
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.404651162790698e-07,
+      "loss": 0.0061,
+      "step": 5170
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.9825918762088974,
+      "eval_loss": 0.15661931037902832,
+      "eval_runtime": 1.5581,
+      "eval_samples_per_second": 663.637,
+      "eval_steps_per_second": 10.911,
+      "step": 5170
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.40453488372093e-07,
+      "loss": 0.0023,
+      "step": 5171
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.4044186046511625e-07,
+      "loss": 0.003,
+      "step": 5172
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.4043023255813953e-07,
+      "loss": 0.0051,
+      "step": 5173
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.4041860465116277e-07,
+      "loss": 0.0033,
+      "step": 5174
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.40406976744186e-07,
+      "loss": 0.0034,
+      "step": 5175
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.403953488372093e-07,
+      "loss": 0.0029,
+      "step": 5176
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.403837209302326e-07,
+      "loss": 0.0028,
+      "step": 5177
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.403720930232558e-07,
+      "loss": 0.0039,
+      "step": 5178
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.4036046511627905e-07,
+      "loss": 0.0036,
+      "step": 5179
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.403488372093023e-07,
+      "loss": 0.0029,
+      "step": 5180
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.403372093023256e-07,
+      "loss": 0.0027,
+      "step": 5181
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.403255813953488e-07,
+      "loss": 0.0029,
+      "step": 5182
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.403139534883721e-07,
+      "loss": 0.0035,
+      "step": 5183
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4030232558139533e-07,
+      "loss": 1.9917,
+      "step": 5184
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.402906976744186e-07,
+      "loss": 0.0131,
+      "step": 5185
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4027906976744186e-07,
+      "loss": 0.005,
+      "step": 5186
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.402674418604651e-07,
+      "loss": 0.0031,
+      "step": 5187
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4025581395348833e-07,
+      "loss": 0.0031,
+      "step": 5188
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.402441860465116e-07,
+      "loss": 0.0065,
+      "step": 5189
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.402325581395349e-07,
+      "loss": 0.0041,
+      "step": 5190
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.4022093023255814e-07,
+      "loss": 0.0036,
+      "step": 5191
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.402093023255814e-07,
+      "loss": 0.0036,
+      "step": 5192
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.4019767441860466e-07,
+      "loss": 0.0059,
+      "step": 5193
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.401860465116279e-07,
+      "loss": 0.0026,
+      "step": 5194
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4017441860465113e-07,
+      "loss": 0.0045,
+      "step": 5195
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.401627906976744e-07,
+      "loss": 0.0029,
+      "step": 5196
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4015116279069766e-07,
+      "loss": 0.003,
+      "step": 5197
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4013953488372094e-07,
+      "loss": 0.0026,
+      "step": 5198
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.401279069767442e-07,
+      "loss": 0.0032,
+      "step": 5199
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.401162790697674e-07,
+      "loss": 0.0036,
+      "step": 5200
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4010465116279065e-07,
+      "loss": 0.0053,
+      "step": 5201
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4009302325581394e-07,
+      "loss": 0.0034,
+      "step": 5202
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.400813953488372e-07,
+      "loss": 0.0393,
+      "step": 5203
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4006976744186046e-07,
+      "loss": 0.004,
+      "step": 5204
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.400581395348837e-07,
+      "loss": 0.0031,
+      "step": 5205
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.40046511627907e-07,
+      "loss": 0.0048,
+      "step": 5206
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.400348837209302e-07,
+      "loss": 0.004,
+      "step": 5207
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4002325581395346e-07,
+      "loss": 0.0044,
+      "step": 5208
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4001162790697674e-07,
+      "loss": 0.0042,
+      "step": 5209
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4e-07,
+      "loss": 0.0075,
+      "step": 5210
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.3998837209302327e-07,
+      "loss": 0.0033,
+      "step": 5211
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.399767441860465e-07,
+      "loss": 0.0048,
+      "step": 5212
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.3996511627906974e-07,
+      "loss": 0.0132,
+      "step": 5213
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.39953488372093e-07,
+      "loss": 0.0041,
+      "step": 5214
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.3994186046511626e-07,
+      "loss": 0.0034,
+      "step": 5215
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.3993023255813955e-07,
+      "loss": 0.6722,
+      "step": 5216
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.399186046511628e-07,
+      "loss": 0.0031,
+      "step": 5217
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.3990697674418607e-07,
+      "loss": 0.0058,
+      "step": 5218
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.398953488372093e-07,
+      "loss": 0.0034,
+      "step": 5219
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.3988372093023254e-07,
+      "loss": 0.0357,
+      "step": 5220
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.398720930232558e-07,
+      "loss": 0.003,
+      "step": 5221
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.3986046511627907e-07,
+      "loss": 0.0026,
+      "step": 5222
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.398488372093023e-07,
+      "loss": 0.0036,
+      "step": 5223
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.398372093023256e-07,
+      "loss": 0.0034,
+      "step": 5224
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.398255813953488e-07,
+      "loss": 0.3484,
+      "step": 5225
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.3981395348837206e-07,
+      "loss": 0.0063,
+      "step": 5226
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.3980232558139535e-07,
+      "loss": 0.0086,
+      "step": 5227
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.397906976744186e-07,
+      "loss": 0.0031,
+      "step": 5228
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.397790697674418e-07,
+      "loss": 0.0039,
+      "step": 5229
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.397674418604651e-07,
+      "loss": 0.0032,
+      "step": 5230
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.397558139534884e-07,
+      "loss": 0.0038,
+      "step": 5231
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.3974418604651163e-07,
+      "loss": 2.0491,
+      "step": 5232
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.3973255813953487e-07,
+      "loss": 0.0032,
+      "step": 5233
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.397209302325581e-07,
+      "loss": 1.1744,
+      "step": 5234
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.397093023255814e-07,
+      "loss": 0.0143,
+      "step": 5235
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.396976744186046e-07,
+      "loss": 0.0052,
+      "step": 5236
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.396860465116279e-07,
+      "loss": 0.0026,
+      "step": 5237
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.3967441860465115e-07,
+      "loss": 0.0033,
+      "step": 5238
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.3966279069767444e-07,
+      "loss": 0.004,
+      "step": 5239
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.3965116279069767e-07,
+      "loss": 0.0039,
+      "step": 5240
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.396395348837209e-07,
+      "loss": 0.0028,
+      "step": 5241
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.3962790697674414e-07,
+      "loss": 1.5093,
+      "step": 5242
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.3961627906976743e-07,
+      "loss": 0.003,
+      "step": 5243
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.396046511627907e-07,
+      "loss": 0.0026,
+      "step": 5244
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.3959302325581395e-07,
+      "loss": 0.0414,
+      "step": 5245
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.395813953488372e-07,
+      "loss": 0.0038,
+      "step": 5246
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.395697674418605e-07,
+      "loss": 0.0839,
+      "step": 5247
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.395581395348837e-07,
+      "loss": 0.6022,
+      "step": 5248
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.3954651162790695e-07,
+      "loss": 0.0041,
+      "step": 5249
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.3953488372093024e-07,
+      "loss": 0.0041,
+      "step": 5250
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.3952325581395347e-07,
+      "loss": 0.0024,
+      "step": 5251
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.3951162790697676e-07,
+      "loss": 0.0044,
+      "step": 5252
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.395e-07,
+      "loss": 0.0033,
+      "step": 5253
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.3948837209302323e-07,
+      "loss": 0.8001,
+      "step": 5254
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.3947674418604646e-07,
+      "loss": 0.0039,
+      "step": 5255
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.3946511627906975e-07,
+      "loss": 0.0039,
+      "step": 5256
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.3945348837209304e-07,
+      "loss": 0.0035,
+      "step": 5257
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.394418604651163e-07,
+      "loss": 0.0044,
+      "step": 5258
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.394302325581395e-07,
+      "loss": 0.0031,
+      "step": 5259
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.394186046511628e-07,
+      "loss": 0.0043,
+      "step": 5260
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.3940697674418603e-07,
+      "loss": 0.0029,
+      "step": 5261
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.3939534883720927e-07,
+      "loss": 0.0029,
+      "step": 5262
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.3938372093023256e-07,
+      "loss": 0.0063,
+      "step": 5263
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.393720930232558e-07,
+      "loss": 0.0025,
+      "step": 5264
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.393604651162791e-07,
+      "loss": 0.0021,
+      "step": 5265
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.393488372093023e-07,
+      "loss": 0.0048,
+      "step": 5266
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.3933720930232555e-07,
+      "loss": 0.0051,
+      "step": 5267
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.3932558139534884e-07,
+      "loss": 0.0039,
+      "step": 5268
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.393139534883721e-07,
+      "loss": 0.0038,
+      "step": 5269
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.393023255813953e-07,
+      "loss": 0.0059,
+      "step": 5270
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.392906976744186e-07,
+      "loss": 0.0023,
+      "step": 5271
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.392790697674419e-07,
+      "loss": 0.0044,
+      "step": 5272
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.392674418604651e-07,
+      "loss": 0.0026,
+      "step": 5273
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.3925581395348836e-07,
+      "loss": 0.0033,
+      "step": 5274
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.392441860465116e-07,
+      "loss": 0.1332,
+      "step": 5275
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.392325581395349e-07,
+      "loss": 0.0027,
+      "step": 5276
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.392209302325581e-07,
+      "loss": 0.0023,
+      "step": 5277
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.392093023255814e-07,
+      "loss": 0.0046,
+      "step": 5278
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.3919767441860464e-07,
+      "loss": 0.0034,
+      "step": 5279
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.391860465116279e-07,
+      "loss": 0.0032,
+      "step": 5280
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.3917441860465116e-07,
+      "loss": 0.003,
+      "step": 5281
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.391627906976744e-07,
+      "loss": 1.7471,
+      "step": 5282
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.3915116279069763e-07,
+      "loss": 0.3402,
+      "step": 5283
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.391395348837209e-07,
+      "loss": 0.0063,
+      "step": 5284
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.391279069767442e-07,
+      "loss": 0.0037,
+      "step": 5285
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.3911627906976745e-07,
+      "loss": 0.0038,
+      "step": 5286
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.391046511627907e-07,
+      "loss": 0.0038,
+      "step": 5287
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.390930232558139e-07,
+      "loss": 0.0029,
+      "step": 5288
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.390813953488372e-07,
+      "loss": 0.0062,
+      "step": 5289
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.3906976744186044e-07,
+      "loss": 0.0042,
+      "step": 5290
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.3905813953488373e-07,
+      "loss": 0.0034,
+      "step": 5291
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.3904651162790696e-07,
+      "loss": 0.0069,
+      "step": 5292
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.3903488372093025e-07,
+      "loss": 0.0037,
+      "step": 5293
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.390232558139535e-07,
+      "loss": 0.0032,
+      "step": 5294
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.390116279069767e-07,
+      "loss": 0.0034,
+      "step": 5295
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.3899999999999996e-07,
+      "loss": 0.0044,
+      "step": 5296
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.3898837209302324e-07,
+      "loss": 0.0044,
+      "step": 5297
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.3897674418604653e-07,
+      "loss": 0.0036,
+      "step": 5298
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.3896511627906977e-07,
+      "loss": 0.0425,
+      "step": 5299
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.38953488372093e-07,
+      "loss": 0.0066,
+      "step": 5300
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.3894186046511624e-07,
+      "loss": 0.0036,
+      "step": 5301
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.3893023255813953e-07,
+      "loss": 1.9399,
+      "step": 5302
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.3891860465116276e-07,
+      "loss": 0.0039,
+      "step": 5303
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.3890697674418605e-07,
+      "loss": 0.0026,
+      "step": 5304
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.388953488372093e-07,
+      "loss": 0.0052,
+      "step": 5305
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.3888372093023257e-07,
+      "loss": 0.0037,
+      "step": 5306
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.388720930232558e-07,
+      "loss": 0.0055,
+      "step": 5307
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.3886046511627904e-07,
+      "loss": 0.0027,
+      "step": 5308
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.388488372093023e-07,
+      "loss": 0.0048,
+      "step": 5309
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.3883720930232557e-07,
+      "loss": 0.0033,
+      "step": 5310
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.3882558139534886e-07,
+      "loss": 0.0142,
+      "step": 5311
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.388139534883721e-07,
+      "loss": 0.0052,
+      "step": 5312
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.388023255813953e-07,
+      "loss": 0.0047,
+      "step": 5313
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.387906976744186e-07,
+      "loss": 0.0041,
+      "step": 5314
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.3877906976744185e-07,
+      "loss": 0.0035,
+      "step": 5315
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.387674418604651e-07,
+      "loss": 0.0035,
+      "step": 5316
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.3875581395348837e-07,
+      "loss": 0.0032,
+      "step": 5317
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.387441860465116e-07,
+      "loss": 0.0041,
+      "step": 5318
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.387325581395349e-07,
+      "loss": 0.0024,
+      "step": 5319
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.3872093023255813e-07,
+      "loss": 0.0028,
+      "step": 5320
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.3870930232558137e-07,
+      "loss": 0.0035,
+      "step": 5321
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.3869767441860465e-07,
+      "loss": 0.0037,
+      "step": 5322
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.386860465116279e-07,
+      "loss": 0.0042,
+      "step": 5323
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.386744186046511e-07,
+      "loss": 0.003,
+      "step": 5324
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.386627906976744e-07,
+      "loss": 0.0133,
+      "step": 5325
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.386511627906977e-07,
+      "loss": 0.0031,
+      "step": 5326
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.3863953488372094e-07,
+      "loss": 0.0051,
+      "step": 5327
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.3862790697674417e-07,
+      "loss": 0.004,
+      "step": 5328
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.386162790697674e-07,
+      "loss": 0.0029,
+      "step": 5329
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.386046511627907e-07,
+      "loss": 0.008,
+      "step": 5330
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.3859302325581393e-07,
+      "loss": 0.0045,
+      "step": 5331
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.385813953488372e-07,
+      "loss": 0.0023,
+      "step": 5332
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.3856976744186045e-07,
+      "loss": 0.003,
+      "step": 5333
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.385581395348837e-07,
+      "loss": 0.0052,
+      "step": 5334
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.38546511627907e-07,
+      "loss": 0.0023,
+      "step": 5335
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.385348837209302e-07,
+      "loss": 0.0025,
+      "step": 5336
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.3852325581395345e-07,
+      "loss": 0.0033,
+      "step": 5337
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.3851162790697674e-07,
+      "loss": 0.0035,
+      "step": 5338
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.385e-07,
+      "loss": 1.354,
+      "step": 5339
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.3848837209302326e-07,
+      "loss": 0.0024,
+      "step": 5340
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.384767441860465e-07,
+      "loss": 0.0037,
+      "step": 5341
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.3846511627906973e-07,
+      "loss": 1.4377,
+      "step": 5342
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.38453488372093e-07,
+      "loss": 0.0034,
+      "step": 5343
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.3844186046511625e-07,
+      "loss": 0.003,
+      "step": 5344
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.3843023255813954e-07,
+      "loss": 0.0051,
+      "step": 5345
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.384186046511628e-07,
+      "loss": 1.2322,
+      "step": 5346
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.3840697674418604e-07,
+      "loss": 0.0025,
+      "step": 5347
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.383953488372093e-07,
+      "loss": 0.2083,
+      "step": 5348
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.3838372093023254e-07,
+      "loss": 0.0054,
+      "step": 5349
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.383720930232558e-07,
+      "loss": 0.003,
+      "step": 5350
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.3836046511627909e-07,
+      "loss": 0.0031,
+      "step": 5351
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.3834883720930232e-07,
+      "loss": 0.0044,
+      "step": 5352
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.3833720930232558e-07,
+      "loss": 0.0027,
+      "step": 5353
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.3832558139534882e-07,
+      "loss": 0.0227,
+      "step": 5354
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.3831395348837205e-07,
+      "loss": 1.8774,
+      "step": 5355
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.3830232558139534e-07,
+      "loss": 0.0151,
+      "step": 5356
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.382906976744186e-07,
+      "loss": 0.0049,
+      "step": 5357
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.3827906976744184e-07,
+      "loss": 0.0033,
+      "step": 5358
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.382674418604651e-07,
+      "loss": 0.003,
+      "step": 5359
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.3825581395348836e-07,
+      "loss": 0.2669,
+      "step": 5360
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.3824418604651162e-07,
+      "loss": 0.0023,
+      "step": 5361
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.3823255813953486e-07,
+      "loss": 0.0287,
+      "step": 5362
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.3822093023255812e-07,
+      "loss": 0.0049,
+      "step": 5363
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.382093023255814e-07,
+      "loss": 0.0027,
+      "step": 5364
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.3819767441860464e-07,
+      "loss": 0.0052,
+      "step": 5365
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.3818604651162788e-07,
+      "loss": 0.0042,
+      "step": 5366
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.3817441860465114e-07,
+      "loss": 0.0052,
+      "step": 5367
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.3816279069767443e-07,
+      "loss": 0.0044,
+      "step": 5368
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.3815116279069766e-07,
+      "loss": 0.005,
+      "step": 5369
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.3813953488372093e-07,
+      "loss": 0.0092,
+      "step": 5370
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.3812790697674416e-07,
+      "loss": 0.0023,
+      "step": 5371
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.3811627906976745e-07,
+      "loss": 0.0051,
+      "step": 5372
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.3810465116279068e-07,
+      "loss": 0.004,
+      "step": 5373
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.3809302325581395e-07,
+      "loss": 0.0035,
+      "step": 5374
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.3808139534883718e-07,
+      "loss": 0.0038,
+      "step": 5375
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.3806976744186047e-07,
+      "loss": 0.0048,
+      "step": 5376
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.380581395348837e-07,
+      "loss": 0.0038,
+      "step": 5377
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.3804651162790697e-07,
+      "loss": 0.0047,
+      "step": 5378
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.380348837209302e-07,
+      "loss": 0.0023,
+      "step": 5379
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.380232558139535e-07,
+      "loss": 0.0023,
+      "step": 5380
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.3801162790697675e-07,
+      "loss": 0.0034,
+      "step": 5381
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.38e-07,
+      "loss": 0.0046,
+      "step": 5382
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.3798837209302325e-07,
+      "loss": 0.0054,
+      "step": 5383
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.3797674418604648e-07,
+      "loss": 0.0045,
+      "step": 5384
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.3796511627906977e-07,
+      "loss": 0.032,
+      "step": 5385
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.37953488372093e-07,
+      "loss": 0.0028,
+      "step": 5386
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.3794186046511627e-07,
+      "loss": 0.0043,
+      "step": 5387
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.379302325581395e-07,
+      "loss": 0.0037,
+      "step": 5388
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.379186046511628e-07,
+      "loss": 0.0114,
+      "step": 5389
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3790697674418603e-07,
+      "loss": 0.0033,
+      "step": 5390
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.378953488372093e-07,
+      "loss": 1.8003,
+      "step": 5391
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3788372093023252e-07,
+      "loss": 0.0059,
+      "step": 5392
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.378720930232558e-07,
+      "loss": 0.0037,
+      "step": 5393
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3786046511627907e-07,
+      "loss": 0.0048,
+      "step": 5394
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.378488372093023e-07,
+      "loss": 0.0031,
+      "step": 5395
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3783720930232557e-07,
+      "loss": 0.0033,
+      "step": 5396
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3782558139534883e-07,
+      "loss": 0.0049,
+      "step": 5397
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.378139534883721e-07,
+      "loss": 0.0037,
+      "step": 5398
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3780232558139533e-07,
+      "loss": 0.003,
+      "step": 5399
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.377906976744186e-07,
+      "loss": 0.0053,
+      "step": 5400
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3777906976744185e-07,
+      "loss": 0.0045,
+      "step": 5401
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3776744186046512e-07,
+      "loss": 0.0044,
+      "step": 5402
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3775581395348835e-07,
+      "loss": 0.1521,
+      "step": 5403
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.377441860465116e-07,
+      "loss": 0.6434,
+      "step": 5404
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.377325581395349e-07,
+      "loss": 0.1698,
+      "step": 5405
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3772093023255814e-07,
+      "loss": 0.0053,
+      "step": 5406
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.377093023255814e-07,
+      "loss": 0.0038,
+      "step": 5407
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3769767441860463e-07,
+      "loss": 0.006,
+      "step": 5408
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3768604651162787e-07,
+      "loss": 0.0023,
+      "step": 5409
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3767441860465116e-07,
+      "loss": 0.0036,
+      "step": 5410
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3766279069767442e-07,
+      "loss": 0.0036,
+      "step": 5411
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3765116279069765e-07,
+      "loss": 0.0038,
+      "step": 5412
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3763953488372091e-07,
+      "loss": 0.0029,
+      "step": 5413
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3762790697674418e-07,
+      "loss": 0.0462,
+      "step": 5414
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3761627906976744e-07,
+      "loss": 1.3882,
+      "step": 5415
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3760465116279067e-07,
+      "loss": 0.0047,
+      "step": 5416
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3759302325581394e-07,
+      "loss": 0.4833,
+      "step": 5417
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3758139534883722e-07,
+      "loss": 0.0029,
+      "step": 5418
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3756976744186046e-07,
+      "loss": 0.0029,
+      "step": 5419
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.375581395348837e-07,
+      "loss": 0.0028,
+      "step": 5420
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3754651162790696e-07,
+      "loss": 0.0033,
+      "step": 5421
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3753488372093024e-07,
+      "loss": 0.0043,
+      "step": 5422
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3752325581395348e-07,
+      "loss": 0.2812,
+      "step": 5423
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3751162790697674e-07,
+      "loss": 0.0056,
+      "step": 5424
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3749999999999998e-07,
+      "loss": 0.3623,
+      "step": 5425
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3748837209302326e-07,
+      "loss": 0.1582,
+      "step": 5426
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.374767441860465e-07,
+      "loss": 0.0031,
+      "step": 5427
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3746511627906976e-07,
+      "loss": 0.0028,
+      "step": 5428
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.37453488372093e-07,
+      "loss": 0.0027,
+      "step": 5429
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3744186046511628e-07,
+      "loss": 0.0052,
+      "step": 5430
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3743023255813952e-07,
+      "loss": 0.0031,
+      "step": 5431
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3741860465116278e-07,
+      "loss": 0.0041,
+      "step": 5432
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3740697674418602e-07,
+      "loss": 0.0029,
+      "step": 5433
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3739534883720928e-07,
+      "loss": 0.0032,
+      "step": 5434
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3738372093023257e-07,
+      "loss": 0.0051,
+      "step": 5435
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.373720930232558e-07,
+      "loss": 0.0029,
+      "step": 5436
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3736046511627906e-07,
+      "loss": 0.0048,
+      "step": 5437
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.373488372093023e-07,
+      "loss": 0.0026,
+      "step": 5438
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.373372093023256e-07,
+      "loss": 0.0027,
+      "step": 5439
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3732558139534882e-07,
+      "loss": 0.0042,
+      "step": 5440
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3731395348837208e-07,
+      "loss": 0.0041,
+      "step": 5441
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3730232558139532e-07,
+      "loss": 0.0034,
+      "step": 5442
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.372906976744186e-07,
+      "loss": 0.0028,
+      "step": 5443
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3727906976744184e-07,
+      "loss": 0.0031,
+      "step": 5444
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.372674418604651e-07,
+      "loss": 0.8352,
+      "step": 5445
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3725581395348834e-07,
+      "loss": 0.0029,
+      "step": 5446
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3724418604651163e-07,
+      "loss": 0.005,
+      "step": 5447
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.372325581395349e-07,
+      "loss": 0.0031,
+      "step": 5448
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3722093023255812e-07,
+      "loss": 0.0069,
+      "step": 5449
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3720930232558136e-07,
+      "loss": 0.0054,
+      "step": 5450
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3719767441860465e-07,
+      "loss": 0.0036,
+      "step": 5451
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.371860465116279e-07,
+      "loss": 0.0022,
+      "step": 5452
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3717441860465115e-07,
+      "loss": 0.0037,
+      "step": 5453
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.371627906976744e-07,
+      "loss": 0.0043,
+      "step": 5454
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3715116279069767e-07,
+      "loss": 0.0131,
+      "step": 5455
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3713953488372093e-07,
+      "loss": 0.0043,
+      "step": 5456
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3712790697674417e-07,
+      "loss": 0.0036,
+      "step": 5457
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3711627906976743e-07,
+      "loss": 0.0051,
+      "step": 5458
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3710465116279072e-07,
+      "loss": 0.0077,
+      "step": 5459
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3709302325581395e-07,
+      "loss": 0.0029,
+      "step": 5460
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3708139534883719e-07,
+      "loss": 0.0029,
+      "step": 5461
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3706976744186045e-07,
+      "loss": 0.0032,
+      "step": 5462
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3705813953488368e-07,
+      "loss": 0.0029,
+      "step": 5463
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3704651162790697e-07,
+      "loss": 0.0043,
+      "step": 5464
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3703488372093023e-07,
+      "loss": 0.0036,
+      "step": 5465
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3702325581395347e-07,
+      "loss": 0.0029,
+      "step": 5466
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3701162790697673e-07,
+      "loss": 0.0029,
+      "step": 5467
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.37e-07,
+      "loss": 0.0029,
+      "step": 5468
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3698837209302325e-07,
+      "loss": 0.0048,
+      "step": 5469
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.369767441860465e-07,
+      "loss": 0.0025,
+      "step": 5470
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3696511627906975e-07,
+      "loss": 1.0858,
+      "step": 5471
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.36953488372093e-07,
+      "loss": 0.0046,
+      "step": 5472
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3694186046511627e-07,
+      "loss": 0.003,
+      "step": 5473
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.369302325581395e-07,
+      "loss": 0.0037,
+      "step": 5474
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3691860465116277e-07,
+      "loss": 0.0028,
+      "step": 5475
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3690697674418606e-07,
+      "loss": 0.0037,
+      "step": 5476
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.368953488372093e-07,
+      "loss": 0.0026,
+      "step": 5477
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3688372093023256e-07,
+      "loss": 0.0044,
+      "step": 5478
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.368720930232558e-07,
+      "loss": 0.0033,
+      "step": 5479
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3686046511627908e-07,
+      "loss": 0.3807,
+      "step": 5480
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3684883720930231e-07,
+      "loss": 0.0029,
+      "step": 5481
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3683720930232558e-07,
+      "loss": 0.0022,
+      "step": 5482
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.368255813953488e-07,
+      "loss": 0.0029,
+      "step": 5483
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.368139534883721e-07,
+      "loss": 0.003,
+      "step": 5484
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3680232558139533e-07,
+      "loss": 0.0028,
+      "step": 5485
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.367906976744186e-07,
+      "loss": 0.0462,
+      "step": 5486
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3677906976744183e-07,
+      "loss": 0.0045,
+      "step": 5487
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.367674418604651e-07,
+      "loss": 0.0048,
+      "step": 5488
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3675581395348838e-07,
+      "loss": 0.0033,
+      "step": 5489
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3674418604651162e-07,
+      "loss": 0.0052,
+      "step": 5490
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3673255813953488e-07,
+      "loss": 1.8307,
+      "step": 5491
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3672093023255811e-07,
+      "loss": 0.003,
+      "step": 5492
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.367093023255814e-07,
+      "loss": 0.0038,
+      "step": 5493
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3669767441860464e-07,
+      "loss": 0.0069,
+      "step": 5494
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.366860465116279e-07,
+      "loss": 0.0625,
+      "step": 5495
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3667441860465113e-07,
+      "loss": 0.0039,
+      "step": 5496
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3666279069767442e-07,
+      "loss": 0.0029,
+      "step": 5497
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3665116279069766e-07,
+      "loss": 0.0045,
+      "step": 5498
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3663953488372092e-07,
+      "loss": 0.012,
+      "step": 5499
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3662790697674415e-07,
+      "loss": 0.0028,
+      "step": 5500
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3661627906976744e-07,
+      "loss": 0.0021,
+      "step": 5501
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.366046511627907e-07,
+      "loss": 0.0036,
+      "step": 5502
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3659302325581394e-07,
+      "loss": 0.0032,
+      "step": 5503
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3658139534883717e-07,
+      "loss": 0.0025,
+      "step": 5504
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3656976744186046e-07,
+      "loss": 0.0157,
+      "step": 5505
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3655813953488372e-07,
+      "loss": 0.006,
+      "step": 5506
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3654651162790696e-07,
+      "loss": 0.0033,
+      "step": 5507
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3653488372093022e-07,
+      "loss": 0.0026,
+      "step": 5508
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3652325581395348e-07,
+      "loss": 2.2376,
+      "step": 5509
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3651162790697675e-07,
+      "loss": 0.0036,
+      "step": 5510
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3649999999999998e-07,
+      "loss": 0.0039,
+      "step": 5511
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3648837209302324e-07,
+      "loss": 0.002,
+      "step": 5512
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3647674418604653e-07,
+      "loss": 0.0065,
+      "step": 5513
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3646511627906977e-07,
+      "loss": 0.0034,
+      "step": 5514
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.36453488372093e-07,
+      "loss": 0.0033,
+      "step": 5515
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3644186046511626e-07,
+      "loss": 0.0026,
+      "step": 5516
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.364302325581395e-07,
+      "loss": 0.0034,
+      "step": 5517
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3641860465116279e-07,
+      "loss": 0.0032,
+      "step": 5518
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3640697674418605e-07,
+      "loss": 0.0039,
+      "step": 5519
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3639534883720928e-07,
+      "loss": 0.0044,
+      "step": 5520
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3638372093023254e-07,
+      "loss": 0.0039,
+      "step": 5521
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.363720930232558e-07,
+      "loss": 0.0031,
+      "step": 5522
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3636046511627907e-07,
+      "loss": 0.0027,
+      "step": 5523
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.363488372093023e-07,
+      "loss": 0.5518,
+      "step": 5524
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3633720930232556e-07,
+      "loss": 0.0025,
+      "step": 5525
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3632558139534883e-07,
+      "loss": 2.0207,
+      "step": 5526
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.363139534883721e-07,
+      "loss": 0.0043,
+      "step": 5527
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3630232558139532e-07,
+      "loss": 0.0084,
+      "step": 5528
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3629069767441859e-07,
+      "loss": 0.0036,
+      "step": 5529
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3627906976744187e-07,
+      "loss": 0.0935,
+      "step": 5530
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.362674418604651e-07,
+      "loss": 0.0206,
+      "step": 5531
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3625581395348837e-07,
+      "loss": 0.0042,
+      "step": 5532
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.362441860465116e-07,
+      "loss": 2.6605,
+      "step": 5533
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.362325581395349e-07,
+      "loss": 0.0027,
+      "step": 5534
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3622093023255813e-07,
+      "loss": 0.004,
+      "step": 5535
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.362093023255814e-07,
+      "loss": 0.0034,
+      "step": 5536
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3619767441860463e-07,
+      "loss": 0.0227,
+      "step": 5537
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3618604651162791e-07,
+      "loss": 0.0024,
+      "step": 5538
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3617441860465115e-07,
+      "loss": 0.0035,
+      "step": 5539
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.361627906976744e-07,
+      "loss": 0.0028,
+      "step": 5540
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3615116279069765e-07,
+      "loss": 0.0026,
+      "step": 5541
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.361395348837209e-07,
+      "loss": 0.0026,
+      "step": 5542
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.361279069767442e-07,
+      "loss": 0.0044,
+      "step": 5543
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3611627906976743e-07,
+      "loss": 1.6166,
+      "step": 5544
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.361046511627907e-07,
+      "loss": 0.0023,
+      "step": 5545
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3609302325581393e-07,
+      "loss": 0.0022,
+      "step": 5546
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3608139534883722e-07,
+      "loss": 0.0029,
+      "step": 5547
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3606976744186045e-07,
+      "loss": 0.004,
+      "step": 5548
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3605813953488371e-07,
+      "loss": 0.0033,
+      "step": 5549
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3604651162790695e-07,
+      "loss": 0.0027,
+      "step": 5550
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3603488372093024e-07,
+      "loss": 0.0027,
+      "step": 5551
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3602325581395347e-07,
+      "loss": 0.0028,
+      "step": 5552
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3601162790697673e-07,
+      "loss": 0.0027,
+      "step": 5553
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3599999999999997e-07,
+      "loss": 0.0028,
+      "step": 5554
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3598837209302326e-07,
+      "loss": 0.0046,
+      "step": 5555
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.359767441860465e-07,
+      "loss": 0.0024,
+      "step": 5556
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3596511627906975e-07,
+      "loss": 0.0034,
+      "step": 5557
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.35953488372093e-07,
+      "loss": 0.0043,
+      "step": 5558
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3594186046511628e-07,
+      "loss": 0.0027,
+      "step": 5559
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3593023255813954e-07,
+      "loss": 0.0045,
+      "step": 5560
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3591860465116277e-07,
+      "loss": 0.0031,
+      "step": 5561
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3590697674418604e-07,
+      "loss": 0.0026,
+      "step": 5562
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.358953488372093e-07,
+      "loss": 0.0036,
+      "step": 5563
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3588372093023256e-07,
+      "loss": 0.0033,
+      "step": 5564
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.358720930232558e-07,
+      "loss": 0.0026,
+      "step": 5565
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3586046511627906e-07,
+      "loss": 0.1383,
+      "step": 5566
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3584883720930232e-07,
+      "loss": 0.0026,
+      "step": 5567
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3583720930232558e-07,
+      "loss": 0.0115,
+      "step": 5568
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3582558139534882e-07,
+      "loss": 0.0027,
+      "step": 5569
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3581395348837208e-07,
+      "loss": 0.0026,
+      "step": 5570
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.358023255813953e-07,
+      "loss": 0.0031,
+      "step": 5571
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.357906976744186e-07,
+      "loss": 0.173,
+      "step": 5572
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3577906976744186e-07,
+      "loss": 0.0035,
+      "step": 5573
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.357674418604651e-07,
+      "loss": 0.0057,
+      "step": 5574
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3575581395348836e-07,
+      "loss": 0.0045,
+      "step": 5575
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3574418604651162e-07,
+      "loss": 0.0578,
+      "step": 5576
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3573255813953488e-07,
+      "loss": 0.0036,
+      "step": 5577
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3572093023255812e-07,
+      "loss": 0.0042,
+      "step": 5578
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3570930232558138e-07,
+      "loss": 0.0273,
+      "step": 5579
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3569767441860464e-07,
+      "loss": 0.0035,
+      "step": 5580
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.356860465116279e-07,
+      "loss": 0.0031,
+      "step": 5581
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3567441860465114e-07,
+      "loss": 0.0789,
+      "step": 5582
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.356627906976744e-07,
+      "loss": 0.0028,
+      "step": 5583
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.356511627906977e-07,
+      "loss": 0.0039,
+      "step": 5584
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3563953488372092e-07,
+      "loss": 0.0033,
+      "step": 5585
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3562790697674419e-07,
+      "loss": 2.6459,
+      "step": 5586
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3561627906976742e-07,
+      "loss": 0.0027,
+      "step": 5587
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.356046511627907e-07,
+      "loss": 0.0032,
+      "step": 5588
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3559302325581394e-07,
+      "loss": 0.0044,
+      "step": 5589
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.355813953488372e-07,
+      "loss": 0.5678,
+      "step": 5590
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3556976744186044e-07,
+      "loss": 0.0108,
+      "step": 5591
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3555813953488373e-07,
+      "loss": 0.0032,
+      "step": 5592
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3554651162790696e-07,
+      "loss": 0.0024,
+      "step": 5593
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3553488372093023e-07,
+      "loss": 0.0023,
+      "step": 5594
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3552325581395346e-07,
+      "loss": 1.985,
+      "step": 5595
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3551162790697672e-07,
+      "loss": 0.0033,
+      "step": 5596
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.355e-07,
+      "loss": 0.1904,
+      "step": 5597
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.3548837209302325e-07,
+      "loss": 0.0036,
+      "step": 5598
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.3547674418604648e-07,
+      "loss": 0.0031,
+      "step": 5599
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.3546511627906974e-07,
+      "loss": 0.0032,
+      "step": 5600
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.3545348837209303e-07,
+      "loss": 0.0072,
+      "step": 5601
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.3544186046511627e-07,
+      "loss": 0.0034,
+      "step": 5602
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.3543023255813953e-07,
+      "loss": 0.0024,
+      "step": 5603
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.3541860465116276e-07,
+      "loss": 0.8059,
+      "step": 5604
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.3540697674418605e-07,
+      "loss": 0.0025,
+      "step": 5605
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.353953488372093e-07,
+      "loss": 0.0051,
+      "step": 5606
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.3538372093023255e-07,
+      "loss": 0.3236,
+      "step": 5607
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.3537209302325578e-07,
+      "loss": 0.0048,
+      "step": 5608
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.3536046511627907e-07,
+      "loss": 0.0019,
+      "step": 5609
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.353488372093023e-07,
+      "loss": 0.0087,
+      "step": 5610
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.3533720930232557e-07,
+      "loss": 0.0022,
+      "step": 5611
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.353255813953488e-07,
+      "loss": 0.0054,
+      "step": 5612
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.353139534883721e-07,
+      "loss": 0.0039,
+      "step": 5613
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.3530232558139535e-07,
+      "loss": 0.002,
+      "step": 5614
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.352906976744186e-07,
+      "loss": 0.021,
+      "step": 5615
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.3527906976744185e-07,
+      "loss": 0.0035,
+      "step": 5616
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.352674418604651e-07,
+      "loss": 0.1561,
+      "step": 5617
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.3525581395348837e-07,
+      "loss": 0.0048,
+      "step": 5618
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.352441860465116e-07,
+      "loss": 0.0036,
+      "step": 5619
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.3523255813953487e-07,
+      "loss": 0.0037,
+      "step": 5620
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.352209302325581e-07,
+      "loss": 2.0557,
+      "step": 5621
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.352093023255814e-07,
+      "loss": 0.038,
+      "step": 5622
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.3519767441860463e-07,
+      "loss": 0.0033,
+      "step": 5623
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.351860465116279e-07,
+      "loss": 0.0025,
+      "step": 5624
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.3517441860465113e-07,
+      "loss": 0.004,
+      "step": 5625
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.3516279069767442e-07,
+      "loss": 0.0045,
+      "step": 5626
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.3515116279069768e-07,
+      "loss": 0.0031,
+      "step": 5627
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.351395348837209e-07,
+      "loss": 0.0019,
+      "step": 5628
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.3512790697674417e-07,
+      "loss": 0.0068,
+      "step": 5629
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.3511627906976744e-07,
+      "loss": 0.0029,
+      "step": 5630
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.351046511627907e-07,
+      "loss": 0.0025,
+      "step": 5631
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.3509302325581393e-07,
+      "loss": 0.0023,
+      "step": 5632
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.350813953488372e-07,
+      "loss": 0.6804,
+      "step": 5633
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.3506976744186046e-07,
+      "loss": 0.0035,
+      "step": 5634
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.3505813953488372e-07,
+      "loss": 0.0026,
+      "step": 5635
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.3504651162790695e-07,
+      "loss": 0.1457,
+      "step": 5636
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.3503488372093021e-07,
+      "loss": 0.004,
+      "step": 5637
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.350232558139535e-07,
+      "loss": 0.0029,
+      "step": 5638
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.3501162790697674e-07,
+      "loss": 0.0038,
+      "step": 5639
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.35e-07,
+      "loss": 0.0028,
+      "step": 5640
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.3498837209302324e-07,
+      "loss": 0.0032,
+      "step": 5641
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.3497674418604652e-07,
+      "loss": 0.0045,
+      "step": 5642
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.3496511627906976e-07,
+      "loss": 0.0029,
+      "step": 5643
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.3495348837209302e-07,
+      "loss": 0.0032,
+      "step": 5644
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.3494186046511626e-07,
+      "loss": 0.023,
+      "step": 5645
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.3493023255813954e-07,
+      "loss": 0.0123,
+      "step": 5646
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.3491860465116278e-07,
+      "loss": 0.0022,
+      "step": 5647
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.3490697674418604e-07,
+      "loss": 0.0028,
+      "step": 5648
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.3489534883720928e-07,
+      "loss": 0.0029,
+      "step": 5649
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.3488372093023254e-07,
+      "loss": 0.0033,
+      "step": 5650
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.3487209302325583e-07,
+      "loss": 0.0047,
+      "step": 5651
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.3486046511627906e-07,
+      "loss": 0.0032,
+      "step": 5652
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.348488372093023e-07,
+      "loss": 0.0024,
+      "step": 5653
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.3483720930232556e-07,
+      "loss": 0.0024,
+      "step": 5654
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.3482558139534885e-07,
+      "loss": 0.003,
+      "step": 5655
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.3481395348837208e-07,
+      "loss": 0.0033,
+      "step": 5656
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.3480232558139534e-07,
+      "loss": 0.003,
+      "step": 5657
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.3479069767441858e-07,
+      "loss": 0.0054,
+      "step": 5658
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.3477906976744187e-07,
+      "loss": 0.0043,
+      "step": 5659
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.347674418604651e-07,
+      "loss": 0.0026,
+      "step": 5660
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.3475581395348836e-07,
+      "loss": 0.0034,
+      "step": 5661
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.347441860465116e-07,
+      "loss": 0.0033,
+      "step": 5662
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.347325581395349e-07,
+      "loss": 0.0021,
+      "step": 5663
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.3472093023255812e-07,
+      "loss": 0.0039,
+      "step": 5664
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.3470930232558138e-07,
+      "loss": 0.0023,
+      "step": 5665
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.3469767441860462e-07,
+      "loss": 0.002,
+      "step": 5666
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.346860465116279e-07,
+      "loss": 0.0035,
+      "step": 5667
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.3467441860465117e-07,
+      "loss": 0.0022,
+      "step": 5668
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.346627906976744e-07,
+      "loss": 0.0024,
+      "step": 5669
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.3465116279069767e-07,
+      "loss": 0.0033,
+      "step": 5670
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.3463953488372093e-07,
+      "loss": 0.0026,
+      "step": 5671
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.346279069767442e-07,
+      "loss": 0.0042,
+      "step": 5672
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.3461627906976742e-07,
+      "loss": 0.0035,
+      "step": 5673
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.3460465116279069e-07,
+      "loss": 0.004,
+      "step": 5674
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.3459302325581392e-07,
+      "loss": 0.0025,
+      "step": 5675
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.345813953488372e-07,
+      "loss": 0.0025,
+      "step": 5676
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.3456976744186045e-07,
+      "loss": 0.0023,
+      "step": 5677
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.345581395348837e-07,
+      "loss": 0.0023,
+      "step": 5678
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.3454651162790694e-07,
+      "loss": 0.0019,
+      "step": 5679
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.3453488372093023e-07,
+      "loss": 0.0043,
+      "step": 5680
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.345232558139535e-07,
+      "loss": 0.0035,
+      "step": 5681
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.3451162790697673e-07,
+      "loss": 0.0036,
+      "step": 5682
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.3449999999999996e-07,
+      "loss": 0.0031,
+      "step": 5683
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.3448837209302325e-07,
+      "loss": 0.1619,
+      "step": 5684
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.344767441860465e-07,
+      "loss": 0.0025,
+      "step": 5685
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.3446511627906975e-07,
+      "loss": 0.0048,
+      "step": 5686
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.34453488372093e-07,
+      "loss": 0.4057,
+      "step": 5687
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.9825918762088974,
+      "eval_loss": 0.1536371111869812,
+      "eval_runtime": 1.5496,
+      "eval_samples_per_second": 667.278,
+      "eval_steps_per_second": 10.971,
+      "step": 5687
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.3444186046511627e-07,
+      "loss": 0.0038,
+      "step": 5688
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.3443023255813953e-07,
+      "loss": 0.2943,
+      "step": 5689
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.3441860465116277e-07,
+      "loss": 1.5334,
+      "step": 5690
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.3440697674418603e-07,
+      "loss": 0.0021,
+      "step": 5691
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.3439534883720932e-07,
+      "loss": 0.0078,
+      "step": 5692
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.3438372093023255e-07,
+      "loss": 0.0024,
+      "step": 5693
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.343720930232558e-07,
+      "loss": 0.0022,
+      "step": 5694
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.3436046511627905e-07,
+      "loss": 0.004,
+      "step": 5695
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.3434883720930234e-07,
+      "loss": 0.0034,
+      "step": 5696
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.3433720930232557e-07,
+      "loss": 0.0283,
+      "step": 5697
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.3432558139534884e-07,
+      "loss": 0.0022,
+      "step": 5698
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.3431395348837207e-07,
+      "loss": 0.003,
+      "step": 5699
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.3430232558139536e-07,
+      "loss": 0.003,
+      "step": 5700
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.342906976744186e-07,
+      "loss": 0.119,
+      "step": 5701
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.3427906976744186e-07,
+      "loss": 0.0027,
+      "step": 5702
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.342674418604651e-07,
+      "loss": 0.0027,
+      "step": 5703
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.3425581395348835e-07,
+      "loss": 0.0024,
+      "step": 5704
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.3424418604651161e-07,
+      "loss": 0.0024,
+      "step": 5705
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.3423255813953488e-07,
+      "loss": 0.0032,
+      "step": 5706
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.342209302325581e-07,
+      "loss": 0.004,
+      "step": 5707
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.3420930232558137e-07,
+      "loss": 0.0042,
+      "step": 5708
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.3419767441860466e-07,
+      "loss": 0.0044,
+      "step": 5709
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.341860465116279e-07,
+      "loss": 0.0026,
+      "step": 5710
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.3417441860465116e-07,
+      "loss": 0.0044,
+      "step": 5711
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.341627906976744e-07,
+      "loss": 0.7707,
+      "step": 5712
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.3415116279069768e-07,
+      "loss": 0.0049,
+      "step": 5713
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.3413953488372092e-07,
+      "loss": 0.0045,
+      "step": 5714
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.3412790697674418e-07,
+      "loss": 0.0027,
+      "step": 5715
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.3411627906976741e-07,
+      "loss": 0.0074,
+      "step": 5716
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.341046511627907e-07,
+      "loss": 0.0036,
+      "step": 5717
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.3409302325581394e-07,
+      "loss": 0.0029,
+      "step": 5718
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.340813953488372e-07,
+      "loss": 0.0024,
+      "step": 5719
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.3406976744186043e-07,
+      "loss": 0.0025,
+      "step": 5720
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.3405813953488372e-07,
+      "loss": 0.0028,
+      "step": 5721
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.3404651162790698e-07,
+      "loss": 0.0028,
+      "step": 5722
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.3403488372093022e-07,
+      "loss": 0.0031,
+      "step": 5723
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.3402325581395348e-07,
+      "loss": 0.0035,
+      "step": 5724
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.3401162790697674e-07,
+      "loss": 0.0019,
+      "step": 5725
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.34e-07,
+      "loss": 0.0037,
+      "step": 5726
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.3398837209302324e-07,
+      "loss": 0.0046,
+      "step": 5727
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.339767441860465e-07,
+      "loss": 0.0022,
+      "step": 5728
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.3396511627906974e-07,
+      "loss": 0.003,
+      "step": 5729
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.3395348837209302e-07,
+      "loss": 0.0038,
+      "step": 5730
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.3394186046511626e-07,
+      "loss": 0.0141,
+      "step": 5731
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.3393023255813952e-07,
+      "loss": 0.0021,
+      "step": 5732
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.3391860465116276e-07,
+      "loss": 0.0589,
+      "step": 5733
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.3390697674418604e-07,
+      "loss": 0.0035,
+      "step": 5734
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.338953488372093e-07,
+      "loss": 0.0031,
+      "step": 5735
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.3388372093023254e-07,
+      "loss": 0.0027,
+      "step": 5736
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.3387209302325578e-07,
+      "loss": 0.0042,
+      "step": 5737
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.3386046511627907e-07,
+      "loss": 0.0029,
+      "step": 5738
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.3384883720930233e-07,
+      "loss": 0.0027,
+      "step": 5739
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.3383720930232556e-07,
+      "loss": 0.0048,
+      "step": 5740
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.3382558139534882e-07,
+      "loss": 0.0031,
+      "step": 5741
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.3381395348837209e-07,
+      "loss": 0.0037,
+      "step": 5742
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.3380232558139535e-07,
+      "loss": 0.0023,
+      "step": 5743
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.3379069767441858e-07,
+      "loss": 0.0028,
+      "step": 5744
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.3377906976744184e-07,
+      "loss": 0.0059,
+      "step": 5745
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.3376744186046513e-07,
+      "loss": 0.0017,
+      "step": 5746
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.3375581395348837e-07,
+      "loss": 0.0022,
+      "step": 5747
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.337441860465116e-07,
+      "loss": 0.0046,
+      "step": 5748
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.3373255813953486e-07,
+      "loss": 0.0027,
+      "step": 5749
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.3372093023255815e-07,
+      "loss": 0.0028,
+      "step": 5750
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.337093023255814e-07,
+      "loss": 0.0069,
+      "step": 5751
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.3369767441860465e-07,
+      "loss": 0.0029,
+      "step": 5752
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.3368604651162789e-07,
+      "loss": 0.0023,
+      "step": 5753
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.3367441860465115e-07,
+      "loss": 0.0035,
+      "step": 5754
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.336627906976744e-07,
+      "loss": 1.5303,
+      "step": 5755
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.3365116279069767e-07,
+      "loss": 0.637,
+      "step": 5756
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.336395348837209e-07,
+      "loss": 0.002,
+      "step": 5757
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.3362790697674417e-07,
+      "loss": 0.0045,
+      "step": 5758
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.3361627906976743e-07,
+      "loss": 0.0035,
+      "step": 5759
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.336046511627907e-07,
+      "loss": 0.0034,
+      "step": 5760
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.3359302325581393e-07,
+      "loss": 0.0018,
+      "step": 5761
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.335813953488372e-07,
+      "loss": 0.0043,
+      "step": 5762
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.3356976744186048e-07,
+      "loss": 0.0043,
+      "step": 5763
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.335581395348837e-07,
+      "loss": 0.0041,
+      "step": 5764
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.3354651162790697e-07,
+      "loss": 0.0042,
+      "step": 5765
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.335348837209302e-07,
+      "loss": 0.0047,
+      "step": 5766
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.335232558139535e-07,
+      "loss": 0.0033,
+      "step": 5767
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.3351162790697673e-07,
+      "loss": 0.0487,
+      "step": 5768
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.335e-07,
+      "loss": 0.0144,
+      "step": 5769
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.3348837209302323e-07,
+      "loss": 0.0032,
+      "step": 5770
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.3347674418604652e-07,
+      "loss": 0.0036,
+      "step": 5771
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.3346511627906975e-07,
+      "loss": 0.0032,
+      "step": 5772
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.3345348837209301e-07,
+      "loss": 0.0036,
+      "step": 5773
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.3344186046511625e-07,
+      "loss": 0.0031,
+      "step": 5774
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.3343023255813954e-07,
+      "loss": 0.0027,
+      "step": 5775
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.334186046511628e-07,
+      "loss": 0.0039,
+      "step": 5776
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.3340697674418603e-07,
+      "loss": 0.0034,
+      "step": 5777
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.333953488372093e-07,
+      "loss": 0.0026,
+      "step": 5778
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.3338372093023256e-07,
+      "loss": 0.0034,
+      "step": 5779
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.3337209302325582e-07,
+      "loss": 0.0037,
+      "step": 5780
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.3336046511627905e-07,
+      "loss": 0.0027,
+      "step": 5781
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.3334883720930232e-07,
+      "loss": 0.0021,
+      "step": 5782
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.3333720930232555e-07,
+      "loss": 0.0036,
+      "step": 5783
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.3332558139534884e-07,
+      "loss": 0.0027,
+      "step": 5784
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.3331395348837207e-07,
+      "loss": 0.0021,
+      "step": 5785
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.3330232558139534e-07,
+      "loss": 1.8359,
+      "step": 5786
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.3329069767441857e-07,
+      "loss": 0.0026,
+      "step": 5787
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.3327906976744186e-07,
+      "loss": 0.0021,
+      "step": 5788
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.3326744186046512e-07,
+      "loss": 0.0037,
+      "step": 5789
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.3325581395348836e-07,
+      "loss": 0.0044,
+      "step": 5790
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.332441860465116e-07,
+      "loss": 0.004,
+      "step": 5791
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.3323255813953488e-07,
+      "loss": 0.0024,
+      "step": 5792
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.3322093023255814e-07,
+      "loss": 0.0024,
+      "step": 5793
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.3320930232558138e-07,
+      "loss": 0.0023,
+      "step": 5794
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.3319767441860464e-07,
+      "loss": 0.0028,
+      "step": 5795
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.331860465116279e-07,
+      "loss": 0.0026,
+      "step": 5796
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.3317441860465116e-07,
+      "loss": 0.2053,
+      "step": 5797
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.331627906976744e-07,
+      "loss": 0.003,
+      "step": 5798
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.3315116279069766e-07,
+      "loss": 0.0037,
+      "step": 5799
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.3313953488372092e-07,
+      "loss": 0.0022,
+      "step": 5800
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.3312790697674418e-07,
+      "loss": 0.0036,
+      "step": 5801
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.3311627906976742e-07,
+      "loss": 0.0028,
+      "step": 5802
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.3310465116279068e-07,
+      "loss": 0.0021,
+      "step": 5803
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.3309302325581397e-07,
+      "loss": 0.0041,
+      "step": 5804
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.330813953488372e-07,
+      "loss": 0.0035,
+      "step": 5805
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.3306976744186046e-07,
+      "loss": 0.003,
+      "step": 5806
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.330581395348837e-07,
+      "loss": 0.0032,
+      "step": 5807
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.3304651162790696e-07,
+      "loss": 0.002,
+      "step": 5808
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.3303488372093022e-07,
+      "loss": 0.0032,
+      "step": 5809
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.3302325581395349e-07,
+      "loss": 0.0025,
+      "step": 5810
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.3301162790697672e-07,
+      "loss": 0.6963,
+      "step": 5811
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.3299999999999998e-07,
+      "loss": 0.0032,
+      "step": 5812
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.3298837209302324e-07,
+      "loss": 0.0025,
+      "step": 5813
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.329767441860465e-07,
+      "loss": 0.0025,
+      "step": 5814
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.3296511627906974e-07,
+      "loss": 0.0023,
+      "step": 5815
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.32953488372093e-07,
+      "loss": 0.0031,
+      "step": 5816
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.329418604651163e-07,
+      "loss": 0.0038,
+      "step": 5817
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.3293023255813953e-07,
+      "loss": 0.0079,
+      "step": 5818
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.329186046511628e-07,
+      "loss": 0.0019,
+      "step": 5819
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.3290697674418602e-07,
+      "loss": 0.1553,
+      "step": 5820
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.328953488372093e-07,
+      "loss": 0.0022,
+      "step": 5821
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.3288372093023255e-07,
+      "loss": 0.0025,
+      "step": 5822
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.328720930232558e-07,
+      "loss": 0.0024,
+      "step": 5823
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.3286046511627904e-07,
+      "loss": 0.0025,
+      "step": 5824
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.3284883720930233e-07,
+      "loss": 2.2742,
+      "step": 5825
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.3283720930232557e-07,
+      "loss": 0.0025,
+      "step": 5826
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.3282558139534883e-07,
+      "loss": 0.0029,
+      "step": 5827
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.3281395348837206e-07,
+      "loss": 2.3964,
+      "step": 5828
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.3280232558139535e-07,
+      "loss": 0.0031,
+      "step": 5829
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.3279069767441861e-07,
+      "loss": 0.003,
+      "step": 5830
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.3277906976744185e-07,
+      "loss": 2.0488,
+      "step": 5831
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.3276744186046508e-07,
+      "loss": 0.0429,
+      "step": 5832
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.3275581395348837e-07,
+      "loss": 0.0027,
+      "step": 5833
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.3274418604651163e-07,
+      "loss": 0.0092,
+      "step": 5834
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.3273255813953487e-07,
+      "loss": 0.0133,
+      "step": 5835
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.3272093023255813e-07,
+      "loss": 0.0088,
+      "step": 5836
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.3270930232558137e-07,
+      "loss": 1.76,
+      "step": 5837
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.3269767441860465e-07,
+      "loss": 0.0026,
+      "step": 5838
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.326860465116279e-07,
+      "loss": 0.0034,
+      "step": 5839
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.3267441860465115e-07,
+      "loss": 0.2317,
+      "step": 5840
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.3266279069767439e-07,
+      "loss": 0.0026,
+      "step": 5841
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.3265116279069767e-07,
+      "loss": 0.0033,
+      "step": 5842
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.326395348837209e-07,
+      "loss": 0.0214,
+      "step": 5843
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.3262790697674417e-07,
+      "loss": 0.0044,
+      "step": 5844
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.326162790697674e-07,
+      "loss": 0.0035,
+      "step": 5845
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.326046511627907e-07,
+      "loss": 0.0026,
+      "step": 5846
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.3259302325581396e-07,
+      "loss": 0.0057,
+      "step": 5847
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.325813953488372e-07,
+      "loss": 0.0037,
+      "step": 5848
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.3256976744186045e-07,
+      "loss": 0.0033,
+      "step": 5849
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.3255813953488372e-07,
+      "loss": 0.0023,
+      "step": 5850
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.3254651162790698e-07,
+      "loss": 0.0135,
+      "step": 5851
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.325348837209302e-07,
+      "loss": 0.0045,
+      "step": 5852
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.3252325581395347e-07,
+      "loss": 0.003,
+      "step": 5853
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.3251162790697674e-07,
+      "loss": 0.0031,
+      "step": 5854
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.325e-07,
+      "loss": 0.0094,
+      "step": 5855
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.3248837209302323e-07,
+      "loss": 0.0021,
+      "step": 5856
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.324767441860465e-07,
+      "loss": 0.004,
+      "step": 5857
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.3246511627906978e-07,
+      "loss": 0.2891,
+      "step": 5858
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.3245348837209302e-07,
+      "loss": 0.0031,
+      "step": 5859
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.3244186046511628e-07,
+      "loss": 0.0037,
+      "step": 5860
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.3243023255813951e-07,
+      "loss": 0.0036,
+      "step": 5861
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.3241860465116278e-07,
+      "loss": 0.0032,
+      "step": 5862
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.3240697674418604e-07,
+      "loss": 0.0029,
+      "step": 5863
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.323953488372093e-07,
+      "loss": 0.0027,
+      "step": 5864
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.3238372093023254e-07,
+      "loss": 0.0029,
+      "step": 5865
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.323720930232558e-07,
+      "loss": 0.0035,
+      "step": 5866
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.3236046511627906e-07,
+      "loss": 0.0032,
+      "step": 5867
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.3234883720930232e-07,
+      "loss": 0.0051,
+      "step": 5868
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.3233720930232556e-07,
+      "loss": 0.0034,
+      "step": 5869
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.3232558139534882e-07,
+      "loss": 0.0194,
+      "step": 5870
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.323139534883721e-07,
+      "loss": 0.0038,
+      "step": 5871
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.3230232558139534e-07,
+      "loss": 0.0051,
+      "step": 5872
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.322906976744186e-07,
+      "loss": 0.0031,
+      "step": 5873
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.3227906976744184e-07,
+      "loss": 0.0029,
+      "step": 5874
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.3226744186046513e-07,
+      "loss": 0.0034,
+      "step": 5875
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.3225581395348836e-07,
+      "loss": 0.0029,
+      "step": 5876
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.3224418604651162e-07,
+      "loss": 0.0016,
+      "step": 5877
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.3223255813953486e-07,
+      "loss": 0.0036,
+      "step": 5878
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.3222093023255815e-07,
+      "loss": 0.0027,
+      "step": 5879
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.3220930232558138e-07,
+      "loss": 0.1554,
+      "step": 5880
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.3219767441860464e-07,
+      "loss": 0.0027,
+      "step": 5881
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.3218604651162788e-07,
+      "loss": 0.0062,
+      "step": 5882
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.3217441860465117e-07,
+      "loss": 0.0022,
+      "step": 5883
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.3216279069767443e-07,
+      "loss": 0.0024,
+      "step": 5884
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.3215116279069766e-07,
+      "loss": 3.0978,
+      "step": 5885
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.321395348837209e-07,
+      "loss": 0.0032,
+      "step": 5886
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.3212790697674416e-07,
+      "loss": 0.0027,
+      "step": 5887
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.3211627906976745e-07,
+      "loss": 0.0041,
+      "step": 5888
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.3210465116279068e-07,
+      "loss": 0.0026,
+      "step": 5889
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.3209302325581395e-07,
+      "loss": 0.0144,
+      "step": 5890
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.3208139534883718e-07,
+      "loss": 0.0079,
+      "step": 5891
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.3206976744186047e-07,
+      "loss": 0.0264,
+      "step": 5892
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.320581395348837e-07,
+      "loss": 0.0033,
+      "step": 5893
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.3204651162790697e-07,
+      "loss": 0.0036,
+      "step": 5894
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.320348837209302e-07,
+      "loss": 0.0029,
+      "step": 5895
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.320232558139535e-07,
+      "loss": 0.0041,
+      "step": 5896
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.3201162790697672e-07,
+      "loss": 0.0023,
+      "step": 5897
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.3199999999999999e-07,
+      "loss": 0.0033,
+      "step": 5898
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.3198837209302322e-07,
+      "loss": 0.0024,
+      "step": 5899
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.319767441860465e-07,
+      "loss": 0.0033,
+      "step": 5900
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.3196511627906977e-07,
+      "loss": 0.0046,
+      "step": 5901
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.31953488372093e-07,
+      "loss": 0.0032,
+      "step": 5902
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.3194186046511627e-07,
+      "loss": 0.0037,
+      "step": 5903
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.3193023255813953e-07,
+      "loss": 0.0027,
+      "step": 5904
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.319186046511628e-07,
+      "loss": 0.0029,
+      "step": 5905
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.3190697674418603e-07,
+      "loss": 0.0036,
+      "step": 5906
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.318953488372093e-07,
+      "loss": 0.0032,
+      "step": 5907
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.3188372093023255e-07,
+      "loss": 0.0035,
+      "step": 5908
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.318720930232558e-07,
+      "loss": 3.1473,
+      "step": 5909
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.3186046511627905e-07,
+      "loss": 0.0037,
+      "step": 5910
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.318488372093023e-07,
+      "loss": 0.0545,
+      "step": 5911
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.318372093023256e-07,
+      "loss": 0.0022,
+      "step": 5912
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.3182558139534883e-07,
+      "loss": 0.0028,
+      "step": 5913
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.318139534883721e-07,
+      "loss": 0.0079,
+      "step": 5914
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.3180232558139533e-07,
+      "loss": 0.0037,
+      "step": 5915
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.317906976744186e-07,
+      "loss": 0.0026,
+      "step": 5916
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.3177906976744185e-07,
+      "loss": 0.005,
+      "step": 5917
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.3176744186046511e-07,
+      "loss": 0.0336,
+      "step": 5918
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.3175581395348835e-07,
+      "loss": 0.0025,
+      "step": 5919
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.317441860465116e-07,
+      "loss": 0.0061,
+      "step": 5920
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.3173255813953487e-07,
+      "loss": 0.003,
+      "step": 5921
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.3172093023255814e-07,
+      "loss": 0.0023,
+      "step": 5922
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.3170930232558137e-07,
+      "loss": 0.2713,
+      "step": 5923
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.3169767441860463e-07,
+      "loss": 0.0029,
+      "step": 5924
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.3168604651162792e-07,
+      "loss": 0.0031,
+      "step": 5925
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.3167441860465116e-07,
+      "loss": 0.0021,
+      "step": 5926
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.3166279069767442e-07,
+      "loss": 0.002,
+      "step": 5927
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.3165116279069765e-07,
+      "loss": 0.701,
+      "step": 5928
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.3163953488372094e-07,
+      "loss": 0.0023,
+      "step": 5929
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.3162790697674418e-07,
+      "loss": 0.0018,
+      "step": 5930
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.3161627906976744e-07,
+      "loss": 0.0023,
+      "step": 5931
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.3160465116279067e-07,
+      "loss": 0.0024,
+      "step": 5932
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.3159302325581396e-07,
+      "loss": 1.4101,
+      "step": 5933
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.315813953488372e-07,
+      "loss": 0.0257,
+      "step": 5934
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.3156976744186046e-07,
+      "loss": 0.0032,
+      "step": 5935
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.315581395348837e-07,
+      "loss": 0.0031,
+      "step": 5936
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.3154651162790698e-07,
+      "loss": 0.0043,
+      "step": 5937
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.3153488372093022e-07,
+      "loss": 0.0022,
+      "step": 5938
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.3152325581395348e-07,
+      "loss": 0.0024,
+      "step": 5939
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.3151162790697671e-07,
+      "loss": 0.0027,
+      "step": 5940
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.3149999999999998e-07,
+      "loss": 0.0028,
+      "step": 5941
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.3148837209302326e-07,
+      "loss": 0.0042,
+      "step": 5942
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.314767441860465e-07,
+      "loss": 0.0033,
+      "step": 5943
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.3146511627906976e-07,
+      "loss": 0.0025,
+      "step": 5944
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.31453488372093e-07,
+      "loss": 0.0018,
+      "step": 5945
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.3144186046511628e-07,
+      "loss": 0.0019,
+      "step": 5946
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.3143023255813952e-07,
+      "loss": 0.0031,
+      "step": 5947
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.3141860465116278e-07,
+      "loss": 0.0032,
+      "step": 5948
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.3140697674418602e-07,
+      "loss": 0.0021,
+      "step": 5949
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.313953488372093e-07,
+      "loss": 0.0025,
+      "step": 5950
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.3138372093023254e-07,
+      "loss": 0.0019,
+      "step": 5951
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.313720930232558e-07,
+      "loss": 0.0033,
+      "step": 5952
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.3136046511627904e-07,
+      "loss": 0.0027,
+      "step": 5953
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.3134883720930232e-07,
+      "loss": 0.0024,
+      "step": 5954
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.3133720930232559e-07,
+      "loss": 0.0021,
+      "step": 5955
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.3132558139534882e-07,
+      "loss": 0.0031,
+      "step": 5956
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.3131395348837208e-07,
+      "loss": 0.0023,
+      "step": 5957
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.3130232558139534e-07,
+      "loss": 0.0029,
+      "step": 5958
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.312906976744186e-07,
+      "loss": 0.0029,
+      "step": 5959
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.3127906976744184e-07,
+      "loss": 0.0021,
+      "step": 5960
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.312674418604651e-07,
+      "loss": 0.5324,
+      "step": 5961
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.3125581395348837e-07,
+      "loss": 0.0025,
+      "step": 5962
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.3124418604651163e-07,
+      "loss": 3.085,
+      "step": 5963
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.3123255813953486e-07,
+      "loss": 0.0024,
+      "step": 5964
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.3122093023255812e-07,
+      "loss": 0.0022,
+      "step": 5965
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.312093023255814e-07,
+      "loss": 0.0096,
+      "step": 5966
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.3119767441860465e-07,
+      "loss": 0.0021,
+      "step": 5967
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.311860465116279e-07,
+      "loss": 0.1264,
+      "step": 5968
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.3117441860465114e-07,
+      "loss": 0.0054,
+      "step": 5969
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.3116279069767438e-07,
+      "loss": 0.0031,
+      "step": 5970
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.3115116279069767e-07,
+      "loss": 0.0028,
+      "step": 5971
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.3113953488372093e-07,
+      "loss": 0.0029,
+      "step": 5972
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.3112790697674416e-07,
+      "loss": 0.0043,
+      "step": 5973
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.3111627906976743e-07,
+      "loss": 0.0023,
+      "step": 5974
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.311046511627907e-07,
+      "loss": 0.0029,
+      "step": 5975
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.3109302325581395e-07,
+      "loss": 0.011,
+      "step": 5976
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.3108139534883719e-07,
+      "loss": 0.0028,
+      "step": 5977
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.3106976744186045e-07,
+      "loss": 0.0022,
+      "step": 5978
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.3105813953488373e-07,
+      "loss": 0.003,
+      "step": 5979
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.3104651162790697e-07,
+      "loss": 0.0031,
+      "step": 5980
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.310348837209302e-07,
+      "loss": 2.0831,
+      "step": 5981
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.3102325581395347e-07,
+      "loss": 0.0145,
+      "step": 5982
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.3101162790697676e-07,
+      "loss": 0.0044,
+      "step": 5983
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.31e-07,
+      "loss": 0.0031,
+      "step": 5984
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.3098837209302325e-07,
+      "loss": 0.0029,
+      "step": 5985
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.309767441860465e-07,
+      "loss": 0.0019,
+      "step": 5986
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.3096511627906978e-07,
+      "loss": 0.0034,
+      "step": 5987
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.30953488372093e-07,
+      "loss": 0.0033,
+      "step": 5988
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.3094186046511627e-07,
+      "loss": 0.0025,
+      "step": 5989
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.309302325581395e-07,
+      "loss": 0.0023,
+      "step": 5990
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.309186046511628e-07,
+      "loss": 0.0026,
+      "step": 5991
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.3090697674418603e-07,
+      "loss": 0.003,
+      "step": 5992
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.308953488372093e-07,
+      "loss": 0.0024,
+      "step": 5993
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.3088372093023253e-07,
+      "loss": 0.0027,
+      "step": 5994
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.308720930232558e-07,
+      "loss": 0.0021,
+      "step": 5995
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.3086046511627908e-07,
+      "loss": 0.0022,
+      "step": 5996
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.3084883720930231e-07,
+      "loss": 0.0019,
+      "step": 5997
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.3083720930232558e-07,
+      "loss": 0.0024,
+      "step": 5998
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.308255813953488e-07,
+      "loss": 0.4293,
+      "step": 5999
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.308139534883721e-07,
+      "loss": 0.0031,
+      "step": 6000
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.3080232558139533e-07,
+      "loss": 0.0027,
+      "step": 6001
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.307906976744186e-07,
+      "loss": 0.0024,
+      "step": 6002
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.3077906976744183e-07,
+      "loss": 0.0018,
+      "step": 6003
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.3076744186046512e-07,
+      "loss": 0.0033,
+      "step": 6004
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.3075581395348835e-07,
+      "loss": 0.0029,
+      "step": 6005
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.3074418604651162e-07,
+      "loss": 0.0032,
+      "step": 6006
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.3073255813953485e-07,
+      "loss": 2.1048,
+      "step": 6007
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.3072093023255814e-07,
+      "loss": 0.002,
+      "step": 6008
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.307093023255814e-07,
+      "loss": 0.0496,
+      "step": 6009
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.3069767441860464e-07,
+      "loss": 0.0033,
+      "step": 6010
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.306860465116279e-07,
+      "loss": 0.003,
+      "step": 6011
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.3067441860465116e-07,
+      "loss": 0.0021,
+      "step": 6012
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.3066279069767442e-07,
+      "loss": 1.6409,
+      "step": 6013
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.3065116279069766e-07,
+      "loss": 0.0022,
+      "step": 6014
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.3063953488372092e-07,
+      "loss": 0.002,
+      "step": 6015
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.3062790697674418e-07,
+      "loss": 0.0082,
+      "step": 6016
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.3061627906976744e-07,
+      "loss": 0.003,
+      "step": 6017
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.3060465116279068e-07,
+      "loss": 0.0029,
+      "step": 6018
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.3059302325581394e-07,
+      "loss": 0.2526,
+      "step": 6019
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.3058139534883723e-07,
+      "loss": 0.5534,
+      "step": 6020
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.3056976744186046e-07,
+      "loss": 0.0022,
+      "step": 6021
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.3055813953488372e-07,
+      "loss": 0.0039,
+      "step": 6022
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.3054651162790696e-07,
+      "loss": 0.0027,
+      "step": 6023
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.305348837209302e-07,
+      "loss": 0.0033,
+      "step": 6024
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.3052325581395348e-07,
+      "loss": 0.0019,
+      "step": 6025
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.3051162790697674e-07,
+      "loss": 0.0024,
+      "step": 6026
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.3049999999999998e-07,
+      "loss": 0.009,
+      "step": 6027
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.3048837209302324e-07,
+      "loss": 0.1623,
+      "step": 6028
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.304767441860465e-07,
+      "loss": 0.1208,
+      "step": 6029
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.3046511627906976e-07,
+      "loss": 0.0759,
+      "step": 6030
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.30453488372093e-07,
+      "loss": 0.002,
+      "step": 6031
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.3044186046511626e-07,
+      "loss": 0.0049,
+      "step": 6032
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.3043023255813955e-07,
+      "loss": 0.0023,
+      "step": 6033
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.3041860465116279e-07,
+      "loss": 0.0318,
+      "step": 6034
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.3040697674418602e-07,
+      "loss": 0.0031,
+      "step": 6035
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.3039534883720928e-07,
+      "loss": 0.0031,
+      "step": 6036
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.3038372093023257e-07,
+      "loss": 0.009,
+      "step": 6037
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.303720930232558e-07,
+      "loss": 0.002,
+      "step": 6038
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.3036046511627907e-07,
+      "loss": 0.0043,
+      "step": 6039
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.303488372093023e-07,
+      "loss": 0.0023,
+      "step": 6040
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.303372093023256e-07,
+      "loss": 0.0037,
+      "step": 6041
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.3032558139534883e-07,
+      "loss": 0.0038,
+      "step": 6042
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.303139534883721e-07,
+      "loss": 0.003,
+      "step": 6043
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.3030232558139532e-07,
+      "loss": 0.2346,
+      "step": 6044
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.302906976744186e-07,
+      "loss": 0.0025,
+      "step": 6045
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.3027906976744185e-07,
+      "loss": 0.0032,
+      "step": 6046
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.302674418604651e-07,
+      "loss": 0.0027,
+      "step": 6047
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.3025581395348834e-07,
+      "loss": 0.0019,
+      "step": 6048
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.302441860465116e-07,
+      "loss": 0.0036,
+      "step": 6049
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.302325581395349e-07,
+      "loss": 0.0031,
+      "step": 6050
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.3022093023255813e-07,
+      "loss": 0.3362,
+      "step": 6051
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.302093023255814e-07,
+      "loss": 0.0026,
+      "step": 6052
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.3019767441860463e-07,
+      "loss": 0.003,
+      "step": 6053
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.3018604651162791e-07,
+      "loss": 0.0018,
+      "step": 6054
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.3017441860465115e-07,
+      "loss": 0.003,
+      "step": 6055
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.301627906976744e-07,
+      "loss": 0.0237,
+      "step": 6056
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.3015116279069765e-07,
+      "loss": 0.0022,
+      "step": 6057
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.3013953488372093e-07,
+      "loss": 0.0054,
+      "step": 6058
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.3012790697674417e-07,
+      "loss": 0.0019,
+      "step": 6059
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.3011627906976743e-07,
+      "loss": 1.7799,
+      "step": 6060
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.3010465116279067e-07,
+      "loss": 0.0045,
+      "step": 6061
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.3009302325581395e-07,
+      "loss": 0.0034,
+      "step": 6062
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.3008139534883722e-07,
+      "loss": 0.0037,
+      "step": 6063
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.3006976744186045e-07,
+      "loss": 0.0159,
+      "step": 6064
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.3005813953488369e-07,
+      "loss": 0.0034,
+      "step": 6065
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.3004651162790697e-07,
+      "loss": 0.0026,
+      "step": 6066
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.3003488372093024e-07,
+      "loss": 0.1811,
+      "step": 6067
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.3002325581395347e-07,
+      "loss": 0.0036,
+      "step": 6068
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.3001162790697673e-07,
+      "loss": 0.0038,
+      "step": 6069
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.3e-07,
+      "loss": 0.0021,
+      "step": 6070
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.2998837209302326e-07,
+      "loss": 0.0033,
+      "step": 6071
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.299767441860465e-07,
+      "loss": 0.0032,
+      "step": 6072
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.2996511627906975e-07,
+      "loss": 0.0033,
+      "step": 6073
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.29953488372093e-07,
+      "loss": 0.0026,
+      "step": 6074
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.2994186046511628e-07,
+      "loss": 0.0028,
+      "step": 6075
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.299302325581395e-07,
+      "loss": 0.0032,
+      "step": 6076
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.2991860465116277e-07,
+      "loss": 0.0029,
+      "step": 6077
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.29906976744186e-07,
+      "loss": 0.0018,
+      "step": 6078
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.298953488372093e-07,
+      "loss": 0.0051,
+      "step": 6079
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.2988372093023256e-07,
+      "loss": 0.0036,
+      "step": 6080
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.298720930232558e-07,
+      "loss": 0.0041,
+      "step": 6081
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.2986046511627906e-07,
+      "loss": 0.0021,
+      "step": 6082
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.2984883720930232e-07,
+      "loss": 0.0074,
+      "step": 6083
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.2983720930232558e-07,
+      "loss": 0.002,
+      "step": 6084
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.2982558139534881e-07,
+      "loss": 2.7017,
+      "step": 6085
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.2981395348837208e-07,
+      "loss": 0.0026,
+      "step": 6086
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.2980232558139534e-07,
+      "loss": 0.002,
+      "step": 6087
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.297906976744186e-07,
+      "loss": 0.0041,
+      "step": 6088
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.2977906976744184e-07,
+      "loss": 0.0043,
+      "step": 6089
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.297674418604651e-07,
+      "loss": 0.0021,
+      "step": 6090
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.2975581395348838e-07,
+      "loss": 0.0022,
+      "step": 6091
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.2974418604651162e-07,
+      "loss": 0.0882,
+      "step": 6092
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.2973255813953488e-07,
+      "loss": 0.0021,
+      "step": 6093
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.2972093023255812e-07,
+      "loss": 0.0034,
+      "step": 6094
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.297093023255814e-07,
+      "loss": 0.0038,
+      "step": 6095
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.2969767441860464e-07,
+      "loss": 0.0028,
+      "step": 6096
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.296860465116279e-07,
+      "loss": 0.0022,
+      "step": 6097
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.2967441860465114e-07,
+      "loss": 0.0021,
+      "step": 6098
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.2966279069767443e-07,
+      "loss": 0.0022,
+      "step": 6099
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.2965116279069766e-07,
+      "loss": 0.1862,
+      "step": 6100
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.2963953488372092e-07,
+      "loss": 0.0024,
+      "step": 6101
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.2962790697674416e-07,
+      "loss": 0.0023,
+      "step": 6102
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.2961627906976742e-07,
+      "loss": 0.0024,
+      "step": 6103
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.296046511627907e-07,
+      "loss": 0.0023,
+      "step": 6104
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.2959302325581394e-07,
+      "loss": 0.0031,
+      "step": 6105
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.295813953488372e-07,
+      "loss": 0.0024,
+      "step": 6106
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.2956976744186044e-07,
+      "loss": 0.0041,
+      "step": 6107
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.2955813953488373e-07,
+      "loss": 0.0031,
+      "step": 6108
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.2954651162790696e-07,
+      "loss": 0.0047,
+      "step": 6109
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.2953488372093023e-07,
+      "loss": 0.0022,
+      "step": 6110
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.2952325581395346e-07,
+      "loss": 0.0044,
+      "step": 6111
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.2951162790697675e-07,
+      "loss": 0.0028,
+      "step": 6112
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.2949999999999998e-07,
+      "loss": 0.0024,
+      "step": 6113
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.2948837209302325e-07,
+      "loss": 2.2617,
+      "step": 6114
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.2947674418604648e-07,
+      "loss": 0.0026,
+      "step": 6115
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.2946511627906977e-07,
+      "loss": 0.023,
+      "step": 6116
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.2945348837209303e-07,
+      "loss": 0.0045,
+      "step": 6117
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.2944186046511627e-07,
+      "loss": 0.0023,
+      "step": 6118
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.294302325581395e-07,
+      "loss": 0.0029,
+      "step": 6119
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.294186046511628e-07,
+      "loss": 0.0025,
+      "step": 6120
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.2940697674418605e-07,
+      "loss": 0.0044,
+      "step": 6121
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.2939534883720929e-07,
+      "loss": 0.0027,
+      "step": 6122
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.2938372093023255e-07,
+      "loss": 0.0039,
+      "step": 6123
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.293720930232558e-07,
+      "loss": 0.0019,
+      "step": 6124
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.2936046511627907e-07,
+      "loss": 0.0024,
+      "step": 6125
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.293488372093023e-07,
+      "loss": 0.002,
+      "step": 6126
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.2933720930232557e-07,
+      "loss": 0.0029,
+      "step": 6127
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.293255813953488e-07,
+      "loss": 0.0023,
+      "step": 6128
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.293139534883721e-07,
+      "loss": 0.0034,
+      "step": 6129
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.2930232558139533e-07,
+      "loss": 0.0019,
+      "step": 6130
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.292906976744186e-07,
+      "loss": 0.0026,
+      "step": 6131
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.2927906976744182e-07,
+      "loss": 0.003,
+      "step": 6132
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.292674418604651e-07,
+      "loss": 0.0071,
+      "step": 6133
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.2925581395348837e-07,
+      "loss": 0.0022,
+      "step": 6134
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.292441860465116e-07,
+      "loss": 0.3873,
+      "step": 6135
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.2923255813953487e-07,
+      "loss": 0.0019,
+      "step": 6136
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.2922093023255813e-07,
+      "loss": 0.0023,
+      "step": 6137
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.292093023255814e-07,
+      "loss": 0.0028,
+      "step": 6138
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.2919767441860463e-07,
+      "loss": 0.0032,
+      "step": 6139
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.291860465116279e-07,
+      "loss": 0.0033,
+      "step": 6140
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.2917441860465115e-07,
+      "loss": 0.0025,
+      "step": 6141
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.2916279069767441e-07,
+      "loss": 0.0163,
+      "step": 6142
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.2915116279069765e-07,
+      "loss": 0.0021,
+      "step": 6143
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.291395348837209e-07,
+      "loss": 0.0033,
+      "step": 6144
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.291279069767442e-07,
+      "loss": 0.0031,
+      "step": 6145
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.2911627906976743e-07,
+      "loss": 0.0023,
+      "step": 6146
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.291046511627907e-07,
+      "loss": 0.0031,
+      "step": 6147
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.2909302325581393e-07,
+      "loss": 0.0024,
+      "step": 6148
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.2908139534883722e-07,
+      "loss": 0.004,
+      "step": 6149
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.2906976744186046e-07,
+      "loss": 0.0086,
+      "step": 6150
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.2905813953488372e-07,
+      "loss": 0.0031,
+      "step": 6151
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.2904651162790695e-07,
+      "loss": 0.0037,
+      "step": 6152
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.2903488372093024e-07,
+      "loss": 0.0025,
+      "step": 6153
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.2902325581395348e-07,
+      "loss": 0.0045,
+      "step": 6154
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.2901162790697674e-07,
+      "loss": 0.0046,
+      "step": 6155
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.2899999999999997e-07,
+      "loss": 0.0019,
+      "step": 6156
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.2898837209302323e-07,
+      "loss": 0.002,
+      "step": 6157
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.2897674418604652e-07,
+      "loss": 0.1115,
+      "step": 6158
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.2896511627906976e-07,
+      "loss": 0.0023,
+      "step": 6159
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.2895348837209302e-07,
+      "loss": 0.0018,
+      "step": 6160
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.2894186046511625e-07,
+      "loss": 0.0029,
+      "step": 6161
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.2893023255813954e-07,
+      "loss": 0.0027,
+      "step": 6162
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.2891860465116278e-07,
+      "loss": 0.0167,
+      "step": 6163
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.2890697674418604e-07,
+      "loss": 0.003,
+      "step": 6164
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.2889534883720928e-07,
+      "loss": 0.0018,
+      "step": 6165
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.2888372093023256e-07,
+      "loss": 0.0026,
+      "step": 6166
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.288720930232558e-07,
+      "loss": 0.0029,
+      "step": 6167
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.2886046511627906e-07,
+      "loss": 0.1875,
+      "step": 6168
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.288488372093023e-07,
+      "loss": 0.003,
+      "step": 6169
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.2883720930232558e-07,
+      "loss": 0.0027,
+      "step": 6170
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.2882558139534885e-07,
+      "loss": 0.0018,
+      "step": 6171
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.2881395348837208e-07,
+      "loss": 0.003,
+      "step": 6172
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.2880232558139532e-07,
+      "loss": 0.0016,
+      "step": 6173
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.287906976744186e-07,
+      "loss": 0.0018,
+      "step": 6174
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.2877906976744187e-07,
+      "loss": 0.0022,
+      "step": 6175
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.287674418604651e-07,
+      "loss": 0.0031,
+      "step": 6176
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.2875581395348836e-07,
+      "loss": 0.0028,
+      "step": 6177
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.2874418604651162e-07,
+      "loss": 0.0029,
+      "step": 6178
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.2873255813953489e-07,
+      "loss": 0.011,
+      "step": 6179
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.2872093023255812e-07,
+      "loss": 0.0027,
+      "step": 6180
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.2870930232558138e-07,
+      "loss": 0.0022,
+      "step": 6181
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.2869767441860462e-07,
+      "loss": 0.0024,
+      "step": 6182
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.286860465116279e-07,
+      "loss": 0.0019,
+      "step": 6183
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.2867441860465114e-07,
+      "loss": 0.0022,
+      "step": 6184
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.286627906976744e-07,
+      "loss": 0.0033,
+      "step": 6185
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.2865116279069764e-07,
+      "loss": 0.0032,
+      "step": 6186
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.2863953488372093e-07,
+      "loss": 0.0028,
+      "step": 6187
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.286279069767442e-07,
+      "loss": 0.0026,
+      "step": 6188
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.2861627906976742e-07,
+      "loss": 0.2314,
+      "step": 6189
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.2860465116279069e-07,
+      "loss": 0.0027,
+      "step": 6190
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.2859302325581395e-07,
+      "loss": 0.0026,
+      "step": 6191
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.285813953488372e-07,
+      "loss": 0.0027,
+      "step": 6192
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.2856976744186044e-07,
+      "loss": 0.0023,
+      "step": 6193
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.285581395348837e-07,
+      "loss": 0.0021,
+      "step": 6194
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.2854651162790697e-07,
+      "loss": 0.0058,
+      "step": 6195
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.2853488372093023e-07,
+      "loss": 0.0035,
+      "step": 6196
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.2852325581395346e-07,
+      "loss": 0.0019,
+      "step": 6197
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.2851162790697673e-07,
+      "loss": 0.0029,
+      "step": 6198
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.2850000000000001e-07,
+      "loss": 0.0017,
+      "step": 6199
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.2848837209302325e-07,
+      "loss": 0.0044,
+      "step": 6200
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.284767441860465e-07,
+      "loss": 0.0029,
+      "step": 6201
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.2846511627906975e-07,
+      "loss": 1.5232,
+      "step": 6202
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.2845348837209303e-07,
+      "loss": 0.0029,
+      "step": 6203
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.2844186046511627e-07,
+      "loss": 0.0045,
+      "step": 6204
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.9835589941972921,
+      "eval_loss": 0.15531936287879944,
+      "eval_runtime": 1.5693,
+      "eval_samples_per_second": 658.903,
+      "eval_steps_per_second": 10.833,
+      "step": 6204
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.2843023255813953e-07,
+      "loss": 0.0023,
+      "step": 6205
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.2841860465116277e-07,
+      "loss": 0.0021,
+      "step": 6206
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.2840697674418603e-07,
+      "loss": 0.0037,
+      "step": 6207
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.283953488372093e-07,
+      "loss": 0.0894,
+      "step": 6208
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.2838372093023255e-07,
+      "loss": 0.003,
+      "step": 6209
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.283720930232558e-07,
+      "loss": 0.0024,
+      "step": 6210
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.2836046511627905e-07,
+      "loss": 0.0017,
+      "step": 6211
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.2834883720930234e-07,
+      "loss": 0.0034,
+      "step": 6212
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.2833720930232557e-07,
+      "loss": 0.0023,
+      "step": 6213
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.283255813953488e-07,
+      "loss": 0.0019,
+      "step": 6214
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.2831395348837207e-07,
+      "loss": 0.0025,
+      "step": 6215
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.2830232558139536e-07,
+      "loss": 0.0018,
+      "step": 6216
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.282906976744186e-07,
+      "loss": 2.1936,
+      "step": 6217
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.2827906976744185e-07,
+      "loss": 0.0035,
+      "step": 6218
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.282674418604651e-07,
+      "loss": 0.0023,
+      "step": 6219
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.2825581395348838e-07,
+      "loss": 0.0027,
+      "step": 6220
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.2824418604651161e-07,
+      "loss": 0.0034,
+      "step": 6221
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.2823255813953488e-07,
+      "loss": 0.0035,
+      "step": 6222
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.282209302325581e-07,
+      "loss": 0.0027,
+      "step": 6223
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.282093023255814e-07,
+      "loss": 0.9529,
+      "step": 6224
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.2819767441860463e-07,
+      "loss": 0.0035,
+      "step": 6225
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.281860465116279e-07,
+      "loss": 0.3961,
+      "step": 6226
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.2817441860465113e-07,
+      "loss": 0.0048,
+      "step": 6227
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.2816279069767442e-07,
+      "loss": 0.0028,
+      "step": 6228
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.2815116279069768e-07,
+      "loss": 0.0028,
+      "step": 6229
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.2813953488372092e-07,
+      "loss": 0.0217,
+      "step": 6230
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.2812790697674418e-07,
+      "loss": 0.0023,
+      "step": 6231
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.2811627906976744e-07,
+      "loss": 0.0017,
+      "step": 6232
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.281046511627907e-07,
+      "loss": 0.0031,
+      "step": 6233
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.2809302325581394e-07,
+      "loss": 0.0036,
+      "step": 6234
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.280813953488372e-07,
+      "loss": 0.0019,
+      "step": 6235
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.2806976744186043e-07,
+      "loss": 0.0029,
+      "step": 6236
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.2805813953488372e-07,
+      "loss": 0.002,
+      "step": 6237
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.2804651162790696e-07,
+      "loss": 0.004,
+      "step": 6238
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.2803488372093022e-07,
+      "loss": 0.0021,
+      "step": 6239
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.2802325581395345e-07,
+      "loss": 0.0023,
+      "step": 6240
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.2801162790697674e-07,
+      "loss": 0.0024,
+      "step": 6241
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.28e-07,
+      "loss": 0.0031,
+      "step": 6242
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.2798837209302324e-07,
+      "loss": 0.0023,
+      "step": 6243
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.279767441860465e-07,
+      "loss": 0.0033,
+      "step": 6244
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.2796511627906976e-07,
+      "loss": 0.0026,
+      "step": 6245
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.2795348837209302e-07,
+      "loss": 0.0027,
+      "step": 6246
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.2794186046511626e-07,
+      "loss": 0.0027,
+      "step": 6247
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.2793023255813952e-07,
+      "loss": 0.0019,
+      "step": 6248
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.2791860465116278e-07,
+      "loss": 0.0022,
+      "step": 6249
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.2790697674418604e-07,
+      "loss": 0.0027,
+      "step": 6250
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.2789534883720928e-07,
+      "loss": 0.0025,
+      "step": 6251
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.2788372093023254e-07,
+      "loss": 0.0024,
+      "step": 6252
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.2787209302325583e-07,
+      "loss": 0.0031,
+      "step": 6253
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.2786046511627906e-07,
+      "loss": 0.0033,
+      "step": 6254
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.2784883720930233e-07,
+      "loss": 0.0025,
+      "step": 6255
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.2783720930232556e-07,
+      "loss": 0.0621,
+      "step": 6256
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.2782558139534885e-07,
+      "loss": 0.004,
+      "step": 6257
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.2781395348837208e-07,
+      "loss": 0.0024,
+      "step": 6258
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.2780232558139535e-07,
+      "loss": 0.0034,
+      "step": 6259
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.2779069767441858e-07,
+      "loss": 0.6879,
+      "step": 6260
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.2777906976744184e-07,
+      "loss": 0.0033,
+      "step": 6261
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.277674418604651e-07,
+      "loss": 0.0019,
+      "step": 6262
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.2775581395348837e-07,
+      "loss": 0.0031,
+      "step": 6263
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.277441860465116e-07,
+      "loss": 0.0022,
+      "step": 6264
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.2773255813953486e-07,
+      "loss": 0.0022,
+      "step": 6265
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.2772093023255815e-07,
+      "loss": 0.0026,
+      "step": 6266
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.277093023255814e-07,
+      "loss": 0.0076,
+      "step": 6267
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.2769767441860462e-07,
+      "loss": 0.0037,
+      "step": 6268
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.2768604651162788e-07,
+      "loss": 0.0025,
+      "step": 6269
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.2767441860465117e-07,
+      "loss": 0.0019,
+      "step": 6270
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.276627906976744e-07,
+      "loss": 0.5028,
+      "step": 6271
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.2765116279069767e-07,
+      "loss": 0.0018,
+      "step": 6272
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.276395348837209e-07,
+      "loss": 0.0019,
+      "step": 6273
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.276279069767442e-07,
+      "loss": 0.0017,
+      "step": 6274
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.2761627906976743e-07,
+      "loss": 1.9878,
+      "step": 6275
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.276046511627907e-07,
+      "loss": 0.0051,
+      "step": 6276
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.2759302325581393e-07,
+      "loss": 0.0027,
+      "step": 6277
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.2758139534883721e-07,
+      "loss": 0.0023,
+      "step": 6278
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.2756976744186045e-07,
+      "loss": 0.0127,
+      "step": 6279
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.275581395348837e-07,
+      "loss": 0.0017,
+      "step": 6280
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.2754651162790695e-07,
+      "loss": 0.0026,
+      "step": 6281
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.2753488372093023e-07,
+      "loss": 0.0685,
+      "step": 6282
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.275232558139535e-07,
+      "loss": 0.0029,
+      "step": 6283
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.2751162790697673e-07,
+      "loss": 0.002,
+      "step": 6284
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.275e-07,
+      "loss": 0.0027,
+      "step": 6285
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.2748837209302325e-07,
+      "loss": 0.0032,
+      "step": 6286
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.2747674418604652e-07,
+      "loss": 0.0029,
+      "step": 6287
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.2746511627906975e-07,
+      "loss": 0.0037,
+      "step": 6288
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.27453488372093e-07,
+      "loss": 0.003,
+      "step": 6289
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.2744186046511625e-07,
+      "loss": 0.0022,
+      "step": 6290
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.2743023255813954e-07,
+      "loss": 0.002,
+      "step": 6291
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.2741860465116277e-07,
+      "loss": 0.002,
+      "step": 6292
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.2740697674418603e-07,
+      "loss": 0.0024,
+      "step": 6293
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.2739534883720927e-07,
+      "loss": 0.1927,
+      "step": 6294
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.2738372093023256e-07,
+      "loss": 0.003,
+      "step": 6295
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.2737209302325582e-07,
+      "loss": 0.0018,
+      "step": 6296
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.2736046511627905e-07,
+      "loss": 0.0156,
+      "step": 6297
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.2734883720930232e-07,
+      "loss": 0.0033,
+      "step": 6298
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.2733720930232558e-07,
+      "loss": 0.0031,
+      "step": 6299
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.2732558139534884e-07,
+      "loss": 2.9494,
+      "step": 6300
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.2731395348837207e-07,
+      "loss": 0.0028,
+      "step": 6301
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.2730232558139534e-07,
+      "loss": 0.0026,
+      "step": 6302
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.272906976744186e-07,
+      "loss": 0.0036,
+      "step": 6303
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.2727906976744186e-07,
+      "loss": 0.0031,
+      "step": 6304
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.272674418604651e-07,
+      "loss": 0.0031,
+      "step": 6305
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.2725581395348836e-07,
+      "loss": 0.0025,
+      "step": 6306
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.2724418604651164e-07,
+      "loss": 0.0019,
+      "step": 6307
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.2723255813953488e-07,
+      "loss": 0.0022,
+      "step": 6308
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.2722093023255814e-07,
+      "loss": 0.0024,
+      "step": 6309
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.2720930232558138e-07,
+      "loss": 0.0022,
+      "step": 6310
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.2719767441860466e-07,
+      "loss": 1.625,
+      "step": 6311
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.271860465116279e-07,
+      "loss": 0.0036,
+      "step": 6312
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.2717441860465116e-07,
+      "loss": 0.0024,
+      "step": 6313
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.271627906976744e-07,
+      "loss": 0.0058,
+      "step": 6314
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.2715116279069766e-07,
+      "loss": 0.0025,
+      "step": 6315
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.2713953488372092e-07,
+      "loss": 0.0043,
+      "step": 6316
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.2712790697674418e-07,
+      "loss": 0.0025,
+      "step": 6317
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.2711627906976742e-07,
+      "loss": 0.0024,
+      "step": 6318
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.2710465116279068e-07,
+      "loss": 0.0027,
+      "step": 6319
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.2709302325581394e-07,
+      "loss": 0.0023,
+      "step": 6320
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.270813953488372e-07,
+      "loss": 0.153,
+      "step": 6321
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.2706976744186044e-07,
+      "loss": 0.0031,
+      "step": 6322
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.270581395348837e-07,
+      "loss": 0.0022,
+      "step": 6323
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.27046511627907e-07,
+      "loss": 0.0028,
+      "step": 6324
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.2703488372093022e-07,
+      "loss": 0.0029,
+      "step": 6325
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.2702325581395348e-07,
+      "loss": 0.002,
+      "step": 6326
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.2701162790697672e-07,
+      "loss": 0.003,
+      "step": 6327
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.27e-07,
+      "loss": 0.0027,
+      "step": 6328
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.2698837209302324e-07,
+      "loss": 0.0023,
+      "step": 6329
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.269767441860465e-07,
+      "loss": 0.0028,
+      "step": 6330
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.2696511627906974e-07,
+      "loss": 2.1715,
+      "step": 6331
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.2695348837209303e-07,
+      "loss": 0.0028,
+      "step": 6332
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.2694186046511626e-07,
+      "loss": 0.0035,
+      "step": 6333
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.2693023255813953e-07,
+      "loss": 0.002,
+      "step": 6334
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.2691860465116276e-07,
+      "loss": 0.0025,
+      "step": 6335
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.2690697674418605e-07,
+      "loss": 0.003,
+      "step": 6336
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.268953488372093e-07,
+      "loss": 0.003,
+      "step": 6337
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.2688372093023255e-07,
+      "loss": 0.0024,
+      "step": 6338
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.268720930232558e-07,
+      "loss": 0.0018,
+      "step": 6339
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.2686046511627907e-07,
+      "loss": 0.002,
+      "step": 6340
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.2684883720930233e-07,
+      "loss": 0.0036,
+      "step": 6341
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.2683720930232557e-07,
+      "loss": 0.002,
+      "step": 6342
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.2682558139534883e-07,
+      "loss": 0.0029,
+      "step": 6343
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.2681395348837206e-07,
+      "loss": 0.0021,
+      "step": 6344
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.2680232558139535e-07,
+      "loss": 0.0024,
+      "step": 6345
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.2679069767441859e-07,
+      "loss": 0.0022,
+      "step": 6346
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.2677906976744185e-07,
+      "loss": 0.0022,
+      "step": 6347
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.2676744186046508e-07,
+      "loss": 0.0019,
+      "step": 6348
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.2675581395348837e-07,
+      "loss": 0.0028,
+      "step": 6349
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.2674418604651163e-07,
+      "loss": 0.0039,
+      "step": 6350
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.2673255813953487e-07,
+      "loss": 0.0022,
+      "step": 6351
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.267209302325581e-07,
+      "loss": 0.0114,
+      "step": 6352
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.267093023255814e-07,
+      "loss": 0.2836,
+      "step": 6353
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.2669767441860465e-07,
+      "loss": 0.0019,
+      "step": 6354
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.266860465116279e-07,
+      "loss": 0.0037,
+      "step": 6355
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.2667441860465115e-07,
+      "loss": 0.0126,
+      "step": 6356
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.266627906976744e-07,
+      "loss": 0.0021,
+      "step": 6357
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.2665116279069767e-07,
+      "loss": 1.5294,
+      "step": 6358
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.266395348837209e-07,
+      "loss": 0.0028,
+      "step": 6359
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.2662790697674417e-07,
+      "loss": 0.0021,
+      "step": 6360
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.2661627906976746e-07,
+      "loss": 0.0024,
+      "step": 6361
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.266046511627907e-07,
+      "loss": 0.011,
+      "step": 6362
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.2659302325581393e-07,
+      "loss": 0.0015,
+      "step": 6363
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.265813953488372e-07,
+      "loss": 0.0028,
+      "step": 6364
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.2656976744186048e-07,
+      "loss": 0.0043,
+      "step": 6365
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.2655813953488371e-07,
+      "loss": 0.0161,
+      "step": 6366
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.2654651162790698e-07,
+      "loss": 0.0031,
+      "step": 6367
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.265348837209302e-07,
+      "loss": 0.0015,
+      "step": 6368
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.2652325581395347e-07,
+      "loss": 0.0021,
+      "step": 6369
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.2651162790697673e-07,
+      "loss": 0.0036,
+      "step": 6370
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.265e-07,
+      "loss": 0.0038,
+      "step": 6371
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.2648837209302323e-07,
+      "loss": 0.1091,
+      "step": 6372
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.264767441860465e-07,
+      "loss": 2.2337,
+      "step": 6373
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.2646511627906976e-07,
+      "loss": 1.6763,
+      "step": 6374
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.2645348837209302e-07,
+      "loss": 0.0023,
+      "step": 6375
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.2644186046511625e-07,
+      "loss": 0.0018,
+      "step": 6376
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.2643023255813951e-07,
+      "loss": 0.002,
+      "step": 6377
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.264186046511628e-07,
+      "loss": 0.0035,
+      "step": 6378
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.2640697674418604e-07,
+      "loss": 0.0026,
+      "step": 6379
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.263953488372093e-07,
+      "loss": 0.0029,
+      "step": 6380
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.2638372093023253e-07,
+      "loss": 0.0029,
+      "step": 6381
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.2637209302325582e-07,
+      "loss": 0.2663,
+      "step": 6382
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.2636046511627906e-07,
+      "loss": 0.0019,
+      "step": 6383
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.2634883720930232e-07,
+      "loss": 0.0026,
+      "step": 6384
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.2633720930232555e-07,
+      "loss": 0.0015,
+      "step": 6385
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.2632558139534884e-07,
+      "loss": 0.0017,
+      "step": 6386
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.2631395348837208e-07,
+      "loss": 1.7652,
+      "step": 6387
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.2630232558139534e-07,
+      "loss": 0.0015,
+      "step": 6388
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.2629069767441858e-07,
+      "loss": 0.0033,
+      "step": 6389
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.2627906976744186e-07,
+      "loss": 0.0034,
+      "step": 6390
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.2626744186046512e-07,
+      "loss": 0.0072,
+      "step": 6391
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.2625581395348836e-07,
+      "loss": 0.002,
+      "step": 6392
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.2624418604651162e-07,
+      "loss": 0.0017,
+      "step": 6393
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.2623255813953486e-07,
+      "loss": 0.0023,
+      "step": 6394
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.2622093023255815e-07,
+      "loss": 0.0017,
+      "step": 6395
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.2620930232558138e-07,
+      "loss": 0.0018,
+      "step": 6396
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.2619767441860464e-07,
+      "loss": 0.0036,
+      "step": 6397
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.2618604651162788e-07,
+      "loss": 0.0031,
+      "step": 6398
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.2617441860465117e-07,
+      "loss": 0.0017,
+      "step": 6399
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.261627906976744e-07,
+      "loss": 0.0027,
+      "step": 6400
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.2615116279069766e-07,
+      "loss": 0.0029,
+      "step": 6401
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.261395348837209e-07,
+      "loss": 2.5081,
+      "step": 6402
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.2612790697674419e-07,
+      "loss": 0.0023,
+      "step": 6403
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.2611627906976745e-07,
+      "loss": 0.0435,
+      "step": 6404
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.2610465116279068e-07,
+      "loss": 0.0061,
+      "step": 6405
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.2609302325581392e-07,
+      "loss": 0.1639,
+      "step": 6406
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.260813953488372e-07,
+      "loss": 0.0037,
+      "step": 6407
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.2606976744186047e-07,
+      "loss": 0.0026,
+      "step": 6408
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.260581395348837e-07,
+      "loss": 0.0025,
+      "step": 6409
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.2604651162790697e-07,
+      "loss": 0.0023,
+      "step": 6410
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.2603488372093023e-07,
+      "loss": 0.0034,
+      "step": 6411
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.260232558139535e-07,
+      "loss": 0.0084,
+      "step": 6412
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.2601162790697672e-07,
+      "loss": 0.0018,
+      "step": 6413
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.2599999999999999e-07,
+      "loss": 0.0017,
+      "step": 6414
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.2598837209302327e-07,
+      "loss": 0.0269,
+      "step": 6415
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.259767441860465e-07,
+      "loss": 0.0026,
+      "step": 6416
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.2596511627906974e-07,
+      "loss": 0.0019,
+      "step": 6417
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.25953488372093e-07,
+      "loss": 0.0025,
+      "step": 6418
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.259418604651163e-07,
+      "loss": 0.0019,
+      "step": 6419
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.2593023255813953e-07,
+      "loss": 0.0017,
+      "step": 6420
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.259186046511628e-07,
+      "loss": 0.0019,
+      "step": 6421
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.2590697674418603e-07,
+      "loss": 0.0025,
+      "step": 6422
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.258953488372093e-07,
+      "loss": 0.0024,
+      "step": 6423
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.2588372093023255e-07,
+      "loss": 0.0096,
+      "step": 6424
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.258720930232558e-07,
+      "loss": 0.0025,
+      "step": 6425
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.2586046511627905e-07,
+      "loss": 0.002,
+      "step": 6426
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.258488372093023e-07,
+      "loss": 0.003,
+      "step": 6427
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.2583720930232557e-07,
+      "loss": 0.0032,
+      "step": 6428
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2582558139534883e-07,
+      "loss": 0.0029,
+      "step": 6429
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2581395348837207e-07,
+      "loss": 0.0023,
+      "step": 6430
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2580232558139533e-07,
+      "loss": 0.0017,
+      "step": 6431
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2579069767441862e-07,
+      "loss": 0.0027,
+      "step": 6432
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2577906976744185e-07,
+      "loss": 0.0018,
+      "step": 6433
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2576744186046511e-07,
+      "loss": 0.0036,
+      "step": 6434
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.2575581395348835e-07,
+      "loss": 0.0025,
+      "step": 6435
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.2574418604651164e-07,
+      "loss": 0.0022,
+      "step": 6436
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.2573255813953487e-07,
+      "loss": 0.1149,
+      "step": 6437
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.2572093023255813e-07,
+      "loss": 0.3245,
+      "step": 6438
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.2570930232558137e-07,
+      "loss": 0.0029,
+      "step": 6439
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.2569767441860466e-07,
+      "loss": 0.0028,
+      "step": 6440
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.256860465116279e-07,
+      "loss": 0.0046,
+      "step": 6441
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.2567441860465115e-07,
+      "loss": 0.0022,
+      "step": 6442
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.256627906976744e-07,
+      "loss": 0.0034,
+      "step": 6443
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.2565116279069768e-07,
+      "loss": 0.0023,
+      "step": 6444
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.2563953488372094e-07,
+      "loss": 0.0019,
+      "step": 6445
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.2562790697674418e-07,
+      "loss": 0.0025,
+      "step": 6446
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.256162790697674e-07,
+      "loss": 0.0026,
+      "step": 6447
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.2560465116279067e-07,
+      "loss": 0.0021,
+      "step": 6448
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.2559302325581396e-07,
+      "loss": 0.0033,
+      "step": 6449
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.255813953488372e-07,
+      "loss": 0.0019,
+      "step": 6450
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.2556976744186046e-07,
+      "loss": 0.0022,
+      "step": 6451
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.255581395348837e-07,
+      "loss": 0.0031,
+      "step": 6452
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.2554651162790698e-07,
+      "loss": 0.0099,
+      "step": 6453
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.2553488372093022e-07,
+      "loss": 0.0029,
+      "step": 6454
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.2552325581395348e-07,
+      "loss": 0.0163,
+      "step": 6455
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.255116279069767e-07,
+      "loss": 0.0014,
+      "step": 6456
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.255e-07,
+      "loss": 0.0073,
+      "step": 6457
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.2548837209302324e-07,
+      "loss": 0.0015,
+      "step": 6458
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.254767441860465e-07,
+      "loss": 0.0033,
+      "step": 6459
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2546511627906973e-07,
+      "loss": 0.0017,
+      "step": 6460
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2545348837209302e-07,
+      "loss": 0.0026,
+      "step": 6461
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2544186046511628e-07,
+      "loss": 0.0061,
+      "step": 6462
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2543023255813952e-07,
+      "loss": 0.0032,
+      "step": 6463
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2541860465116278e-07,
+      "loss": 0.0033,
+      "step": 6464
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2540697674418604e-07,
+      "loss": 0.0017,
+      "step": 6465
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.253953488372093e-07,
+      "loss": 0.0024,
+      "step": 6466
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.2538372093023254e-07,
+      "loss": 1.7525,
+      "step": 6467
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.253720930232558e-07,
+      "loss": 0.0027,
+      "step": 6468
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.2536046511627906e-07,
+      "loss": 0.0023,
+      "step": 6469
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.2534883720930232e-07,
+      "loss": 0.002,
+      "step": 6470
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.2533720930232556e-07,
+      "loss": 0.0025,
+      "step": 6471
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.2532558139534882e-07,
+      "loss": 0.0056,
+      "step": 6472
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.253139534883721e-07,
+      "loss": 0.0024,
+      "step": 6473
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.2530232558139534e-07,
+      "loss": 0.0032,
+      "step": 6474
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.252906976744186e-07,
+      "loss": 0.0019,
+      "step": 6475
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.2527906976744184e-07,
+      "loss": 0.0023,
+      "step": 6476
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.252674418604651e-07,
+      "loss": 0.0039,
+      "step": 6477
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.2525581395348836e-07,
+      "loss": 0.0052,
+      "step": 6478
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.2524418604651163e-07,
+      "loss": 0.0048,
+      "step": 6479
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.2523255813953486e-07,
+      "loss": 0.0022,
+      "step": 6480
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.2522093023255812e-07,
+      "loss": 0.0029,
+      "step": 6481
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.2520930232558138e-07,
+      "loss": 0.0027,
+      "step": 6482
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.2519767441860465e-07,
+      "loss": 0.0017,
+      "step": 6483
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.2518604651162788e-07,
+      "loss": 0.003,
+      "step": 6484
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.2517441860465114e-07,
+      "loss": 0.0028,
+      "step": 6485
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.2516279069767443e-07,
+      "loss": 0.0026,
+      "step": 6486
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.2515116279069767e-07,
+      "loss": 0.0104,
+      "step": 6487
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.2513953488372093e-07,
+      "loss": 0.0031,
+      "step": 6488
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.2512790697674416e-07,
+      "loss": 0.002,
+      "step": 6489
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.2511627906976745e-07,
+      "loss": 0.3443,
+      "step": 6490
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.251046511627907e-07,
+      "loss": 0.0028,
+      "step": 6491
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.2509302325581395e-07,
+      "loss": 1.9566,
+      "step": 6492
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.2508139534883718e-07,
+      "loss": 0.0023,
+      "step": 6493
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.2506976744186047e-07,
+      "loss": 0.0028,
+      "step": 6494
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.250581395348837e-07,
+      "loss": 0.0024,
+      "step": 6495
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.2504651162790697e-07,
+      "loss": 0.0049,
+      "step": 6496
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.250348837209302e-07,
+      "loss": 0.0032,
+      "step": 6497
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.250232558139535e-07,
+      "loss": 0.0023,
+      "step": 6498
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.2501162790697675e-07,
+      "loss": 0.0293,
+      "step": 6499
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.25e-07,
+      "loss": 0.0021,
+      "step": 6500
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.2498837209302323e-07,
+      "loss": 0.0027,
+      "step": 6501
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.249767441860465e-07,
+      "loss": 0.0021,
+      "step": 6502
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.2496511627906977e-07,
+      "loss": 0.0016,
+      "step": 6503
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.24953488372093e-07,
+      "loss": 0.0022,
+      "step": 6504
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.2494186046511627e-07,
+      "loss": 0.0042,
+      "step": 6505
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.249302325581395e-07,
+      "loss": 0.0032,
+      "step": 6506
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.249186046511628e-07,
+      "loss": 0.0018,
+      "step": 6507
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.2490697674418603e-07,
+      "loss": 0.0027,
+      "step": 6508
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.248953488372093e-07,
+      "loss": 0.0025,
+      "step": 6509
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.2488372093023253e-07,
+      "loss": 0.0018,
+      "step": 6510
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.2487209302325582e-07,
+      "loss": 0.0036,
+      "step": 6511
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.2486046511627905e-07,
+      "loss": 0.0019,
+      "step": 6512
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.248488372093023e-07,
+      "loss": 0.002,
+      "step": 6513
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.2483720930232555e-07,
+      "loss": 0.0022,
+      "step": 6514
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.2482558139534884e-07,
+      "loss": 0.0033,
+      "step": 6515
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.248139534883721e-07,
+      "loss": 0.002,
+      "step": 6516
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.2480232558139533e-07,
+      "loss": 0.0028,
+      "step": 6517
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.247906976744186e-07,
+      "loss": 0.0021,
+      "step": 6518
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.2477906976744186e-07,
+      "loss": 0.0154,
+      "step": 6519
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.2476744186046512e-07,
+      "loss": 0.0027,
+      "step": 6520
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.2475581395348835e-07,
+      "loss": 0.0016,
+      "step": 6521
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.2474418604651162e-07,
+      "loss": 0.0023,
+      "step": 6522
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.2473255813953488e-07,
+      "loss": 0.0032,
+      "step": 6523
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.2472093023255814e-07,
+      "loss": 0.0021,
+      "step": 6524
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.2470930232558137e-07,
+      "loss": 0.0023,
+      "step": 6525
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.2469767441860464e-07,
+      "loss": 0.0047,
+      "step": 6526
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.2468604651162787e-07,
+      "loss": 0.0014,
+      "step": 6527
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.2467441860465116e-07,
+      "loss": 0.0014,
+      "step": 6528
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.2466279069767442e-07,
+      "loss": 0.0032,
+      "step": 6529
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.2465116279069766e-07,
+      "loss": 0.0027,
+      "step": 6530
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.2463953488372092e-07,
+      "loss": 0.0027,
+      "step": 6531
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.2462790697674418e-07,
+      "loss": 0.002,
+      "step": 6532
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.2461627906976744e-07,
+      "loss": 0.0026,
+      "step": 6533
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.2460465116279068e-07,
+      "loss": 0.0036,
+      "step": 6534
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.2459302325581394e-07,
+      "loss": 0.0023,
+      "step": 6535
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.245813953488372e-07,
+      "loss": 0.0017,
+      "step": 6536
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.2456976744186046e-07,
+      "loss": 0.002,
+      "step": 6537
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.245581395348837e-07,
+      "loss": 0.0023,
+      "step": 6538
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.2454651162790696e-07,
+      "loss": 0.0072,
+      "step": 6539
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.2453488372093025e-07,
+      "loss": 0.003,
+      "step": 6540
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.2452325581395348e-07,
+      "loss": 0.0018,
+      "step": 6541
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.2451162790697674e-07,
+      "loss": 0.0057,
+      "step": 6542
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.2449999999999998e-07,
+      "loss": 0.003,
+      "step": 6543
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.2448837209302327e-07,
+      "loss": 0.0032,
+      "step": 6544
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.244767441860465e-07,
+      "loss": 0.9466,
+      "step": 6545
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.2446511627906976e-07,
+      "loss": 0.0034,
+      "step": 6546
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.24453488372093e-07,
+      "loss": 0.0024,
+      "step": 6547
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.244418604651163e-07,
+      "loss": 0.002,
+      "step": 6548
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.2443023255813952e-07,
+      "loss": 0.0015,
+      "step": 6549
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.2441860465116278e-07,
+      "loss": 0.0022,
+      "step": 6550
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.2440697674418602e-07,
+      "loss": 0.0036,
+      "step": 6551
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.243953488372093e-07,
+      "loss": 0.0054,
+      "step": 6552
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.2438372093023257e-07,
+      "loss": 0.004,
+      "step": 6553
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.243720930232558e-07,
+      "loss": 0.002,
+      "step": 6554
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.2436046511627904e-07,
+      "loss": 0.0032,
+      "step": 6555
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.243488372093023e-07,
+      "loss": 0.0079,
+      "step": 6556
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.243372093023256e-07,
+      "loss": 0.0033,
+      "step": 6557
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.2432558139534882e-07,
+      "loss": 0.0032,
+      "step": 6558
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.2431395348837209e-07,
+      "loss": 0.0017,
+      "step": 6559
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.2430232558139532e-07,
+      "loss": 0.0021,
+      "step": 6560
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.242906976744186e-07,
+      "loss": 0.0056,
+      "step": 6561
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.2427906976744185e-07,
+      "loss": 0.0016,
+      "step": 6562
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.242674418604651e-07,
+      "loss": 0.0068,
+      "step": 6563
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.2425581395348834e-07,
+      "loss": 0.0018,
+      "step": 6564
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.2424418604651163e-07,
+      "loss": 0.0048,
+      "step": 6565
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.2423255813953487e-07,
+      "loss": 0.0016,
+      "step": 6566
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.2422093023255813e-07,
+      "loss": 0.0021,
+      "step": 6567
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.2420930232558136e-07,
+      "loss": 0.0019,
+      "step": 6568
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.2419767441860465e-07,
+      "loss": 1.3078,
+      "step": 6569
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.241860465116279e-07,
+      "loss": 0.003,
+      "step": 6570
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.2417441860465115e-07,
+      "loss": 0.0058,
+      "step": 6571
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.241627906976744e-07,
+      "loss": 0.002,
+      "step": 6572
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.2415116279069767e-07,
+      "loss": 0.0023,
+      "step": 6573
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.2413953488372093e-07,
+      "loss": 0.0025,
+      "step": 6574
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.2412790697674417e-07,
+      "loss": 0.0023,
+      "step": 6575
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.2411627906976743e-07,
+      "loss": 0.0023,
+      "step": 6576
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.241046511627907e-07,
+      "loss": 0.0372,
+      "step": 6577
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.2409302325581395e-07,
+      "loss": 1.3908,
+      "step": 6578
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.240813953488372e-07,
+      "loss": 0.0017,
+      "step": 6579
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.2406976744186045e-07,
+      "loss": 0.0018,
+      "step": 6580
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.2405813953488369e-07,
+      "loss": 0.0028,
+      "step": 6581
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.2404651162790697e-07,
+      "loss": 0.002,
+      "step": 6582
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.2403488372093024e-07,
+      "loss": 0.0019,
+      "step": 6583
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.2402325581395347e-07,
+      "loss": 0.0026,
+      "step": 6584
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.240116279069767e-07,
+      "loss": 0.0021,
+      "step": 6585
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.24e-07,
+      "loss": 0.002,
+      "step": 6586
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.2398837209302326e-07,
+      "loss": 0.002,
+      "step": 6587
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.239767441860465e-07,
+      "loss": 0.0026,
+      "step": 6588
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.2396511627906975e-07,
+      "loss": 1.7445,
+      "step": 6589
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.2395348837209301e-07,
+      "loss": 0.0036,
+      "step": 6590
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.2394186046511628e-07,
+      "loss": 0.1904,
+      "step": 6591
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.239302325581395e-07,
+      "loss": 0.0015,
+      "step": 6592
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.2391860465116277e-07,
+      "loss": 0.0026,
+      "step": 6593
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.2390697674418606e-07,
+      "loss": 0.0021,
+      "step": 6594
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.238953488372093e-07,
+      "loss": 0.0018,
+      "step": 6595
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.2388372093023253e-07,
+      "loss": 0.0021,
+      "step": 6596
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.238720930232558e-07,
+      "loss": 0.0025,
+      "step": 6597
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.2386046511627908e-07,
+      "loss": 0.003,
+      "step": 6598
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.2384883720930232e-07,
+      "loss": 0.0018,
+      "step": 6599
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.2383720930232558e-07,
+      "loss": 0.002,
+      "step": 6600
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.2382558139534881e-07,
+      "loss": 0.0025,
+      "step": 6601
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.238139534883721e-07,
+      "loss": 0.0025,
+      "step": 6602
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.2380232558139534e-07,
+      "loss": 0.0023,
+      "step": 6603
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.237906976744186e-07,
+      "loss": 1.1216,
+      "step": 6604
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.2377906976744183e-07,
+      "loss": 0.002,
+      "step": 6605
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.2376744186046512e-07,
+      "loss": 0.0031,
+      "step": 6606
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.2375581395348836e-07,
+      "loss": 0.0022,
+      "step": 6607
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.2374418604651162e-07,
+      "loss": 0.0037,
+      "step": 6608
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.2373255813953485e-07,
+      "loss": 0.002,
+      "step": 6609
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.2372093023255812e-07,
+      "loss": 0.0015,
+      "step": 6610
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.237093023255814e-07,
+      "loss": 0.0021,
+      "step": 6611
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.2369767441860464e-07,
+      "loss": 0.1099,
+      "step": 6612
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.236860465116279e-07,
+      "loss": 0.0022,
+      "step": 6613
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.2367441860465114e-07,
+      "loss": 0.0019,
+      "step": 6614
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.2366279069767442e-07,
+      "loss": 0.0024,
+      "step": 6615
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.2365116279069766e-07,
+      "loss": 0.0024,
+      "step": 6616
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.2363953488372092e-07,
+      "loss": 0.002,
+      "step": 6617
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.2362790697674416e-07,
+      "loss": 0.0027,
+      "step": 6618
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.2361627906976745e-07,
+      "loss": 0.0022,
+      "step": 6619
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.2360465116279068e-07,
+      "loss": 0.0027,
+      "step": 6620
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.2359302325581394e-07,
+      "loss": 0.0019,
+      "step": 6621
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.2358139534883718e-07,
+      "loss": 0.0025,
+      "step": 6622
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.2356976744186047e-07,
+      "loss": 0.002,
+      "step": 6623
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.2355813953488373e-07,
+      "loss": 0.0036,
+      "step": 6624
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.2354651162790696e-07,
+      "loss": 0.0025,
+      "step": 6625
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.2353488372093022e-07,
+      "loss": 0.0031,
+      "step": 6626
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.2352325581395349e-07,
+      "loss": 0.0013,
+      "step": 6627
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.2351162790697675e-07,
+      "loss": 0.0018,
+      "step": 6628
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.2349999999999998e-07,
+      "loss": 0.0026,
+      "step": 6629
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.2348837209302324e-07,
+      "loss": 0.0017,
+      "step": 6630
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.234767441860465e-07,
+      "loss": 0.0027,
+      "step": 6631
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.2346511627906977e-07,
+      "loss": 0.0026,
+      "step": 6632
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.23453488372093e-07,
+      "loss": 0.0021,
+      "step": 6633
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.2344186046511627e-07,
+      "loss": 0.0024,
+      "step": 6634
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.234302325581395e-07,
+      "loss": 0.0023,
+      "step": 6635
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.234186046511628e-07,
+      "loss": 0.0022,
+      "step": 6636
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.2340697674418605e-07,
+      "loss": 0.3015,
+      "step": 6637
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.2339534883720929e-07,
+      "loss": 0.1343,
+      "step": 6638
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.2338372093023252e-07,
+      "loss": 0.0018,
+      "step": 6639
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.233720930232558e-07,
+      "loss": 0.0031,
+      "step": 6640
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2336046511627907e-07,
+      "loss": 0.002,
+      "step": 6641
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.233488372093023e-07,
+      "loss": 0.0026,
+      "step": 6642
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2333720930232557e-07,
+      "loss": 0.0028,
+      "step": 6643
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2332558139534883e-07,
+      "loss": 0.0018,
+      "step": 6644
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.233139534883721e-07,
+      "loss": 0.2592,
+      "step": 6645
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2330232558139533e-07,
+      "loss": 0.0026,
+      "step": 6646
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.232906976744186e-07,
+      "loss": 0.0029,
+      "step": 6647
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.2327906976744188e-07,
+      "loss": 0.0064,
+      "step": 6648
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.232674418604651e-07,
+      "loss": 0.0024,
+      "step": 6649
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.2325581395348835e-07,
+      "loss": 0.0035,
+      "step": 6650
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.232441860465116e-07,
+      "loss": 0.0029,
+      "step": 6651
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.232325581395349e-07,
+      "loss": 0.0024,
+      "step": 6652
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.2322093023255813e-07,
+      "loss": 0.0017,
+      "step": 6653
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.232093023255814e-07,
+      "loss": 0.0022,
+      "step": 6654
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.2319767441860463e-07,
+      "loss": 0.048,
+      "step": 6655
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.2318604651162792e-07,
+      "loss": 0.0033,
+      "step": 6656
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.2317441860465115e-07,
+      "loss": 0.0025,
+      "step": 6657
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.2316279069767441e-07,
+      "loss": 0.0021,
+      "step": 6658
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.2315116279069765e-07,
+      "loss": 0.0019,
+      "step": 6659
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.231395348837209e-07,
+      "loss": 0.0047,
+      "step": 6660
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.2312790697674417e-07,
+      "loss": 0.003,
+      "step": 6661
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2311627906976743e-07,
+      "loss": 0.007,
+      "step": 6662
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2310465116279067e-07,
+      "loss": 0.0017,
+      "step": 6663
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2309302325581393e-07,
+      "loss": 0.3606,
+      "step": 6664
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2308139534883722e-07,
+      "loss": 0.0031,
+      "step": 6665
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2306976744186045e-07,
+      "loss": 0.0024,
+      "step": 6666
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2305813953488372e-07,
+      "loss": 0.002,
+      "step": 6667
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2304651162790695e-07,
+      "loss": 0.0024,
+      "step": 6668
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2303488372093024e-07,
+      "loss": 0.0048,
+      "step": 6669
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2302325581395347e-07,
+      "loss": 0.002,
+      "step": 6670
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2301162790697674e-07,
+      "loss": 0.002,
+      "step": 6671
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2299999999999997e-07,
+      "loss": 0.005,
+      "step": 6672
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2298837209302326e-07,
+      "loss": 0.002,
+      "step": 6673
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.229767441860465e-07,
+      "loss": 0.0024,
+      "step": 6674
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2296511627906976e-07,
+      "loss": 0.018,
+      "step": 6675
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.22953488372093e-07,
+      "loss": 0.0015,
+      "step": 6676
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2294186046511628e-07,
+      "loss": 0.0015,
+      "step": 6677
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.2293023255813954e-07,
+      "loss": 0.0022,
+      "step": 6678
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.2291860465116278e-07,
+      "loss": 0.0018,
+      "step": 6679
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.2290697674418604e-07,
+      "loss": 0.0032,
+      "step": 6680
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.228953488372093e-07,
+      "loss": 0.0086,
+      "step": 6681
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.2288372093023256e-07,
+      "loss": 0.0018,
+      "step": 6682
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.228720930232558e-07,
+      "loss": 0.0025,
+      "step": 6683
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2286046511627906e-07,
+      "loss": 0.002,
+      "step": 6684
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2284883720930232e-07,
+      "loss": 0.0763,
+      "step": 6685
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2283720930232558e-07,
+      "loss": 0.3695,
+      "step": 6686
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2282558139534882e-07,
+      "loss": 0.0033,
+      "step": 6687
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.2281395348837208e-07,
+      "loss": 1.1473,
+      "step": 6688
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.2280232558139532e-07,
+      "loss": 0.0025,
+      "step": 6689
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.227906976744186e-07,
+      "loss": 0.0026,
+      "step": 6690
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.2277906976744184e-07,
+      "loss": 0.0016,
+      "step": 6691
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.227674418604651e-07,
+      "loss": 1.4206,
+      "step": 6692
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2275581395348834e-07,
+      "loss": 0.002,
+      "step": 6693
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2274418604651162e-07,
+      "loss": 0.0019,
+      "step": 6694
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2273255813953489e-07,
+      "loss": 0.2526,
+      "step": 6695
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2272093023255812e-07,
+      "loss": 0.0033,
+      "step": 6696
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2270930232558138e-07,
+      "loss": 0.0022,
+      "step": 6697
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2269767441860464e-07,
+      "loss": 0.81,
+      "step": 6698
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.226860465116279e-07,
+      "loss": 0.0019,
+      "step": 6699
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2267441860465114e-07,
+      "loss": 0.0017,
+      "step": 6700
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.226627906976744e-07,
+      "loss": 0.0034,
+      "step": 6701
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2265116279069766e-07,
+      "loss": 0.006,
+      "step": 6702
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2263953488372093e-07,
+      "loss": 0.0127,
+      "step": 6703
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2262790697674416e-07,
+      "loss": 0.0022,
+      "step": 6704
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2261627906976742e-07,
+      "loss": 0.0015,
+      "step": 6705
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.226046511627907e-07,
+      "loss": 0.0036,
+      "step": 6706
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2259302325581395e-07,
+      "loss": 0.0031,
+      "step": 6707
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.225813953488372e-07,
+      "loss": 0.025,
+      "step": 6708
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2256976744186044e-07,
+      "loss": 0.0019,
+      "step": 6709
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2255813953488373e-07,
+      "loss": 0.0025,
+      "step": 6710
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2254651162790697e-07,
+      "loss": 0.0033,
+      "step": 6711
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2253488372093023e-07,
+      "loss": 0.0184,
+      "step": 6712
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2252325581395346e-07,
+      "loss": 0.0024,
+      "step": 6713
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2251162790697673e-07,
+      "loss": 0.0024,
+      "step": 6714
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.225e-07,
+      "loss": 0.002,
+      "step": 6715
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2248837209302325e-07,
+      "loss": 0.0029,
+      "step": 6716
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2247674418604648e-07,
+      "loss": 0.0015,
+      "step": 6717
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2246511627906975e-07,
+      "loss": 1.9261,
+      "step": 6718
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2245348837209303e-07,
+      "loss": 0.0022,
+      "step": 6719
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2244186046511627e-07,
+      "loss": 0.0023,
+      "step": 6720
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2243023255813953e-07,
+      "loss": 0.0019,
+      "step": 6721
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9835589941972921,
+      "eval_loss": 0.15013490617275238,
+      "eval_runtime": 1.5501,
+      "eval_samples_per_second": 667.059,
+      "eval_steps_per_second": 10.967,
+      "step": 6721
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2241860465116277e-07,
+      "loss": 0.0024,
+      "step": 6722
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2240697674418605e-07,
+      "loss": 0.0027,
+      "step": 6723
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.223953488372093e-07,
+      "loss": 0.0053,
+      "step": 6724
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2238372093023255e-07,
+      "loss": 0.0021,
+      "step": 6725
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2237209302325579e-07,
+      "loss": 0.0024,
+      "step": 6726
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2236046511627907e-07,
+      "loss": 0.0023,
+      "step": 6727
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.223488372093023e-07,
+      "loss": 0.4223,
+      "step": 6728
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.2233720930232557e-07,
+      "loss": 0.0018,
+      "step": 6729
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.223255813953488e-07,
+      "loss": 0.0029,
+      "step": 6730
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.223139534883721e-07,
+      "loss": 0.0017,
+      "step": 6731
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.2230232558139536e-07,
+      "loss": 0.0038,
+      "step": 6732
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.222906976744186e-07,
+      "loss": 0.0671,
+      "step": 6733
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2227906976744183e-07,
+      "loss": 0.0022,
+      "step": 6734
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2226744186046512e-07,
+      "loss": 0.0017,
+      "step": 6735
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2225581395348838e-07,
+      "loss": 0.0026,
+      "step": 6736
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.222441860465116e-07,
+      "loss": 0.0154,
+      "step": 6737
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2223255813953487e-07,
+      "loss": 0.0022,
+      "step": 6738
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2222093023255814e-07,
+      "loss": 0.0034,
+      "step": 6739
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.222093023255814e-07,
+      "loss": 0.0029,
+      "step": 6740
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2219767441860463e-07,
+      "loss": 0.0018,
+      "step": 6741
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.221860465116279e-07,
+      "loss": 0.0021,
+      "step": 6742
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2217441860465113e-07,
+      "loss": 0.0023,
+      "step": 6743
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2216279069767442e-07,
+      "loss": 0.0017,
+      "step": 6744
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2215116279069765e-07,
+      "loss": 0.0023,
+      "step": 6745
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2213953488372092e-07,
+      "loss": 0.0013,
+      "step": 6746
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2212790697674415e-07,
+      "loss": 0.0052,
+      "step": 6747
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2211627906976744e-07,
+      "loss": 0.0024,
+      "step": 6748
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.221046511627907e-07,
+      "loss": 0.0029,
+      "step": 6749
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2209302325581394e-07,
+      "loss": 1.9282,
+      "step": 6750
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.220813953488372e-07,
+      "loss": 0.2136,
+      "step": 6751
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2206976744186046e-07,
+      "loss": 0.0018,
+      "step": 6752
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2205813953488372e-07,
+      "loss": 0.2807,
+      "step": 6753
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2204651162790696e-07,
+      "loss": 0.0965,
+      "step": 6754
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2203488372093022e-07,
+      "loss": 0.0034,
+      "step": 6755
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2202325581395348e-07,
+      "loss": 0.0017,
+      "step": 6756
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2201162790697674e-07,
+      "loss": 0.0017,
+      "step": 6757
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2199999999999998e-07,
+      "loss": 0.0039,
+      "step": 6758
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2198837209302324e-07,
+      "loss": 0.0022,
+      "step": 6759
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2197674418604653e-07,
+      "loss": 0.0027,
+      "step": 6760
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2196511627906976e-07,
+      "loss": 0.0021,
+      "step": 6761
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2195348837209302e-07,
+      "loss": 0.0035,
+      "step": 6762
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2194186046511626e-07,
+      "loss": 0.0015,
+      "step": 6763
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2193023255813955e-07,
+      "loss": 0.0018,
+      "step": 6764
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2191860465116278e-07,
+      "loss": 0.0027,
+      "step": 6765
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2190697674418604e-07,
+      "loss": 0.0019,
+      "step": 6766
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2189534883720928e-07,
+      "loss": 0.003,
+      "step": 6767
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2188372093023254e-07,
+      "loss": 0.0018,
+      "step": 6768
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.218720930232558e-07,
+      "loss": 0.0022,
+      "step": 6769
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2186046511627906e-07,
+      "loss": 0.003,
+      "step": 6770
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.218488372093023e-07,
+      "loss": 0.0014,
+      "step": 6771
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2183720930232556e-07,
+      "loss": 0.0022,
+      "step": 6772
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2182558139534885e-07,
+      "loss": 0.0024,
+      "step": 6773
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2181395348837208e-07,
+      "loss": 0.0025,
+      "step": 6774
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2180232558139535e-07,
+      "loss": 0.0014,
+      "step": 6775
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2179069767441858e-07,
+      "loss": 0.4889,
+      "step": 6776
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2177906976744187e-07,
+      "loss": 0.0037,
+      "step": 6777
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.217674418604651e-07,
+      "loss": 0.0042,
+      "step": 6778
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2175581395348837e-07,
+      "loss": 0.003,
+      "step": 6779
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.217441860465116e-07,
+      "loss": 0.0019,
+      "step": 6780
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.217325581395349e-07,
+      "loss": 0.0021,
+      "step": 6781
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.2172093023255812e-07,
+      "loss": 0.0018,
+      "step": 6782
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.2170930232558139e-07,
+      "loss": 0.0021,
+      "step": 6783
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.2169767441860462e-07,
+      "loss": 0.0014,
+      "step": 6784
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.216860465116279e-07,
+      "loss": 0.0016,
+      "step": 6785
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.2167441860465117e-07,
+      "loss": 0.0028,
+      "step": 6786
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.216627906976744e-07,
+      "loss": 0.0024,
+      "step": 6787
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.2165116279069764e-07,
+      "loss": 0.002,
+      "step": 6788
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.2163953488372093e-07,
+      "loss": 0.0022,
+      "step": 6789
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.216279069767442e-07,
+      "loss": 0.0024,
+      "step": 6790
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.2161627906976743e-07,
+      "loss": 0.0047,
+      "step": 6791
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.216046511627907e-07,
+      "loss": 0.0024,
+      "step": 6792
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.2159302325581395e-07,
+      "loss": 0.0022,
+      "step": 6793
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.215813953488372e-07,
+      "loss": 0.0016,
+      "step": 6794
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.2156976744186045e-07,
+      "loss": 1.9429,
+      "step": 6795
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.215581395348837e-07,
+      "loss": 0.0017,
+      "step": 6796
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2154651162790694e-07,
+      "loss": 0.0015,
+      "step": 6797
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2153488372093023e-07,
+      "loss": 0.0024,
+      "step": 6798
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2152325581395347e-07,
+      "loss": 0.0019,
+      "step": 6799
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2151162790697673e-07,
+      "loss": 0.0028,
+      "step": 6800
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2149999999999997e-07,
+      "loss": 0.0023,
+      "step": 6801
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2148837209302325e-07,
+      "loss": 0.0029,
+      "step": 6802
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2147674418604651e-07,
+      "loss": 0.0027,
+      "step": 6803
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2146511627906975e-07,
+      "loss": 0.0017,
+      "step": 6804
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.21453488372093e-07,
+      "loss": 0.0023,
+      "step": 6805
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2144186046511627e-07,
+      "loss": 0.0019,
+      "step": 6806
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2143023255813954e-07,
+      "loss": 0.0022,
+      "step": 6807
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2141860465116277e-07,
+      "loss": 0.0022,
+      "step": 6808
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2140697674418603e-07,
+      "loss": 0.003,
+      "step": 6809
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.213953488372093e-07,
+      "loss": 0.0021,
+      "step": 6810
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2138372093023256e-07,
+      "loss": 0.0078,
+      "step": 6811
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.213720930232558e-07,
+      "loss": 0.0018,
+      "step": 6812
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2136046511627905e-07,
+      "loss": 0.0035,
+      "step": 6813
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2134883720930234e-07,
+      "loss": 1.6292,
+      "step": 6814
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2133720930232558e-07,
+      "loss": 0.0026,
+      "step": 6815
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2132558139534884e-07,
+      "loss": 0.6023,
+      "step": 6816
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.2131395348837207e-07,
+      "loss": 0.0022,
+      "step": 6817
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.2130232558139536e-07,
+      "loss": 0.0025,
+      "step": 6818
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.212906976744186e-07,
+      "loss": 0.0169,
+      "step": 6819
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.2127906976744186e-07,
+      "loss": 0.0041,
+      "step": 6820
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.212674418604651e-07,
+      "loss": 0.0024,
+      "step": 6821
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.2125581395348836e-07,
+      "loss": 0.0019,
+      "step": 6822
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.2124418604651162e-07,
+      "loss": 0.0015,
+      "step": 6823
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.2123255813953488e-07,
+      "loss": 0.0023,
+      "step": 6824
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.2122093023255811e-07,
+      "loss": 0.0026,
+      "step": 6825
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.2120930232558138e-07,
+      "loss": 0.002,
+      "step": 6826
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2119767441860466e-07,
+      "loss": 0.0017,
+      "step": 6827
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.211860465116279e-07,
+      "loss": 0.0022,
+      "step": 6828
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2117441860465113e-07,
+      "loss": 0.0253,
+      "step": 6829
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.211627906976744e-07,
+      "loss": 0.0027,
+      "step": 6830
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2115116279069768e-07,
+      "loss": 0.0014,
+      "step": 6831
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2113953488372092e-07,
+      "loss": 0.0022,
+      "step": 6832
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.2112790697674418e-07,
+      "loss": 0.0048,
+      "step": 6833
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.2111627906976742e-07,
+      "loss": 0.002,
+      "step": 6834
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.211046511627907e-07,
+      "loss": 0.0019,
+      "step": 6835
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.2109302325581394e-07,
+      "loss": 0.0017,
+      "step": 6836
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.210813953488372e-07,
+      "loss": 0.0017,
+      "step": 6837
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2106976744186044e-07,
+      "loss": 0.002,
+      "step": 6838
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2105813953488372e-07,
+      "loss": 0.0019,
+      "step": 6839
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2104651162790696e-07,
+      "loss": 0.0748,
+      "step": 6840
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2103488372093022e-07,
+      "loss": 0.0022,
+      "step": 6841
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2102325581395346e-07,
+      "loss": 0.0018,
+      "step": 6842
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2101162790697675e-07,
+      "loss": 0.0024,
+      "step": 6843
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.21e-07,
+      "loss": 0.0015,
+      "step": 6844
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2098837209302324e-07,
+      "loss": 0.0061,
+      "step": 6845
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.209767441860465e-07,
+      "loss": 0.0833,
+      "step": 6846
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2096511627906974e-07,
+      "loss": 0.0023,
+      "step": 6847
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2095348837209303e-07,
+      "loss": 0.0036,
+      "step": 6848
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2094186046511626e-07,
+      "loss": 0.0028,
+      "step": 6849
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2093023255813952e-07,
+      "loss": 0.0018,
+      "step": 6850
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2091860465116276e-07,
+      "loss": 0.002,
+      "step": 6851
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2090697674418605e-07,
+      "loss": 0.0021,
+      "step": 6852
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2089534883720928e-07,
+      "loss": 0.0087,
+      "step": 6853
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2088372093023254e-07,
+      "loss": 0.0076,
+      "step": 6854
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2087209302325578e-07,
+      "loss": 0.9216,
+      "step": 6855
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2086046511627907e-07,
+      "loss": 0.0013,
+      "step": 6856
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2084883720930233e-07,
+      "loss": 0.0017,
+      "step": 6857
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2083720930232557e-07,
+      "loss": 0.0037,
+      "step": 6858
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.2082558139534883e-07,
+      "loss": 0.0019,
+      "step": 6859
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.208139534883721e-07,
+      "loss": 0.0018,
+      "step": 6860
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.2080232558139535e-07,
+      "loss": 0.0018,
+      "step": 6861
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.2079069767441859e-07,
+      "loss": 0.0021,
+      "step": 6862
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.2077906976744185e-07,
+      "loss": 0.0042,
+      "step": 6863
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.207674418604651e-07,
+      "loss": 0.0018,
+      "step": 6864
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2075581395348837e-07,
+      "loss": 0.002,
+      "step": 6865
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.207441860465116e-07,
+      "loss": 0.004,
+      "step": 6866
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2073255813953487e-07,
+      "loss": 0.0022,
+      "step": 6867
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2072093023255816e-07,
+      "loss": 0.0019,
+      "step": 6868
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.207093023255814e-07,
+      "loss": 0.0026,
+      "step": 6869
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.2069767441860465e-07,
+      "loss": 0.0031,
+      "step": 6870
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.206860465116279e-07,
+      "loss": 0.0024,
+      "step": 6871
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.2067441860465118e-07,
+      "loss": 0.0037,
+      "step": 6872
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.206627906976744e-07,
+      "loss": 0.0021,
+      "step": 6873
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.2065116279069767e-07,
+      "loss": 0.0022,
+      "step": 6874
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.206395348837209e-07,
+      "loss": 0.0031,
+      "step": 6875
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.2062790697674417e-07,
+      "loss": 0.0032,
+      "step": 6876
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.2061627906976743e-07,
+      "loss": 0.2025,
+      "step": 6877
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.206046511627907e-07,
+      "loss": 0.0015,
+      "step": 6878
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2059302325581393e-07,
+      "loss": 0.0022,
+      "step": 6879
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.205813953488372e-07,
+      "loss": 0.0015,
+      "step": 6880
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2056976744186048e-07,
+      "loss": 0.0017,
+      "step": 6881
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2055813953488371e-07,
+      "loss": 0.0027,
+      "step": 6882
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2054651162790695e-07,
+      "loss": 0.0022,
+      "step": 6883
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.205348837209302e-07,
+      "loss": 0.0023,
+      "step": 6884
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.205232558139535e-07,
+      "loss": 0.002,
+      "step": 6885
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.2051162790697673e-07,
+      "loss": 0.002,
+      "step": 6886
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.205e-07,
+      "loss": 0.0047,
+      "step": 6887
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.2048837209302323e-07,
+      "loss": 0.0025,
+      "step": 6888
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.2047674418604652e-07,
+      "loss": 0.0027,
+      "step": 6889
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2046511627906975e-07,
+      "loss": 0.0017,
+      "step": 6890
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2045348837209302e-07,
+      "loss": 0.039,
+      "step": 6891
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2044186046511625e-07,
+      "loss": 0.0018,
+      "step": 6892
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2043023255813954e-07,
+      "loss": 0.0019,
+      "step": 6893
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2041860465116277e-07,
+      "loss": 0.7573,
+      "step": 6894
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2040697674418604e-07,
+      "loss": 0.0016,
+      "step": 6895
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2039534883720927e-07,
+      "loss": 0.0019,
+      "step": 6896
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2038372093023256e-07,
+      "loss": 0.0016,
+      "step": 6897
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2037209302325582e-07,
+      "loss": 0.0018,
+      "step": 6898
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2036046511627906e-07,
+      "loss": 1.6319,
+      "step": 6899
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.2034883720930232e-07,
+      "loss": 0.0022,
+      "step": 6900
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.2033720930232555e-07,
+      "loss": 0.0023,
+      "step": 6901
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.2032558139534884e-07,
+      "loss": 0.0019,
+      "step": 6902
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.2031395348837208e-07,
+      "loss": 0.0018,
+      "step": 6903
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.2030232558139534e-07,
+      "loss": 0.0026,
+      "step": 6904
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2029069767441857e-07,
+      "loss": 0.002,
+      "step": 6905
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2027906976744186e-07,
+      "loss": 0.0015,
+      "step": 6906
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.202674418604651e-07,
+      "loss": 0.0019,
+      "step": 6907
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2025581395348836e-07,
+      "loss": 0.0026,
+      "step": 6908
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.202441860465116e-07,
+      "loss": 0.0018,
+      "step": 6909
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.2023255813953488e-07,
+      "loss": 0.0028,
+      "step": 6910
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.2022093023255814e-07,
+      "loss": 0.0045,
+      "step": 6911
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.2020930232558138e-07,
+      "loss": 0.0149,
+      "step": 6912
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.2019767441860464e-07,
+      "loss": 0.0034,
+      "step": 6913
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.201860465116279e-07,
+      "loss": 0.0026,
+      "step": 6914
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2017441860465116e-07,
+      "loss": 0.0023,
+      "step": 6915
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.201627906976744e-07,
+      "loss": 0.0032,
+      "step": 6916
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2015116279069766e-07,
+      "loss": 0.0023,
+      "step": 6917
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2013953488372092e-07,
+      "loss": 0.0034,
+      "step": 6918
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2012790697674419e-07,
+      "loss": 0.0512,
+      "step": 6919
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2011627906976742e-07,
+      "loss": 0.0026,
+      "step": 6920
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.2010465116279068e-07,
+      "loss": 0.34,
+      "step": 6921
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.2009302325581397e-07,
+      "loss": 0.0029,
+      "step": 6922
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.200813953488372e-07,
+      "loss": 0.002,
+      "step": 6923
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.2006976744186047e-07,
+      "loss": 0.5399,
+      "step": 6924
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.200581395348837e-07,
+      "loss": 2.0721,
+      "step": 6925
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.20046511627907e-07,
+      "loss": 0.0018,
+      "step": 6926
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.2003488372093023e-07,
+      "loss": 0.0019,
+      "step": 6927
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.200232558139535e-07,
+      "loss": 0.0037,
+      "step": 6928
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.2001162790697672e-07,
+      "loss": 0.0021,
+      "step": 6929
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.1999999999999998e-07,
+      "loss": 0.0014,
+      "step": 6930
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.1998837209302325e-07,
+      "loss": 0.0025,
+      "step": 6931
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.199767441860465e-07,
+      "loss": 0.0034,
+      "step": 6932
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.1996511627906974e-07,
+      "loss": 0.0016,
+      "step": 6933
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.19953488372093e-07,
+      "loss": 0.0018,
+      "step": 6934
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.199418604651163e-07,
+      "loss": 0.0062,
+      "step": 6935
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.1993023255813953e-07,
+      "loss": 1.5529,
+      "step": 6936
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.1991860465116276e-07,
+      "loss": 0.0022,
+      "step": 6937
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.1990697674418603e-07,
+      "loss": 0.0019,
+      "step": 6938
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.1989534883720931e-07,
+      "loss": 0.0019,
+      "step": 6939
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.1988372093023255e-07,
+      "loss": 0.0832,
+      "step": 6940
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.198720930232558e-07,
+      "loss": 0.0042,
+      "step": 6941
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.1986046511627905e-07,
+      "loss": 0.1179,
+      "step": 6942
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.1984883720930233e-07,
+      "loss": 0.0029,
+      "step": 6943
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.1983720930232557e-07,
+      "loss": 0.0019,
+      "step": 6944
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.1982558139534883e-07,
+      "loss": 0.0028,
+      "step": 6945
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.1981395348837207e-07,
+      "loss": 0.0016,
+      "step": 6946
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.1980232558139535e-07,
+      "loss": 0.0023,
+      "step": 6947
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.197906976744186e-07,
+      "loss": 0.1871,
+      "step": 6948
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.1977906976744185e-07,
+      "loss": 0.2486,
+      "step": 6949
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.1976744186046509e-07,
+      "loss": 0.0023,
+      "step": 6950
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.1975581395348837e-07,
+      "loss": 0.0026,
+      "step": 6951
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.1974418604651164e-07,
+      "loss": 0.3674,
+      "step": 6952
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.1973255813953487e-07,
+      "loss": 0.003,
+      "step": 6953
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.1972093023255813e-07,
+      "loss": 0.0017,
+      "step": 6954
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.1970930232558137e-07,
+      "loss": 0.0022,
+      "step": 6955
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.1969767441860466e-07,
+      "loss": 0.0024,
+      "step": 6956
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.196860465116279e-07,
+      "loss": 0.0016,
+      "step": 6957
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.1967441860465115e-07,
+      "loss": 0.5249,
+      "step": 6958
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.196627906976744e-07,
+      "loss": 0.0024,
+      "step": 6959
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.1965116279069768e-07,
+      "loss": 0.0026,
+      "step": 6960
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.196395348837209e-07,
+      "loss": 0.0023,
+      "step": 6961
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.1962790697674417e-07,
+      "loss": 0.0015,
+      "step": 6962
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.196162790697674e-07,
+      "loss": 0.0028,
+      "step": 6963
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.196046511627907e-07,
+      "loss": 0.0017,
+      "step": 6964
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.1959302325581396e-07,
+      "loss": 0.002,
+      "step": 6965
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.195813953488372e-07,
+      "loss": 0.0027,
+      "step": 6966
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.1956976744186043e-07,
+      "loss": 0.0015,
+      "step": 6967
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.1955813953488372e-07,
+      "loss": 0.0105,
+      "step": 6968
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.1954651162790698e-07,
+      "loss": 0.0053,
+      "step": 6969
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.1953488372093021e-07,
+      "loss": 0.0024,
+      "step": 6970
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.1952325581395348e-07,
+      "loss": 0.0018,
+      "step": 6971
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.1951162790697674e-07,
+      "loss": 0.0024,
+      "step": 6972
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.195e-07,
+      "loss": 0.0057,
+      "step": 6973
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.1948837209302324e-07,
+      "loss": 0.0049,
+      "step": 6974
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.194767441860465e-07,
+      "loss": 0.0057,
+      "step": 6975
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.1946511627906979e-07,
+      "loss": 0.0021,
+      "step": 6976
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.1945348837209302e-07,
+      "loss": 0.0031,
+      "step": 6977
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.1944186046511626e-07,
+      "loss": 0.0048,
+      "step": 6978
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.1943023255813952e-07,
+      "loss": 0.0022,
+      "step": 6979
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.1941860465116275e-07,
+      "loss": 0.0017,
+      "step": 6980
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.1940697674418604e-07,
+      "loss": 0.0027,
+      "step": 6981
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.193953488372093e-07,
+      "loss": 0.0015,
+      "step": 6982
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.1938372093023254e-07,
+      "loss": 0.0021,
+      "step": 6983
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.193720930232558e-07,
+      "loss": 0.0033,
+      "step": 6984
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.1936046511627906e-07,
+      "loss": 0.3769,
+      "step": 6985
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.1934883720930232e-07,
+      "loss": 0.0028,
+      "step": 6986
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.1933720930232556e-07,
+      "loss": 0.0032,
+      "step": 6987
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.1932558139534882e-07,
+      "loss": 0.0072,
+      "step": 6988
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.1931395348837208e-07,
+      "loss": 0.0023,
+      "step": 6989
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.1930232558139534e-07,
+      "loss": 0.0021,
+      "step": 6990
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.1929069767441858e-07,
+      "loss": 0.0017,
+      "step": 6991
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.1927906976744184e-07,
+      "loss": 0.0032,
+      "step": 6992
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.1926744186046513e-07,
+      "loss": 0.0018,
+      "step": 6993
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.1925581395348836e-07,
+      "loss": 0.0315,
+      "step": 6994
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.1924418604651163e-07,
+      "loss": 0.0016,
+      "step": 6995
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.1923255813953486e-07,
+      "loss": 0.0023,
+      "step": 6996
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.1922093023255815e-07,
+      "loss": 0.0034,
+      "step": 6997
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.1920930232558138e-07,
+      "loss": 0.0016,
+      "step": 6998
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.1919767441860465e-07,
+      "loss": 0.0017,
+      "step": 6999
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.1918604651162788e-07,
+      "loss": 0.0022,
+      "step": 7000
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.1917441860465117e-07,
+      "loss": 0.0015,
+      "step": 7001
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.191627906976744e-07,
+      "loss": 0.0019,
+      "step": 7002
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.1915116279069767e-07,
+      "loss": 0.0027,
+      "step": 7003
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.191395348837209e-07,
+      "loss": 0.0026,
+      "step": 7004
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.191279069767442e-07,
+      "loss": 0.0019,
+      "step": 7005
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.1911627906976745e-07,
+      "loss": 0.0016,
+      "step": 7006
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.1910465116279069e-07,
+      "loss": 0.0018,
+      "step": 7007
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.1909302325581395e-07,
+      "loss": 0.003,
+      "step": 7008
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.1908139534883718e-07,
+      "loss": 0.0024,
+      "step": 7009
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.1906976744186047e-07,
+      "loss": 0.1596,
+      "step": 7010
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.190581395348837e-07,
+      "loss": 0.0032,
+      "step": 7011
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.1904651162790697e-07,
+      "loss": 0.0027,
+      "step": 7012
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.190348837209302e-07,
+      "loss": 0.0016,
+      "step": 7013
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.190232558139535e-07,
+      "loss": 0.0022,
+      "step": 7014
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.1901162790697673e-07,
+      "loss": 0.0019,
+      "step": 7015
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.19e-07,
+      "loss": 0.0036,
+      "step": 7016
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.1898837209302322e-07,
+      "loss": 0.002,
+      "step": 7017
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.189767441860465e-07,
+      "loss": 0.002,
+      "step": 7018
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.1896511627906977e-07,
+      "loss": 0.9754,
+      "step": 7019
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.18953488372093e-07,
+      "loss": 0.0028,
+      "step": 7020
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.1894186046511624e-07,
+      "loss": 0.0029,
+      "step": 7021
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.1893023255813953e-07,
+      "loss": 0.0016,
+      "step": 7022
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.189186046511628e-07,
+      "loss": 0.004,
+      "step": 7023
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1890697674418603e-07,
+      "loss": 0.0037,
+      "step": 7024
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.188953488372093e-07,
+      "loss": 0.0027,
+      "step": 7025
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1888372093023255e-07,
+      "loss": 0.0026,
+      "step": 7026
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1887209302325581e-07,
+      "loss": 0.0019,
+      "step": 7027
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1886046511627905e-07,
+      "loss": 0.0016,
+      "step": 7028
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.188488372093023e-07,
+      "loss": 0.0709,
+      "step": 7029
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.188372093023256e-07,
+      "loss": 0.0024,
+      "step": 7030
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.1882558139534884e-07,
+      "loss": 0.0026,
+      "step": 7031
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.1881395348837207e-07,
+      "loss": 0.0037,
+      "step": 7032
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.1880232558139533e-07,
+      "loss": 0.0023,
+      "step": 7033
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1879069767441857e-07,
+      "loss": 0.0025,
+      "step": 7034
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1877906976744186e-07,
+      "loss": 0.0017,
+      "step": 7035
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1876744186046512e-07,
+      "loss": 0.0023,
+      "step": 7036
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1875581395348835e-07,
+      "loss": 0.0014,
+      "step": 7037
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1874418604651161e-07,
+      "loss": 0.003,
+      "step": 7038
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1873255813953488e-07,
+      "loss": 0.0031,
+      "step": 7039
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1872093023255814e-07,
+      "loss": 0.2912,
+      "step": 7040
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1870930232558137e-07,
+      "loss": 0.0016,
+      "step": 7041
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1869767441860463e-07,
+      "loss": 0.002,
+      "step": 7042
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.186860465116279e-07,
+      "loss": 0.003,
+      "step": 7043
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1867441860465116e-07,
+      "loss": 0.0017,
+      "step": 7044
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.186627906976744e-07,
+      "loss": 0.0049,
+      "step": 7045
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.1865116279069766e-07,
+      "loss": 0.0016,
+      "step": 7046
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.1863953488372094e-07,
+      "loss": 0.7417,
+      "step": 7047
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.1862790697674418e-07,
+      "loss": 0.0019,
+      "step": 7048
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.1861627906976744e-07,
+      "loss": 0.0018,
+      "step": 7049
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.1860465116279068e-07,
+      "loss": 0.0017,
+      "step": 7050
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.1859302325581396e-07,
+      "loss": 0.004,
+      "step": 7051
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.185813953488372e-07,
+      "loss": 0.0023,
+      "step": 7052
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.1856976744186046e-07,
+      "loss": 0.0049,
+      "step": 7053
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.185581395348837e-07,
+      "loss": 0.0017,
+      "step": 7054
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1854651162790698e-07,
+      "loss": 0.0016,
+      "step": 7055
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1853488372093022e-07,
+      "loss": 0.0015,
+      "step": 7056
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1852325581395348e-07,
+      "loss": 0.0018,
+      "step": 7057
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1851162790697672e-07,
+      "loss": 0.0017,
+      "step": 7058
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.185e-07,
+      "loss": 0.0037,
+      "step": 7059
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.1848837209302327e-07,
+      "loss": 0.0021,
+      "step": 7060
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.184767441860465e-07,
+      "loss": 0.0017,
+      "step": 7061
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.1846511627906976e-07,
+      "loss": 0.0013,
+      "step": 7062
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.18453488372093e-07,
+      "loss": 0.9662,
+      "step": 7063
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.1844186046511629e-07,
+      "loss": 0.0021,
+      "step": 7064
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1843023255813952e-07,
+      "loss": 0.0021,
+      "step": 7065
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1841860465116278e-07,
+      "loss": 0.0023,
+      "step": 7066
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1840697674418602e-07,
+      "loss": 0.002,
+      "step": 7067
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.183953488372093e-07,
+      "loss": 0.0853,
+      "step": 7068
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1838372093023254e-07,
+      "loss": 0.0014,
+      "step": 7069
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.183720930232558e-07,
+      "loss": 0.0017,
+      "step": 7070
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1836046511627904e-07,
+      "loss": 0.0019,
+      "step": 7071
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1834883720930233e-07,
+      "loss": 0.0107,
+      "step": 7072
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1833720930232556e-07,
+      "loss": 0.0028,
+      "step": 7073
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1832558139534882e-07,
+      "loss": 0.0159,
+      "step": 7074
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1831395348837206e-07,
+      "loss": 0.002,
+      "step": 7075
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.1830232558139535e-07,
+      "loss": 0.4722,
+      "step": 7076
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.182906976744186e-07,
+      "loss": 0.0026,
+      "step": 7077
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.1827906976744184e-07,
+      "loss": 0.0029,
+      "step": 7078
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.182674418604651e-07,
+      "loss": 0.1774,
+      "step": 7079
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.1825581395348837e-07,
+      "loss": 0.0024,
+      "step": 7080
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.1824418604651163e-07,
+      "loss": 0.0023,
+      "step": 7081
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.1823255813953486e-07,
+      "loss": 0.0015,
+      "step": 7082
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.1822093023255813e-07,
+      "loss": 0.0029,
+      "step": 7083
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.182093023255814e-07,
+      "loss": 0.0022,
+      "step": 7084
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.1819767441860465e-07,
+      "loss": 0.0018,
+      "step": 7085
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1818604651162789e-07,
+      "loss": 0.0142,
+      "step": 7086
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1817441860465115e-07,
+      "loss": 0.0016,
+      "step": 7087
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1816279069767438e-07,
+      "loss": 0.0018,
+      "step": 7088
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1815116279069767e-07,
+      "loss": 0.0017,
+      "step": 7089
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1813953488372093e-07,
+      "loss": 0.002,
+      "step": 7090
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.1812790697674417e-07,
+      "loss": 0.0031,
+      "step": 7091
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.1811627906976743e-07,
+      "loss": 0.0018,
+      "step": 7092
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.181046511627907e-07,
+      "loss": 0.0022,
+      "step": 7093
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.1809302325581395e-07,
+      "loss": 2.149,
+      "step": 7094
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.180813953488372e-07,
+      "loss": 0.0022,
+      "step": 7095
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.1806976744186045e-07,
+      "loss": 0.0032,
+      "step": 7096
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.180581395348837e-07,
+      "loss": 0.7475,
+      "step": 7097
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.1804651162790697e-07,
+      "loss": 0.0015,
+      "step": 7098
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.180348837209302e-07,
+      "loss": 0.0021,
+      "step": 7099
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.1802325581395347e-07,
+      "loss": 0.0023,
+      "step": 7100
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1801162790697676e-07,
+      "loss": 0.0027,
+      "step": 7101
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.18e-07,
+      "loss": 0.0025,
+      "step": 7102
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1798837209302326e-07,
+      "loss": 0.0026,
+      "step": 7103
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.179767441860465e-07,
+      "loss": 0.0012,
+      "step": 7104
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1796511627906978e-07,
+      "loss": 0.002,
+      "step": 7105
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1795348837209301e-07,
+      "loss": 0.0016,
+      "step": 7106
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.1794186046511628e-07,
+      "loss": 0.0024,
+      "step": 7107
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.179302325581395e-07,
+      "loss": 0.0015,
+      "step": 7108
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.179186046511628e-07,
+      "loss": 0.0039,
+      "step": 7109
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.1790697674418603e-07,
+      "loss": 0.0023,
+      "step": 7110
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.178953488372093e-07,
+      "loss": 0.0033,
+      "step": 7111
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1788372093023253e-07,
+      "loss": 0.0029,
+      "step": 7112
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1787209302325582e-07,
+      "loss": 0.002,
+      "step": 7113
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1786046511627908e-07,
+      "loss": 0.0022,
+      "step": 7114
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1784883720930232e-07,
+      "loss": 0.0023,
+      "step": 7115
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1783720930232555e-07,
+      "loss": 0.002,
+      "step": 7116
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.178255813953488e-07,
+      "loss": 0.002,
+      "step": 7117
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.178139534883721e-07,
+      "loss": 0.0016,
+      "step": 7118
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.1780232558139534e-07,
+      "loss": 0.002,
+      "step": 7119
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.177906976744186e-07,
+      "loss": 0.0016,
+      "step": 7120
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.1777906976744183e-07,
+      "loss": 0.0017,
+      "step": 7121
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1776744186046512e-07,
+      "loss": 0.0014,
+      "step": 7122
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1775581395348836e-07,
+      "loss": 0.0023,
+      "step": 7123
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1774418604651162e-07,
+      "loss": 0.0096,
+      "step": 7124
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1773255813953485e-07,
+      "loss": 0.0105,
+      "step": 7125
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1772093023255814e-07,
+      "loss": 0.0014,
+      "step": 7126
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1770930232558138e-07,
+      "loss": 0.0015,
+      "step": 7127
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1769767441860464e-07,
+      "loss": 0.0024,
+      "step": 7128
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1768604651162787e-07,
+      "loss": 0.0021,
+      "step": 7129
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1767441860465116e-07,
+      "loss": 0.0018,
+      "step": 7130
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1766279069767442e-07,
+      "loss": 0.0015,
+      "step": 7131
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1765116279069766e-07,
+      "loss": 0.0017,
+      "step": 7132
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1763953488372092e-07,
+      "loss": 0.0019,
+      "step": 7133
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1762790697674418e-07,
+      "loss": 0.3833,
+      "step": 7134
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1761627906976744e-07,
+      "loss": 0.0058,
+      "step": 7135
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1760465116279068e-07,
+      "loss": 0.0023,
+      "step": 7136
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1759302325581394e-07,
+      "loss": 0.0016,
+      "step": 7137
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.175813953488372e-07,
+      "loss": 0.0019,
+      "step": 7138
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.1756976744186046e-07,
+      "loss": 0.0024,
+      "step": 7139
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.175581395348837e-07,
+      "loss": 0.0024,
+      "step": 7140
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.1754651162790696e-07,
+      "loss": 0.0024,
+      "step": 7141
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.175348837209302e-07,
+      "loss": 0.0029,
+      "step": 7142
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1752325581395349e-07,
+      "loss": 0.0015,
+      "step": 7143
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1751162790697675e-07,
+      "loss": 0.0045,
+      "step": 7144
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1749999999999998e-07,
+      "loss": 0.002,
+      "step": 7145
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1748837209302324e-07,
+      "loss": 0.0024,
+      "step": 7146
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.174767441860465e-07,
+      "loss": 0.0073,
+      "step": 7147
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.1746511627906977e-07,
+      "loss": 0.0017,
+      "step": 7148
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.17453488372093e-07,
+      "loss": 0.0041,
+      "step": 7149
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.1744186046511626e-07,
+      "loss": 0.2612,
+      "step": 7150
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.1743023255813953e-07,
+      "loss": 0.0021,
+      "step": 7151
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.174186046511628e-07,
+      "loss": 0.0019,
+      "step": 7152
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1740697674418602e-07,
+      "loss": 0.0015,
+      "step": 7153
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1739534883720928e-07,
+      "loss": 0.0028,
+      "step": 7154
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1738372093023257e-07,
+      "loss": 0.0019,
+      "step": 7155
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.173720930232558e-07,
+      "loss": 0.0014,
+      "step": 7156
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1736046511627907e-07,
+      "loss": 0.0016,
+      "step": 7157
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.173488372093023e-07,
+      "loss": 0.0037,
+      "step": 7158
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.173372093023256e-07,
+      "loss": 0.0363,
+      "step": 7159
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.1732558139534883e-07,
+      "loss": 0.0016,
+      "step": 7160
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.173139534883721e-07,
+      "loss": 0.0015,
+      "step": 7161
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.1730232558139533e-07,
+      "loss": 0.0014,
+      "step": 7162
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.1729069767441861e-07,
+      "loss": 0.0019,
+      "step": 7163
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.1727906976744185e-07,
+      "loss": 0.0024,
+      "step": 7164
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.172674418604651e-07,
+      "loss": 0.0016,
+      "step": 7165
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.1725581395348835e-07,
+      "loss": 1.7166,
+      "step": 7166
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.172441860465116e-07,
+      "loss": 0.0052,
+      "step": 7167
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.172325581395349e-07,
+      "loss": 0.0016,
+      "step": 7168
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1722093023255813e-07,
+      "loss": 0.0019,
+      "step": 7169
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1720930232558137e-07,
+      "loss": 0.0013,
+      "step": 7170
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1719767441860463e-07,
+      "loss": 0.0023,
+      "step": 7171
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1718604651162792e-07,
+      "loss": 0.0017,
+      "step": 7172
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1717441860465115e-07,
+      "loss": 0.0052,
+      "step": 7173
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.171627906976744e-07,
+      "loss": 0.0023,
+      "step": 7174
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.1715116279069765e-07,
+      "loss": 0.8028,
+      "step": 7175
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.1713953488372094e-07,
+      "loss": 0.0019,
+      "step": 7176
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.1712790697674417e-07,
+      "loss": 0.0024,
+      "step": 7177
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.1711627906976743e-07,
+      "loss": 0.0017,
+      "step": 7178
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.1710465116279067e-07,
+      "loss": 0.0019,
+      "step": 7179
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.1709302325581396e-07,
+      "loss": 0.0015,
+      "step": 7180
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.170813953488372e-07,
+      "loss": 0.0014,
+      "step": 7181
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.1706976744186045e-07,
+      "loss": 0.002,
+      "step": 7182
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.170581395348837e-07,
+      "loss": 0.0054,
+      "step": 7183
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.1704651162790698e-07,
+      "loss": 0.0119,
+      "step": 7184
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.1703488372093024e-07,
+      "loss": 0.0019,
+      "step": 7185
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.1702325581395347e-07,
+      "loss": 0.0127,
+      "step": 7186
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.1701162790697674e-07,
+      "loss": 0.0312,
+      "step": 7187
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.17e-07,
+      "loss": 0.0022,
+      "step": 7188
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1698837209302326e-07,
+      "loss": 0.002,
+      "step": 7189
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.169767441860465e-07,
+      "loss": 0.0022,
+      "step": 7190
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1696511627906976e-07,
+      "loss": 0.0019,
+      "step": 7191
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1695348837209302e-07,
+      "loss": 0.0021,
+      "step": 7192
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1694186046511628e-07,
+      "loss": 0.0015,
+      "step": 7193
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1693023255813951e-07,
+      "loss": 0.0014,
+      "step": 7194
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.1691860465116278e-07,
+      "loss": 0.0033,
+      "step": 7195
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.16906976744186e-07,
+      "loss": 1.5293,
+      "step": 7196
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.168953488372093e-07,
+      "loss": 0.0023,
+      "step": 7197
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.1688372093023256e-07,
+      "loss": 0.0024,
+      "step": 7198
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.168720930232558e-07,
+      "loss": 3.0098,
+      "step": 7199
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1686046511627906e-07,
+      "loss": 0.0015,
+      "step": 7200
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1684883720930232e-07,
+      "loss": 0.0012,
+      "step": 7201
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1683720930232558e-07,
+      "loss": 0.0022,
+      "step": 7202
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1682558139534882e-07,
+      "loss": 2.2095,
+      "step": 7203
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1681395348837208e-07,
+      "loss": 0.0109,
+      "step": 7204
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1680232558139534e-07,
+      "loss": 0.0023,
+      "step": 7205
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.167906976744186e-07,
+      "loss": 0.0023,
+      "step": 7206
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1677906976744184e-07,
+      "loss": 0.0024,
+      "step": 7207
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.167674418604651e-07,
+      "loss": 0.002,
+      "step": 7208
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.167558139534884e-07,
+      "loss": 0.0016,
+      "step": 7209
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.1674418604651162e-07,
+      "loss": 0.0025,
+      "step": 7210
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.1673255813953486e-07,
+      "loss": 0.0016,
+      "step": 7211
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.1672093023255812e-07,
+      "loss": 1.7815,
+      "step": 7212
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.167093023255814e-07,
+      "loss": 0.0016,
+      "step": 7213
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.1669767441860464e-07,
+      "loss": 0.0056,
+      "step": 7214
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.166860465116279e-07,
+      "loss": 0.002,
+      "step": 7215
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1667441860465114e-07,
+      "loss": 0.0081,
+      "step": 7216
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1666279069767443e-07,
+      "loss": 0.0029,
+      "step": 7217
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1665116279069766e-07,
+      "loss": 0.0024,
+      "step": 7218
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1663953488372093e-07,
+      "loss": 0.0023,
+      "step": 7219
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1662790697674416e-07,
+      "loss": 0.0018,
+      "step": 7220
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1661627906976742e-07,
+      "loss": 0.0015,
+      "step": 7221
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1660465116279068e-07,
+      "loss": 0.0021,
+      "step": 7222
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1659302325581395e-07,
+      "loss": 0.0014,
+      "step": 7223
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1658139534883718e-07,
+      "loss": 0.002,
+      "step": 7224
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1656976744186044e-07,
+      "loss": 0.0029,
+      "step": 7225
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1655813953488373e-07,
+      "loss": 0.0023,
+      "step": 7226
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1654651162790697e-07,
+      "loss": 0.0027,
+      "step": 7227
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1653488372093023e-07,
+      "loss": 0.0027,
+      "step": 7228
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1652325581395346e-07,
+      "loss": 0.0029,
+      "step": 7229
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1651162790697675e-07,
+      "loss": 0.0017,
+      "step": 7230
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1649999999999999e-07,
+      "loss": 0.0017,
+      "step": 7231
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1648837209302325e-07,
+      "loss": 0.0013,
+      "step": 7232
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1647674418604648e-07,
+      "loss": 0.0032,
+      "step": 7233
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1646511627906977e-07,
+      "loss": 0.0013,
+      "step": 7234
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.16453488372093e-07,
+      "loss": 0.0029,
+      "step": 7235
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.1644186046511627e-07,
+      "loss": 0.002,
+      "step": 7236
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.164302325581395e-07,
+      "loss": 0.0016,
+      "step": 7237
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.164186046511628e-07,
+      "loss": 0.0045,
+      "step": 7238
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.9864603481624759,
+      "eval_loss": 0.15736442804336548,
+      "eval_runtime": 1.5601,
+      "eval_samples_per_second": 662.775,
+      "eval_steps_per_second": 10.897,
+      "step": 7238
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.1640697674418605e-07,
+      "loss": 0.0037,
+      "step": 7239
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.163953488372093e-07,
+      "loss": 0.0021,
+      "step": 7240
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.1638372093023255e-07,
+      "loss": 0.0029,
+      "step": 7241
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.163720930232558e-07,
+      "loss": 0.0024,
+      "step": 7242
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.1636046511627907e-07,
+      "loss": 0.0048,
+      "step": 7243
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.163488372093023e-07,
+      "loss": 0.0025,
+      "step": 7244
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.1633720930232557e-07,
+      "loss": 0.0021,
+      "step": 7245
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1632558139534883e-07,
+      "loss": 0.0017,
+      "step": 7246
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.163139534883721e-07,
+      "loss": 0.0027,
+      "step": 7247
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1630232558139533e-07,
+      "loss": 0.0016,
+      "step": 7248
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.162906976744186e-07,
+      "loss": 0.0026,
+      "step": 7249
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1627906976744183e-07,
+      "loss": 0.0014,
+      "step": 7250
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.1626744186046511e-07,
+      "loss": 0.0018,
+      "step": 7251
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.1625581395348838e-07,
+      "loss": 0.0015,
+      "step": 7252
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.162441860465116e-07,
+      "loss": 0.0025,
+      "step": 7253
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.1623255813953485e-07,
+      "loss": 0.0028,
+      "step": 7254
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.1622093023255814e-07,
+      "loss": 0.0028,
+      "step": 7255
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.162093023255814e-07,
+      "loss": 0.0023,
+      "step": 7256
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1619767441860463e-07,
+      "loss": 0.0019,
+      "step": 7257
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.161860465116279e-07,
+      "loss": 0.0016,
+      "step": 7258
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1617441860465116e-07,
+      "loss": 0.0016,
+      "step": 7259
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1616279069767442e-07,
+      "loss": 0.0033,
+      "step": 7260
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1615116279069765e-07,
+      "loss": 0.0016,
+      "step": 7261
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.1613953488372091e-07,
+      "loss": 0.0027,
+      "step": 7262
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.161279069767442e-07,
+      "loss": 1.8661,
+      "step": 7263
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.1611627906976744e-07,
+      "loss": 0.0012,
+      "step": 7264
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.1610465116279067e-07,
+      "loss": 0.0023,
+      "step": 7265
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.1609302325581393e-07,
+      "loss": 0.0019,
+      "step": 7266
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1608139534883722e-07,
+      "loss": 0.0016,
+      "step": 7267
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1606976744186046e-07,
+      "loss": 0.0017,
+      "step": 7268
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1605813953488372e-07,
+      "loss": 0.0021,
+      "step": 7269
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1604651162790696e-07,
+      "loss": 0.0016,
+      "step": 7270
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1603488372093024e-07,
+      "loss": 0.0018,
+      "step": 7271
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.1602325581395348e-07,
+      "loss": 0.0016,
+      "step": 7272
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.1601162790697674e-07,
+      "loss": 0.0032,
+      "step": 7273
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.1599999999999998e-07,
+      "loss": 0.0336,
+      "step": 7274
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.1598837209302324e-07,
+      "loss": 0.0065,
+      "step": 7275
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.159767441860465e-07,
+      "loss": 0.0013,
+      "step": 7276
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.1596511627906976e-07,
+      "loss": 0.0027,
+      "step": 7277
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.15953488372093e-07,
+      "loss": 0.0019,
+      "step": 7278
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.1594186046511626e-07,
+      "loss": 0.0015,
+      "step": 7279
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.1593023255813955e-07,
+      "loss": 0.0018,
+      "step": 7280
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.1591860465116278e-07,
+      "loss": 0.0021,
+      "step": 7281
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1590697674418604e-07,
+      "loss": 0.0014,
+      "step": 7282
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1589534883720928e-07,
+      "loss": 0.002,
+      "step": 7283
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1588372093023257e-07,
+      "loss": 0.0021,
+      "step": 7284
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.158720930232558e-07,
+      "loss": 0.0015,
+      "step": 7285
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1586046511627906e-07,
+      "loss": 0.0022,
+      "step": 7286
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.158488372093023e-07,
+      "loss": 0.0016,
+      "step": 7287
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.1583720930232559e-07,
+      "loss": 0.0014,
+      "step": 7288
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.1582558139534882e-07,
+      "loss": 0.0075,
+      "step": 7289
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.1581395348837208e-07,
+      "loss": 0.0029,
+      "step": 7290
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.1580232558139532e-07,
+      "loss": 0.0014,
+      "step": 7291
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.157906976744186e-07,
+      "loss": 0.0019,
+      "step": 7292
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.1577906976744187e-07,
+      "loss": 0.0022,
+      "step": 7293
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.157674418604651e-07,
+      "loss": 0.0021,
+      "step": 7294
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.1575581395348837e-07,
+      "loss": 0.0022,
+      "step": 7295
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.1574418604651163e-07,
+      "loss": 0.0018,
+      "step": 7296
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.157325581395349e-07,
+      "loss": 0.0017,
+      "step": 7297
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1572093023255812e-07,
+      "loss": 0.0025,
+      "step": 7298
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1570930232558139e-07,
+      "loss": 0.0024,
+      "step": 7299
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1569767441860462e-07,
+      "loss": 0.0014,
+      "step": 7300
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.156860465116279e-07,
+      "loss": 0.0016,
+      "step": 7301
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1567441860465114e-07,
+      "loss": 0.0021,
+      "step": 7302
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.156627906976744e-07,
+      "loss": 0.0025,
+      "step": 7303
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.1565116279069764e-07,
+      "loss": 0.0027,
+      "step": 7304
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.1563953488372093e-07,
+      "loss": 0.0018,
+      "step": 7305
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.156279069767442e-07,
+      "loss": 0.0071,
+      "step": 7306
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.1561627906976743e-07,
+      "loss": 0.0013,
+      "step": 7307
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1560465116279066e-07,
+      "loss": 0.0015,
+      "step": 7308
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1559302325581395e-07,
+      "loss": 0.0024,
+      "step": 7309
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.155813953488372e-07,
+      "loss": 0.0018,
+      "step": 7310
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1556976744186045e-07,
+      "loss": 0.0013,
+      "step": 7311
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.155581395348837e-07,
+      "loss": 0.0019,
+      "step": 7312
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1554651162790697e-07,
+      "loss": 0.002,
+      "step": 7313
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1553488372093023e-07,
+      "loss": 0.0016,
+      "step": 7314
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1552325581395347e-07,
+      "loss": 0.0024,
+      "step": 7315
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1551162790697673e-07,
+      "loss": 0.0017,
+      "step": 7316
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1550000000000002e-07,
+      "loss": 0.0014,
+      "step": 7317
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1548837209302325e-07,
+      "loss": 0.3242,
+      "step": 7318
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.154767441860465e-07,
+      "loss": 0.0451,
+      "step": 7319
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1546511627906975e-07,
+      "loss": 0.0026,
+      "step": 7320
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1545348837209304e-07,
+      "loss": 0.0014,
+      "step": 7321
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1544186046511627e-07,
+      "loss": 0.0021,
+      "step": 7322
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1543023255813953e-07,
+      "loss": 0.0032,
+      "step": 7323
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1541860465116277e-07,
+      "loss": 0.0021,
+      "step": 7324
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1540697674418606e-07,
+      "loss": 0.002,
+      "step": 7325
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.153953488372093e-07,
+      "loss": 0.002,
+      "step": 7326
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1538372093023255e-07,
+      "loss": 0.0028,
+      "step": 7327
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.153720930232558e-07,
+      "loss": 0.0021,
+      "step": 7328
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.1536046511627905e-07,
+      "loss": 0.002,
+      "step": 7329
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.1534883720930231e-07,
+      "loss": 0.0017,
+      "step": 7330
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.1533720930232558e-07,
+      "loss": 0.0027,
+      "step": 7331
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.153255813953488e-07,
+      "loss": 0.0014,
+      "step": 7332
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.1531395348837207e-07,
+      "loss": 0.0019,
+      "step": 7333
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1530232558139536e-07,
+      "loss": 0.0456,
+      "step": 7334
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.152906976744186e-07,
+      "loss": 0.0016,
+      "step": 7335
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1527906976744186e-07,
+      "loss": 0.0014,
+      "step": 7336
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.152674418604651e-07,
+      "loss": 0.0018,
+      "step": 7337
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1525581395348838e-07,
+      "loss": 0.0016,
+      "step": 7338
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.1524418604651162e-07,
+      "loss": 0.0112,
+      "step": 7339
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.1523255813953488e-07,
+      "loss": 0.0016,
+      "step": 7340
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.152209302325581e-07,
+      "loss": 0.002,
+      "step": 7341
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.152093023255814e-07,
+      "loss": 0.0024,
+      "step": 7342
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.1519767441860464e-07,
+      "loss": 0.002,
+      "step": 7343
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.151860465116279e-07,
+      "loss": 0.0018,
+      "step": 7344
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1517441860465113e-07,
+      "loss": 0.002,
+      "step": 7345
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1516279069767442e-07,
+      "loss": 0.0015,
+      "step": 7346
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1515116279069768e-07,
+      "loss": 0.0012,
+      "step": 7347
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1513953488372092e-07,
+      "loss": 0.0021,
+      "step": 7348
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1512790697674415e-07,
+      "loss": 0.0016,
+      "step": 7349
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.1511627906976744e-07,
+      "loss": 0.0024,
+      "step": 7350
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.151046511627907e-07,
+      "loss": 0.002,
+      "step": 7351
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.1509302325581394e-07,
+      "loss": 0.0031,
+      "step": 7352
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.150813953488372e-07,
+      "loss": 1.3875,
+      "step": 7353
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.1506976744186044e-07,
+      "loss": 2.7546,
+      "step": 7354
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1505813953488372e-07,
+      "loss": 0.007,
+      "step": 7355
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1504651162790696e-07,
+      "loss": 0.0021,
+      "step": 7356
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1503488372093022e-07,
+      "loss": 0.002,
+      "step": 7357
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1502325581395346e-07,
+      "loss": 0.0021,
+      "step": 7358
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1501162790697674e-07,
+      "loss": 0.0024,
+      "step": 7359
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1499999999999998e-07,
+      "loss": 0.0017,
+      "step": 7360
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1498837209302324e-07,
+      "loss": 0.0028,
+      "step": 7361
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1497674418604648e-07,
+      "loss": 0.0019,
+      "step": 7362
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1496511627906976e-07,
+      "loss": 0.0025,
+      "step": 7363
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1495348837209303e-07,
+      "loss": 0.3969,
+      "step": 7364
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1494186046511626e-07,
+      "loss": 0.22,
+      "step": 7365
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1493023255813952e-07,
+      "loss": 0.0013,
+      "step": 7366
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1491860465116279e-07,
+      "loss": 0.002,
+      "step": 7367
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1490697674418605e-07,
+      "loss": 0.002,
+      "step": 7368
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1489534883720928e-07,
+      "loss": 0.0021,
+      "step": 7369
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1488372093023254e-07,
+      "loss": 0.0025,
+      "step": 7370
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.148720930232558e-07,
+      "loss": 0.0013,
+      "step": 7371
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1486046511627907e-07,
+      "loss": 1.2698,
+      "step": 7372
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.148488372093023e-07,
+      "loss": 0.0083,
+      "step": 7373
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1483720930232556e-07,
+      "loss": 0.002,
+      "step": 7374
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1482558139534885e-07,
+      "loss": 0.0017,
+      "step": 7375
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.148139534883721e-07,
+      "loss": 0.0016,
+      "step": 7376
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1480232558139535e-07,
+      "loss": 0.0022,
+      "step": 7377
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1479069767441858e-07,
+      "loss": 0.0034,
+      "step": 7378
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1477906976744187e-07,
+      "loss": 0.0014,
+      "step": 7379
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.147674418604651e-07,
+      "loss": 0.0021,
+      "step": 7380
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1475581395348837e-07,
+      "loss": 0.0018,
+      "step": 7381
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.147441860465116e-07,
+      "loss": 0.0017,
+      "step": 7382
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1473255813953487e-07,
+      "loss": 0.0026,
+      "step": 7383
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1472093023255813e-07,
+      "loss": 2.22,
+      "step": 7384
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.147093023255814e-07,
+      "loss": 1.7273,
+      "step": 7385
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1469767441860463e-07,
+      "loss": 0.0016,
+      "step": 7386
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.146860465116279e-07,
+      "loss": 0.0022,
+      "step": 7387
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1467441860465118e-07,
+      "loss": 0.0023,
+      "step": 7388
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.146627906976744e-07,
+      "loss": 0.0015,
+      "step": 7389
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1465116279069767e-07,
+      "loss": 0.0018,
+      "step": 7390
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.146395348837209e-07,
+      "loss": 0.0022,
+      "step": 7391
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.146279069767442e-07,
+      "loss": 0.0017,
+      "step": 7392
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.1461627906976743e-07,
+      "loss": 0.0027,
+      "step": 7393
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.146046511627907e-07,
+      "loss": 0.0013,
+      "step": 7394
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.1459302325581393e-07,
+      "loss": 0.0021,
+      "step": 7395
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1458139534883722e-07,
+      "loss": 0.002,
+      "step": 7396
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1456976744186045e-07,
+      "loss": 0.0035,
+      "step": 7397
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.145581395348837e-07,
+      "loss": 0.0019,
+      "step": 7398
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1454651162790695e-07,
+      "loss": 0.0016,
+      "step": 7399
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1453488372093024e-07,
+      "loss": 0.0017,
+      "step": 7400
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.145232558139535e-07,
+      "loss": 0.0037,
+      "step": 7401
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1451162790697673e-07,
+      "loss": 0.0018,
+      "step": 7402
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1449999999999997e-07,
+      "loss": 0.0016,
+      "step": 7403
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1448837209302326e-07,
+      "loss": 0.0033,
+      "step": 7404
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1447674418604652e-07,
+      "loss": 0.0021,
+      "step": 7405
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1446511627906975e-07,
+      "loss": 0.0046,
+      "step": 7406
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.1445348837209302e-07,
+      "loss": 0.0017,
+      "step": 7407
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.1444186046511625e-07,
+      "loss": 0.0019,
+      "step": 7408
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.1443023255813954e-07,
+      "loss": 0.0021,
+      "step": 7409
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.1441860465116277e-07,
+      "loss": 0.0042,
+      "step": 7410
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.1440697674418604e-07,
+      "loss": 0.0028,
+      "step": 7411
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.1439534883720927e-07,
+      "loss": 0.0021,
+      "step": 7412
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.1438372093023256e-07,
+      "loss": 0.0039,
+      "step": 7413
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.143720930232558e-07,
+      "loss": 0.0019,
+      "step": 7414
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.1436046511627906e-07,
+      "loss": 0.0015,
+      "step": 7415
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.143488372093023e-07,
+      "loss": 0.0013,
+      "step": 7416
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.1433720930232558e-07,
+      "loss": 0.0015,
+      "step": 7417
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.1432558139534884e-07,
+      "loss": 2.5705,
+      "step": 7418
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.1431395348837208e-07,
+      "loss": 0.0024,
+      "step": 7419
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.1430232558139534e-07,
+      "loss": 0.002,
+      "step": 7420
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.142906976744186e-07,
+      "loss": 0.0027,
+      "step": 7421
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.1427906976744186e-07,
+      "loss": 0.0017,
+      "step": 7422
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.142674418604651e-07,
+      "loss": 0.0013,
+      "step": 7423
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.1425581395348836e-07,
+      "loss": 0.0018,
+      "step": 7424
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.1424418604651162e-07,
+      "loss": 0.0017,
+      "step": 7425
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.1423255813953488e-07,
+      "loss": 0.0013,
+      "step": 7426
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.1422093023255812e-07,
+      "loss": 0.0036,
+      "step": 7427
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.1420930232558138e-07,
+      "loss": 2.246,
+      "step": 7428
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.1419767441860467e-07,
+      "loss": 0.0023,
+      "step": 7429
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.141860465116279e-07,
+      "loss": 0.0021,
+      "step": 7430
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.1417441860465116e-07,
+      "loss": 0.0017,
+      "step": 7431
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.141627906976744e-07,
+      "loss": 0.0016,
+      "step": 7432
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.1415116279069766e-07,
+      "loss": 0.0015,
+      "step": 7433
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.1413953488372092e-07,
+      "loss": 0.377,
+      "step": 7434
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.1412790697674418e-07,
+      "loss": 0.0019,
+      "step": 7435
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.1411627906976742e-07,
+      "loss": 0.0015,
+      "step": 7436
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.1410465116279068e-07,
+      "loss": 0.0015,
+      "step": 7437
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.1409302325581394e-07,
+      "loss": 0.002,
+      "step": 7438
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.140813953488372e-07,
+      "loss": 0.0075,
+      "step": 7439
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.1406976744186044e-07,
+      "loss": 0.0019,
+      "step": 7440
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.140581395348837e-07,
+      "loss": 0.0035,
+      "step": 7441
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.14046511627907e-07,
+      "loss": 0.0029,
+      "step": 7442
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.1403488372093023e-07,
+      "loss": 2.2086,
+      "step": 7443
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.140232558139535e-07,
+      "loss": 0.0015,
+      "step": 7444
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.1401162790697672e-07,
+      "loss": 0.0014,
+      "step": 7445
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.14e-07,
+      "loss": 0.0015,
+      "step": 7446
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.1398837209302325e-07,
+      "loss": 0.0019,
+      "step": 7447
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.139767441860465e-07,
+      "loss": 0.0019,
+      "step": 7448
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.1396511627906974e-07,
+      "loss": 0.0023,
+      "step": 7449
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.1395348837209303e-07,
+      "loss": 0.0147,
+      "step": 7450
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.1394186046511627e-07,
+      "loss": 0.4035,
+      "step": 7451
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.1393023255813953e-07,
+      "loss": 0.0018,
+      "step": 7452
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.1391860465116276e-07,
+      "loss": 0.0028,
+      "step": 7453
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.1390697674418605e-07,
+      "loss": 0.0024,
+      "step": 7454
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.1389534883720929e-07,
+      "loss": 0.0025,
+      "step": 7455
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.1388372093023255e-07,
+      "loss": 0.0026,
+      "step": 7456
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.1387209302325578e-07,
+      "loss": 0.0023,
+      "step": 7457
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.1386046511627907e-07,
+      "loss": 3.0955,
+      "step": 7458
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.1384883720930233e-07,
+      "loss": 0.0013,
+      "step": 7459
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.1383720930232557e-07,
+      "loss": 0.0022,
+      "step": 7460
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.1382558139534883e-07,
+      "loss": 0.0017,
+      "step": 7461
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.1381395348837207e-07,
+      "loss": 0.0016,
+      "step": 7462
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.1380232558139535e-07,
+      "loss": 0.0013,
+      "step": 7463
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.137906976744186e-07,
+      "loss": 0.1378,
+      "step": 7464
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.1377906976744185e-07,
+      "loss": 0.0018,
+      "step": 7465
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.1376744186046509e-07,
+      "loss": 0.0021,
+      "step": 7466
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.1375581395348837e-07,
+      "loss": 0.002,
+      "step": 7467
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.137441860465116e-07,
+      "loss": 0.0015,
+      "step": 7468
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.1373255813953487e-07,
+      "loss": 0.0018,
+      "step": 7469
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.137209302325581e-07,
+      "loss": 0.0019,
+      "step": 7470
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.137093023255814e-07,
+      "loss": 0.0023,
+      "step": 7471
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.1369767441860466e-07,
+      "loss": 0.0022,
+      "step": 7472
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.136860465116279e-07,
+      "loss": 0.0037,
+      "step": 7473
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.1367441860465115e-07,
+      "loss": 0.3663,
+      "step": 7474
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.1366279069767441e-07,
+      "loss": 0.0021,
+      "step": 7475
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.1365116279069768e-07,
+      "loss": 0.0021,
+      "step": 7476
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.136395348837209e-07,
+      "loss": 0.0023,
+      "step": 7477
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.1362790697674417e-07,
+      "loss": 0.0015,
+      "step": 7478
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.1361627906976744e-07,
+      "loss": 0.0021,
+      "step": 7479
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.136046511627907e-07,
+      "loss": 0.0273,
+      "step": 7480
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.1359302325581393e-07,
+      "loss": 0.0021,
+      "step": 7481
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.135813953488372e-07,
+      "loss": 0.0022,
+      "step": 7482
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.1356976744186048e-07,
+      "loss": 0.0014,
+      "step": 7483
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.1355813953488372e-07,
+      "loss": 0.0023,
+      "step": 7484
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.1354651162790698e-07,
+      "loss": 0.0015,
+      "step": 7485
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.1353488372093021e-07,
+      "loss": 0.0019,
+      "step": 7486
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.1352325581395345e-07,
+      "loss": 0.0017,
+      "step": 7487
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.1351162790697674e-07,
+      "loss": 0.0019,
+      "step": 7488
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.135e-07,
+      "loss": 0.0018,
+      "step": 7489
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.1348837209302323e-07,
+      "loss": 0.003,
+      "step": 7490
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.134767441860465e-07,
+      "loss": 0.0023,
+      "step": 7491
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.1346511627906976e-07,
+      "loss": 0.0015,
+      "step": 7492
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.1345348837209302e-07,
+      "loss": 0.0022,
+      "step": 7493
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.1344186046511625e-07,
+      "loss": 0.0031,
+      "step": 7494
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.1343023255813952e-07,
+      "loss": 0.0016,
+      "step": 7495
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.134186046511628e-07,
+      "loss": 0.0021,
+      "step": 7496
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.1340697674418604e-07,
+      "loss": 0.0036,
+      "step": 7497
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.1339534883720928e-07,
+      "loss": 0.0017,
+      "step": 7498
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.1338372093023254e-07,
+      "loss": 0.002,
+      "step": 7499
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.1337209302325583e-07,
+      "loss": 0.0016,
+      "step": 7500
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.1336046511627906e-07,
+      "loss": 0.002,
+      "step": 7501
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.1334883720930232e-07,
+      "loss": 0.0026,
+      "step": 7502
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.1333720930232556e-07,
+      "loss": 0.0019,
+      "step": 7503
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.1332558139534885e-07,
+      "loss": 0.0062,
+      "step": 7504
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.1331395348837208e-07,
+      "loss": 0.0029,
+      "step": 7505
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.1330232558139534e-07,
+      "loss": 0.0029,
+      "step": 7506
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.1329069767441858e-07,
+      "loss": 0.0013,
+      "step": 7507
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.1327906976744187e-07,
+      "loss": 0.0015,
+      "step": 7508
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.132674418604651e-07,
+      "loss": 1.8882,
+      "step": 7509
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.1325581395348836e-07,
+      "loss": 0.002,
+      "step": 7510
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.132441860465116e-07,
+      "loss": 0.0081,
+      "step": 7511
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.1323255813953489e-07,
+      "loss": 0.0023,
+      "step": 7512
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.1322093023255815e-07,
+      "loss": 0.0014,
+      "step": 7513
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.1320930232558138e-07,
+      "loss": 0.002,
+      "step": 7514
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.1319767441860465e-07,
+      "loss": 0.003,
+      "step": 7515
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.1318604651162788e-07,
+      "loss": 0.0346,
+      "step": 7516
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.1317441860465117e-07,
+      "loss": 0.1922,
+      "step": 7517
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.131627906976744e-07,
+      "loss": 0.0084,
+      "step": 7518
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.1315116279069767e-07,
+      "loss": 0.0018,
+      "step": 7519
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.131395348837209e-07,
+      "loss": 0.0022,
+      "step": 7520
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.131279069767442e-07,
+      "loss": 0.0014,
+      "step": 7521
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.1311627906976742e-07,
+      "loss": 0.0015,
+      "step": 7522
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.1310465116279069e-07,
+      "loss": 0.0018,
+      "step": 7523
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.1309302325581392e-07,
+      "loss": 0.0032,
+      "step": 7524
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.130813953488372e-07,
+      "loss": 0.197,
+      "step": 7525
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.1306976744186047e-07,
+      "loss": 0.0015,
+      "step": 7526
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.130581395348837e-07,
+      "loss": 0.0017,
+      "step": 7527
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.1304651162790697e-07,
+      "loss": 0.0027,
+      "step": 7528
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.1303488372093023e-07,
+      "loss": 0.6074,
+      "step": 7529
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.130232558139535e-07,
+      "loss": 0.0027,
+      "step": 7530
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.1301162790697673e-07,
+      "loss": 0.1773,
+      "step": 7531
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.13e-07,
+      "loss": 0.0021,
+      "step": 7532
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.1298837209302325e-07,
+      "loss": 0.0039,
+      "step": 7533
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.129767441860465e-07,
+      "loss": 0.0015,
+      "step": 7534
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.1296511627906975e-07,
+      "loss": 0.0016,
+      "step": 7535
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.12953488372093e-07,
+      "loss": 0.0014,
+      "step": 7536
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.129418604651163e-07,
+      "loss": 0.0034,
+      "step": 7537
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.1293023255813953e-07,
+      "loss": 0.145,
+      "step": 7538
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.129186046511628e-07,
+      "loss": 0.0017,
+      "step": 7539
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.1290697674418603e-07,
+      "loss": 0.0015,
+      "step": 7540
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.1289534883720926e-07,
+      "loss": 0.0018,
+      "step": 7541
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.1288372093023255e-07,
+      "loss": 0.0015,
+      "step": 7542
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.1287209302325581e-07,
+      "loss": 0.0033,
+      "step": 7543
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.1286046511627905e-07,
+      "loss": 0.0022,
+      "step": 7544
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.128488372093023e-07,
+      "loss": 0.0017,
+      "step": 7545
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.1283720930232557e-07,
+      "loss": 0.0025,
+      "step": 7546
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.1282558139534883e-07,
+      "loss": 0.0013,
+      "step": 7547
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.1281395348837207e-07,
+      "loss": 0.0016,
+      "step": 7548
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.1280232558139533e-07,
+      "loss": 0.0017,
+      "step": 7549
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.1279069767441862e-07,
+      "loss": 0.0018,
+      "step": 7550
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.1277906976744185e-07,
+      "loss": 0.0033,
+      "step": 7551
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.127674418604651e-07,
+      "loss": 0.0022,
+      "step": 7552
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.1275581395348835e-07,
+      "loss": 0.0027,
+      "step": 7553
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.1274418604651164e-07,
+      "loss": 0.0029,
+      "step": 7554
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.1273255813953488e-07,
+      "loss": 0.0018,
+      "step": 7555
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.1272093023255814e-07,
+      "loss": 0.0023,
+      "step": 7556
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.1270930232558137e-07,
+      "loss": 0.0018,
+      "step": 7557
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.1269767441860466e-07,
+      "loss": 0.0068,
+      "step": 7558
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.126860465116279e-07,
+      "loss": 0.0015,
+      "step": 7559
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.1267441860465116e-07,
+      "loss": 0.0026,
+      "step": 7560
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.126627906976744e-07,
+      "loss": 0.0015,
+      "step": 7561
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.1265116279069768e-07,
+      "loss": 0.0016,
+      "step": 7562
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.1263953488372092e-07,
+      "loss": 0.0018,
+      "step": 7563
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.1262790697674418e-07,
+      "loss": 0.0015,
+      "step": 7564
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.126162790697674e-07,
+      "loss": 0.0234,
+      "step": 7565
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.126046511627907e-07,
+      "loss": 0.0021,
+      "step": 7566
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.1259302325581396e-07,
+      "loss": 0.0017,
+      "step": 7567
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.125813953488372e-07,
+      "loss": 2.1628,
+      "step": 7568
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.1256976744186046e-07,
+      "loss": 0.0015,
+      "step": 7569
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.125581395348837e-07,
+      "loss": 0.0013,
+      "step": 7570
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.1254651162790698e-07,
+      "loss": 0.0033,
+      "step": 7571
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.1253488372093022e-07,
+      "loss": 0.0026,
+      "step": 7572
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.1252325581395348e-07,
+      "loss": 0.0018,
+      "step": 7573
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.1251162790697672e-07,
+      "loss": 0.0014,
+      "step": 7574
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.125e-07,
+      "loss": 0.0014,
+      "step": 7575
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.1248837209302324e-07,
+      "loss": 0.0016,
+      "step": 7576
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.124767441860465e-07,
+      "loss": 0.0019,
+      "step": 7577
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.1246511627906974e-07,
+      "loss": 0.0013,
+      "step": 7578
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.1245348837209302e-07,
+      "loss": 0.0015,
+      "step": 7579
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.1244186046511629e-07,
+      "loss": 0.0023,
+      "step": 7580
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.1243023255813952e-07,
+      "loss": 0.0023,
+      "step": 7581
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.1241860465116276e-07,
+      "loss": 0.0203,
+      "step": 7582
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.1240697674418604e-07,
+      "loss": 2.2226,
+      "step": 7583
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.123953488372093e-07,
+      "loss": 0.0038,
+      "step": 7584
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.1238372093023254e-07,
+      "loss": 0.0035,
+      "step": 7585
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.123720930232558e-07,
+      "loss": 0.002,
+      "step": 7586
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.1236046511627906e-07,
+      "loss": 0.0015,
+      "step": 7587
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.1234883720930233e-07,
+      "loss": 0.0017,
+      "step": 7588
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.1233720930232556e-07,
+      "loss": 0.0032,
+      "step": 7589
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.1232558139534882e-07,
+      "loss": 0.0012,
+      "step": 7590
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.123139534883721e-07,
+      "loss": 0.0019,
+      "step": 7591
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.1230232558139535e-07,
+      "loss": 0.0021,
+      "step": 7592
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.1229069767441858e-07,
+      "loss": 0.0014,
+      "step": 7593
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.1227906976744184e-07,
+      "loss": 0.0013,
+      "step": 7594
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.1226744186046508e-07,
+      "loss": 0.0028,
+      "step": 7595
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.1225581395348837e-07,
+      "loss": 0.0225,
+      "step": 7596
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.1224418604651163e-07,
+      "loss": 0.0012,
+      "step": 7597
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.1223255813953486e-07,
+      "loss": 0.0022,
+      "step": 7598
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.1222093023255813e-07,
+      "loss": 0.0017,
+      "step": 7599
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.122093023255814e-07,
+      "loss": 0.0016,
+      "step": 7600
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.1219767441860465e-07,
+      "loss": 0.0017,
+      "step": 7601
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.1218604651162788e-07,
+      "loss": 0.1035,
+      "step": 7602
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.1217441860465115e-07,
+      "loss": 0.0011,
+      "step": 7603
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.121627906976744e-07,
+      "loss": 0.0023,
+      "step": 7604
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.1215116279069767e-07,
+      "loss": 0.0015,
+      "step": 7605
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.121395348837209e-07,
+      "loss": 0.0017,
+      "step": 7606
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.1212790697674417e-07,
+      "loss": 0.0012,
+      "step": 7607
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.1211627906976745e-07,
+      "loss": 0.0018,
+      "step": 7608
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.121046511627907e-07,
+      "loss": 0.0018,
+      "step": 7609
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.1209302325581395e-07,
+      "loss": 0.0019,
+      "step": 7610
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.120813953488372e-07,
+      "loss": 0.002,
+      "step": 7611
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.1206976744186048e-07,
+      "loss": 0.0024,
+      "step": 7612
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.120581395348837e-07,
+      "loss": 0.3107,
+      "step": 7613
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.1204651162790697e-07,
+      "loss": 0.0012,
+      "step": 7614
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.120348837209302e-07,
+      "loss": 0.0022,
+      "step": 7615
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.120232558139535e-07,
+      "loss": 0.0073,
+      "step": 7616
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.1201162790697673e-07,
+      "loss": 0.0021,
+      "step": 7617
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.12e-07,
+      "loss": 0.0022,
+      "step": 7618
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.1198837209302323e-07,
+      "loss": 0.0019,
+      "step": 7619
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.119767441860465e-07,
+      "loss": 0.0011,
+      "step": 7620
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.1196511627906978e-07,
+      "loss": 0.0012,
+      "step": 7621
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.11953488372093e-07,
+      "loss": 0.0056,
+      "step": 7622
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.1194186046511627e-07,
+      "loss": 0.0023,
+      "step": 7623
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.119302325581395e-07,
+      "loss": 1.0939,
+      "step": 7624
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.119186046511628e-07,
+      "loss": 0.0018,
+      "step": 7625
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.1190697674418603e-07,
+      "loss": 0.0012,
+      "step": 7626
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.118953488372093e-07,
+      "loss": 0.002,
+      "step": 7627
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.1188372093023253e-07,
+      "loss": 0.0017,
+      "step": 7628
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.1187209302325582e-07,
+      "loss": 0.0012,
+      "step": 7629
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.1186046511627905e-07,
+      "loss": 0.0023,
+      "step": 7630
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.1184883720930232e-07,
+      "loss": 0.0013,
+      "step": 7631
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.1183720930232555e-07,
+      "loss": 0.0016,
+      "step": 7632
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.1182558139534884e-07,
+      "loss": 0.0027,
+      "step": 7633
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.118139534883721e-07,
+      "loss": 0.0028,
+      "step": 7634
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.1180232558139534e-07,
+      "loss": 0.0022,
+      "step": 7635
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.1179069767441857e-07,
+      "loss": 0.0022,
+      "step": 7636
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.1177906976744186e-07,
+      "loss": 0.0017,
+      "step": 7637
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.1176744186046512e-07,
+      "loss": 0.002,
+      "step": 7638
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.1175581395348836e-07,
+      "loss": 0.0784,
+      "step": 7639
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.1174418604651162e-07,
+      "loss": 0.0015,
+      "step": 7640
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.1173255813953488e-07,
+      "loss": 0.0021,
+      "step": 7641
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.1172093023255814e-07,
+      "loss": 0.0026,
+      "step": 7642
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.1170930232558138e-07,
+      "loss": 0.0019,
+      "step": 7643
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.1169767441860464e-07,
+      "loss": 0.0017,
+      "step": 7644
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.1168604651162793e-07,
+      "loss": 0.0017,
+      "step": 7645
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.1167441860465116e-07,
+      "loss": 0.0025,
+      "step": 7646
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.116627906976744e-07,
+      "loss": 0.0016,
+      "step": 7647
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.1165116279069766e-07,
+      "loss": 0.0019,
+      "step": 7648
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.116395348837209e-07,
+      "loss": 0.0016,
+      "step": 7649
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.1162790697674418e-07,
+      "loss": 0.0029,
+      "step": 7650
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.1161627906976744e-07,
+      "loss": 0.0017,
+      "step": 7651
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.1160465116279068e-07,
+      "loss": 0.0016,
+      "step": 7652
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.1159302325581394e-07,
+      "loss": 0.018,
+      "step": 7653
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.115813953488372e-07,
+      "loss": 0.0021,
+      "step": 7654
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.1156976744186046e-07,
+      "loss": 0.0013,
+      "step": 7655
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.115581395348837e-07,
+      "loss": 0.0016,
+      "step": 7656
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.1154651162790696e-07,
+      "loss": 0.7867,
+      "step": 7657
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.1153488372093022e-07,
+      "loss": 0.0011,
+      "step": 7658
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.1152325581395348e-07,
+      "loss": 0.0017,
+      "step": 7659
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.1151162790697672e-07,
+      "loss": 0.0015,
+      "step": 7660
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.1149999999999998e-07,
+      "loss": 0.0015,
+      "step": 7661
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.1148837209302327e-07,
+      "loss": 0.0014,
+      "step": 7662
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.114767441860465e-07,
+      "loss": 0.0018,
+      "step": 7663
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.1146511627906977e-07,
+      "loss": 0.0019,
+      "step": 7664
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.11453488372093e-07,
+      "loss": 0.0016,
+      "step": 7665
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.114418604651163e-07,
+      "loss": 0.5554,
+      "step": 7666
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.1143023255813953e-07,
+      "loss": 0.0012,
+      "step": 7667
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.114186046511628e-07,
+      "loss": 0.0019,
+      "step": 7668
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.1140697674418602e-07,
+      "loss": 0.002,
+      "step": 7669
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.113953488372093e-07,
+      "loss": 0.0013,
+      "step": 7670
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.1138372093023255e-07,
+      "loss": 0.0019,
+      "step": 7671
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.113720930232558e-07,
+      "loss": 0.0015,
+      "step": 7672
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.1136046511627904e-07,
+      "loss": 0.0017,
+      "step": 7673
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.113488372093023e-07,
+      "loss": 0.0013,
+      "step": 7674
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.113372093023256e-07,
+      "loss": 0.0018,
+      "step": 7675
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.1132558139534883e-07,
+      "loss": 0.0019,
+      "step": 7676
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.113139534883721e-07,
+      "loss": 0.2245,
+      "step": 7677
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.1130232558139532e-07,
+      "loss": 0.0021,
+      "step": 7678
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.112906976744186e-07,
+      "loss": 0.0155,
+      "step": 7679
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.1127906976744185e-07,
+      "loss": 0.115,
+      "step": 7680
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.112674418604651e-07,
+      "loss": 0.0115,
+      "step": 7681
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.1125581395348835e-07,
+      "loss": 0.0018,
+      "step": 7682
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.1124418604651163e-07,
+      "loss": 0.0021,
+      "step": 7683
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.1123255813953487e-07,
+      "loss": 0.0019,
+      "step": 7684
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.1122093023255813e-07,
+      "loss": 0.0033,
+      "step": 7685
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.1120930232558137e-07,
+      "loss": 0.0015,
+      "step": 7686
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.1119767441860465e-07,
+      "loss": 0.1622,
+      "step": 7687
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.1118604651162792e-07,
+      "loss": 0.0011,
+      "step": 7688
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.1117441860465115e-07,
+      "loss": 2.1974,
+      "step": 7689
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.1116279069767439e-07,
+      "loss": 0.0018,
+      "step": 7690
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.1115116279069767e-07,
+      "loss": 0.002,
+      "step": 7691
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.1113953488372094e-07,
+      "loss": 0.1072,
+      "step": 7692
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.1112790697674417e-07,
+      "loss": 2.0901,
+      "step": 7693
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.1111627906976743e-07,
+      "loss": 0.0012,
+      "step": 7694
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.111046511627907e-07,
+      "loss": 0.0017,
+      "step": 7695
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.1109302325581396e-07,
+      "loss": 0.0018,
+      "step": 7696
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.110813953488372e-07,
+      "loss": 0.0014,
+      "step": 7697
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.1106976744186045e-07,
+      "loss": 0.0012,
+      "step": 7698
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.1105813953488371e-07,
+      "loss": 1.2898,
+      "step": 7699
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.1104651162790698e-07,
+      "loss": 0.0022,
+      "step": 7700
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.110348837209302e-07,
+      "loss": 0.0076,
+      "step": 7701
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.1102325581395347e-07,
+      "loss": 0.0016,
+      "step": 7702
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.110116279069767e-07,
+      "loss": 0.0039,
+      "step": 7703
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.11e-07,
+      "loss": 0.0016,
+      "step": 7704
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.1098837209302326e-07,
+      "loss": 0.0018,
+      "step": 7705
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.109767441860465e-07,
+      "loss": 0.0013,
+      "step": 7706
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.1096511627906976e-07,
+      "loss": 0.0015,
+      "step": 7707
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.1095348837209302e-07,
+      "loss": 0.0027,
+      "step": 7708
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.1094186046511628e-07,
+      "loss": 0.0548,
+      "step": 7709
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.1093023255813951e-07,
+      "loss": 0.0017,
+      "step": 7710
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.1091860465116278e-07,
+      "loss": 0.1979,
+      "step": 7711
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.1090697674418604e-07,
+      "loss": 0.0016,
+      "step": 7712
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.108953488372093e-07,
+      "loss": 0.0021,
+      "step": 7713
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.1088372093023253e-07,
+      "loss": 0.0014,
+      "step": 7714
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.108720930232558e-07,
+      "loss": 0.0015,
+      "step": 7715
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.1086046511627908e-07,
+      "loss": 0.0015,
+      "step": 7716
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.1084883720930232e-07,
+      "loss": 0.0043,
+      "step": 7717
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.1083720930232558e-07,
+      "loss": 0.2963,
+      "step": 7718
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.1082558139534882e-07,
+      "loss": 0.0015,
+      "step": 7719
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.108139534883721e-07,
+      "loss": 0.0012,
+      "step": 7720
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.1080232558139534e-07,
+      "loss": 0.0019,
+      "step": 7721
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.107906976744186e-07,
+      "loss": 0.0021,
+      "step": 7722
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.1077906976744184e-07,
+      "loss": 0.0027,
+      "step": 7723
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.1076744186046513e-07,
+      "loss": 0.0018,
+      "step": 7724
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.1075581395348836e-07,
+      "loss": 0.0151,
+      "step": 7725
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.1074418604651162e-07,
+      "loss": 0.0018,
+      "step": 7726
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.1073255813953486e-07,
+      "loss": 0.0021,
+      "step": 7727
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.1072093023255812e-07,
+      "loss": 0.0013,
+      "step": 7728
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.107093023255814e-07,
+      "loss": 0.0024,
+      "step": 7729
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.1069767441860464e-07,
+      "loss": 0.0019,
+      "step": 7730
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.1068604651162788e-07,
+      "loss": 0.0022,
+      "step": 7731
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.1067441860465114e-07,
+      "loss": 0.002,
+      "step": 7732
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.1066279069767443e-07,
+      "loss": 0.0023,
+      "step": 7733
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.1065116279069766e-07,
+      "loss": 0.0014,
+      "step": 7734
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.1063953488372092e-07,
+      "loss": 0.0011,
+      "step": 7735
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.1062790697674416e-07,
+      "loss": 0.0037,
+      "step": 7736
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.1061627906976745e-07,
+      "loss": 0.002,
+      "step": 7737
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.1060465116279068e-07,
+      "loss": 0.0013,
+      "step": 7738
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.1059302325581394e-07,
+      "loss": 0.0026,
+      "step": 7739
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.1058139534883718e-07,
+      "loss": 0.0017,
+      "step": 7740
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.1056976744186047e-07,
+      "loss": 0.0021,
+      "step": 7741
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.105581395348837e-07,
+      "loss": 0.0017,
+      "step": 7742
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.1054651162790697e-07,
+      "loss": 0.0017,
+      "step": 7743
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.105348837209302e-07,
+      "loss": 0.0019,
+      "step": 7744
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.105232558139535e-07,
+      "loss": 0.0018,
+      "step": 7745
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.1051162790697675e-07,
+      "loss": 0.0018,
+      "step": 7746
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.1049999999999999e-07,
+      "loss": 0.0013,
+      "step": 7747
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.1048837209302325e-07,
+      "loss": 0.0018,
+      "step": 7748
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.104767441860465e-07,
+      "loss": 0.0022,
+      "step": 7749
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.1046511627906977e-07,
+      "loss": 0.0072,
+      "step": 7750
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.10453488372093e-07,
+      "loss": 0.0014,
+      "step": 7751
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.1044186046511627e-07,
+      "loss": 0.0019,
+      "step": 7752
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.104302325581395e-07,
+      "loss": 1.325,
+      "step": 7753
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.104186046511628e-07,
+      "loss": 0.0016,
+      "step": 7754
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.1040697674418603e-07,
+      "loss": 0.0013,
+      "step": 7755
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.9845261121856866,
+      "eval_loss": 0.15089702606201172,
+      "eval_runtime": 1.5494,
+      "eval_samples_per_second": 667.335,
+      "eval_steps_per_second": 10.972,
+      "step": 7755
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.103953488372093e-07,
+      "loss": 0.0012,
+      "step": 7756
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.1038372093023252e-07,
+      "loss": 0.0021,
+      "step": 7757
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.103720930232558e-07,
+      "loss": 0.0023,
+      "step": 7758
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.1036046511627907e-07,
+      "loss": 1.6497,
+      "step": 7759
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.103488372093023e-07,
+      "loss": 0.0015,
+      "step": 7760
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.1033720930232557e-07,
+      "loss": 0.0017,
+      "step": 7761
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.1032558139534883e-07,
+      "loss": 0.0149,
+      "step": 7762
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.103139534883721e-07,
+      "loss": 0.0035,
+      "step": 7763
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.1030232558139533e-07,
+      "loss": 0.0025,
+      "step": 7764
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.102906976744186e-07,
+      "loss": 0.0024,
+      "step": 7765
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.1027906976744185e-07,
+      "loss": 0.0021,
+      "step": 7766
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.1026744186046511e-07,
+      "loss": 0.0025,
+      "step": 7767
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.1025581395348835e-07,
+      "loss": 0.0017,
+      "step": 7768
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.102441860465116e-07,
+      "loss": 0.0015,
+      "step": 7769
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.102325581395349e-07,
+      "loss": 0.0016,
+      "step": 7770
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.1022093023255813e-07,
+      "loss": 0.0018,
+      "step": 7771
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.102093023255814e-07,
+      "loss": 0.0015,
+      "step": 7772
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.1019767441860463e-07,
+      "loss": 0.0029,
+      "step": 7773
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.1018604651162792e-07,
+      "loss": 0.0033,
+      "step": 7774
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.1017441860465115e-07,
+      "loss": 0.001,
+      "step": 7775
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.1016279069767442e-07,
+      "loss": 0.0021,
+      "step": 7776
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.1015116279069765e-07,
+      "loss": 2.8814,
+      "step": 7777
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.1013953488372094e-07,
+      "loss": 0.0013,
+      "step": 7778
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.1012790697674418e-07,
+      "loss": 0.0016,
+      "step": 7779
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.1011627906976744e-07,
+      "loss": 0.0018,
+      "step": 7780
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.1010465116279067e-07,
+      "loss": 0.0016,
+      "step": 7781
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.1009302325581393e-07,
+      "loss": 0.0015,
+      "step": 7782
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.1008139534883722e-07,
+      "loss": 0.0019,
+      "step": 7783
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.1006976744186046e-07,
+      "loss": 0.0021,
+      "step": 7784
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.100581395348837e-07,
+      "loss": 0.0019,
+      "step": 7785
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.1004651162790695e-07,
+      "loss": 0.0018,
+      "step": 7786
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.1003488372093024e-07,
+      "loss": 0.0014,
+      "step": 7787
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.1002325581395348e-07,
+      "loss": 0.0014,
+      "step": 7788
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.1001162790697674e-07,
+      "loss": 0.0014,
+      "step": 7789
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.0999999999999997e-07,
+      "loss": 0.0039,
+      "step": 7790
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.0998837209302326e-07,
+      "loss": 0.0023,
+      "step": 7791
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.099767441860465e-07,
+      "loss": 0.0016,
+      "step": 7792
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.0996511627906976e-07,
+      "loss": 0.002,
+      "step": 7793
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.09953488372093e-07,
+      "loss": 0.0018,
+      "step": 7794
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.0994186046511628e-07,
+      "loss": 0.0015,
+      "step": 7795
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.0993023255813952e-07,
+      "loss": 0.415,
+      "step": 7796
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.0991860465116278e-07,
+      "loss": 0.0017,
+      "step": 7797
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.0990697674418602e-07,
+      "loss": 0.0015,
+      "step": 7798
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 2.098953488372093e-07,
+      "loss": 0.0016,
+      "step": 7799
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 2.0988372093023257e-07,
+      "loss": 0.0029,
+      "step": 7800
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 2.098720930232558e-07,
+      "loss": 0.0053,
+      "step": 7801
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 2.0986046511627906e-07,
+      "loss": 0.0018,
+      "step": 7802
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 2.0984883720930232e-07,
+      "loss": 1.1737,
+      "step": 7803
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 2.0983720930232559e-07,
+      "loss": 0.0012,
+      "step": 7804
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2.0982558139534882e-07,
+      "loss": 0.0017,
+      "step": 7805
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2.0981395348837208e-07,
+      "loss": 0.0041,
+      "step": 7806
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2.0980232558139532e-07,
+      "loss": 0.0017,
+      "step": 7807
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2.097906976744186e-07,
+      "loss": 0.0015,
+      "step": 7808
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2.0977906976744184e-07,
+      "loss": 0.1093,
+      "step": 7809
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 2.097674418604651e-07,
+      "loss": 0.0016,
+      "step": 7810
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 2.0975581395348834e-07,
+      "loss": 0.0019,
+      "step": 7811
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 2.0974418604651163e-07,
+      "loss": 0.0575,
+      "step": 7812
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 2.097325581395349e-07,
+      "loss": 0.0018,
+      "step": 7813
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 2.0972093023255812e-07,
+      "loss": 0.002,
+      "step": 7814
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 2.0970930232558139e-07,
+      "loss": 0.0022,
+      "step": 7815
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 2.0969767441860465e-07,
+      "loss": 0.0079,
+      "step": 7816
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 2.096860465116279e-07,
+      "loss": 0.0016,
+      "step": 7817
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 2.0967441860465114e-07,
+      "loss": 0.0016,
+      "step": 7818
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 2.096627906976744e-07,
+      "loss": 0.0338,
+      "step": 7819
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 2.0965116279069767e-07,
+      "loss": 0.0019,
+      "step": 7820
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 2.0963953488372093e-07,
+      "loss": 0.0018,
+      "step": 7821
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 2.0962790697674416e-07,
+      "loss": 0.0032,
+      "step": 7822
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 2.0961627906976743e-07,
+      "loss": 0.0014,
+      "step": 7823
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 2.0960465116279071e-07,
+      "loss": 0.004,
+      "step": 7824
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 2.0959302325581395e-07,
+      "loss": 0.0022,
+      "step": 7825
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 2.095813953488372e-07,
+      "loss": 1.0613,
+      "step": 7826
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 2.0956976744186045e-07,
+      "loss": 0.0023,
+      "step": 7827
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 2.0955813953488373e-07,
+      "loss": 0.0019,
+      "step": 7828
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 2.0954651162790697e-07,
+      "loss": 0.0014,
+      "step": 7829
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.0953488372093023e-07,
+      "loss": 0.002,
+      "step": 7830
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.0952325581395347e-07,
+      "loss": 0.0022,
+      "step": 7831
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.0951162790697675e-07,
+      "loss": 0.0019,
+      "step": 7832
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.095e-07,
+      "loss": 0.0016,
+      "step": 7833
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.0948837209302325e-07,
+      "loss": 0.0013,
+      "step": 7834
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.094767441860465e-07,
+      "loss": 0.0019,
+      "step": 7835
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 2.0946511627906975e-07,
+      "loss": 0.0018,
+      "step": 7836
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 2.09453488372093e-07,
+      "loss": 0.0024,
+      "step": 7837
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 2.0944186046511627e-07,
+      "loss": 0.0018,
+      "step": 7838
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 2.094302325581395e-07,
+      "loss": 0.0017,
+      "step": 7839
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 2.0941860465116277e-07,
+      "loss": 0.0011,
+      "step": 7840
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 2.0940697674418606e-07,
+      "loss": 0.6461,
+      "step": 7841
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 2.093953488372093e-07,
+      "loss": 0.0017,
+      "step": 7842
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 2.0938372093023255e-07,
+      "loss": 0.0021,
+      "step": 7843
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 2.093720930232558e-07,
+      "loss": 1.877,
+      "step": 7844
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 2.0936046511627908e-07,
+      "loss": 0.0012,
+      "step": 7845
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 2.093488372093023e-07,
+      "loss": 0.0021,
+      "step": 7846
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 2.0933720930232557e-07,
+      "loss": 0.0012,
+      "step": 7847
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 2.093255813953488e-07,
+      "loss": 0.0106,
+      "step": 7848
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 2.093139534883721e-07,
+      "loss": 0.0019,
+      "step": 7849
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 2.0930232558139533e-07,
+      "loss": 0.0018,
+      "step": 7850
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 2.092906976744186e-07,
+      "loss": 0.0017,
+      "step": 7851
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 2.0927906976744183e-07,
+      "loss": 0.0019,
+      "step": 7852
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 2.0926744186046512e-07,
+      "loss": 0.0032,
+      "step": 7853
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 2.0925581395348838e-07,
+      "loss": 0.0017,
+      "step": 7854
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 2.0924418604651162e-07,
+      "loss": 0.0033,
+      "step": 7855
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 2.0923255813953488e-07,
+      "loss": 0.0015,
+      "step": 7856
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 2.0922093023255814e-07,
+      "loss": 0.0026,
+      "step": 7857
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 2.092093023255814e-07,
+      "loss": 0.0013,
+      "step": 7858
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 2.0919767441860464e-07,
+      "loss": 0.0033,
+      "step": 7859
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 2.091860465116279e-07,
+      "loss": 0.0903,
+      "step": 7860
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.0917441860465113e-07,
+      "loss": 0.0023,
+      "step": 7861
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.0916279069767442e-07,
+      "loss": 1.7209,
+      "step": 7862
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.0915116279069766e-07,
+      "loss": 0.0021,
+      "step": 7863
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.0913953488372092e-07,
+      "loss": 0.0019,
+      "step": 7864
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.0912790697674415e-07,
+      "loss": 0.0014,
+      "step": 7865
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.0911627906976744e-07,
+      "loss": 0.8213,
+      "step": 7866
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 2.091046511627907e-07,
+      "loss": 0.0011,
+      "step": 7867
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 2.0909302325581394e-07,
+      "loss": 0.0021,
+      "step": 7868
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 2.0908139534883717e-07,
+      "loss": 0.0015,
+      "step": 7869
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 2.0906976744186046e-07,
+      "loss": 0.0029,
+      "step": 7870
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 2.0905813953488372e-07,
+      "loss": 0.0014,
+      "step": 7871
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 2.0904651162790696e-07,
+      "loss": 0.0015,
+      "step": 7872
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 2.0903488372093022e-07,
+      "loss": 0.0019,
+      "step": 7873
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 2.0902325581395348e-07,
+      "loss": 0.0013,
+      "step": 7874
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 2.0901162790697674e-07,
+      "loss": 0.0014,
+      "step": 7875
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 2.0899999999999998e-07,
+      "loss": 0.0015,
+      "step": 7876
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 2.0898837209302324e-07,
+      "loss": 0.0015,
+      "step": 7877
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 2.0897674418604653e-07,
+      "loss": 0.0012,
+      "step": 7878
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 2.0896511627906976e-07,
+      "loss": 0.0025,
+      "step": 7879
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 2.08953488372093e-07,
+      "loss": 0.0019,
+      "step": 7880
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 2.0894186046511626e-07,
+      "loss": 0.0016,
+      "step": 7881
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 2.0893023255813955e-07,
+      "loss": 0.0028,
+      "step": 7882
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 2.0891860465116278e-07,
+      "loss": 0.0015,
+      "step": 7883
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 2.0890697674418605e-07,
+      "loss": 0.0017,
+      "step": 7884
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 2.0889534883720928e-07,
+      "loss": 0.0026,
+      "step": 7885
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 2.0888372093023254e-07,
+      "loss": 0.0013,
+      "step": 7886
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.088720930232558e-07,
+      "loss": 0.0022,
+      "step": 7887
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.0886046511627907e-07,
+      "loss": 0.0013,
+      "step": 7888
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.088488372093023e-07,
+      "loss": 0.0016,
+      "step": 7889
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.0883720930232556e-07,
+      "loss": 0.0013,
+      "step": 7890
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.0882558139534883e-07,
+      "loss": 0.0014,
+      "step": 7891
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.088139534883721e-07,
+      "loss": 0.0016,
+      "step": 7892
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 2.0880232558139532e-07,
+      "loss": 0.0015,
+      "step": 7893
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 2.0879069767441858e-07,
+      "loss": 0.0047,
+      "step": 7894
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 2.0877906976744187e-07,
+      "loss": 0.0016,
+      "step": 7895
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 2.087674418604651e-07,
+      "loss": 0.0015,
+      "step": 7896
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 2.0875581395348837e-07,
+      "loss": 0.0015,
+      "step": 7897
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 2.087441860465116e-07,
+      "loss": 0.0021,
+      "step": 7898
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 2.087325581395349e-07,
+      "loss": 0.004,
+      "step": 7899
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 2.0872093023255813e-07,
+      "loss": 0.002,
+      "step": 7900
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 2.087093023255814e-07,
+      "loss": 0.0016,
+      "step": 7901
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 2.0869767441860462e-07,
+      "loss": 0.0018,
+      "step": 7902
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 2.086860465116279e-07,
+      "loss": 0.7872,
+      "step": 7903
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 2.0867441860465115e-07,
+      "loss": 0.0017,
+      "step": 7904
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 2.086627906976744e-07,
+      "loss": 0.0016,
+      "step": 7905
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 2.0865116279069764e-07,
+      "loss": 0.0016,
+      "step": 7906
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 2.0863953488372093e-07,
+      "loss": 0.0018,
+      "step": 7907
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 2.086279069767442e-07,
+      "loss": 0.0015,
+      "step": 7908
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 2.0861627906976743e-07,
+      "loss": 0.0019,
+      "step": 7909
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 2.086046511627907e-07,
+      "loss": 0.0025,
+      "step": 7910
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 2.0859302325581395e-07,
+      "loss": 0.0012,
+      "step": 7911
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 2.0858139534883722e-07,
+      "loss": 0.0014,
+      "step": 7912
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 2.0856976744186045e-07,
+      "loss": 0.0021,
+      "step": 7913
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 2.085581395348837e-07,
+      "loss": 0.0026,
+      "step": 7914
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 2.0854651162790695e-07,
+      "loss": 0.0018,
+      "step": 7915
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 2.0853488372093024e-07,
+      "loss": 0.0022,
+      "step": 7916
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 2.0852325581395347e-07,
+      "loss": 0.0025,
+      "step": 7917
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.0851162790697673e-07,
+      "loss": 0.002,
+      "step": 7918
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.0849999999999997e-07,
+      "loss": 0.0028,
+      "step": 7919
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.0848837209302326e-07,
+      "loss": 0.0499,
+      "step": 7920
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.0847674418604652e-07,
+      "loss": 0.0021,
+      "step": 7921
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.0846511627906975e-07,
+      "loss": 0.003,
+      "step": 7922
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.08453488372093e-07,
+      "loss": 0.0015,
+      "step": 7923
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 2.0844186046511628e-07,
+      "loss": 0.0021,
+      "step": 7924
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 2.0843023255813954e-07,
+      "loss": 0.0011,
+      "step": 7925
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 2.0841860465116277e-07,
+      "loss": 0.0018,
+      "step": 7926
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 2.0840697674418604e-07,
+      "loss": 0.0013,
+      "step": 7927
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 2.083953488372093e-07,
+      "loss": 0.0015,
+      "step": 7928
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 2.0838372093023256e-07,
+      "loss": 0.0034,
+      "step": 7929
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 2.083720930232558e-07,
+      "loss": 0.0023,
+      "step": 7930
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 2.0836046511627906e-07,
+      "loss": 0.0021,
+      "step": 7931
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 2.0834883720930234e-07,
+      "loss": 0.0041,
+      "step": 7932
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 2.0833720930232558e-07,
+      "loss": 0.0021,
+      "step": 7933
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 2.0832558139534881e-07,
+      "loss": 0.0021,
+      "step": 7934
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 2.0831395348837208e-07,
+      "loss": 0.0278,
+      "step": 7935
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 2.0830232558139536e-07,
+      "loss": 0.0016,
+      "step": 7936
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 2.082906976744186e-07,
+      "loss": 0.0014,
+      "step": 7937
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 2.0827906976744186e-07,
+      "loss": 0.0017,
+      "step": 7938
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 2.082674418604651e-07,
+      "loss": 0.0017,
+      "step": 7939
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 2.0825581395348836e-07,
+      "loss": 0.0014,
+      "step": 7940
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 2.0824418604651162e-07,
+      "loss": 0.0035,
+      "step": 7941
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 2.0823255813953488e-07,
+      "loss": 0.0018,
+      "step": 7942
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 2.0822093023255812e-07,
+      "loss": 0.0022,
+      "step": 7943
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 2.0820930232558138e-07,
+      "loss": 0.0022,
+      "step": 7944
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 2.0819767441860464e-07,
+      "loss": 0.0029,
+      "step": 7945
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 2.081860465116279e-07,
+      "loss": 0.0016,
+      "step": 7946
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 2.0817441860465114e-07,
+      "loss": 0.0019,
+      "step": 7947
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 2.081627906976744e-07,
+      "loss": 0.0014,
+      "step": 7948
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 2.081511627906977e-07,
+      "loss": 0.0059,
+      "step": 7949
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 2.0813953488372092e-07,
+      "loss": 0.0014,
+      "step": 7950
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 2.0812790697674418e-07,
+      "loss": 0.0017,
+      "step": 7951
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 2.0811627906976742e-07,
+      "loss": 0.0021,
+      "step": 7952
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 2.081046511627907e-07,
+      "loss": 0.0013,
+      "step": 7953
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 2.0809302325581394e-07,
+      "loss": 0.0021,
+      "step": 7954
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 2.080813953488372e-07,
+      "loss": 0.0018,
+      "step": 7955
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 2.0806976744186044e-07,
+      "loss": 0.0021,
+      "step": 7956
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 2.0805813953488373e-07,
+      "loss": 0.0012,
+      "step": 7957
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 2.0804651162790696e-07,
+      "loss": 0.0022,
+      "step": 7958
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 2.0803488372093022e-07,
+      "loss": 0.0017,
+      "step": 7959
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 2.0802325581395346e-07,
+      "loss": 0.0016,
+      "step": 7960
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 2.0801162790697675e-07,
+      "loss": 0.0023,
+      "step": 7961
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 2.08e-07,
+      "loss": 0.0011,
+      "step": 7962
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 2.0798837209302324e-07,
+      "loss": 0.0104,
+      "step": 7963
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 2.0797674418604648e-07,
+      "loss": 0.0017,
+      "step": 7964
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2.0796511627906977e-07,
+      "loss": 0.0015,
+      "step": 7965
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2.0795348837209303e-07,
+      "loss": 0.0014,
+      "step": 7966
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2.0794186046511627e-07,
+      "loss": 0.001,
+      "step": 7967
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2.0793023255813953e-07,
+      "loss": 0.0014,
+      "step": 7968
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2.0791860465116276e-07,
+      "loss": 0.0014,
+      "step": 7969
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 2.0790697674418605e-07,
+      "loss": 0.0014,
+      "step": 7970
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 2.0789534883720929e-07,
+      "loss": 0.0014,
+      "step": 7971
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 2.0788372093023255e-07,
+      "loss": 0.0047,
+      "step": 7972
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 2.0787209302325578e-07,
+      "loss": 0.0016,
+      "step": 7973
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 2.0786046511627907e-07,
+      "loss": 0.1645,
+      "step": 7974
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 2.078488372093023e-07,
+      "loss": 0.0012,
+      "step": 7975
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 2.0783720930232557e-07,
+      "loss": 0.0017,
+      "step": 7976
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 2.078255813953488e-07,
+      "loss": 0.0048,
+      "step": 7977
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 2.078139534883721e-07,
+      "loss": 0.0023,
+      "step": 7978
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 2.0780232558139535e-07,
+      "loss": 0.0023,
+      "step": 7979
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.077906976744186e-07,
+      "loss": 0.0012,
+      "step": 7980
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.0777906976744185e-07,
+      "loss": 0.0011,
+      "step": 7981
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.077674418604651e-07,
+      "loss": 0.0015,
+      "step": 7982
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.0775581395348837e-07,
+      "loss": 0.0021,
+      "step": 7983
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.077441860465116e-07,
+      "loss": 0.0023,
+      "step": 7984
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.0773255813953487e-07,
+      "loss": 0.0021,
+      "step": 7985
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 2.0772093023255813e-07,
+      "loss": 0.0024,
+      "step": 7986
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 2.077093023255814e-07,
+      "loss": 0.0019,
+      "step": 7987
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 2.0769767441860463e-07,
+      "loss": 0.0023,
+      "step": 7988
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 2.076860465116279e-07,
+      "loss": 0.0019,
+      "step": 7989
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 2.0767441860465118e-07,
+      "loss": 0.0012,
+      "step": 7990
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2.0766279069767441e-07,
+      "loss": 0.0015,
+      "step": 7991
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2.0765116279069768e-07,
+      "loss": 0.0017,
+      "step": 7992
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2.076395348837209e-07,
+      "loss": 0.0017,
+      "step": 7993
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2.0762790697674417e-07,
+      "loss": 0.0022,
+      "step": 7994
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2.0761627906976743e-07,
+      "loss": 0.0015,
+      "step": 7995
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 2.076046511627907e-07,
+      "loss": 0.0016,
+      "step": 7996
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 2.0759302325581393e-07,
+      "loss": 0.1032,
+      "step": 7997
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 2.075813953488372e-07,
+      "loss": 0.0012,
+      "step": 7998
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 2.0756976744186045e-07,
+      "loss": 0.0157,
+      "step": 7999
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 2.0755813953488372e-07,
+      "loss": 0.0014,
+      "step": 8000
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 2.0754651162790695e-07,
+      "loss": 0.0046,
+      "step": 8001
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 2.0753488372093021e-07,
+      "loss": 0.0014,
+      "step": 8002
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 2.075232558139535e-07,
+      "loss": 0.0015,
+      "step": 8003
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 2.0751162790697674e-07,
+      "loss": 0.0019,
+      "step": 8004
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 2.075e-07,
+      "loss": 0.0022,
+      "step": 8005
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 2.0748837209302323e-07,
+      "loss": 0.0015,
+      "step": 8006
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 2.0747674418604652e-07,
+      "loss": 0.0015,
+      "step": 8007
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 2.0746511627906976e-07,
+      "loss": 0.0018,
+      "step": 8008
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 2.0745348837209302e-07,
+      "loss": 3.155,
+      "step": 8009
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 2.0744186046511625e-07,
+      "loss": 0.0014,
+      "step": 8010
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.0743023255813954e-07,
+      "loss": 0.0012,
+      "step": 8011
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.0741860465116278e-07,
+      "loss": 0.0017,
+      "step": 8012
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.0740697674418604e-07,
+      "loss": 0.0017,
+      "step": 8013
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.0739534883720927e-07,
+      "loss": 0.0331,
+      "step": 8014
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.0738372093023256e-07,
+      "loss": 0.0032,
+      "step": 8015
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.0737209302325582e-07,
+      "loss": 0.0326,
+      "step": 8016
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 2.0736046511627906e-07,
+      "loss": 0.0016,
+      "step": 8017
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 2.073488372093023e-07,
+      "loss": 0.0019,
+      "step": 8018
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 2.0733720930232558e-07,
+      "loss": 0.0059,
+      "step": 8019
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 2.0732558139534884e-07,
+      "loss": 0.003,
+      "step": 8020
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 2.0731395348837208e-07,
+      "loss": 0.0016,
+      "step": 8021
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 2.0730232558139534e-07,
+      "loss": 0.0013,
+      "step": 8022
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 2.0729069767441858e-07,
+      "loss": 0.0022,
+      "step": 8023
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 2.0727906976744187e-07,
+      "loss": 0.0013,
+      "step": 8024
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 2.072674418604651e-07,
+      "loss": 0.0013,
+      "step": 8025
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 2.0725581395348836e-07,
+      "loss": 0.0019,
+      "step": 8026
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 2.072441860465116e-07,
+      "loss": 0.0015,
+      "step": 8027
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 2.0723255813953489e-07,
+      "loss": 0.0013,
+      "step": 8028
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 2.0722093023255812e-07,
+      "loss": 0.0013,
+      "step": 8029
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 2.0720930232558138e-07,
+      "loss": 0.0026,
+      "step": 8030
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 2.0719767441860462e-07,
+      "loss": 0.0018,
+      "step": 8031
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 2.071860465116279e-07,
+      "loss": 0.0016,
+      "step": 8032
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 2.0717441860465117e-07,
+      "loss": 0.2398,
+      "step": 8033
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 2.071627906976744e-07,
+      "loss": 0.0017,
+      "step": 8034
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 2.0715116279069766e-07,
+      "loss": 0.0017,
+      "step": 8035
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 2.0713953488372093e-07,
+      "loss": 0.0037,
+      "step": 8036
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 2.071279069767442e-07,
+      "loss": 0.0014,
+      "step": 8037
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 2.0711627906976742e-07,
+      "loss": 0.0021,
+      "step": 8038
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 2.0710465116279068e-07,
+      "loss": 0.0016,
+      "step": 8039
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 2.0709302325581395e-07,
+      "loss": 0.0016,
+      "step": 8040
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 2.070813953488372e-07,
+      "loss": 0.0015,
+      "step": 8041
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2.0706976744186044e-07,
+      "loss": 0.0011,
+      "step": 8042
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2.070581395348837e-07,
+      "loss": 0.0022,
+      "step": 8043
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2.07046511627907e-07,
+      "loss": 0.0013,
+      "step": 8044
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2.0703488372093023e-07,
+      "loss": 0.0014,
+      "step": 8045
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2.070232558139535e-07,
+      "loss": 0.0021,
+      "step": 8046
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2.0701162790697673e-07,
+      "loss": 0.0022,
+      "step": 8047
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 2.07e-07,
+      "loss": 0.002,
+      "step": 8048
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 2.0698837209302325e-07,
+      "loss": 0.1773,
+      "step": 8049
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 2.069767441860465e-07,
+      "loss": 0.0012,
+      "step": 8050
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 2.0696511627906975e-07,
+      "loss": 0.0014,
+      "step": 8051
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 2.06953488372093e-07,
+      "loss": 0.2066,
+      "step": 8052
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 2.0694186046511627e-07,
+      "loss": 0.0012,
+      "step": 8053
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 2.0693023255813953e-07,
+      "loss": 0.002,
+      "step": 8054
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 2.0691860465116277e-07,
+      "loss": 0.0027,
+      "step": 8055
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 2.0690697674418603e-07,
+      "loss": 0.0042,
+      "step": 8056
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 2.0689534883720932e-07,
+      "loss": 0.0013,
+      "step": 8057
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 2.0688372093023255e-07,
+      "loss": 0.0012,
+      "step": 8058
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 2.0687209302325581e-07,
+      "loss": 0.0019,
+      "step": 8059
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 2.0686046511627905e-07,
+      "loss": 0.0013,
+      "step": 8060
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 2.0684883720930234e-07,
+      "loss": 0.0026,
+      "step": 8061
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 2.0683720930232557e-07,
+      "loss": 0.0015,
+      "step": 8062
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 2.0682558139534883e-07,
+      "loss": 0.0011,
+      "step": 8063
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 2.0681395348837207e-07,
+      "loss": 0.0014,
+      "step": 8064
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 2.0680232558139536e-07,
+      "loss": 0.0025,
+      "step": 8065
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 2.067906976744186e-07,
+      "loss": 1.7944,
+      "step": 8066
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 2.0677906976744185e-07,
+      "loss": 0.0068,
+      "step": 8067
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 2.067674418604651e-07,
+      "loss": 0.295,
+      "step": 8068
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 2.0675581395348838e-07,
+      "loss": 3.3633,
+      "step": 8069
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 2.0674418604651164e-07,
+      "loss": 0.0011,
+      "step": 8070
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 2.0673255813953487e-07,
+      "loss": 0.293,
+      "step": 8071
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 2.067209302325581e-07,
+      "loss": 0.0014,
+      "step": 8072
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2.0670930232558137e-07,
+      "loss": 0.0018,
+      "step": 8073
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2.0669767441860466e-07,
+      "loss": 0.0013,
+      "step": 8074
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2.066860465116279e-07,
+      "loss": 0.0016,
+      "step": 8075
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2.0667441860465116e-07,
+      "loss": 0.0016,
+      "step": 8076
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2.066627906976744e-07,
+      "loss": 0.0015,
+      "step": 8077
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2.0665116279069768e-07,
+      "loss": 0.0018,
+      "step": 8078
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 2.0663953488372092e-07,
+      "loss": 0.0013,
+      "step": 8079
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 2.0662790697674418e-07,
+      "loss": 0.0015,
+      "step": 8080
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 2.066162790697674e-07,
+      "loss": 0.0052,
+      "step": 8081
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 2.066046511627907e-07,
+      "loss": 0.0901,
+      "step": 8082
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 2.0659302325581394e-07,
+      "loss": 0.0016,
+      "step": 8083
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 2.065813953488372e-07,
+      "loss": 0.0018,
+      "step": 8084
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 2.0656976744186043e-07,
+      "loss": 0.0017,
+      "step": 8085
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 2.0655813953488372e-07,
+      "loss": 0.6243,
+      "step": 8086
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 2.0654651162790698e-07,
+      "loss": 0.0418,
+      "step": 8087
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 2.0653488372093022e-07,
+      "loss": 0.0021,
+      "step": 8088
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 2.0652325581395348e-07,
+      "loss": 0.0033,
+      "step": 8089
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 2.0651162790697674e-07,
+      "loss": 0.0013,
+      "step": 8090
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 2.065e-07,
+      "loss": 2.2242,
+      "step": 8091
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 2.0648837209302324e-07,
+      "loss": 0.0041,
+      "step": 8092
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 2.064767441860465e-07,
+      "loss": 0.0016,
+      "step": 8093
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 2.0646511627906976e-07,
+      "loss": 0.0016,
+      "step": 8094
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 2.0645348837209302e-07,
+      "loss": 0.0015,
+      "step": 8095
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 2.0644186046511626e-07,
+      "loss": 0.0016,
+      "step": 8096
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 2.0643023255813952e-07,
+      "loss": 0.0021,
+      "step": 8097
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 2.064186046511628e-07,
+      "loss": 0.0032,
+      "step": 8098
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 2.0640697674418604e-07,
+      "loss": 0.4649,
+      "step": 8099
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 2.063953488372093e-07,
+      "loss": 0.0016,
+      "step": 8100
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 2.0638372093023254e-07,
+      "loss": 0.0045,
+      "step": 8101
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 2.0637209302325578e-07,
+      "loss": 0.0015,
+      "step": 8102
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 2.0636046511627906e-07,
+      "loss": 0.0016,
+      "step": 8103
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.0634883720930233e-07,
+      "loss": 0.0013,
+      "step": 8104
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.0633720930232556e-07,
+      "loss": 0.002,
+      "step": 8105
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.0632558139534882e-07,
+      "loss": 0.0015,
+      "step": 8106
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.0631395348837208e-07,
+      "loss": 0.0013,
+      "step": 8107
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.0630232558139535e-07,
+      "loss": 0.0016,
+      "step": 8108
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.0629069767441858e-07,
+      "loss": 0.0016,
+      "step": 8109
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 2.0627906976744184e-07,
+      "loss": 0.0012,
+      "step": 8110
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 2.0626744186046513e-07,
+      "loss": 0.0034,
+      "step": 8111
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 2.0625581395348837e-07,
+      "loss": 0.0016,
+      "step": 8112
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 2.062441860465116e-07,
+      "loss": 0.0017,
+      "step": 8113
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 2.0623255813953486e-07,
+      "loss": 0.0013,
+      "step": 8114
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 2.0622093023255815e-07,
+      "loss": 0.0014,
+      "step": 8115
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 2.062093023255814e-07,
+      "loss": 2.207,
+      "step": 8116
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 2.0619767441860465e-07,
+      "loss": 0.0288,
+      "step": 8117
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 2.0618604651162788e-07,
+      "loss": 0.0012,
+      "step": 8118
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 2.0617441860465117e-07,
+      "loss": 0.0011,
+      "step": 8119
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 2.061627906976744e-07,
+      "loss": 0.0018,
+      "step": 8120
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 2.0615116279069767e-07,
+      "loss": 0.0012,
+      "step": 8121
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 2.061395348837209e-07,
+      "loss": 0.0015,
+      "step": 8122
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 2.061279069767442e-07,
+      "loss": 0.0053,
+      "step": 8123
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 2.0611627906976743e-07,
+      "loss": 3.3282,
+      "step": 8124
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2.061046511627907e-07,
+      "loss": 0.0031,
+      "step": 8125
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2.0609302325581392e-07,
+      "loss": 0.7358,
+      "step": 8126
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2.0608139534883719e-07,
+      "loss": 0.0017,
+      "step": 8127
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2.0606976744186047e-07,
+      "loss": 0.154,
+      "step": 8128
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2.060581395348837e-07,
+      "loss": 0.0017,
+      "step": 8129
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 2.0604651162790697e-07,
+      "loss": 0.002,
+      "step": 8130
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 2.060348837209302e-07,
+      "loss": 0.0032,
+      "step": 8131
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 2.060232558139535e-07,
+      "loss": 0.0014,
+      "step": 8132
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 2.0601162790697673e-07,
+      "loss": 0.0015,
+      "step": 8133
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 2.06e-07,
+      "loss": 0.0062,
+      "step": 8134
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 2.0598837209302323e-07,
+      "loss": 0.0066,
+      "step": 8135
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 2.0597674418604652e-07,
+      "loss": 0.0013,
+      "step": 8136
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 2.0596511627906975e-07,
+      "loss": 0.0014,
+      "step": 8137
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 2.05953488372093e-07,
+      "loss": 0.002,
+      "step": 8138
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 2.0594186046511625e-07,
+      "loss": 0.0014,
+      "step": 8139
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 2.0593023255813954e-07,
+      "loss": 0.0015,
+      "step": 8140
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 2.059186046511628e-07,
+      "loss": 0.0014,
+      "step": 8141
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 2.0590697674418603e-07,
+      "loss": 2.0321,
+      "step": 8142
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 2.058953488372093e-07,
+      "loss": 0.0014,
+      "step": 8143
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 2.0588372093023256e-07,
+      "loss": 0.0015,
+      "step": 8144
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 2.0587209302325582e-07,
+      "loss": 0.0013,
+      "step": 8145
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 2.0586046511627905e-07,
+      "loss": 0.0016,
+      "step": 8146
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 2.0584883720930231e-07,
+      "loss": 0.0013,
+      "step": 8147
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 2.0583720930232558e-07,
+      "loss": 0.0029,
+      "step": 8148
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 2.0582558139534884e-07,
+      "loss": 0.0017,
+      "step": 8149
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 2.0581395348837207e-07,
+      "loss": 0.0022,
+      "step": 8150
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2.0580232558139533e-07,
+      "loss": 0.0016,
+      "step": 8151
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2.0579069767441862e-07,
+      "loss": 0.0012,
+      "step": 8152
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2.0577906976744186e-07,
+      "loss": 0.0014,
+      "step": 8153
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2.0576744186046512e-07,
+      "loss": 0.002,
+      "step": 8154
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2.0575581395348836e-07,
+      "loss": 0.002,
+      "step": 8155
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 2.057441860465116e-07,
+      "loss": 0.0014,
+      "step": 8156
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 2.0573255813953488e-07,
+      "loss": 0.0028,
+      "step": 8157
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 2.0572093023255814e-07,
+      "loss": 0.0018,
+      "step": 8158
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 2.0570930232558138e-07,
+      "loss": 0.0023,
+      "step": 8159
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 2.0569767441860464e-07,
+      "loss": 0.0097,
+      "step": 8160
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.056860465116279e-07,
+      "loss": 0.0011,
+      "step": 8161
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.0567441860465116e-07,
+      "loss": 0.0073,
+      "step": 8162
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.056627906976744e-07,
+      "loss": 0.0017,
+      "step": 8163
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.0565116279069766e-07,
+      "loss": 0.0017,
+      "step": 8164
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.0563953488372095e-07,
+      "loss": 0.0015,
+      "step": 8165
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.0562790697674418e-07,
+      "loss": 0.0015,
+      "step": 8166
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 2.0561627906976742e-07,
+      "loss": 0.0018,
+      "step": 8167
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 2.0560465116279068e-07,
+      "loss": 0.1131,
+      "step": 8168
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 2.0559302325581397e-07,
+      "loss": 0.0026,
+      "step": 8169
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 2.055813953488372e-07,
+      "loss": 0.0019,
+      "step": 8170
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 2.0556976744186046e-07,
+      "loss": 0.0028,
+      "step": 8171
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 2.055581395348837e-07,
+      "loss": 0.0013,
+      "step": 8172
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 2.0554651162790699e-07,
+      "loss": 0.006,
+      "step": 8173
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 2.0553488372093022e-07,
+      "loss": 0.0019,
+      "step": 8174
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 2.0552325581395348e-07,
+      "loss": 0.0024,
+      "step": 8175
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 2.0551162790697672e-07,
+      "loss": 0.0113,
+      "step": 8176
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2.055e-07,
+      "loss": 0.007,
+      "step": 8177
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2.0548837209302324e-07,
+      "loss": 0.0016,
+      "step": 8178
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2.054767441860465e-07,
+      "loss": 0.0017,
+      "step": 8179
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2.0546511627906974e-07,
+      "loss": 0.0013,
+      "step": 8180
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2.05453488372093e-07,
+      "loss": 0.0017,
+      "step": 8181
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 2.054418604651163e-07,
+      "loss": 0.0013,
+      "step": 8182
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 2.0543023255813952e-07,
+      "loss": 0.0015,
+      "step": 8183
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 2.0541860465116279e-07,
+      "loss": 0.0017,
+      "step": 8184
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 2.0540697674418602e-07,
+      "loss": 0.0019,
+      "step": 8185
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 2.053953488372093e-07,
+      "loss": 0.0013,
+      "step": 8186
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 2.0538372093023254e-07,
+      "loss": 0.0013,
+      "step": 8187
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 2.053720930232558e-07,
+      "loss": 0.0022,
+      "step": 8188
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 2.0536046511627904e-07,
+      "loss": 0.0166,
+      "step": 8189
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 2.0534883720930233e-07,
+      "loss": 0.0013,
+      "step": 8190
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 2.0533720930232557e-07,
+      "loss": 0.0012,
+      "step": 8191
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 2.0532558139534883e-07,
+      "loss": 0.0013,
+      "step": 8192
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 2.0531395348837206e-07,
+      "loss": 0.0022,
+      "step": 8193
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 2.0530232558139535e-07,
+      "loss": 0.0011,
+      "step": 8194
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 2.052906976744186e-07,
+      "loss": 0.0013,
+      "step": 8195
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 2.0527906976744185e-07,
+      "loss": 0.002,
+      "step": 8196
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 2.052674418604651e-07,
+      "loss": 0.002,
+      "step": 8197
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 2.0525581395348837e-07,
+      "loss": 0.0018,
+      "step": 8198
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 2.0524418604651163e-07,
+      "loss": 0.0014,
+      "step": 8199
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 2.0523255813953487e-07,
+      "loss": 0.0031,
+      "step": 8200
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 2.0522093023255813e-07,
+      "loss": 0.0116,
+      "step": 8201
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 2.052093023255814e-07,
+      "loss": 0.0014,
+      "step": 8202
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 2.0519767441860465e-07,
+      "loss": 0.0017,
+      "step": 8203
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 2.051860465116279e-07,
+      "loss": 0.0016,
+      "step": 8204
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 2.0517441860465115e-07,
+      "loss": 0.6999,
+      "step": 8205
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 2.0516279069767438e-07,
+      "loss": 0.0013,
+      "step": 8206
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 2.0515116279069767e-07,
+      "loss": 0.0636,
+      "step": 8207
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 2.051395348837209e-07,
+      "loss": 0.0019,
+      "step": 8208
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 2.0512790697674417e-07,
+      "loss": 0.0079,
+      "step": 8209
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 2.051162790697674e-07,
+      "loss": 0.0027,
+      "step": 8210
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 2.051046511627907e-07,
+      "loss": 0.0013,
+      "step": 8211
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 2.0509302325581396e-07,
+      "loss": 0.0012,
+      "step": 8212
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 2.050813953488372e-07,
+      "loss": 2.0306,
+      "step": 8213
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 2.0506976744186045e-07,
+      "loss": 0.0139,
+      "step": 8214
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 2.0505813953488371e-07,
+      "loss": 0.0018,
+      "step": 8215
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 2.0504651162790698e-07,
+      "loss": 0.0015,
+      "step": 8216
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 2.050348837209302e-07,
+      "loss": 0.0022,
+      "step": 8217
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 2.0502325581395347e-07,
+      "loss": 0.0012,
+      "step": 8218
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 2.0501162790697673e-07,
+      "loss": 0.0015,
+      "step": 8219
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 2.05e-07,
+      "loss": 0.0012,
+      "step": 8220
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 2.0498837209302323e-07,
+      "loss": 0.0015,
+      "step": 8221
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 2.049767441860465e-07,
+      "loss": 0.0022,
+      "step": 8222
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 2.0496511627906978e-07,
+      "loss": 0.0011,
+      "step": 8223
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 2.0495348837209302e-07,
+      "loss": 0.0015,
+      "step": 8224
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 2.0494186046511628e-07,
+      "loss": 0.0015,
+      "step": 8225
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 2.0493023255813951e-07,
+      "loss": 0.0016,
+      "step": 8226
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 2.049186046511628e-07,
+      "loss": 0.0015,
+      "step": 8227
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 2.0490697674418604e-07,
+      "loss": 0.4588,
+      "step": 8228
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2.048953488372093e-07,
+      "loss": 0.0018,
+      "step": 8229
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2.0488372093023253e-07,
+      "loss": 0.0013,
+      "step": 8230
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2.0487209302325582e-07,
+      "loss": 0.0016,
+      "step": 8231
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2.0486046511627906e-07,
+      "loss": 0.0122,
+      "step": 8232
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2.0484883720930232e-07,
+      "loss": 0.0012,
+      "step": 8233
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 2.0483720930232555e-07,
+      "loss": 0.0014,
+      "step": 8234
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 2.0482558139534882e-07,
+      "loss": 0.0012,
+      "step": 8235
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 2.048139534883721e-07,
+      "loss": 0.0043,
+      "step": 8236
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 2.0480232558139534e-07,
+      "loss": 2.9082,
+      "step": 8237
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 2.047906976744186e-07,
+      "loss": 0.0021,
+      "step": 8238
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 2.0477906976744184e-07,
+      "loss": 0.0023,
+      "step": 8239
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 2.0476744186046512e-07,
+      "loss": 0.0013,
+      "step": 8240
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 2.0475581395348836e-07,
+      "loss": 0.0022,
+      "step": 8241
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 2.0474418604651162e-07,
+      "loss": 0.0017,
+      "step": 8242
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 2.0473255813953486e-07,
+      "loss": 0.0013,
+      "step": 8243
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 2.0472093023255814e-07,
+      "loss": 0.0017,
+      "step": 8244
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 2.0470930232558138e-07,
+      "loss": 0.0014,
+      "step": 8245
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 2.0469767441860464e-07,
+      "loss": 0.0017,
+      "step": 8246
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 2.0468604651162788e-07,
+      "loss": 0.5802,
+      "step": 8247
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 2.0467441860465117e-07,
+      "loss": 0.0018,
+      "step": 8248
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 2.0466279069767443e-07,
+      "loss": 0.0019,
+      "step": 8249
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 2.0465116279069766e-07,
+      "loss": 0.0013,
+      "step": 8250
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 2.046395348837209e-07,
+      "loss": 0.0014,
+      "step": 8251
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 2.0462790697674419e-07,
+      "loss": 0.0017,
+      "step": 8252
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 2.0461627906976745e-07,
+      "loss": 0.0018,
+      "step": 8253
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.0460465116279068e-07,
+      "loss": 0.0021,
+      "step": 8254
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.0459302325581394e-07,
+      "loss": 0.0015,
+      "step": 8255
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.045813953488372e-07,
+      "loss": 0.0013,
+      "step": 8256
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.0456976744186047e-07,
+      "loss": 0.0017,
+      "step": 8257
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.045581395348837e-07,
+      "loss": 0.0012,
+      "step": 8258
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.0454651162790696e-07,
+      "loss": 0.1728,
+      "step": 8259
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 2.045348837209302e-07,
+      "loss": 0.0016,
+      "step": 8260
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 2.045232558139535e-07,
+      "loss": 0.0019,
+      "step": 8261
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 2.0451162790697672e-07,
+      "loss": 0.0011,
+      "step": 8262
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 2.0449999999999998e-07,
+      "loss": 0.0018,
+      "step": 8263
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 2.0448837209302322e-07,
+      "loss": 0.0018,
+      "step": 8264
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 2.044767441860465e-07,
+      "loss": 0.0013,
+      "step": 8265
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 2.0446511627906977e-07,
+      "loss": 0.0012,
+      "step": 8266
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 2.04453488372093e-07,
+      "loss": 0.0017,
+      "step": 8267
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 2.0444186046511627e-07,
+      "loss": 0.0055,
+      "step": 8268
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 2.0443023255813953e-07,
+      "loss": 0.0019,
+      "step": 8269
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 2.044186046511628e-07,
+      "loss": 0.0014,
+      "step": 8270
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 2.0440697674418603e-07,
+      "loss": 0.0014,
+      "step": 8271
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 2.043953488372093e-07,
+      "loss": 0.0018,
+      "step": 8272
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.988394584139265,
+      "eval_loss": 0.1583469659090042,
+      "eval_runtime": 1.5649,
+      "eval_samples_per_second": 660.76,
+      "eval_steps_per_second": 10.864,
+      "step": 8272
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 2.0438372093023255e-07,
+      "loss": 0.0473,
+      "step": 8273
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 2.043720930232558e-07,
+      "loss": 0.0013,
+      "step": 8274
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 2.0436046511627905e-07,
+      "loss": 0.0016,
+      "step": 8275
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 2.043488372093023e-07,
+      "loss": 0.5621,
+      "step": 8276
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 2.043372093023256e-07,
+      "loss": 0.0092,
+      "step": 8277
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 2.0432558139534883e-07,
+      "loss": 0.0012,
+      "step": 8278
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 2.043139534883721e-07,
+      "loss": 0.0018,
+      "step": 8279
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 2.0430232558139533e-07,
+      "loss": 0.0012,
+      "step": 8280
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 2.0429069767441862e-07,
+      "loss": 0.0022,
+      "step": 8281
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 2.0427906976744185e-07,
+      "loss": 0.0108,
+      "step": 8282
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 2.0426744186046511e-07,
+      "loss": 0.0013,
+      "step": 8283
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 2.0425581395348835e-07,
+      "loss": 0.0013,
+      "step": 8284
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.0424418604651164e-07,
+      "loss": 0.0013,
+      "step": 8285
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.0423255813953487e-07,
+      "loss": 0.0022,
+      "step": 8286
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.0422093023255813e-07,
+      "loss": 0.0017,
+      "step": 8287
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.0420930232558137e-07,
+      "loss": 0.0014,
+      "step": 8288
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.0419767441860463e-07,
+      "loss": 0.0013,
+      "step": 8289
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.0418604651162792e-07,
+      "loss": 0.6521,
+      "step": 8290
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 2.0417441860465115e-07,
+      "loss": 0.002,
+      "step": 8291
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 2.0416279069767442e-07,
+      "loss": 0.0014,
+      "step": 8292
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 2.0415116279069765e-07,
+      "loss": 0.0017,
+      "step": 8293
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 2.0413953488372094e-07,
+      "loss": 0.0019,
+      "step": 8294
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 2.0412790697674417e-07,
+      "loss": 0.0011,
+      "step": 8295
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 2.0411627906976744e-07,
+      "loss": 0.0013,
+      "step": 8296
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 2.0410465116279067e-07,
+      "loss": 0.0014,
+      "step": 8297
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 2.0409302325581396e-07,
+      "loss": 0.0013,
+      "step": 8298
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 2.040813953488372e-07,
+      "loss": 0.0068,
+      "step": 8299
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 2.0406976744186046e-07,
+      "loss": 0.0013,
+      "step": 8300
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 2.040581395348837e-07,
+      "loss": 0.0011,
+      "step": 8301
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 2.0404651162790698e-07,
+      "loss": 0.0015,
+      "step": 8302
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 2.0403488372093024e-07,
+      "loss": 0.0014,
+      "step": 8303
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 2.0402325581395348e-07,
+      "loss": 0.0013,
+      "step": 8304
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 2.040116279069767e-07,
+      "loss": 0.0013,
+      "step": 8305
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 2.04e-07,
+      "loss": 0.0018,
+      "step": 8306
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 2.0398837209302326e-07,
+      "loss": 0.0019,
+      "step": 8307
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 2.039767441860465e-07,
+      "loss": 0.0013,
+      "step": 8308
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 2.0396511627906976e-07,
+      "loss": 0.0012,
+      "step": 8309
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 2.0395348837209302e-07,
+      "loss": 0.0018,
+      "step": 8310
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2.0394186046511628e-07,
+      "loss": 0.0023,
+      "step": 8311
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2.0393023255813952e-07,
+      "loss": 0.001,
+      "step": 8312
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2.0391860465116278e-07,
+      "loss": 0.0034,
+      "step": 8313
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2.0390697674418601e-07,
+      "loss": 0.0163,
+      "step": 8314
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2.038953488372093e-07,
+      "loss": 0.0019,
+      "step": 8315
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 2.0388372093023254e-07,
+      "loss": 0.0016,
+      "step": 8316
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 2.038720930232558e-07,
+      "loss": 0.1845,
+      "step": 8317
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 2.0386046511627903e-07,
+      "loss": 0.0017,
+      "step": 8318
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 2.0384883720930232e-07,
+      "loss": 0.0014,
+      "step": 8319
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 2.0383720930232558e-07,
+      "loss": 0.002,
+      "step": 8320
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 2.0382558139534882e-07,
+      "loss": 0.0013,
+      "step": 8321
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 2.0381395348837208e-07,
+      "loss": 0.0206,
+      "step": 8322
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 2.0380232558139534e-07,
+      "loss": 0.0015,
+      "step": 8323
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 2.037906976744186e-07,
+      "loss": 0.0012,
+      "step": 8324
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 2.0377906976744184e-07,
+      "loss": 0.0031,
+      "step": 8325
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 2.037674418604651e-07,
+      "loss": 0.1462,
+      "step": 8326
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 2.0375581395348836e-07,
+      "loss": 0.0069,
+      "step": 8327
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 2.0374418604651163e-07,
+      "loss": 0.0014,
+      "step": 8328
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 2.0373255813953486e-07,
+      "loss": 0.0014,
+      "step": 8329
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 2.0372093023255812e-07,
+      "loss": 0.0013,
+      "step": 8330
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 2.037093023255814e-07,
+      "loss": 0.0015,
+      "step": 8331
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 2.0369767441860465e-07,
+      "loss": 0.0051,
+      "step": 8332
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 2.036860465116279e-07,
+      "loss": 1.3568,
+      "step": 8333
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 2.0367441860465114e-07,
+      "loss": 0.0024,
+      "step": 8334
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 2.0366279069767443e-07,
+      "loss": 0.0013,
+      "step": 8335
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 2.0365116279069767e-07,
+      "loss": 0.0012,
+      "step": 8336
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2.0363953488372093e-07,
+      "loss": 0.0022,
+      "step": 8337
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2.0362790697674416e-07,
+      "loss": 0.0014,
+      "step": 8338
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2.0361627906976745e-07,
+      "loss": 0.0018,
+      "step": 8339
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2.0360465116279069e-07,
+      "loss": 0.0025,
+      "step": 8340
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2.0359302325581395e-07,
+      "loss": 0.0013,
+      "step": 8341
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 2.0358139534883718e-07,
+      "loss": 0.0048,
+      "step": 8342
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 2.0356976744186045e-07,
+      "loss": 0.0014,
+      "step": 8343
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 2.0355813953488373e-07,
+      "loss": 0.002,
+      "step": 8344
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 2.0354651162790697e-07,
+      "loss": 0.0016,
+      "step": 8345
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 2.035348837209302e-07,
+      "loss": 0.0017,
+      "step": 8346
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.0352325581395347e-07,
+      "loss": 0.0022,
+      "step": 8347
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.0351162790697675e-07,
+      "loss": 0.0045,
+      "step": 8348
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.035e-07,
+      "loss": 0.0017,
+      "step": 8349
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.0348837209302325e-07,
+      "loss": 0.0012,
+      "step": 8350
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.0347674418604649e-07,
+      "loss": 0.0012,
+      "step": 8351
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.0346511627906977e-07,
+      "loss": 0.0014,
+      "step": 8352
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 2.03453488372093e-07,
+      "loss": 0.0016,
+      "step": 8353
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 2.0344186046511627e-07,
+      "loss": 0.0014,
+      "step": 8354
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 2.034302325581395e-07,
+      "loss": 0.0023,
+      "step": 8355
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 2.034186046511628e-07,
+      "loss": 0.0015,
+      "step": 8356
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 2.0340697674418603e-07,
+      "loss": 0.0019,
+      "step": 8357
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 2.033953488372093e-07,
+      "loss": 0.0013,
+      "step": 8358
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 2.0338372093023253e-07,
+      "loss": 0.0025,
+      "step": 8359
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 2.0337209302325582e-07,
+      "loss": 0.002,
+      "step": 8360
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 2.0336046511627908e-07,
+      "loss": 0.0017,
+      "step": 8361
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 2.033488372093023e-07,
+      "loss": 0.0015,
+      "step": 8362
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 2.0333720930232557e-07,
+      "loss": 0.0012,
+      "step": 8363
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 2.0332558139534884e-07,
+      "loss": 0.0014,
+      "step": 8364
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 2.033139534883721e-07,
+      "loss": 0.0023,
+      "step": 8365
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 2.0330232558139533e-07,
+      "loss": 0.0407,
+      "step": 8366
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 2.032906976744186e-07,
+      "loss": 0.0016,
+      "step": 8367
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 2.0327906976744183e-07,
+      "loss": 0.001,
+      "step": 8368
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 2.0326744186046512e-07,
+      "loss": 0.0228,
+      "step": 8369
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 2.0325581395348835e-07,
+      "loss": 0.0026,
+      "step": 8370
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 2.0324418604651161e-07,
+      "loss": 0.0037,
+      "step": 8371
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 2.0323255813953485e-07,
+      "loss": 0.0022,
+      "step": 8372
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 2.0322093023255814e-07,
+      "loss": 0.0014,
+      "step": 8373
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 2.032093023255814e-07,
+      "loss": 0.0586,
+      "step": 8374
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 2.0319767441860463e-07,
+      "loss": 0.0015,
+      "step": 8375
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 2.031860465116279e-07,
+      "loss": 0.0018,
+      "step": 8376
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 2.0317441860465116e-07,
+      "loss": 0.0015,
+      "step": 8377
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.0316279069767442e-07,
+      "loss": 0.0019,
+      "step": 8378
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.0315116279069766e-07,
+      "loss": 0.0014,
+      "step": 8379
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.0313953488372092e-07,
+      "loss": 0.0014,
+      "step": 8380
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.0312790697674418e-07,
+      "loss": 0.0011,
+      "step": 8381
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.0311627906976744e-07,
+      "loss": 0.0011,
+      "step": 8382
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.0310465116279068e-07,
+      "loss": 0.0014,
+      "step": 8383
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 2.0309302325581394e-07,
+      "loss": 0.002,
+      "step": 8384
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 2.0308139534883723e-07,
+      "loss": 0.0022,
+      "step": 8385
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 2.0306976744186046e-07,
+      "loss": 0.0011,
+      "step": 8386
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 2.0305813953488372e-07,
+      "loss": 0.0015,
+      "step": 8387
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 2.0304651162790696e-07,
+      "loss": 0.3028,
+      "step": 8388
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 2.0303488372093025e-07,
+      "loss": 0.0014,
+      "step": 8389
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 2.0302325581395348e-07,
+      "loss": 0.0012,
+      "step": 8390
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 2.0301162790697674e-07,
+      "loss": 0.0015,
+      "step": 8391
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 2.0299999999999998e-07,
+      "loss": 0.0008,
+      "step": 8392
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 2.0298837209302324e-07,
+      "loss": 0.0017,
+      "step": 8393
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2.029767441860465e-07,
+      "loss": 0.0017,
+      "step": 8394
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2.0296511627906976e-07,
+      "loss": 0.0021,
+      "step": 8395
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2.02953488372093e-07,
+      "loss": 0.001,
+      "step": 8396
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2.0294186046511626e-07,
+      "loss": 0.0026,
+      "step": 8397
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2.0293023255813955e-07,
+      "loss": 0.003,
+      "step": 8398
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 2.0291860465116278e-07,
+      "loss": 0.001,
+      "step": 8399
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 2.0290697674418602e-07,
+      "loss": 0.0019,
+      "step": 8400
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 2.0289534883720928e-07,
+      "loss": 0.007,
+      "step": 8401
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 2.0288372093023257e-07,
+      "loss": 0.0019,
+      "step": 8402
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 2.028720930232558e-07,
+      "loss": 0.0014,
+      "step": 8403
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.0286046511627907e-07,
+      "loss": 0.0011,
+      "step": 8404
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.028488372093023e-07,
+      "loss": 0.0013,
+      "step": 8405
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.028372093023256e-07,
+      "loss": 0.0012,
+      "step": 8406
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.0282558139534882e-07,
+      "loss": 2.9345,
+      "step": 8407
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.0281395348837209e-07,
+      "loss": 0.0013,
+      "step": 8408
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.0280232558139532e-07,
+      "loss": 0.002,
+      "step": 8409
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 2.027906976744186e-07,
+      "loss": 0.0009,
+      "step": 8410
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 2.0277906976744184e-07,
+      "loss": 0.0054,
+      "step": 8411
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 2.027674418604651e-07,
+      "loss": 0.0017,
+      "step": 8412
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 2.0275581395348834e-07,
+      "loss": 0.0014,
+      "step": 8413
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 2.0274418604651163e-07,
+      "loss": 0.0015,
+      "step": 8414
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 2.027325581395349e-07,
+      "loss": 0.0019,
+      "step": 8415
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 2.0272093023255813e-07,
+      "loss": 0.0017,
+      "step": 8416
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 2.027093023255814e-07,
+      "loss": 0.0021,
+      "step": 8417
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 2.0269767441860465e-07,
+      "loss": 0.0012,
+      "step": 8418
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 2.026860465116279e-07,
+      "loss": 0.0144,
+      "step": 8419
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 2.0267441860465115e-07,
+      "loss": 0.002,
+      "step": 8420
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 2.026627906976744e-07,
+      "loss": 0.0014,
+      "step": 8421
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 2.0265116279069764e-07,
+      "loss": 3.4555,
+      "step": 8422
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 2.0263953488372093e-07,
+      "loss": 0.0013,
+      "step": 8423
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 2.0262790697674417e-07,
+      "loss": 0.002,
+      "step": 8424
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 2.0261627906976743e-07,
+      "loss": 0.002,
+      "step": 8425
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 2.0260465116279066e-07,
+      "loss": 0.0019,
+      "step": 8426
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 2.0259302325581395e-07,
+      "loss": 0.0028,
+      "step": 8427
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 2.0258139534883721e-07,
+      "loss": 0.0019,
+      "step": 8428
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 2.0256976744186045e-07,
+      "loss": 0.0021,
+      "step": 8429
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 2.025581395348837e-07,
+      "loss": 0.0018,
+      "step": 8430
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 2.0254651162790697e-07,
+      "loss": 0.0013,
+      "step": 8431
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 2.0253488372093023e-07,
+      "loss": 0.0018,
+      "step": 8432
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 2.0252325581395347e-07,
+      "loss": 0.0014,
+      "step": 8433
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 2.0251162790697673e-07,
+      "loss": 0.0015,
+      "step": 8434
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.025e-07,
+      "loss": 0.5024,
+      "step": 8435
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.0248837209302326e-07,
+      "loss": 0.0013,
+      "step": 8436
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.024767441860465e-07,
+      "loss": 0.0013,
+      "step": 8437
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.0246511627906975e-07,
+      "loss": 0.0012,
+      "step": 8438
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.0245348837209304e-07,
+      "loss": 0.6292,
+      "step": 8439
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.0244186046511628e-07,
+      "loss": 0.0019,
+      "step": 8440
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 2.0243023255813954e-07,
+      "loss": 0.0016,
+      "step": 8441
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 2.0241860465116277e-07,
+      "loss": 0.0011,
+      "step": 8442
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 2.0240697674418606e-07,
+      "loss": 0.0017,
+      "step": 8443
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 2.023953488372093e-07,
+      "loss": 0.0015,
+      "step": 8444
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 2.0238372093023256e-07,
+      "loss": 0.0012,
+      "step": 8445
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 2.023720930232558e-07,
+      "loss": 0.0012,
+      "step": 8446
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 2.0236046511627905e-07,
+      "loss": 0.4892,
+      "step": 8447
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 2.0234883720930232e-07,
+      "loss": 0.0021,
+      "step": 8448
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 2.0233720930232558e-07,
+      "loss": 0.0015,
+      "step": 8449
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 2.0232558139534881e-07,
+      "loss": 0.0011,
+      "step": 8450
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 2.0231395348837207e-07,
+      "loss": 0.0013,
+      "step": 8451
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 2.0230232558139536e-07,
+      "loss": 0.0014,
+      "step": 8452
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 2.022906976744186e-07,
+      "loss": 0.006,
+      "step": 8453
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 2.0227906976744183e-07,
+      "loss": 0.0014,
+      "step": 8454
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 2.022674418604651e-07,
+      "loss": 0.0013,
+      "step": 8455
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 2.0225581395348838e-07,
+      "loss": 0.0024,
+      "step": 8456
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 2.0224418604651162e-07,
+      "loss": 0.0013,
+      "step": 8457
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 2.0223255813953488e-07,
+      "loss": 0.0024,
+      "step": 8458
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 2.0222093023255812e-07,
+      "loss": 0.0012,
+      "step": 8459
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 2.022093023255814e-07,
+      "loss": 0.0022,
+      "step": 8460
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 2.0219767441860464e-07,
+      "loss": 0.0013,
+      "step": 8461
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 2.021860465116279e-07,
+      "loss": 0.0012,
+      "step": 8462
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 2.0217441860465114e-07,
+      "loss": 0.0014,
+      "step": 8463
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 2.0216279069767442e-07,
+      "loss": 0.0012,
+      "step": 8464
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 2.0215116279069766e-07,
+      "loss": 0.0017,
+      "step": 8465
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 2.0213953488372092e-07,
+      "loss": 0.0021,
+      "step": 8466
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 2.0212790697674416e-07,
+      "loss": 0.2625,
+      "step": 8467
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 2.0211627906976744e-07,
+      "loss": 0.0053,
+      "step": 8468
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 2.021046511627907e-07,
+      "loss": 0.0008,
+      "step": 8469
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 2.0209302325581394e-07,
+      "loss": 0.002,
+      "step": 8470
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 2.020813953488372e-07,
+      "loss": 3.2209,
+      "step": 8471
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2.0206976744186047e-07,
+      "loss": 0.0021,
+      "step": 8472
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2.0205813953488373e-07,
+      "loss": 0.0014,
+      "step": 8473
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2.0204651162790696e-07,
+      "loss": 0.0017,
+      "step": 8474
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2.0203488372093022e-07,
+      "loss": 0.0022,
+      "step": 8475
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2.0202325581395346e-07,
+      "loss": 0.0024,
+      "step": 8476
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 2.0201162790697675e-07,
+      "loss": 0.0021,
+      "step": 8477
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 2.0199999999999998e-07,
+      "loss": 0.0017,
+      "step": 8478
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 2.0198837209302324e-07,
+      "loss": 0.0015,
+      "step": 8479
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 2.0197674418604648e-07,
+      "loss": 0.0012,
+      "step": 8480
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 2.0196511627906977e-07,
+      "loss": 0.0014,
+      "step": 8481
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 2.0195348837209303e-07,
+      "loss": 0.0028,
+      "step": 8482
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 2.0194186046511626e-07,
+      "loss": 0.0019,
+      "step": 8483
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 2.019302325581395e-07,
+      "loss": 0.0015,
+      "step": 8484
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 2.019186046511628e-07,
+      "loss": 0.0232,
+      "step": 8485
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 2.0190697674418605e-07,
+      "loss": 0.002,
+      "step": 8486
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 2.0189534883720928e-07,
+      "loss": 0.0032,
+      "step": 8487
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 2.0188372093023255e-07,
+      "loss": 0.0014,
+      "step": 8488
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 2.018720930232558e-07,
+      "loss": 0.0012,
+      "step": 8489
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 2.0186046511627907e-07,
+      "loss": 0.001,
+      "step": 8490
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 2.018488372093023e-07,
+      "loss": 0.0013,
+      "step": 8491
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 2.0183720930232557e-07,
+      "loss": 0.0156,
+      "step": 8492
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 2.0182558139534886e-07,
+      "loss": 0.0017,
+      "step": 8493
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 2.018139534883721e-07,
+      "loss": 0.0016,
+      "step": 8494
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 2.0180232558139533e-07,
+      "loss": 0.0012,
+      "step": 8495
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 2.017906976744186e-07,
+      "loss": 0.0014,
+      "step": 8496
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.0177906976744188e-07,
+      "loss": 0.0012,
+      "step": 8497
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.017674418604651e-07,
+      "loss": 0.0013,
+      "step": 8498
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.0175581395348837e-07,
+      "loss": 0.0012,
+      "step": 8499
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.017441860465116e-07,
+      "loss": 0.005,
+      "step": 8500
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.0173255813953487e-07,
+      "loss": 0.0035,
+      "step": 8501
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.0172093023255813e-07,
+      "loss": 0.0044,
+      "step": 8502
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 2.017093023255814e-07,
+      "loss": 0.0018,
+      "step": 8503
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 2.0169767441860463e-07,
+      "loss": 0.0011,
+      "step": 8504
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 2.016860465116279e-07,
+      "loss": 0.0012,
+      "step": 8505
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 2.0167441860465115e-07,
+      "loss": 0.001,
+      "step": 8506
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 2.016627906976744e-07,
+      "loss": 0.0013,
+      "step": 8507
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 2.0165116279069765e-07,
+      "loss": 0.3016,
+      "step": 8508
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 2.016395348837209e-07,
+      "loss": 0.0033,
+      "step": 8509
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 2.016279069767442e-07,
+      "loss": 1.5552,
+      "step": 8510
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 2.0161627906976743e-07,
+      "loss": 1.694,
+      "step": 8511
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 2.016046511627907e-07,
+      "loss": 0.0731,
+      "step": 8512
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 2.0159302325581393e-07,
+      "loss": 0.0016,
+      "step": 8513
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 2.0158139534883722e-07,
+      "loss": 0.0012,
+      "step": 8514
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 2.0156976744186045e-07,
+      "loss": 0.0036,
+      "step": 8515
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 2.0155813953488372e-07,
+      "loss": 0.0015,
+      "step": 8516
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 2.0154651162790695e-07,
+      "loss": 0.0013,
+      "step": 8517
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 2.0153488372093024e-07,
+      "loss": 0.0014,
+      "step": 8518
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 2.0152325581395347e-07,
+      "loss": 0.0015,
+      "step": 8519
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 2.0151162790697674e-07,
+      "loss": 1.0046,
+      "step": 8520
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 2.0149999999999997e-07,
+      "loss": 0.0012,
+      "step": 8521
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 2.0148837209302326e-07,
+      "loss": 0.7637,
+      "step": 8522
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 2.0147674418604652e-07,
+      "loss": 0.0011,
+      "step": 8523
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 2.0146511627906976e-07,
+      "loss": 1.4736,
+      "step": 8524
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 2.0145348837209302e-07,
+      "loss": 0.0021,
+      "step": 8525
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 2.0144186046511625e-07,
+      "loss": 0.0021,
+      "step": 8526
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 2.0143023255813954e-07,
+      "loss": 0.0023,
+      "step": 8527
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.0141860465116278e-07,
+      "loss": 0.0017,
+      "step": 8528
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.0140697674418604e-07,
+      "loss": 0.0012,
+      "step": 8529
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.0139534883720927e-07,
+      "loss": 0.0012,
+      "step": 8530
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.0138372093023256e-07,
+      "loss": 0.0013,
+      "step": 8531
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.013720930232558e-07,
+      "loss": 0.0015,
+      "step": 8532
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.0136046511627906e-07,
+      "loss": 0.0011,
+      "step": 8533
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 2.013488372093023e-07,
+      "loss": 0.0022,
+      "step": 8534
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 2.0133720930232558e-07,
+      "loss": 0.0013,
+      "step": 8535
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 2.0132558139534884e-07,
+      "loss": 0.0018,
+      "step": 8536
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 2.0131395348837208e-07,
+      "loss": 0.0041,
+      "step": 8537
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 2.0130232558139531e-07,
+      "loss": 0.0014,
+      "step": 8538
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 2.012906976744186e-07,
+      "loss": 0.7947,
+      "step": 8539
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 2.0127906976744186e-07,
+      "loss": 0.0014,
+      "step": 8540
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 2.012674418604651e-07,
+      "loss": 0.0016,
+      "step": 8541
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 2.0125581395348836e-07,
+      "loss": 0.0012,
+      "step": 8542
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 2.0124418604651162e-07,
+      "loss": 0.0014,
+      "step": 8543
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 2.0123255813953488e-07,
+      "loss": 0.0012,
+      "step": 8544
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 2.0122093023255812e-07,
+      "loss": 1.6257,
+      "step": 8545
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 2.0120930232558138e-07,
+      "loss": 0.0018,
+      "step": 8546
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 2.0119767441860467e-07,
+      "loss": 0.0016,
+      "step": 8547
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 2.011860465116279e-07,
+      "loss": 0.0039,
+      "step": 8548
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 2.0117441860465114e-07,
+      "loss": 0.0011,
+      "step": 8549
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 2.011627906976744e-07,
+      "loss": 0.0011,
+      "step": 8550
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 2.011511627906977e-07,
+      "loss": 0.0014,
+      "step": 8551
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 2.0113953488372093e-07,
+      "loss": 0.0008,
+      "step": 8552
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 2.011279069767442e-07,
+      "loss": 0.0009,
+      "step": 8553
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 2.0111627906976742e-07,
+      "loss": 0.0016,
+      "step": 8554
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 2.0110465116279068e-07,
+      "loss": 0.0047,
+      "step": 8555
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 2.0109302325581395e-07,
+      "loss": 0.0013,
+      "step": 8556
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 2.010813953488372e-07,
+      "loss": 0.0027,
+      "step": 8557
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 2.0106976744186044e-07,
+      "loss": 0.0014,
+      "step": 8558
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 2.010581395348837e-07,
+      "loss": 0.0011,
+      "step": 8559
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 2.0104651162790697e-07,
+      "loss": 0.0034,
+      "step": 8560
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 2.0103488372093023e-07,
+      "loss": 0.0015,
+      "step": 8561
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 2.0102325581395346e-07,
+      "loss": 0.0021,
+      "step": 8562
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 2.0101162790697672e-07,
+      "loss": 0.0015,
+      "step": 8563
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 2.01e-07,
+      "loss": 0.0015,
+      "step": 8564
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 2.0098837209302325e-07,
+      "loss": 0.001,
+      "step": 8565
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 2.009767441860465e-07,
+      "loss": 0.0019,
+      "step": 8566
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 2.0096511627906975e-07,
+      "loss": 0.0015,
+      "step": 8567
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 2.0095348837209303e-07,
+      "loss": 0.0022,
+      "step": 8568
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 2.0094186046511627e-07,
+      "loss": 0.0019,
+      "step": 8569
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 2.0093023255813953e-07,
+      "loss": 0.0014,
+      "step": 8570
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 2.0091860465116277e-07,
+      "loss": 0.0012,
+      "step": 8571
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 2.0090697674418605e-07,
+      "loss": 0.0013,
+      "step": 8572
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 2.008953488372093e-07,
+      "loss": 0.0013,
+      "step": 8573
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 2.0088372093023255e-07,
+      "loss": 0.002,
+      "step": 8574
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 2.0087209302325579e-07,
+      "loss": 0.0028,
+      "step": 8575
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 2.0086046511627907e-07,
+      "loss": 0.0016,
+      "step": 8576
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 2.0084883720930234e-07,
+      "loss": 0.0021,
+      "step": 8577
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 2.0083720930232557e-07,
+      "loss": 0.0012,
+      "step": 8578
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 2.0082558139534883e-07,
+      "loss": 0.0024,
+      "step": 8579
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 2.0081395348837207e-07,
+      "loss": 0.0067,
+      "step": 8580
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 2.0080232558139536e-07,
+      "loss": 0.0015,
+      "step": 8581
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 2.007906976744186e-07,
+      "loss": 0.0011,
+      "step": 8582
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 2.0077906976744185e-07,
+      "loss": 2.0606,
+      "step": 8583
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 2.007674418604651e-07,
+      "loss": 0.0019,
+      "step": 8584
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 2.0075581395348838e-07,
+      "loss": 0.0012,
+      "step": 8585
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 2.007441860465116e-07,
+      "loss": 0.0021,
+      "step": 8586
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 2.0073255813953487e-07,
+      "loss": 0.0013,
+      "step": 8587
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 2.007209302325581e-07,
+      "loss": 0.0013,
+      "step": 8588
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 2.007093023255814e-07,
+      "loss": 0.0013,
+      "step": 8589
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.0069767441860463e-07,
+      "loss": 0.0025,
+      "step": 8590
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.006860465116279e-07,
+      "loss": 0.0017,
+      "step": 8591
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.0067441860465113e-07,
+      "loss": 0.0013,
+      "step": 8592
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.0066279069767442e-07,
+      "loss": 0.0015,
+      "step": 8593
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.0065116279069768e-07,
+      "loss": 0.0016,
+      "step": 8594
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.0063953488372091e-07,
+      "loss": 0.0016,
+      "step": 8595
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 2.0062790697674418e-07,
+      "loss": 0.0011,
+      "step": 8596
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 2.0061627906976744e-07,
+      "loss": 0.0012,
+      "step": 8597
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 2.006046511627907e-07,
+      "loss": 0.0013,
+      "step": 8598
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 2.0059302325581393e-07,
+      "loss": 0.0011,
+      "step": 8599
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 2.005813953488372e-07,
+      "loss": 0.0014,
+      "step": 8600
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 2.0056976744186046e-07,
+      "loss": 2.4011,
+      "step": 8601
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 2.0055813953488372e-07,
+      "loss": 0.0012,
+      "step": 8602
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 2.0054651162790696e-07,
+      "loss": 0.0016,
+      "step": 8603
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 2.0053488372093022e-07,
+      "loss": 0.0029,
+      "step": 8604
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 2.005232558139535e-07,
+      "loss": 0.0014,
+      "step": 8605
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 2.0051162790697674e-07,
+      "loss": 0.6794,
+      "step": 8606
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 2.005e-07,
+      "loss": 0.0022,
+      "step": 8607
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 2.0048837209302324e-07,
+      "loss": 0.0017,
+      "step": 8608
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 2.004767441860465e-07,
+      "loss": 0.0018,
+      "step": 8609
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 2.0046511627906976e-07,
+      "loss": 0.0009,
+      "step": 8610
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 2.0045348837209302e-07,
+      "loss": 2.8213,
+      "step": 8611
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 2.0044186046511626e-07,
+      "loss": 0.0013,
+      "step": 8612
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 2.0043023255813952e-07,
+      "loss": 0.0021,
+      "step": 8613
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 2.0041860465116278e-07,
+      "loss": 0.0012,
+      "step": 8614
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 2.0040697674418604e-07,
+      "loss": 0.0026,
+      "step": 8615
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 2.0039534883720928e-07,
+      "loss": 0.0015,
+      "step": 8616
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 2.0038372093023254e-07,
+      "loss": 0.001,
+      "step": 8617
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 2.0037209302325583e-07,
+      "loss": 0.0015,
+      "step": 8618
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 2.0036046511627906e-07,
+      "loss": 0.0012,
+      "step": 8619
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 2.0034883720930232e-07,
+      "loss": 0.0012,
+      "step": 8620
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.0033720930232556e-07,
+      "loss": 0.0013,
+      "step": 8621
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.0032558139534885e-07,
+      "loss": 0.0021,
+      "step": 8622
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.0031395348837208e-07,
+      "loss": 0.002,
+      "step": 8623
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.0030232558139535e-07,
+      "loss": 0.3878,
+      "step": 8624
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.0029069767441858e-07,
+      "loss": 0.0012,
+      "step": 8625
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.0027906976744187e-07,
+      "loss": 0.0015,
+      "step": 8626
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 2.002674418604651e-07,
+      "loss": 0.0009,
+      "step": 8627
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 2.0025581395348837e-07,
+      "loss": 0.0012,
+      "step": 8628
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 2.002441860465116e-07,
+      "loss": 0.0017,
+      "step": 8629
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 2.002325581395349e-07,
+      "loss": 0.0021,
+      "step": 8630
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 2.0022093023255815e-07,
+      "loss": 0.0013,
+      "step": 8631
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2.0020930232558139e-07,
+      "loss": 0.0012,
+      "step": 8632
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2.0019767441860462e-07,
+      "loss": 0.0012,
+      "step": 8633
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2.0018604651162788e-07,
+      "loss": 0.0012,
+      "step": 8634
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2.0017441860465117e-07,
+      "loss": 0.0015,
+      "step": 8635
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2.001627906976744e-07,
+      "loss": 0.0016,
+      "step": 8636
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.0015116279069767e-07,
+      "loss": 0.001,
+      "step": 8637
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.001395348837209e-07,
+      "loss": 0.0014,
+      "step": 8638
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.001279069767442e-07,
+      "loss": 0.0015,
+      "step": 8639
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.0011627906976743e-07,
+      "loss": 0.0016,
+      "step": 8640
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.001046511627907e-07,
+      "loss": 0.0011,
+      "step": 8641
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 2.0009302325581392e-07,
+      "loss": 0.0013,
+      "step": 8642
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 2.000813953488372e-07,
+      "loss": 0.0013,
+      "step": 8643
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 2.0006976744186045e-07,
+      "loss": 0.0014,
+      "step": 8644
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 2.000581395348837e-07,
+      "loss": 0.0017,
+      "step": 8645
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 2.0004651162790694e-07,
+      "loss": 0.002,
+      "step": 8646
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 2.0003488372093023e-07,
+      "loss": 0.0017,
+      "step": 8647
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 2.000232558139535e-07,
+      "loss": 0.0016,
+      "step": 8648
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 2.0001162790697673e-07,
+      "loss": 0.06,
+      "step": 8649
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 2e-07,
+      "loss": 0.0025,
+      "step": 8650
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 1.9998837209302325e-07,
+      "loss": 0.0017,
+      "step": 8651
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.9997674418604651e-07,
+      "loss": 0.001,
+      "step": 8652
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.9996511627906975e-07,
+      "loss": 0.0013,
+      "step": 8653
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.99953488372093e-07,
+      "loss": 0.0015,
+      "step": 8654
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.9994186046511627e-07,
+      "loss": 0.0013,
+      "step": 8655
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.9993023255813953e-07,
+      "loss": 0.0013,
+      "step": 8656
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.9991860465116277e-07,
+      "loss": 0.0029,
+      "step": 8657
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.9990697674418603e-07,
+      "loss": 0.0029,
+      "step": 8658
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.9989534883720927e-07,
+      "loss": 0.0018,
+      "step": 8659
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.9988372093023256e-07,
+      "loss": 0.0019,
+      "step": 8660
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.9987209302325582e-07,
+      "loss": 0.0024,
+      "step": 8661
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.9986046511627905e-07,
+      "loss": 0.0011,
+      "step": 8662
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.9984883720930231e-07,
+      "loss": 0.0014,
+      "step": 8663
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.9983720930232558e-07,
+      "loss": 0.0011,
+      "step": 8664
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.9982558139534884e-07,
+      "loss": 0.002,
+      "step": 8665
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.9981395348837207e-07,
+      "loss": 0.0031,
+      "step": 8666
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.9980232558139533e-07,
+      "loss": 0.0039,
+      "step": 8667
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.997906976744186e-07,
+      "loss": 0.0015,
+      "step": 8668
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.9977906976744186e-07,
+      "loss": 0.0012,
+      "step": 8669
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.997674418604651e-07,
+      "loss": 0.0022,
+      "step": 8670
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.9975581395348835e-07,
+      "loss": 0.0021,
+      "step": 8671
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.9974418604651164e-07,
+      "loss": 0.0014,
+      "step": 8672
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.9973255813953488e-07,
+      "loss": 0.0012,
+      "step": 8673
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.9972093023255814e-07,
+      "loss": 0.0024,
+      "step": 8674
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.9970930232558137e-07,
+      "loss": 0.0012,
+      "step": 8675
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.9969767441860466e-07,
+      "loss": 0.0011,
+      "step": 8676
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.996860465116279e-07,
+      "loss": 0.0017,
+      "step": 8677
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.9967441860465116e-07,
+      "loss": 0.001,
+      "step": 8678
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.996627906976744e-07,
+      "loss": 0.0014,
+      "step": 8679
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.9965116279069768e-07,
+      "loss": 0.0013,
+      "step": 8680
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.9963953488372092e-07,
+      "loss": 0.6633,
+      "step": 8681
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.9962790697674418e-07,
+      "loss": 0.0058,
+      "step": 8682
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.9961627906976742e-07,
+      "loss": 0.0013,
+      "step": 8683
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.996046511627907e-07,
+      "loss": 0.001,
+      "step": 8684
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.9959302325581397e-07,
+      "loss": 0.0021,
+      "step": 8685
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.995813953488372e-07,
+      "loss": 0.2434,
+      "step": 8686
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.9956976744186044e-07,
+      "loss": 0.0018,
+      "step": 8687
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.995581395348837e-07,
+      "loss": 0.001,
+      "step": 8688
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.9954651162790699e-07,
+      "loss": 0.0016,
+      "step": 8689
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.9953488372093022e-07,
+      "loss": 0.002,
+      "step": 8690
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.9952325581395348e-07,
+      "loss": 0.0013,
+      "step": 8691
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.9951162790697672e-07,
+      "loss": 0.0012,
+      "step": 8692
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.995e-07,
+      "loss": 0.3645,
+      "step": 8693
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.9948837209302324e-07,
+      "loss": 0.0556,
+      "step": 8694
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.994767441860465e-07,
+      "loss": 0.2082,
+      "step": 8695
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.9946511627906974e-07,
+      "loss": 0.0015,
+      "step": 8696
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.9945348837209303e-07,
+      "loss": 0.0018,
+      "step": 8697
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.9944186046511626e-07,
+      "loss": 0.001,
+      "step": 8698
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.9943023255813952e-07,
+      "loss": 0.0016,
+      "step": 8699
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.9941860465116276e-07,
+      "loss": 0.0011,
+      "step": 8700
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.9940697674418605e-07,
+      "loss": 0.0016,
+      "step": 8701
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.993953488372093e-07,
+      "loss": 0.0016,
+      "step": 8702
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.9938372093023254e-07,
+      "loss": 0.001,
+      "step": 8703
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.993720930232558e-07,
+      "loss": 0.0031,
+      "step": 8704
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.9936046511627907e-07,
+      "loss": 0.0017,
+      "step": 8705
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.9934883720930233e-07,
+      "loss": 0.0013,
+      "step": 8706
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.9933720930232556e-07,
+      "loss": 0.0014,
+      "step": 8707
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.9932558139534883e-07,
+      "loss": 0.001,
+      "step": 8708
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.993139534883721e-07,
+      "loss": 0.0019,
+      "step": 8709
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.9930232558139535e-07,
+      "loss": 1.2167,
+      "step": 8710
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.9929069767441858e-07,
+      "loss": 0.0019,
+      "step": 8711
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.9927906976744185e-07,
+      "loss": 0.0018,
+      "step": 8712
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.9926744186046508e-07,
+      "loss": 0.0014,
+      "step": 8713
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.9925581395348837e-07,
+      "loss": 0.0016,
+      "step": 8714
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.9924418604651163e-07,
+      "loss": 1.142,
+      "step": 8715
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.9923255813953487e-07,
+      "loss": 0.0016,
+      "step": 8716
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.9922093023255813e-07,
+      "loss": 0.0015,
+      "step": 8717
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.992093023255814e-07,
+      "loss": 0.0014,
+      "step": 8718
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.9919767441860465e-07,
+      "loss": 0.0013,
+      "step": 8719
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.991860465116279e-07,
+      "loss": 0.0012,
+      "step": 8720
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.9917441860465115e-07,
+      "loss": 0.0017,
+      "step": 8721
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.991627906976744e-07,
+      "loss": 0.0012,
+      "step": 8722
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.9915116279069767e-07,
+      "loss": 0.0011,
+      "step": 8723
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.991395348837209e-07,
+      "loss": 0.0011,
+      "step": 8724
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.9912790697674417e-07,
+      "loss": 0.0011,
+      "step": 8725
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.9911627906976746e-07,
+      "loss": 0.0038,
+      "step": 8726
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.991046511627907e-07,
+      "loss": 0.0016,
+      "step": 8727
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.9909302325581393e-07,
+      "loss": 0.0012,
+      "step": 8728
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.990813953488372e-07,
+      "loss": 0.0011,
+      "step": 8729
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.9906976744186048e-07,
+      "loss": 0.0013,
+      "step": 8730
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.990581395348837e-07,
+      "loss": 0.0013,
+      "step": 8731
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.9904651162790697e-07,
+      "loss": 0.0051,
+      "step": 8732
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.990348837209302e-07,
+      "loss": 0.0015,
+      "step": 8733
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.990232558139535e-07,
+      "loss": 0.0018,
+      "step": 8734
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.9901162790697673e-07,
+      "loss": 0.0014,
+      "step": 8735
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.99e-07,
+      "loss": 0.0013,
+      "step": 8736
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.9898837209302323e-07,
+      "loss": 0.0013,
+      "step": 8737
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.9897674418604652e-07,
+      "loss": 0.0143,
+      "step": 8738
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.9896511627906975e-07,
+      "loss": 0.0014,
+      "step": 8739
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.9895348837209302e-07,
+      "loss": 0.0019,
+      "step": 8740
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.9894186046511625e-07,
+      "loss": 0.0014,
+      "step": 8741
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.989302325581395e-07,
+      "loss": 0.0049,
+      "step": 8742
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.989186046511628e-07,
+      "loss": 0.0012,
+      "step": 8743
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.9890697674418604e-07,
+      "loss": 0.0018,
+      "step": 8744
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.988953488372093e-07,
+      "loss": 0.0017,
+      "step": 8745
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.9888372093023253e-07,
+      "loss": 0.0015,
+      "step": 8746
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.9887209302325582e-07,
+      "loss": 0.0016,
+      "step": 8747
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.9886046511627906e-07,
+      "loss": 0.0015,
+      "step": 8748
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.9884883720930232e-07,
+      "loss": 0.0013,
+      "step": 8749
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.9883720930232555e-07,
+      "loss": 0.0025,
+      "step": 8750
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.9882558139534884e-07,
+      "loss": 0.001,
+      "step": 8751
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.9881395348837208e-07,
+      "loss": 0.0018,
+      "step": 8752
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.9880232558139534e-07,
+      "loss": 0.0013,
+      "step": 8753
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.9879069767441857e-07,
+      "loss": 0.0011,
+      "step": 8754
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.9877906976744186e-07,
+      "loss": 0.0016,
+      "step": 8755
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.9876744186046512e-07,
+      "loss": 0.0015,
+      "step": 8756
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.9875581395348836e-07,
+      "loss": 0.0017,
+      "step": 8757
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.9874418604651162e-07,
+      "loss": 0.0013,
+      "step": 8758
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.9873255813953488e-07,
+      "loss": 0.0011,
+      "step": 8759
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.9872093023255814e-07,
+      "loss": 0.0014,
+      "step": 8760
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.9870930232558138e-07,
+      "loss": 0.0299,
+      "step": 8761
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.9869767441860464e-07,
+      "loss": 0.0016,
+      "step": 8762
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.986860465116279e-07,
+      "loss": 0.0013,
+      "step": 8763
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.9867441860465116e-07,
+      "loss": 0.0013,
+      "step": 8764
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.986627906976744e-07,
+      "loss": 0.0014,
+      "step": 8765
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.9865116279069766e-07,
+      "loss": 0.0011,
+      "step": 8766
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.986395348837209e-07,
+      "loss": 0.0014,
+      "step": 8767
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.9862790697674418e-07,
+      "loss": 0.2325,
+      "step": 8768
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.9861627906976745e-07,
+      "loss": 0.0013,
+      "step": 8769
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.9860465116279068e-07,
+      "loss": 0.0016,
+      "step": 8770
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.9859302325581392e-07,
+      "loss": 0.0018,
+      "step": 8771
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.985813953488372e-07,
+      "loss": 0.0011,
+      "step": 8772
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.9856976744186047e-07,
+      "loss": 0.0012,
+      "step": 8773
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.985581395348837e-07,
+      "loss": 0.0012,
+      "step": 8774
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.9854651162790696e-07,
+      "loss": 0.0018,
+      "step": 8775
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.9853488372093023e-07,
+      "loss": 0.0018,
+      "step": 8776
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.985232558139535e-07,
+      "loss": 0.0012,
+      "step": 8777
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.9851162790697672e-07,
+      "loss": 0.0012,
+      "step": 8778
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.9849999999999998e-07,
+      "loss": 0.0018,
+      "step": 8779
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.9848837209302327e-07,
+      "loss": 1.9984,
+      "step": 8780
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.984767441860465e-07,
+      "loss": 0.0016,
+      "step": 8781
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.9846511627906974e-07,
+      "loss": 0.001,
+      "step": 8782
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.98453488372093e-07,
+      "loss": 0.0015,
+      "step": 8783
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.984418604651163e-07,
+      "loss": 0.0184,
+      "step": 8784
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.9843023255813953e-07,
+      "loss": 0.001,
+      "step": 8785
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.984186046511628e-07,
+      "loss": 0.0023,
+      "step": 8786
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.9840697674418602e-07,
+      "loss": 0.0013,
+      "step": 8787
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.983953488372093e-07,
+      "loss": 0.001,
+      "step": 8788
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.9838372093023255e-07,
+      "loss": 0.001,
+      "step": 8789
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.988394584139265,
+      "eval_loss": 0.15435631573200226,
+      "eval_runtime": 1.5582,
+      "eval_samples_per_second": 663.587,
+      "eval_steps_per_second": 10.91,
+      "step": 8789
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.983720930232558e-07,
+      "loss": 0.0014,
+      "step": 8790
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.9836046511627905e-07,
+      "loss": 0.0033,
+      "step": 8791
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.9834883720930233e-07,
+      "loss": 0.0045,
+      "step": 8792
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.9833720930232557e-07,
+      "loss": 0.0014,
+      "step": 8793
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.9832558139534883e-07,
+      "loss": 0.0035,
+      "step": 8794
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.9831395348837207e-07,
+      "loss": 0.0015,
+      "step": 8795
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.9830232558139533e-07,
+      "loss": 0.0022,
+      "step": 8796
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.9829069767441862e-07,
+      "loss": 0.0016,
+      "step": 8797
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.9827906976744185e-07,
+      "loss": 0.0018,
+      "step": 8798
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.982674418604651e-07,
+      "loss": 0.0012,
+      "step": 8799
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.9825581395348835e-07,
+      "loss": 0.0015,
+      "step": 8800
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.9824418604651164e-07,
+      "loss": 0.0024,
+      "step": 8801
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.9823255813953487e-07,
+      "loss": 0.0016,
+      "step": 8802
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.9822093023255813e-07,
+      "loss": 0.0013,
+      "step": 8803
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.9820930232558137e-07,
+      "loss": 0.0011,
+      "step": 8804
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.9819767441860466e-07,
+      "loss": 0.0014,
+      "step": 8805
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.981860465116279e-07,
+      "loss": 0.0015,
+      "step": 8806
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.9817441860465115e-07,
+      "loss": 0.0013,
+      "step": 8807
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.981627906976744e-07,
+      "loss": 0.0014,
+      "step": 8808
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.9815116279069768e-07,
+      "loss": 0.0015,
+      "step": 8809
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.9813953488372094e-07,
+      "loss": 0.0012,
+      "step": 8810
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.9812790697674417e-07,
+      "loss": 0.0033,
+      "step": 8811
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.9811627906976744e-07,
+      "loss": 0.0012,
+      "step": 8812
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.981046511627907e-07,
+      "loss": 0.0009,
+      "step": 8813
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.9809302325581396e-07,
+      "loss": 0.0014,
+      "step": 8814
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.980813953488372e-07,
+      "loss": 0.0025,
+      "step": 8815
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.9806976744186046e-07,
+      "loss": 0.0012,
+      "step": 8816
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.9805813953488372e-07,
+      "loss": 0.0013,
+      "step": 8817
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.9804651162790698e-07,
+      "loss": 0.0017,
+      "step": 8818
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.9803488372093021e-07,
+      "loss": 0.0013,
+      "step": 8819
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.9802325581395348e-07,
+      "loss": 0.0024,
+      "step": 8820
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.980116279069767e-07,
+      "loss": 0.0018,
+      "step": 8821
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.98e-07,
+      "loss": 0.0014,
+      "step": 8822
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.9798837209302326e-07,
+      "loss": 0.0018,
+      "step": 8823
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.979767441860465e-07,
+      "loss": 0.0012,
+      "step": 8824
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.9796511627906973e-07,
+      "loss": 0.0014,
+      "step": 8825
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.9795348837209302e-07,
+      "loss": 0.0013,
+      "step": 8826
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.9794186046511628e-07,
+      "loss": 0.0021,
+      "step": 8827
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.9793023255813952e-07,
+      "loss": 0.0015,
+      "step": 8828
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.9791860465116278e-07,
+      "loss": 0.0056,
+      "step": 8829
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.9790697674418604e-07,
+      "loss": 0.0015,
+      "step": 8830
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.978953488372093e-07,
+      "loss": 0.0018,
+      "step": 8831
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.9788372093023254e-07,
+      "loss": 0.1809,
+      "step": 8832
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.978720930232558e-07,
+      "loss": 0.0011,
+      "step": 8833
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.978604651162791e-07,
+      "loss": 0.0015,
+      "step": 8834
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.9784883720930232e-07,
+      "loss": 0.0015,
+      "step": 8835
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.9783720930232556e-07,
+      "loss": 0.0012,
+      "step": 8836
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.9782558139534882e-07,
+      "loss": 0.1896,
+      "step": 8837
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.978139534883721e-07,
+      "loss": 0.0016,
+      "step": 8838
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.9780232558139534e-07,
+      "loss": 0.001,
+      "step": 8839
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.977906976744186e-07,
+      "loss": 0.0016,
+      "step": 8840
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.9777906976744184e-07,
+      "loss": 0.0019,
+      "step": 8841
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.9776744186046513e-07,
+      "loss": 0.0043,
+      "step": 8842
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.9775581395348836e-07,
+      "loss": 0.0013,
+      "step": 8843
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.9774418604651162e-07,
+      "loss": 0.0014,
+      "step": 8844
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.9773255813953486e-07,
+      "loss": 0.0014,
+      "step": 8845
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.9772093023255812e-07,
+      "loss": 0.0013,
+      "step": 8846
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.9770930232558138e-07,
+      "loss": 0.0009,
+      "step": 8847
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.9769767441860465e-07,
+      "loss": 0.0012,
+      "step": 8848
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.9768604651162788e-07,
+      "loss": 0.0012,
+      "step": 8849
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.9767441860465114e-07,
+      "loss": 0.0013,
+      "step": 8850
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.9766279069767443e-07,
+      "loss": 0.0014,
+      "step": 8851
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.9765116279069767e-07,
+      "loss": 0.0016,
+      "step": 8852
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.9763953488372093e-07,
+      "loss": 1.0323,
+      "step": 8853
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.9762790697674416e-07,
+      "loss": 0.0015,
+      "step": 8854
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.9761627906976745e-07,
+      "loss": 0.0017,
+      "step": 8855
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.9760465116279069e-07,
+      "loss": 0.0015,
+      "step": 8856
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.9759302325581395e-07,
+      "loss": 0.0012,
+      "step": 8857
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.9758139534883718e-07,
+      "loss": 0.0012,
+      "step": 8858
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.9756976744186047e-07,
+      "loss": 0.0016,
+      "step": 8859
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.975581395348837e-07,
+      "loss": 0.0013,
+      "step": 8860
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.9754651162790697e-07,
+      "loss": 0.0018,
+      "step": 8861
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.975348837209302e-07,
+      "loss": 0.0016,
+      "step": 8862
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.975232558139535e-07,
+      "loss": 0.001,
+      "step": 8863
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.9751162790697675e-07,
+      "loss": 0.0017,
+      "step": 8864
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.975e-07,
+      "loss": 0.002,
+      "step": 8865
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.9748837209302322e-07,
+      "loss": 0.001,
+      "step": 8866
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.974767441860465e-07,
+      "loss": 0.0012,
+      "step": 8867
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.9746511627906977e-07,
+      "loss": 0.0015,
+      "step": 8868
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.97453488372093e-07,
+      "loss": 0.2902,
+      "step": 8869
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.9744186046511627e-07,
+      "loss": 0.0014,
+      "step": 8870
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.9743023255813953e-07,
+      "loss": 0.001,
+      "step": 8871
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.974186046511628e-07,
+      "loss": 3.1701,
+      "step": 8872
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.9740697674418603e-07,
+      "loss": 0.0016,
+      "step": 8873
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.973953488372093e-07,
+      "loss": 0.0077,
+      "step": 8874
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.9738372093023253e-07,
+      "loss": 2.3403,
+      "step": 8875
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.9737209302325581e-07,
+      "loss": 0.0012,
+      "step": 8876
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.9736046511627905e-07,
+      "loss": 0.7962,
+      "step": 8877
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.973488372093023e-07,
+      "loss": 0.0018,
+      "step": 8878
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.9733720930232555e-07,
+      "loss": 0.0022,
+      "step": 8879
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.9732558139534883e-07,
+      "loss": 0.0012,
+      "step": 8880
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.973139534883721e-07,
+      "loss": 3.2256,
+      "step": 8881
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.9730232558139533e-07,
+      "loss": 0.0014,
+      "step": 8882
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.972906976744186e-07,
+      "loss": 0.0013,
+      "step": 8883
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.9727906976744186e-07,
+      "loss": 0.001,
+      "step": 8884
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.9726744186046512e-07,
+      "loss": 0.001,
+      "step": 8885
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.9725581395348835e-07,
+      "loss": 0.001,
+      "step": 8886
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.9724418604651161e-07,
+      "loss": 0.0015,
+      "step": 8887
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.9723255813953488e-07,
+      "loss": 0.0011,
+      "step": 8888
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.9722093023255814e-07,
+      "loss": 0.0015,
+      "step": 8889
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.9720930232558137e-07,
+      "loss": 0.0011,
+      "step": 8890
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.9719767441860463e-07,
+      "loss": 0.0018,
+      "step": 8891
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.9718604651162792e-07,
+      "loss": 0.0011,
+      "step": 8892
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.9717441860465116e-07,
+      "loss": 0.0022,
+      "step": 8893
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.9716279069767442e-07,
+      "loss": 0.0014,
+      "step": 8894
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.9715116279069765e-07,
+      "loss": 0.0015,
+      "step": 8895
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.9713953488372094e-07,
+      "loss": 0.0019,
+      "step": 8896
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.9712790697674418e-07,
+      "loss": 0.001,
+      "step": 8897
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.9711627906976744e-07,
+      "loss": 0.0014,
+      "step": 8898
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.9710465116279067e-07,
+      "loss": 0.0009,
+      "step": 8899
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.9709302325581394e-07,
+      "loss": 0.0016,
+      "step": 8900
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.970813953488372e-07,
+      "loss": 0.0015,
+      "step": 8901
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.9706976744186046e-07,
+      "loss": 0.0015,
+      "step": 8902
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.970581395348837e-07,
+      "loss": 0.0015,
+      "step": 8903
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.9704651162790696e-07,
+      "loss": 0.0015,
+      "step": 8904
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.9703488372093025e-07,
+      "loss": 0.0012,
+      "step": 8905
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.9702325581395348e-07,
+      "loss": 0.0015,
+      "step": 8906
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.9701162790697674e-07,
+      "loss": 0.0013,
+      "step": 8907
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.9699999999999998e-07,
+      "loss": 0.0032,
+      "step": 8908
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.9698837209302327e-07,
+      "loss": 0.0015,
+      "step": 8909
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.969767441860465e-07,
+      "loss": 0.0011,
+      "step": 8910
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.9696511627906976e-07,
+      "loss": 0.0013,
+      "step": 8911
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.96953488372093e-07,
+      "loss": 0.0015,
+      "step": 8912
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.9694186046511629e-07,
+      "loss": 0.0016,
+      "step": 8913
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.9693023255813952e-07,
+      "loss": 0.002,
+      "step": 8914
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.9691860465116278e-07,
+      "loss": 0.0011,
+      "step": 8915
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.9690697674418602e-07,
+      "loss": 0.0024,
+      "step": 8916
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.968953488372093e-07,
+      "loss": 0.0014,
+      "step": 8917
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.9688372093023257e-07,
+      "loss": 0.002,
+      "step": 8918
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.968720930232558e-07,
+      "loss": 0.0011,
+      "step": 8919
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.9686046511627904e-07,
+      "loss": 0.0013,
+      "step": 8920
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.9684883720930233e-07,
+      "loss": 0.0015,
+      "step": 8921
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.968372093023256e-07,
+      "loss": 0.0012,
+      "step": 8922
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.9682558139534882e-07,
+      "loss": 0.0014,
+      "step": 8923
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.9681395348837209e-07,
+      "loss": 0.0019,
+      "step": 8924
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.9680232558139535e-07,
+      "loss": 0.0013,
+      "step": 8925
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.967906976744186e-07,
+      "loss": 0.0022,
+      "step": 8926
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.9677906976744184e-07,
+      "loss": 0.002,
+      "step": 8927
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.967674418604651e-07,
+      "loss": 0.0013,
+      "step": 8928
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.9675581395348834e-07,
+      "loss": 0.0018,
+      "step": 8929
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.9674418604651163e-07,
+      "loss": 0.0014,
+      "step": 8930
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.9673255813953486e-07,
+      "loss": 0.0058,
+      "step": 8931
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.9672093023255813e-07,
+      "loss": 0.0013,
+      "step": 8932
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.9670930232558136e-07,
+      "loss": 0.0011,
+      "step": 8933
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.9669767441860465e-07,
+      "loss": 0.0014,
+      "step": 8934
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.966860465116279e-07,
+      "loss": 0.0012,
+      "step": 8935
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.9667441860465115e-07,
+      "loss": 0.002,
+      "step": 8936
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.966627906976744e-07,
+      "loss": 0.0017,
+      "step": 8937
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.9665116279069767e-07,
+      "loss": 0.0009,
+      "step": 8938
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.9663953488372093e-07,
+      "loss": 0.0015,
+      "step": 8939
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.9662790697674417e-07,
+      "loss": 0.0018,
+      "step": 8940
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.9661627906976743e-07,
+      "loss": 0.0102,
+      "step": 8941
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.966046511627907e-07,
+      "loss": 0.9731,
+      "step": 8942
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.9659302325581395e-07,
+      "loss": 0.0012,
+      "step": 8943
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.965813953488372e-07,
+      "loss": 0.0425,
+      "step": 8944
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.9656976744186045e-07,
+      "loss": 0.0008,
+      "step": 8945
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.9655813953488374e-07,
+      "loss": 0.0018,
+      "step": 8946
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.9654651162790697e-07,
+      "loss": 0.0013,
+      "step": 8947
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.9653488372093023e-07,
+      "loss": 0.0011,
+      "step": 8948
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.9652325581395347e-07,
+      "loss": 0.002,
+      "step": 8949
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.9651162790697676e-07,
+      "loss": 0.0014,
+      "step": 8950
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.965e-07,
+      "loss": 0.0015,
+      "step": 8951
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.9648837209302325e-07,
+      "loss": 0.0013,
+      "step": 8952
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.964767441860465e-07,
+      "loss": 0.0017,
+      "step": 8953
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.9646511627906975e-07,
+      "loss": 1.1588,
+      "step": 8954
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.96453488372093e-07,
+      "loss": 0.06,
+      "step": 8955
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.9644186046511627e-07,
+      "loss": 0.0012,
+      "step": 8956
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.964302325581395e-07,
+      "loss": 0.0017,
+      "step": 8957
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.9641860465116277e-07,
+      "loss": 0.1167,
+      "step": 8958
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.9640697674418606e-07,
+      "loss": 0.0016,
+      "step": 8959
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.963953488372093e-07,
+      "loss": 0.0009,
+      "step": 8960
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.9638372093023256e-07,
+      "loss": 0.0014,
+      "step": 8961
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.963720930232558e-07,
+      "loss": 0.0014,
+      "step": 8962
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.9636046511627908e-07,
+      "loss": 0.0015,
+      "step": 8963
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.9634883720930232e-07,
+      "loss": 0.0014,
+      "step": 8964
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.9633720930232558e-07,
+      "loss": 0.0011,
+      "step": 8965
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.963255813953488e-07,
+      "loss": 0.0013,
+      "step": 8966
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.963139534883721e-07,
+      "loss": 0.0015,
+      "step": 8967
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.9630232558139534e-07,
+      "loss": 0.0011,
+      "step": 8968
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.962906976744186e-07,
+      "loss": 0.0014,
+      "step": 8969
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.9627906976744183e-07,
+      "loss": 0.0018,
+      "step": 8970
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.9626744186046512e-07,
+      "loss": 0.0011,
+      "step": 8971
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.9625581395348836e-07,
+      "loss": 0.0013,
+      "step": 8972
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.9624418604651162e-07,
+      "loss": 0.3254,
+      "step": 8973
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.9623255813953485e-07,
+      "loss": 0.0147,
+      "step": 8974
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.9622093023255814e-07,
+      "loss": 0.001,
+      "step": 8975
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.962093023255814e-07,
+      "loss": 0.0013,
+      "step": 8976
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.9619767441860464e-07,
+      "loss": 0.0014,
+      "step": 8977
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.961860465116279e-07,
+      "loss": 1.7224,
+      "step": 8978
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.9617441860465114e-07,
+      "loss": 0.0018,
+      "step": 8979
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.9616279069767442e-07,
+      "loss": 1.3458,
+      "step": 8980
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.9615116279069766e-07,
+      "loss": 0.001,
+      "step": 8981
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.9613953488372092e-07,
+      "loss": 0.0012,
+      "step": 8982
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.9612790697674416e-07,
+      "loss": 0.001,
+      "step": 8983
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.9611627906976744e-07,
+      "loss": 0.001,
+      "step": 8984
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.9610465116279068e-07,
+      "loss": 0.0013,
+      "step": 8985
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.9609302325581394e-07,
+      "loss": 0.0013,
+      "step": 8986
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.9608139534883718e-07,
+      "loss": 0.0019,
+      "step": 8987
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.9606976744186046e-07,
+      "loss": 0.3515,
+      "step": 8988
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.9605813953488373e-07,
+      "loss": 0.0019,
+      "step": 8989
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.9604651162790696e-07,
+      "loss": 0.0011,
+      "step": 8990
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.9603488372093022e-07,
+      "loss": 0.0009,
+      "step": 8991
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.9602325581395348e-07,
+      "loss": 0.0019,
+      "step": 8992
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.9601162790697675e-07,
+      "loss": 0.0008,
+      "step": 8993
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.9599999999999998e-07,
+      "loss": 0.0013,
+      "step": 8994
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.9598837209302324e-07,
+      "loss": 0.0018,
+      "step": 8995
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.959767441860465e-07,
+      "loss": 0.0012,
+      "step": 8996
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.9596511627906977e-07,
+      "loss": 0.001,
+      "step": 8997
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.95953488372093e-07,
+      "loss": 0.0012,
+      "step": 8998
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.9594186046511626e-07,
+      "loss": 0.0013,
+      "step": 8999
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.9593023255813955e-07,
+      "loss": 0.0037,
+      "step": 9000
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.959186046511628e-07,
+      "loss": 0.0013,
+      "step": 9001
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.9590697674418605e-07,
+      "loss": 0.0019,
+      "step": 9002
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.9589534883720928e-07,
+      "loss": 0.0012,
+      "step": 9003
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.9588372093023257e-07,
+      "loss": 0.0013,
+      "step": 9004
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.958720930232558e-07,
+      "loss": 0.0014,
+      "step": 9005
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.9586046511627907e-07,
+      "loss": 0.0011,
+      "step": 9006
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.958488372093023e-07,
+      "loss": 0.0014,
+      "step": 9007
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.9583720930232557e-07,
+      "loss": 0.0011,
+      "step": 9008
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.9582558139534883e-07,
+      "loss": 0.001,
+      "step": 9009
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.958139534883721e-07,
+      "loss": 0.0015,
+      "step": 9010
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.9580232558139532e-07,
+      "loss": 0.2172,
+      "step": 9011
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.9579069767441859e-07,
+      "loss": 0.0011,
+      "step": 9012
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.9577906976744187e-07,
+      "loss": 0.0015,
+      "step": 9013
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.957674418604651e-07,
+      "loss": 0.0016,
+      "step": 9014
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.9575581395348835e-07,
+      "loss": 0.0011,
+      "step": 9015
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.957441860465116e-07,
+      "loss": 0.0017,
+      "step": 9016
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.957325581395349e-07,
+      "loss": 0.0116,
+      "step": 9017
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.9572093023255813e-07,
+      "loss": 0.0014,
+      "step": 9018
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.957093023255814e-07,
+      "loss": 0.001,
+      "step": 9019
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.9569767441860463e-07,
+      "loss": 0.0024,
+      "step": 9020
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.9568604651162792e-07,
+      "loss": 0.0009,
+      "step": 9021
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.9567441860465115e-07,
+      "loss": 0.0014,
+      "step": 9022
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.956627906976744e-07,
+      "loss": 0.0018,
+      "step": 9023
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.9565116279069765e-07,
+      "loss": 0.002,
+      "step": 9024
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.9563953488372094e-07,
+      "loss": 0.0116,
+      "step": 9025
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.9562790697674417e-07,
+      "loss": 0.0012,
+      "step": 9026
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.9561627906976743e-07,
+      "loss": 0.0012,
+      "step": 9027
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.9560465116279067e-07,
+      "loss": 0.0014,
+      "step": 9028
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.9559302325581396e-07,
+      "loss": 0.001,
+      "step": 9029
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.9558139534883722e-07,
+      "loss": 0.5985,
+      "step": 9030
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.9556976744186045e-07,
+      "loss": 0.0024,
+      "step": 9031
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.9555813953488371e-07,
+      "loss": 0.0018,
+      "step": 9032
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.9554651162790695e-07,
+      "loss": 0.0013,
+      "step": 9033
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.9553488372093024e-07,
+      "loss": 0.001,
+      "step": 9034
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.9552325581395347e-07,
+      "loss": 0.0012,
+      "step": 9035
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.9551162790697674e-07,
+      "loss": 0.0014,
+      "step": 9036
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.9549999999999997e-07,
+      "loss": 0.0011,
+      "step": 9037
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.9548837209302326e-07,
+      "loss": 0.002,
+      "step": 9038
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.954767441860465e-07,
+      "loss": 0.0012,
+      "step": 9039
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.9546511627906976e-07,
+      "loss": 0.0012,
+      "step": 9040
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.95453488372093e-07,
+      "loss": 0.001,
+      "step": 9041
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.9544186046511628e-07,
+      "loss": 0.0014,
+      "step": 9042
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.9543023255813954e-07,
+      "loss": 0.0014,
+      "step": 9043
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.9541860465116278e-07,
+      "loss": 2.7142,
+      "step": 9044
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.9540697674418604e-07,
+      "loss": 0.0014,
+      "step": 9045
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.953953488372093e-07,
+      "loss": 0.0014,
+      "step": 9046
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.9538372093023256e-07,
+      "loss": 0.0018,
+      "step": 9047
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.953720930232558e-07,
+      "loss": 0.001,
+      "step": 9048
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.9536046511627906e-07,
+      "loss": 0.0013,
+      "step": 9049
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.9534883720930232e-07,
+      "loss": 0.0018,
+      "step": 9050
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.9533720930232558e-07,
+      "loss": 0.001,
+      "step": 9051
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.9532558139534882e-07,
+      "loss": 0.0011,
+      "step": 9052
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.9531395348837208e-07,
+      "loss": 0.0013,
+      "step": 9053
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.9530232558139537e-07,
+      "loss": 0.0014,
+      "step": 9054
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.952906976744186e-07,
+      "loss": 0.0012,
+      "step": 9055
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.9527906976744186e-07,
+      "loss": 0.0011,
+      "step": 9056
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.952674418604651e-07,
+      "loss": 0.0019,
+      "step": 9057
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.952558139534884e-07,
+      "loss": 0.0013,
+      "step": 9058
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.9524418604651162e-07,
+      "loss": 0.0013,
+      "step": 9059
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.9523255813953488e-07,
+      "loss": 0.0012,
+      "step": 9060
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.9522093023255812e-07,
+      "loss": 0.0013,
+      "step": 9061
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.9520930232558138e-07,
+      "loss": 0.0013,
+      "step": 9062
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.9519767441860464e-07,
+      "loss": 0.0011,
+      "step": 9063
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.951860465116279e-07,
+      "loss": 0.002,
+      "step": 9064
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.9517441860465114e-07,
+      "loss": 3.1236,
+      "step": 9065
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.951627906976744e-07,
+      "loss": 0.0013,
+      "step": 9066
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.951511627906977e-07,
+      "loss": 0.0011,
+      "step": 9067
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.9513953488372092e-07,
+      "loss": 0.2955,
+      "step": 9068
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.9512790697674416e-07,
+      "loss": 0.0016,
+      "step": 9069
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.9511627906976742e-07,
+      "loss": 0.0017,
+      "step": 9070
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.951046511627907e-07,
+      "loss": 0.0012,
+      "step": 9071
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.9509302325581395e-07,
+      "loss": 0.0009,
+      "step": 9072
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.950813953488372e-07,
+      "loss": 3.0026,
+      "step": 9073
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.9506976744186044e-07,
+      "loss": 0.0018,
+      "step": 9074
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.9505813953488373e-07,
+      "loss": 0.0019,
+      "step": 9075
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.9504651162790697e-07,
+      "loss": 0.0012,
+      "step": 9076
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.9503488372093023e-07,
+      "loss": 0.0012,
+      "step": 9077
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.9502325581395346e-07,
+      "loss": 0.0012,
+      "step": 9078
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.9501162790697675e-07,
+      "loss": 0.0027,
+      "step": 9079
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.9499999999999999e-07,
+      "loss": 0.0012,
+      "step": 9080
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.9498837209302325e-07,
+      "loss": 0.0012,
+      "step": 9081
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.9497674418604648e-07,
+      "loss": 0.0012,
+      "step": 9082
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.9496511627906977e-07,
+      "loss": 0.0009,
+      "step": 9083
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.9495348837209303e-07,
+      "loss": 0.0017,
+      "step": 9084
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.9494186046511627e-07,
+      "loss": 0.0012,
+      "step": 9085
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.9493023255813953e-07,
+      "loss": 0.0013,
+      "step": 9086
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.9491860465116276e-07,
+      "loss": 0.0011,
+      "step": 9087
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.9490697674418605e-07,
+      "loss": 0.0104,
+      "step": 9088
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.948953488372093e-07,
+      "loss": 2.0213,
+      "step": 9089
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.9488372093023255e-07,
+      "loss": 0.0011,
+      "step": 9090
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.9487209302325579e-07,
+      "loss": 0.0017,
+      "step": 9091
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.9486046511627907e-07,
+      "loss": 0.0013,
+      "step": 9092
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.948488372093023e-07,
+      "loss": 0.0012,
+      "step": 9093
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.9483720930232557e-07,
+      "loss": 0.0012,
+      "step": 9094
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.948255813953488e-07,
+      "loss": 0.002,
+      "step": 9095
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.948139534883721e-07,
+      "loss": 0.0014,
+      "step": 9096
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.9480232558139536e-07,
+      "loss": 0.0012,
+      "step": 9097
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.947906976744186e-07,
+      "loss": 0.0022,
+      "step": 9098
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.9477906976744183e-07,
+      "loss": 0.0031,
+      "step": 9099
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.9476744186046511e-07,
+      "loss": 0.0011,
+      "step": 9100
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.9475581395348838e-07,
+      "loss": 0.0025,
+      "step": 9101
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.947441860465116e-07,
+      "loss": 0.0024,
+      "step": 9102
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.9473255813953487e-07,
+      "loss": 0.0014,
+      "step": 9103
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.9472093023255813e-07,
+      "loss": 0.001,
+      "step": 9104
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.947093023255814e-07,
+      "loss": 0.0016,
+      "step": 9105
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.9469767441860463e-07,
+      "loss": 0.0023,
+      "step": 9106
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.946860465116279e-07,
+      "loss": 0.084,
+      "step": 9107
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.9467441860465118e-07,
+      "loss": 0.0012,
+      "step": 9108
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.9466279069767442e-07,
+      "loss": 0.0013,
+      "step": 9109
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.9465116279069765e-07,
+      "loss": 0.3368,
+      "step": 9110
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.9463953488372091e-07,
+      "loss": 0.001,
+      "step": 9111
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.9462790697674415e-07,
+      "loss": 0.0009,
+      "step": 9112
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.9461627906976744e-07,
+      "loss": 0.0012,
+      "step": 9113
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.946046511627907e-07,
+      "loss": 0.001,
+      "step": 9114
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.9459302325581393e-07,
+      "loss": 0.002,
+      "step": 9115
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.945813953488372e-07,
+      "loss": 0.0029,
+      "step": 9116
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.9456976744186046e-07,
+      "loss": 0.0013,
+      "step": 9117
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.9455813953488372e-07,
+      "loss": 0.0013,
+      "step": 9118
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.9454651162790695e-07,
+      "loss": 0.0011,
+      "step": 9119
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.9453488372093022e-07,
+      "loss": 0.0016,
+      "step": 9120
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.9452325581395348e-07,
+      "loss": 0.0016,
+      "step": 9121
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.9451162790697674e-07,
+      "loss": 0.0237,
+      "step": 9122
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.9449999999999997e-07,
+      "loss": 0.0013,
+      "step": 9123
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.9448837209302324e-07,
+      "loss": 0.0012,
+      "step": 9124
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.9447674418604652e-07,
+      "loss": 0.0014,
+      "step": 9125
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.9446511627906976e-07,
+      "loss": 0.0015,
+      "step": 9126
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.9445348837209302e-07,
+      "loss": 0.2118,
+      "step": 9127
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.9444186046511626e-07,
+      "loss": 0.0016,
+      "step": 9128
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.9443023255813955e-07,
+      "loss": 0.0019,
+      "step": 9129
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.9441860465116278e-07,
+      "loss": 0.0016,
+      "step": 9130
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.9440697674418604e-07,
+      "loss": 0.0011,
+      "step": 9131
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.9439534883720928e-07,
+      "loss": 3.0718,
+      "step": 9132
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.9438372093023257e-07,
+      "loss": 0.0012,
+      "step": 9133
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.943720930232558e-07,
+      "loss": 0.1413,
+      "step": 9134
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.9436046511627906e-07,
+      "loss": 0.0011,
+      "step": 9135
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.943488372093023e-07,
+      "loss": 0.0025,
+      "step": 9136
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.9433720930232559e-07,
+      "loss": 0.001,
+      "step": 9137
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.9432558139534885e-07,
+      "loss": 0.0016,
+      "step": 9138
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.9431395348837208e-07,
+      "loss": 0.001,
+      "step": 9139
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.9430232558139534e-07,
+      "loss": 0.0111,
+      "step": 9140
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.9429069767441858e-07,
+      "loss": 0.0011,
+      "step": 9141
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.9427906976744187e-07,
+      "loss": 0.0012,
+      "step": 9142
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.942674418604651e-07,
+      "loss": 0.0009,
+      "step": 9143
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.9425581395348836e-07,
+      "loss": 0.001,
+      "step": 9144
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.942441860465116e-07,
+      "loss": 0.0014,
+      "step": 9145
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.942325581395349e-07,
+      "loss": 0.0017,
+      "step": 9146
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.9422093023255812e-07,
+      "loss": 0.6386,
+      "step": 9147
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.9420930232558139e-07,
+      "loss": 0.0013,
+      "step": 9148
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.9419767441860462e-07,
+      "loss": 0.0012,
+      "step": 9149
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.941860465116279e-07,
+      "loss": 0.001,
+      "step": 9150
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.9417441860465117e-07,
+      "loss": 0.0011,
+      "step": 9151
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.941627906976744e-07,
+      "loss": 0.0107,
+      "step": 9152
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.9415116279069764e-07,
+      "loss": 0.0011,
+      "step": 9153
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.9413953488372093e-07,
+      "loss": 0.0872,
+      "step": 9154
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.941279069767442e-07,
+      "loss": 0.0011,
+      "step": 9155
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.9411627906976743e-07,
+      "loss": 0.0032,
+      "step": 9156
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.941046511627907e-07,
+      "loss": 0.0014,
+      "step": 9157
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.9409302325581395e-07,
+      "loss": 0.0015,
+      "step": 9158
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.940813953488372e-07,
+      "loss": 0.0334,
+      "step": 9159
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.9406976744186045e-07,
+      "loss": 0.002,
+      "step": 9160
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.940581395348837e-07,
+      "loss": 0.0012,
+      "step": 9161
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.94046511627907e-07,
+      "loss": 0.0014,
+      "step": 9162
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.9403488372093023e-07,
+      "loss": 0.3032,
+      "step": 9163
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.9402325581395347e-07,
+      "loss": 0.0017,
+      "step": 9164
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.9401162790697673e-07,
+      "loss": 0.002,
+      "step": 9165
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.9399999999999996e-07,
+      "loss": 0.0433,
+      "step": 9166
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.9398837209302325e-07,
+      "loss": 0.0014,
+      "step": 9167
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.9397674418604651e-07,
+      "loss": 0.7618,
+      "step": 9168
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.9396511627906975e-07,
+      "loss": 0.0016,
+      "step": 9169
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.93953488372093e-07,
+      "loss": 0.0017,
+      "step": 9170
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.9394186046511627e-07,
+      "loss": 0.3615,
+      "step": 9171
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.9393023255813953e-07,
+      "loss": 0.0012,
+      "step": 9172
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.9391860465116277e-07,
+      "loss": 0.0024,
+      "step": 9173
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.9390697674418603e-07,
+      "loss": 0.0064,
+      "step": 9174
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.938953488372093e-07,
+      "loss": 0.0015,
+      "step": 9175
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.9388372093023255e-07,
+      "loss": 0.0012,
+      "step": 9176
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.938720930232558e-07,
+      "loss": 0.001,
+      "step": 9177
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.9386046511627905e-07,
+      "loss": 0.0012,
+      "step": 9178
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.9384883720930234e-07,
+      "loss": 0.0012,
+      "step": 9179
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.9383720930232557e-07,
+      "loss": 0.0015,
+      "step": 9180
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.9382558139534884e-07,
+      "loss": 0.0015,
+      "step": 9181
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.9381395348837207e-07,
+      "loss": 0.0015,
+      "step": 9182
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.9380232558139536e-07,
+      "loss": 0.0011,
+      "step": 9183
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.937906976744186e-07,
+      "loss": 0.0019,
+      "step": 9184
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.9377906976744186e-07,
+      "loss": 0.001,
+      "step": 9185
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.937674418604651e-07,
+      "loss": 0.005,
+      "step": 9186
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.9375581395348838e-07,
+      "loss": 0.0014,
+      "step": 9187
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.9374418604651162e-07,
+      "loss": 0.0014,
+      "step": 9188
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.9373255813953488e-07,
+      "loss": 0.0014,
+      "step": 9189
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.937209302325581e-07,
+      "loss": 0.0013,
+      "step": 9190
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.937093023255814e-07,
+      "loss": 0.0012,
+      "step": 9191
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.9369767441860466e-07,
+      "loss": 0.0015,
+      "step": 9192
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.936860465116279e-07,
+      "loss": 0.0012,
+      "step": 9193
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.9367441860465116e-07,
+      "loss": 0.0011,
+      "step": 9194
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.936627906976744e-07,
+      "loss": 0.0008,
+      "step": 9195
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.9365116279069768e-07,
+      "loss": 0.0012,
+      "step": 9196
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.9363953488372092e-07,
+      "loss": 0.0016,
+      "step": 9197
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.9362790697674418e-07,
+      "loss": 0.0021,
+      "step": 9198
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.9361627906976741e-07,
+      "loss": 0.0267,
+      "step": 9199
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.936046511627907e-07,
+      "loss": 0.0106,
+      "step": 9200
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.9359302325581394e-07,
+      "loss": 0.0012,
+      "step": 9201
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.935813953488372e-07,
+      "loss": 0.0492,
+      "step": 9202
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.9356976744186044e-07,
+      "loss": 0.0023,
+      "step": 9203
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.9355813953488372e-07,
+      "loss": 0.0012,
+      "step": 9204
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.9354651162790699e-07,
+      "loss": 0.0012,
+      "step": 9205
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.9353488372093022e-07,
+      "loss": 0.0042,
+      "step": 9206
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.9352325581395346e-07,
+      "loss": 0.0011,
+      "step": 9207
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.9351162790697674e-07,
+      "loss": 0.0014,
+      "step": 9208
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.935e-07,
+      "loss": 0.001,
+      "step": 9209
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.9348837209302324e-07,
+      "loss": 0.0018,
+      "step": 9210
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.934767441860465e-07,
+      "loss": 0.0013,
+      "step": 9211
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.9346511627906976e-07,
+      "loss": 0.0014,
+      "step": 9212
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.9345348837209303e-07,
+      "loss": 0.0019,
+      "step": 9213
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.9344186046511626e-07,
+      "loss": 0.0009,
+      "step": 9214
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.9343023255813952e-07,
+      "loss": 0.001,
+      "step": 9215
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.9341860465116278e-07,
+      "loss": 0.0011,
+      "step": 9216
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.9340697674418605e-07,
+      "loss": 0.0015,
+      "step": 9217
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.9339534883720928e-07,
+      "loss": 0.0009,
+      "step": 9218
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.9338372093023254e-07,
+      "loss": 0.0012,
+      "step": 9219
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.9337209302325578e-07,
+      "loss": 0.0014,
+      "step": 9220
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.9336046511627907e-07,
+      "loss": 0.0018,
+      "step": 9221
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.9334883720930233e-07,
+      "loss": 0.0011,
+      "step": 9222
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.9333720930232556e-07,
+      "loss": 0.001,
+      "step": 9223
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.9332558139534883e-07,
+      "loss": 0.002,
+      "step": 9224
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.933139534883721e-07,
+      "loss": 0.0018,
+      "step": 9225
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.9330232558139535e-07,
+      "loss": 0.0014,
+      "step": 9226
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.9329069767441858e-07,
+      "loss": 0.0017,
+      "step": 9227
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.9327906976744185e-07,
+      "loss": 0.001,
+      "step": 9228
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.932674418604651e-07,
+      "loss": 0.0011,
+      "step": 9229
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.9325581395348837e-07,
+      "loss": 0.0012,
+      "step": 9230
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.932441860465116e-07,
+      "loss": 0.0014,
+      "step": 9231
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.9323255813953487e-07,
+      "loss": 0.0015,
+      "step": 9232
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.9322093023255815e-07,
+      "loss": 0.0013,
+      "step": 9233
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.932093023255814e-07,
+      "loss": 0.0014,
+      "step": 9234
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.9319767441860465e-07,
+      "loss": 0.001,
+      "step": 9235
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.9318604651162789e-07,
+      "loss": 0.0009,
+      "step": 9236
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.9317441860465117e-07,
+      "loss": 0.0012,
+      "step": 9237
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.931627906976744e-07,
+      "loss": 0.0022,
+      "step": 9238
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.9315116279069767e-07,
+      "loss": 0.0016,
+      "step": 9239
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.931395348837209e-07,
+      "loss": 0.0017,
+      "step": 9240
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.931279069767442e-07,
+      "loss": 2.362,
+      "step": 9241
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.9311627906976743e-07,
+      "loss": 0.0011,
+      "step": 9242
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.931046511627907e-07,
+      "loss": 0.0012,
+      "step": 9243
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.9309302325581393e-07,
+      "loss": 0.0012,
+      "step": 9244
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.9308139534883722e-07,
+      "loss": 0.0014,
+      "step": 9245
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.9306976744186048e-07,
+      "loss": 0.001,
+      "step": 9246
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.930581395348837e-07,
+      "loss": 0.0016,
+      "step": 9247
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.9304651162790695e-07,
+      "loss": 0.0028,
+      "step": 9248
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.930348837209302e-07,
+      "loss": 0.001,
+      "step": 9249
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.930232558139535e-07,
+      "loss": 0.0012,
+      "step": 9250
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.9301162790697673e-07,
+      "loss": 0.0017,
+      "step": 9251
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.93e-07,
+      "loss": 0.0014,
+      "step": 9252
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.9298837209302323e-07,
+      "loss": 0.258,
+      "step": 9253
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.9297674418604652e-07,
+      "loss": 0.001,
+      "step": 9254
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.9296511627906975e-07,
+      "loss": 0.006,
+      "step": 9255
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.9295348837209301e-07,
+      "loss": 0.004,
+      "step": 9256
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.9294186046511625e-07,
+      "loss": 0.0013,
+      "step": 9257
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.9293023255813954e-07,
+      "loss": 0.0015,
+      "step": 9258
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.9291860465116277e-07,
+      "loss": 0.0011,
+      "step": 9259
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.9290697674418604e-07,
+      "loss": 0.0012,
+      "step": 9260
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.9289534883720927e-07,
+      "loss": 0.0177,
+      "step": 9261
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.9288372093023256e-07,
+      "loss": 0.219,
+      "step": 9262
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.9287209302325582e-07,
+      "loss": 0.3059,
+      "step": 9263
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.9286046511627906e-07,
+      "loss": 0.0012,
+      "step": 9264
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.9284883720930232e-07,
+      "loss": 0.0017,
+      "step": 9265
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.9283720930232558e-07,
+      "loss": 0.0149,
+      "step": 9266
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.9282558139534884e-07,
+      "loss": 0.0012,
+      "step": 9267
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.9281395348837208e-07,
+      "loss": 0.0013,
+      "step": 9268
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.9280232558139534e-07,
+      "loss": 0.0016,
+      "step": 9269
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.927906976744186e-07,
+      "loss": 0.0012,
+      "step": 9270
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.9277906976744186e-07,
+      "loss": 0.0013,
+      "step": 9271
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.927674418604651e-07,
+      "loss": 0.0038,
+      "step": 9272
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.9275581395348836e-07,
+      "loss": 0.002,
+      "step": 9273
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.927441860465116e-07,
+      "loss": 1.3324,
+      "step": 9274
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.9273255813953488e-07,
+      "loss": 0.0011,
+      "step": 9275
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.9272093023255814e-07,
+      "loss": 0.001,
+      "step": 9276
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.9270930232558138e-07,
+      "loss": 0.0013,
+      "step": 9277
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.9269767441860464e-07,
+      "loss": 0.0081,
+      "step": 9278
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.926860465116279e-07,
+      "loss": 0.0008,
+      "step": 9279
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.9267441860465116e-07,
+      "loss": 0.2807,
+      "step": 9280
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.926627906976744e-07,
+      "loss": 0.0011,
+      "step": 9281
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.9265116279069766e-07,
+      "loss": 0.0013,
+      "step": 9282
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.9263953488372092e-07,
+      "loss": 0.0209,
+      "step": 9283
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.9262790697674418e-07,
+      "loss": 0.0015,
+      "step": 9284
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.9261627906976742e-07,
+      "loss": 0.0011,
+      "step": 9285
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.9260465116279068e-07,
+      "loss": 0.0021,
+      "step": 9286
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.9259302325581397e-07,
+      "loss": 0.0011,
+      "step": 9287
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.925813953488372e-07,
+      "loss": 0.0059,
+      "step": 9288
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.9256976744186047e-07,
+      "loss": 0.0011,
+      "step": 9289
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.925581395348837e-07,
+      "loss": 0.001,
+      "step": 9290
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.92546511627907e-07,
+      "loss": 0.0013,
+      "step": 9291
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.9253488372093022e-07,
+      "loss": 0.0127,
+      "step": 9292
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.9252325581395349e-07,
+      "loss": 0.001,
+      "step": 9293
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.9251162790697672e-07,
+      "loss": 0.0008,
+      "step": 9294
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.925e-07,
+      "loss": 0.001,
+      "step": 9295
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.9248837209302325e-07,
+      "loss": 3.217,
+      "step": 9296
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.924767441860465e-07,
+      "loss": 0.0018,
+      "step": 9297
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.9246511627906974e-07,
+      "loss": 0.001,
+      "step": 9298
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.92453488372093e-07,
+      "loss": 0.0016,
+      "step": 9299
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.924418604651163e-07,
+      "loss": 0.001,
+      "step": 9300
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.9243023255813953e-07,
+      "loss": 0.0011,
+      "step": 9301
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.9241860465116276e-07,
+      "loss": 0.0013,
+      "step": 9302
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.9240697674418602e-07,
+      "loss": 0.0014,
+      "step": 9303
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.923953488372093e-07,
+      "loss": 0.001,
+      "step": 9304
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.9238372093023255e-07,
+      "loss": 0.0016,
+      "step": 9305
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.923720930232558e-07,
+      "loss": 0.0011,
+      "step": 9306
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.988394584139265,
+      "eval_loss": 0.15373294055461884,
+      "eval_runtime": 1.5607,
+      "eval_samples_per_second": 662.523,
+      "eval_steps_per_second": 10.893,
+      "step": 9306
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.9236046511627904e-07,
+      "loss": 0.0014,
+      "step": 9307
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.9234883720930233e-07,
+      "loss": 0.0013,
+      "step": 9308
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.9233720930232557e-07,
+      "loss": 0.0014,
+      "step": 9309
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.9232558139534883e-07,
+      "loss": 0.001,
+      "step": 9310
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.9231395348837206e-07,
+      "loss": 0.0012,
+      "step": 9311
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.9230232558139535e-07,
+      "loss": 0.0009,
+      "step": 9312
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.922906976744186e-07,
+      "loss": 0.0016,
+      "step": 9313
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.9227906976744185e-07,
+      "loss": 0.0014,
+      "step": 9314
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.9226744186046509e-07,
+      "loss": 0.0012,
+      "step": 9315
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.9225581395348837e-07,
+      "loss": 0.0011,
+      "step": 9316
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.9224418604651164e-07,
+      "loss": 0.0013,
+      "step": 9317
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.9223255813953487e-07,
+      "loss": 0.0034,
+      "step": 9318
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.9222093023255813e-07,
+      "loss": 0.0311,
+      "step": 9319
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.922093023255814e-07,
+      "loss": 0.001,
+      "step": 9320
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.9219767441860466e-07,
+      "loss": 0.5405,
+      "step": 9321
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.921860465116279e-07,
+      "loss": 0.0068,
+      "step": 9322
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.9217441860465115e-07,
+      "loss": 0.0012,
+      "step": 9323
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.9216279069767441e-07,
+      "loss": 0.0022,
+      "step": 9324
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.9215116279069768e-07,
+      "loss": 0.0012,
+      "step": 9325
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.921395348837209e-07,
+      "loss": 0.0017,
+      "step": 9326
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.9212790697674417e-07,
+      "loss": 0.0017,
+      "step": 9327
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.921162790697674e-07,
+      "loss": 0.0023,
+      "step": 9328
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.921046511627907e-07,
+      "loss": 0.0013,
+      "step": 9329
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.9209302325581396e-07,
+      "loss": 0.0015,
+      "step": 9330
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.920813953488372e-07,
+      "loss": 0.0013,
+      "step": 9331
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.9206976744186045e-07,
+      "loss": 0.001,
+      "step": 9332
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.9205813953488372e-07,
+      "loss": 0.0009,
+      "step": 9333
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.9204651162790698e-07,
+      "loss": 0.0009,
+      "step": 9334
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.9203488372093021e-07,
+      "loss": 0.0019,
+      "step": 9335
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.9202325581395348e-07,
+      "loss": 0.0011,
+      "step": 9336
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.9201162790697674e-07,
+      "loss": 0.002,
+      "step": 9337
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.92e-07,
+      "loss": 0.0013,
+      "step": 9338
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.9198837209302323e-07,
+      "loss": 0.0018,
+      "step": 9339
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.919767441860465e-07,
+      "loss": 0.0598,
+      "step": 9340
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.9196511627906978e-07,
+      "loss": 0.0014,
+      "step": 9341
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.9195348837209302e-07,
+      "loss": 0.0009,
+      "step": 9342
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.9194186046511628e-07,
+      "loss": 0.0012,
+      "step": 9343
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.9193023255813952e-07,
+      "loss": 0.0012,
+      "step": 9344
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.919186046511628e-07,
+      "loss": 0.0011,
+      "step": 9345
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.9190697674418604e-07,
+      "loss": 0.0016,
+      "step": 9346
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.918953488372093e-07,
+      "loss": 0.0014,
+      "step": 9347
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.9188372093023254e-07,
+      "loss": 0.0015,
+      "step": 9348
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.9187209302325582e-07,
+      "loss": 0.0215,
+      "step": 9349
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.9186046511627906e-07,
+      "loss": 0.0016,
+      "step": 9350
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.9184883720930232e-07,
+      "loss": 0.0013,
+      "step": 9351
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.9183720930232556e-07,
+      "loss": 0.0014,
+      "step": 9352
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.9182558139534882e-07,
+      "loss": 0.0013,
+      "step": 9353
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.9181395348837208e-07,
+      "loss": 0.0013,
+      "step": 9354
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.9180232558139534e-07,
+      "loss": 0.0015,
+      "step": 9355
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.9179069767441858e-07,
+      "loss": 0.0013,
+      "step": 9356
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.9177906976744184e-07,
+      "loss": 0.0246,
+      "step": 9357
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.9176744186046513e-07,
+      "loss": 0.0011,
+      "step": 9358
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.9175581395348836e-07,
+      "loss": 0.0008,
+      "step": 9359
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.9174418604651162e-07,
+      "loss": 0.0011,
+      "step": 9360
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.9173255813953486e-07,
+      "loss": 0.0012,
+      "step": 9361
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.9172093023255815e-07,
+      "loss": 0.001,
+      "step": 9362
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.9170930232558138e-07,
+      "loss": 0.0008,
+      "step": 9363
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.9169767441860464e-07,
+      "loss": 0.001,
+      "step": 9364
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.9168604651162788e-07,
+      "loss": 0.0011,
+      "step": 9365
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.9167441860465117e-07,
+      "loss": 0.9864,
+      "step": 9366
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.916627906976744e-07,
+      "loss": 0.0009,
+      "step": 9367
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.9165116279069766e-07,
+      "loss": 0.0126,
+      "step": 9368
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.916395348837209e-07,
+      "loss": 0.0011,
+      "step": 9369
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.916279069767442e-07,
+      "loss": 0.0011,
+      "step": 9370
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.9161627906976745e-07,
+      "loss": 0.0013,
+      "step": 9371
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.9160465116279069e-07,
+      "loss": 0.0012,
+      "step": 9372
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.9159302325581395e-07,
+      "loss": 0.0012,
+      "step": 9373
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.915813953488372e-07,
+      "loss": 0.0015,
+      "step": 9374
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.9156976744186047e-07,
+      "loss": 0.0011,
+      "step": 9375
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.915581395348837e-07,
+      "loss": 0.0015,
+      "step": 9376
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.9154651162790697e-07,
+      "loss": 0.002,
+      "step": 9377
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.9153488372093023e-07,
+      "loss": 2.0944,
+      "step": 9378
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.915232558139535e-07,
+      "loss": 0.0015,
+      "step": 9379
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.9151162790697673e-07,
+      "loss": 0.0013,
+      "step": 9380
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.915e-07,
+      "loss": 0.0014,
+      "step": 9381
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.9148837209302322e-07,
+      "loss": 0.0014,
+      "step": 9382
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.914767441860465e-07,
+      "loss": 0.0011,
+      "step": 9383
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.9146511627906977e-07,
+      "loss": 0.0015,
+      "step": 9384
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.91453488372093e-07,
+      "loss": 0.0015,
+      "step": 9385
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.9144186046511624e-07,
+      "loss": 0.0014,
+      "step": 9386
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.9143023255813953e-07,
+      "loss": 0.2172,
+      "step": 9387
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.914186046511628e-07,
+      "loss": 0.0012,
+      "step": 9388
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.9140697674418603e-07,
+      "loss": 0.0016,
+      "step": 9389
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.913953488372093e-07,
+      "loss": 0.0012,
+      "step": 9390
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.9138372093023255e-07,
+      "loss": 0.0014,
+      "step": 9391
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.9137209302325581e-07,
+      "loss": 0.0036,
+      "step": 9392
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.9136046511627905e-07,
+      "loss": 0.001,
+      "step": 9393
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.913488372093023e-07,
+      "loss": 0.0021,
+      "step": 9394
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.913372093023256e-07,
+      "loss": 0.001,
+      "step": 9395
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.9132558139534883e-07,
+      "loss": 0.0031,
+      "step": 9396
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.9131395348837207e-07,
+      "loss": 0.001,
+      "step": 9397
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.9130232558139533e-07,
+      "loss": 0.0018,
+      "step": 9398
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.9129069767441862e-07,
+      "loss": 0.001,
+      "step": 9399
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.9127906976744185e-07,
+      "loss": 0.001,
+      "step": 9400
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.9126744186046512e-07,
+      "loss": 0.0014,
+      "step": 9401
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.9125581395348835e-07,
+      "loss": 0.0049,
+      "step": 9402
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.9124418604651164e-07,
+      "loss": 0.0011,
+      "step": 9403
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.9123255813953487e-07,
+      "loss": 0.0018,
+      "step": 9404
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.9122093023255814e-07,
+      "loss": 0.0028,
+      "step": 9405
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.9120930232558137e-07,
+      "loss": 0.0009,
+      "step": 9406
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.9119767441860463e-07,
+      "loss": 0.5497,
+      "step": 9407
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.911860465116279e-07,
+      "loss": 0.0014,
+      "step": 9408
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.9117441860465116e-07,
+      "loss": 0.1032,
+      "step": 9409
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.911627906976744e-07,
+      "loss": 0.0023,
+      "step": 9410
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.9115116279069765e-07,
+      "loss": 0.0017,
+      "step": 9411
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.9113953488372094e-07,
+      "loss": 0.0013,
+      "step": 9412
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.9112790697674418e-07,
+      "loss": 0.0024,
+      "step": 9413
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.9111627906976744e-07,
+      "loss": 0.0025,
+      "step": 9414
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.9110465116279067e-07,
+      "loss": 0.0072,
+      "step": 9415
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.9109302325581396e-07,
+      "loss": 0.0015,
+      "step": 9416
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.910813953488372e-07,
+      "loss": 0.0024,
+      "step": 9417
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.9106976744186046e-07,
+      "loss": 0.0012,
+      "step": 9418
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.910581395348837e-07,
+      "loss": 0.0009,
+      "step": 9419
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.9104651162790698e-07,
+      "loss": 0.0015,
+      "step": 9420
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.9103488372093022e-07,
+      "loss": 1.3356,
+      "step": 9421
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.9102325581395348e-07,
+      "loss": 0.0015,
+      "step": 9422
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.9101162790697671e-07,
+      "loss": 0.0013,
+      "step": 9423
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.91e-07,
+      "loss": 0.0008,
+      "step": 9424
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.9098837209302326e-07,
+      "loss": 0.001,
+      "step": 9425
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.909767441860465e-07,
+      "loss": 0.0014,
+      "step": 9426
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.9096511627906976e-07,
+      "loss": 0.0024,
+      "step": 9427
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.9095348837209302e-07,
+      "loss": 0.0078,
+      "step": 9428
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.9094186046511629e-07,
+      "loss": 0.0023,
+      "step": 9429
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.9093023255813952e-07,
+      "loss": 0.0013,
+      "step": 9430
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.9091860465116278e-07,
+      "loss": 0.0014,
+      "step": 9431
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.9090697674418602e-07,
+      "loss": 0.002,
+      "step": 9432
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.908953488372093e-07,
+      "loss": 0.0012,
+      "step": 9433
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.9088372093023254e-07,
+      "loss": 0.0009,
+      "step": 9434
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.908720930232558e-07,
+      "loss": 0.5762,
+      "step": 9435
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.9086046511627904e-07,
+      "loss": 0.0017,
+      "step": 9436
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.9084883720930233e-07,
+      "loss": 0.0012,
+      "step": 9437
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.908372093023256e-07,
+      "loss": 0.0033,
+      "step": 9438
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.9082558139534882e-07,
+      "loss": 0.0009,
+      "step": 9439
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.9081395348837206e-07,
+      "loss": 0.0013,
+      "step": 9440
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.9080232558139535e-07,
+      "loss": 0.0013,
+      "step": 9441
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.907906976744186e-07,
+      "loss": 0.0012,
+      "step": 9442
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.9077906976744184e-07,
+      "loss": 0.0008,
+      "step": 9443
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.907674418604651e-07,
+      "loss": 0.0012,
+      "step": 9444
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.9075581395348837e-07,
+      "loss": 0.0011,
+      "step": 9445
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.9074418604651163e-07,
+      "loss": 0.0011,
+      "step": 9446
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.9073255813953486e-07,
+      "loss": 0.0012,
+      "step": 9447
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.9072093023255813e-07,
+      "loss": 0.001,
+      "step": 9448
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.9070930232558141e-07,
+      "loss": 0.0012,
+      "step": 9449
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.9069767441860465e-07,
+      "loss": 0.0016,
+      "step": 9450
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.9068604651162788e-07,
+      "loss": 0.0029,
+      "step": 9451
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.9067441860465115e-07,
+      "loss": 0.0476,
+      "step": 9452
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.9066279069767443e-07,
+      "loss": 0.0013,
+      "step": 9453
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.9065116279069767e-07,
+      "loss": 0.0011,
+      "step": 9454
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.9063953488372093e-07,
+      "loss": 0.0008,
+      "step": 9455
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.9062790697674417e-07,
+      "loss": 0.0011,
+      "step": 9456
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.9061627906976745e-07,
+      "loss": 0.0014,
+      "step": 9457
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.906046511627907e-07,
+      "loss": 0.0019,
+      "step": 9458
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.9059302325581395e-07,
+      "loss": 0.009,
+      "step": 9459
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.9058139534883719e-07,
+      "loss": 0.0012,
+      "step": 9460
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.9056976744186045e-07,
+      "loss": 0.0011,
+      "step": 9461
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.905581395348837e-07,
+      "loss": 0.0011,
+      "step": 9462
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.9054651162790697e-07,
+      "loss": 0.0012,
+      "step": 9463
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.905348837209302e-07,
+      "loss": 0.0012,
+      "step": 9464
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.9052325581395347e-07,
+      "loss": 0.0012,
+      "step": 9465
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.9051162790697676e-07,
+      "loss": 0.0013,
+      "step": 9466
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.905e-07,
+      "loss": 0.0014,
+      "step": 9467
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.9048837209302325e-07,
+      "loss": 0.001,
+      "step": 9468
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.904767441860465e-07,
+      "loss": 0.0017,
+      "step": 9469
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.9046511627906978e-07,
+      "loss": 0.0011,
+      "step": 9470
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.90453488372093e-07,
+      "loss": 0.0011,
+      "step": 9471
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.9044186046511627e-07,
+      "loss": 0.0009,
+      "step": 9472
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.904302325581395e-07,
+      "loss": 0.0017,
+      "step": 9473
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.904186046511628e-07,
+      "loss": 3.0662,
+      "step": 9474
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.9040697674418603e-07,
+      "loss": 0.0016,
+      "step": 9475
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.903953488372093e-07,
+      "loss": 0.0021,
+      "step": 9476
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.9038372093023253e-07,
+      "loss": 0.0016,
+      "step": 9477
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.9037209302325582e-07,
+      "loss": 0.0011,
+      "step": 9478
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.9036046511627908e-07,
+      "loss": 0.0025,
+      "step": 9479
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.9034883720930231e-07,
+      "loss": 0.0011,
+      "step": 9480
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.9033720930232555e-07,
+      "loss": 0.0014,
+      "step": 9481
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.9032558139534884e-07,
+      "loss": 0.0017,
+      "step": 9482
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.903139534883721e-07,
+      "loss": 0.001,
+      "step": 9483
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.9030232558139534e-07,
+      "loss": 0.001,
+      "step": 9484
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.902906976744186e-07,
+      "loss": 0.0013,
+      "step": 9485
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.9027906976744183e-07,
+      "loss": 0.0012,
+      "step": 9486
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.9026744186046512e-07,
+      "loss": 0.7394,
+      "step": 9487
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.9025581395348836e-07,
+      "loss": 0.0018,
+      "step": 9488
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.9024418604651162e-07,
+      "loss": 0.0008,
+      "step": 9489
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.9023255813953485e-07,
+      "loss": 0.0011,
+      "step": 9490
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.9022093023255814e-07,
+      "loss": 0.1581,
+      "step": 9491
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.9020930232558138e-07,
+      "loss": 0.0014,
+      "step": 9492
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.9019767441860464e-07,
+      "loss": 0.0009,
+      "step": 9493
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.9018604651162787e-07,
+      "loss": 0.0008,
+      "step": 9494
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.9017441860465116e-07,
+      "loss": 0.0017,
+      "step": 9495
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.9016279069767442e-07,
+      "loss": 0.0011,
+      "step": 9496
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.9015116279069766e-07,
+      "loss": 0.0034,
+      "step": 9497
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.9013953488372092e-07,
+      "loss": 0.0018,
+      "step": 9498
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.9012790697674418e-07,
+      "loss": 0.0011,
+      "step": 9499
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.9011627906976744e-07,
+      "loss": 0.0014,
+      "step": 9500
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.9010465116279068e-07,
+      "loss": 0.0008,
+      "step": 9501
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.9009302325581394e-07,
+      "loss": 0.001,
+      "step": 9502
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.900813953488372e-07,
+      "loss": 0.0012,
+      "step": 9503
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.9006976744186046e-07,
+      "loss": 0.0011,
+      "step": 9504
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.900581395348837e-07,
+      "loss": 0.0012,
+      "step": 9505
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.9004651162790696e-07,
+      "loss": 0.0021,
+      "step": 9506
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.9003488372093025e-07,
+      "loss": 0.0011,
+      "step": 9507
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.9002325581395348e-07,
+      "loss": 0.0013,
+      "step": 9508
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.9001162790697675e-07,
+      "loss": 0.001,
+      "step": 9509
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.8999999999999998e-07,
+      "loss": 0.0011,
+      "step": 9510
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.8998837209302327e-07,
+      "loss": 0.0019,
+      "step": 9511
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.899767441860465e-07,
+      "loss": 0.6834,
+      "step": 9512
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.8996511627906977e-07,
+      "loss": 0.0894,
+      "step": 9513
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.89953488372093e-07,
+      "loss": 0.0014,
+      "step": 9514
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.8994186046511626e-07,
+      "loss": 0.0011,
+      "step": 9515
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.8993023255813952e-07,
+      "loss": 0.0017,
+      "step": 9516
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.8991860465116279e-07,
+      "loss": 0.0012,
+      "step": 9517
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.8990697674418602e-07,
+      "loss": 2.4493,
+      "step": 9518
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.8989534883720928e-07,
+      "loss": 0.0013,
+      "step": 9519
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.8988372093023257e-07,
+      "loss": 0.0011,
+      "step": 9520
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.898720930232558e-07,
+      "loss": 0.0012,
+      "step": 9521
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.8986046511627907e-07,
+      "loss": 0.0011,
+      "step": 9522
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.898488372093023e-07,
+      "loss": 0.0013,
+      "step": 9523
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.898372093023256e-07,
+      "loss": 0.0014,
+      "step": 9524
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.8982558139534883e-07,
+      "loss": 0.0009,
+      "step": 9525
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.898139534883721e-07,
+      "loss": 0.0009,
+      "step": 9526
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.8980232558139532e-07,
+      "loss": 0.0043,
+      "step": 9527
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.897906976744186e-07,
+      "loss": 0.1002,
+      "step": 9528
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.8977906976744185e-07,
+      "loss": 0.0026,
+      "step": 9529
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.897674418604651e-07,
+      "loss": 1.0419,
+      "step": 9530
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.8975581395348834e-07,
+      "loss": 0.001,
+      "step": 9531
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.8974418604651163e-07,
+      "loss": 0.0012,
+      "step": 9532
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.897325581395349e-07,
+      "loss": 0.0008,
+      "step": 9533
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.8972093023255813e-07,
+      "loss": 0.0013,
+      "step": 9534
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.8970930232558136e-07,
+      "loss": 0.0011,
+      "step": 9535
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.8969767441860465e-07,
+      "loss": 3.1786,
+      "step": 9536
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.8968604651162791e-07,
+      "loss": 0.0011,
+      "step": 9537
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.8967441860465115e-07,
+      "loss": 0.0015,
+      "step": 9538
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.896627906976744e-07,
+      "loss": 0.001,
+      "step": 9539
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.8965116279069765e-07,
+      "loss": 0.0011,
+      "step": 9540
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.8963953488372094e-07,
+      "loss": 0.0017,
+      "step": 9541
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.8962790697674417e-07,
+      "loss": 0.0012,
+      "step": 9542
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.8961627906976743e-07,
+      "loss": 0.001,
+      "step": 9543
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.8960465116279067e-07,
+      "loss": 0.0014,
+      "step": 9544
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.8959302325581396e-07,
+      "loss": 0.0016,
+      "step": 9545
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.895813953488372e-07,
+      "loss": 0.0081,
+      "step": 9546
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.8956976744186045e-07,
+      "loss": 0.0012,
+      "step": 9547
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.895581395348837e-07,
+      "loss": 0.0011,
+      "step": 9548
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.8954651162790698e-07,
+      "loss": 0.0009,
+      "step": 9549
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.8953488372093024e-07,
+      "loss": 0.0012,
+      "step": 9550
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.8952325581395347e-07,
+      "loss": 0.0013,
+      "step": 9551
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.8951162790697673e-07,
+      "loss": 0.0021,
+      "step": 9552
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.895e-07,
+      "loss": 0.0014,
+      "step": 9553
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.8948837209302326e-07,
+      "loss": 0.0012,
+      "step": 9554
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.894767441860465e-07,
+      "loss": 0.0013,
+      "step": 9555
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.8946511627906975e-07,
+      "loss": 0.0024,
+      "step": 9556
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.8945348837209302e-07,
+      "loss": 0.0014,
+      "step": 9557
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.8944186046511628e-07,
+      "loss": 0.0013,
+      "step": 9558
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.8943023255813951e-07,
+      "loss": 0.0012,
+      "step": 9559
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.8941860465116278e-07,
+      "loss": 0.0018,
+      "step": 9560
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.8940697674418606e-07,
+      "loss": 0.0017,
+      "step": 9561
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.893953488372093e-07,
+      "loss": 0.0011,
+      "step": 9562
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.8938372093023256e-07,
+      "loss": 0.0043,
+      "step": 9563
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.893720930232558e-07,
+      "loss": 0.0014,
+      "step": 9564
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.8936046511627908e-07,
+      "loss": 0.0024,
+      "step": 9565
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.8934883720930232e-07,
+      "loss": 0.0009,
+      "step": 9566
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.8933720930232558e-07,
+      "loss": 0.0014,
+      "step": 9567
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.8932558139534882e-07,
+      "loss": 0.0037,
+      "step": 9568
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.8931395348837208e-07,
+      "loss": 0.0015,
+      "step": 9569
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.8930232558139534e-07,
+      "loss": 0.0011,
+      "step": 9570
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.892906976744186e-07,
+      "loss": 0.0014,
+      "step": 9571
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.8927906976744184e-07,
+      "loss": 0.0141,
+      "step": 9572
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.892674418604651e-07,
+      "loss": 0.0009,
+      "step": 9573
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.8925581395348839e-07,
+      "loss": 0.0036,
+      "step": 9574
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.8924418604651162e-07,
+      "loss": 0.0017,
+      "step": 9575
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.8923255813953488e-07,
+      "loss": 1.3366,
+      "step": 9576
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.8922093023255812e-07,
+      "loss": 0.0014,
+      "step": 9577
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.892093023255814e-07,
+      "loss": 0.0009,
+      "step": 9578
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.8919767441860464e-07,
+      "loss": 0.0012,
+      "step": 9579
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.891860465116279e-07,
+      "loss": 0.0015,
+      "step": 9580
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.8917441860465114e-07,
+      "loss": 0.007,
+      "step": 9581
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.8916279069767443e-07,
+      "loss": 0.0014,
+      "step": 9582
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.8915116279069766e-07,
+      "loss": 0.1459,
+      "step": 9583
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.8913953488372092e-07,
+      "loss": 0.001,
+      "step": 9584
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.8912790697674416e-07,
+      "loss": 0.0013,
+      "step": 9585
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.8911627906976745e-07,
+      "loss": 0.0012,
+      "step": 9586
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.891046511627907e-07,
+      "loss": 0.1177,
+      "step": 9587
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.8909302325581394e-07,
+      "loss": 0.0013,
+      "step": 9588
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.8908139534883718e-07,
+      "loss": 0.0011,
+      "step": 9589
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.8906976744186047e-07,
+      "loss": 0.364,
+      "step": 9590
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.8905813953488373e-07,
+      "loss": 0.0011,
+      "step": 9591
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.8904651162790696e-07,
+      "loss": 0.001,
+      "step": 9592
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.8903488372093023e-07,
+      "loss": 0.0013,
+      "step": 9593
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.8902325581395346e-07,
+      "loss": 0.0017,
+      "step": 9594
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.8901162790697675e-07,
+      "loss": 0.0011,
+      "step": 9595
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.8899999999999999e-07,
+      "loss": 0.0014,
+      "step": 9596
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.8898837209302325e-07,
+      "loss": 0.0014,
+      "step": 9597
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.8897674418604648e-07,
+      "loss": 0.0012,
+      "step": 9598
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.8896511627906977e-07,
+      "loss": 0.0009,
+      "step": 9599
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.88953488372093e-07,
+      "loss": 0.0014,
+      "step": 9600
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.8894186046511627e-07,
+      "loss": 0.0013,
+      "step": 9601
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.889302325581395e-07,
+      "loss": 0.0012,
+      "step": 9602
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.889186046511628e-07,
+      "loss": 0.0024,
+      "step": 9603
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.8890697674418605e-07,
+      "loss": 0.001,
+      "step": 9604
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.888953488372093e-07,
+      "loss": 0.001,
+      "step": 9605
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.8888372093023255e-07,
+      "loss": 0.1956,
+      "step": 9606
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.888720930232558e-07,
+      "loss": 0.002,
+      "step": 9607
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.8886046511627907e-07,
+      "loss": 0.0009,
+      "step": 9608
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.888488372093023e-07,
+      "loss": 0.0008,
+      "step": 9609
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.8883720930232557e-07,
+      "loss": 0.0011,
+      "step": 9610
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.8882558139534883e-07,
+      "loss": 0.0013,
+      "step": 9611
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.888139534883721e-07,
+      "loss": 0.0047,
+      "step": 9612
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.8880232558139533e-07,
+      "loss": 0.001,
+      "step": 9613
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.887906976744186e-07,
+      "loss": 0.0047,
+      "step": 9614
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.8877906976744188e-07,
+      "loss": 0.0673,
+      "step": 9615
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.8876744186046511e-07,
+      "loss": 0.4435,
+      "step": 9616
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.8875581395348838e-07,
+      "loss": 0.0013,
+      "step": 9617
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.887441860465116e-07,
+      "loss": 0.0011,
+      "step": 9618
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.8873255813953485e-07,
+      "loss": 0.0013,
+      "step": 9619
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.8872093023255813e-07,
+      "loss": 0.0012,
+      "step": 9620
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.887093023255814e-07,
+      "loss": 0.0014,
+      "step": 9621
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.8869767441860463e-07,
+      "loss": 0.0012,
+      "step": 9622
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.886860465116279e-07,
+      "loss": 0.0012,
+      "step": 9623
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.8867441860465115e-07,
+      "loss": 0.0013,
+      "step": 9624
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.8866279069767442e-07,
+      "loss": 0.0014,
+      "step": 9625
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.8865116279069765e-07,
+      "loss": 0.0012,
+      "step": 9626
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.886395348837209e-07,
+      "loss": 0.0023,
+      "step": 9627
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.886279069767442e-07,
+      "loss": 0.001,
+      "step": 9628
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.8861627906976744e-07,
+      "loss": 0.0021,
+      "step": 9629
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.8860465116279067e-07,
+      "loss": 0.0011,
+      "step": 9630
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.8859302325581393e-07,
+      "loss": 0.0015,
+      "step": 9631
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.8858139534883722e-07,
+      "loss": 0.002,
+      "step": 9632
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.8856976744186046e-07,
+      "loss": 0.0012,
+      "step": 9633
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.8855813953488372e-07,
+      "loss": 0.0014,
+      "step": 9634
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.8854651162790695e-07,
+      "loss": 0.0015,
+      "step": 9635
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.8853488372093024e-07,
+      "loss": 0.001,
+      "step": 9636
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.8852325581395348e-07,
+      "loss": 0.0014,
+      "step": 9637
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.8851162790697674e-07,
+      "loss": 0.0016,
+      "step": 9638
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.8849999999999997e-07,
+      "loss": 0.001,
+      "step": 9639
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.8848837209302326e-07,
+      "loss": 2.9157,
+      "step": 9640
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.884767441860465e-07,
+      "loss": 0.0297,
+      "step": 9641
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.8846511627906976e-07,
+      "loss": 0.0035,
+      "step": 9642
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.88453488372093e-07,
+      "loss": 0.0017,
+      "step": 9643
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.8844186046511628e-07,
+      "loss": 0.0012,
+      "step": 9644
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.8843023255813954e-07,
+      "loss": 0.0013,
+      "step": 9645
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.8841860465116278e-07,
+      "loss": 0.0011,
+      "step": 9646
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.8840697674418604e-07,
+      "loss": 0.0012,
+      "step": 9647
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.8839534883720928e-07,
+      "loss": 0.001,
+      "step": 9648
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.8838372093023256e-07,
+      "loss": 0.0012,
+      "step": 9649
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.883720930232558e-07,
+      "loss": 0.0008,
+      "step": 9650
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.8836046511627906e-07,
+      "loss": 0.0014,
+      "step": 9651
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.883488372093023e-07,
+      "loss": 0.0012,
+      "step": 9652
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.8833720930232559e-07,
+      "loss": 0.0011,
+      "step": 9653
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.8832558139534882e-07,
+      "loss": 0.002,
+      "step": 9654
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.8831395348837208e-07,
+      "loss": 0.0032,
+      "step": 9655
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.8830232558139532e-07,
+      "loss": 0.0014,
+      "step": 9656
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.882906976744186e-07,
+      "loss": 0.0022,
+      "step": 9657
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.8827906976744187e-07,
+      "loss": 0.0022,
+      "step": 9658
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.882674418604651e-07,
+      "loss": 0.0012,
+      "step": 9659
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.8825581395348836e-07,
+      "loss": 0.001,
+      "step": 9660
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.8824418604651163e-07,
+      "loss": 0.0016,
+      "step": 9661
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.882325581395349e-07,
+      "loss": 0.0015,
+      "step": 9662
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.8822093023255812e-07,
+      "loss": 0.0056,
+      "step": 9663
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.8820930232558138e-07,
+      "loss": 0.0051,
+      "step": 9664
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.8819767441860465e-07,
+      "loss": 0.003,
+      "step": 9665
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.881860465116279e-07,
+      "loss": 0.0009,
+      "step": 9666
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.8817441860465114e-07,
+      "loss": 0.0015,
+      "step": 9667
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.881627906976744e-07,
+      "loss": 0.0012,
+      "step": 9668
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.881511627906977e-07,
+      "loss": 0.0013,
+      "step": 9669
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.8813953488372093e-07,
+      "loss": 0.0015,
+      "step": 9670
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.881279069767442e-07,
+      "loss": 0.0014,
+      "step": 9671
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.8811627906976743e-07,
+      "loss": 0.0015,
+      "step": 9672
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.8810465116279066e-07,
+      "loss": 0.0011,
+      "step": 9673
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.8809302325581395e-07,
+      "loss": 0.0012,
+      "step": 9674
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.880813953488372e-07,
+      "loss": 0.0013,
+      "step": 9675
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.8806976744186045e-07,
+      "loss": 0.0013,
+      "step": 9676
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.880581395348837e-07,
+      "loss": 0.0011,
+      "step": 9677
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.8804651162790697e-07,
+      "loss": 0.0012,
+      "step": 9678
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.8803488372093023e-07,
+      "loss": 0.001,
+      "step": 9679
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.8802325581395347e-07,
+      "loss": 0.0013,
+      "step": 9680
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.8801162790697673e-07,
+      "loss": 0.0011,
+      "step": 9681
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.8800000000000002e-07,
+      "loss": 0.0033,
+      "step": 9682
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.8798837209302325e-07,
+      "loss": 0.3847,
+      "step": 9683
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.8797674418604649e-07,
+      "loss": 0.0011,
+      "step": 9684
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.8796511627906975e-07,
+      "loss": 0.001,
+      "step": 9685
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.8795348837209304e-07,
+      "loss": 0.0013,
+      "step": 9686
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.8794186046511627e-07,
+      "loss": 0.0009,
+      "step": 9687
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.8793023255813953e-07,
+      "loss": 0.0013,
+      "step": 9688
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.8791860465116277e-07,
+      "loss": 0.0009,
+      "step": 9689
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.8790697674418606e-07,
+      "loss": 0.001,
+      "step": 9690
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.878953488372093e-07,
+      "loss": 0.0014,
+      "step": 9691
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.8788372093023255e-07,
+      "loss": 0.002,
+      "step": 9692
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.878720930232558e-07,
+      "loss": 0.0018,
+      "step": 9693
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.8786046511627908e-07,
+      "loss": 0.0016,
+      "step": 9694
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.878488372093023e-07,
+      "loss": 0.0015,
+      "step": 9695
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.8783720930232557e-07,
+      "loss": 0.0013,
+      "step": 9696
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.878255813953488e-07,
+      "loss": 0.0012,
+      "step": 9697
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.878139534883721e-07,
+      "loss": 0.001,
+      "step": 9698
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.8780232558139536e-07,
+      "loss": 0.0116,
+      "step": 9699
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.877906976744186e-07,
+      "loss": 0.0011,
+      "step": 9700
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.8777906976744186e-07,
+      "loss": 0.1299,
+      "step": 9701
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.877674418604651e-07,
+      "loss": 1.6189,
+      "step": 9702
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.8775581395348838e-07,
+      "loss": 0.0013,
+      "step": 9703
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.8774418604651161e-07,
+      "loss": 0.0018,
+      "step": 9704
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.8773255813953488e-07,
+      "loss": 0.0012,
+      "step": 9705
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.877209302325581e-07,
+      "loss": 0.0012,
+      "step": 9706
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.877093023255814e-07,
+      "loss": 0.0012,
+      "step": 9707
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.8769767441860464e-07,
+      "loss": 0.0021,
+      "step": 9708
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.876860465116279e-07,
+      "loss": 0.0008,
+      "step": 9709
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.8767441860465113e-07,
+      "loss": 0.2051,
+      "step": 9710
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.8766279069767442e-07,
+      "loss": 0.0011,
+      "step": 9711
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.8765116279069768e-07,
+      "loss": 0.0009,
+      "step": 9712
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.8763953488372092e-07,
+      "loss": 0.0016,
+      "step": 9713
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.8762790697674418e-07,
+      "loss": 0.0011,
+      "step": 9714
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.8761627906976744e-07,
+      "loss": 1.8358,
+      "step": 9715
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.876046511627907e-07,
+      "loss": 0.0009,
+      "step": 9716
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.8759302325581394e-07,
+      "loss": 0.0033,
+      "step": 9717
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.875813953488372e-07,
+      "loss": 0.0019,
+      "step": 9718
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.8756976744186046e-07,
+      "loss": 0.0014,
+      "step": 9719
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.8755813953488372e-07,
+      "loss": 0.0019,
+      "step": 9720
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.8754651162790696e-07,
+      "loss": 0.2079,
+      "step": 9721
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.8753488372093022e-07,
+      "loss": 0.7949,
+      "step": 9722
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.875232558139535e-07,
+      "loss": 0.0395,
+      "step": 9723
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8751162790697674e-07,
+      "loss": 0.0015,
+      "step": 9724
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.875e-07,
+      "loss": 0.001,
+      "step": 9725
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8748837209302324e-07,
+      "loss": 0.0012,
+      "step": 9726
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8747674418604648e-07,
+      "loss": 0.0011,
+      "step": 9727
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8746511627906976e-07,
+      "loss": 0.0017,
+      "step": 9728
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8745348837209303e-07,
+      "loss": 0.0011,
+      "step": 9729
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8744186046511626e-07,
+      "loss": 0.0011,
+      "step": 9730
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8743023255813952e-07,
+      "loss": 0.001,
+      "step": 9731
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.8741860465116278e-07,
+      "loss": 0.0013,
+      "step": 9732
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8740697674418605e-07,
+      "loss": 0.0009,
+      "step": 9733
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8739534883720928e-07,
+      "loss": 0.0012,
+      "step": 9734
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8738372093023254e-07,
+      "loss": 0.0012,
+      "step": 9735
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.873720930232558e-07,
+      "loss": 0.4163,
+      "step": 9736
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8736046511627907e-07,
+      "loss": 0.001,
+      "step": 9737
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.873488372093023e-07,
+      "loss": 0.001,
+      "step": 9738
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8733720930232556e-07,
+      "loss": 0.0011,
+      "step": 9739
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8732558139534885e-07,
+      "loss": 0.0034,
+      "step": 9740
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8731395348837209e-07,
+      "loss": 0.0011,
+      "step": 9741
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.8730232558139535e-07,
+      "loss": 0.0012,
+      "step": 9742
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.8729069767441858e-07,
+      "loss": 0.0013,
+      "step": 9743
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.8727906976744187e-07,
+      "loss": 0.0105,
+      "step": 9744
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.872674418604651e-07,
+      "loss": 0.0008,
+      "step": 9745
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.8725581395348837e-07,
+      "loss": 1.3606,
+      "step": 9746
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.872441860465116e-07,
+      "loss": 0.0009,
+      "step": 9747
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.872325581395349e-07,
+      "loss": 0.0011,
+      "step": 9748
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8722093023255813e-07,
+      "loss": 0.0013,
+      "step": 9749
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.872093023255814e-07,
+      "loss": 0.0011,
+      "step": 9750
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8719767441860462e-07,
+      "loss": 0.001,
+      "step": 9751
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8718604651162789e-07,
+      "loss": 0.007,
+      "step": 9752
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.8717441860465117e-07,
+      "loss": 0.0016,
+      "step": 9753
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.871627906976744e-07,
+      "loss": 0.001,
+      "step": 9754
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.8715116279069767e-07,
+      "loss": 0.0013,
+      "step": 9755
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.871395348837209e-07,
+      "loss": 0.0011,
+      "step": 9756
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.871279069767442e-07,
+      "loss": 0.0012,
+      "step": 9757
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.8711627906976743e-07,
+      "loss": 0.0011,
+      "step": 9758
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.871046511627907e-07,
+      "loss": 0.001,
+      "step": 9759
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.8709302325581393e-07,
+      "loss": 0.0013,
+      "step": 9760
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.8708139534883721e-07,
+      "loss": 0.0009,
+      "step": 9761
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.8706976744186045e-07,
+      "loss": 0.0012,
+      "step": 9762
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.870581395348837e-07,
+      "loss": 0.0013,
+      "step": 9763
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.8704651162790695e-07,
+      "loss": 0.0009,
+      "step": 9764
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.8703488372093023e-07,
+      "loss": 0.0019,
+      "step": 9765
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.870232558139535e-07,
+      "loss": 1.411,
+      "step": 9766
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.8701162790697673e-07,
+      "loss": 0.0018,
+      "step": 9767
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.8699999999999997e-07,
+      "loss": 0.0014,
+      "step": 9768
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.8698837209302326e-07,
+      "loss": 0.0024,
+      "step": 9769
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.8697674418604652e-07,
+      "loss": 0.0038,
+      "step": 9770
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.8696511627906975e-07,
+      "loss": 0.001,
+      "step": 9771
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.8695348837209301e-07,
+      "loss": 0.001,
+      "step": 9772
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.8694186046511628e-07,
+      "loss": 0.0008,
+      "step": 9773
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.8693023255813954e-07,
+      "loss": 0.001,
+      "step": 9774
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.8691860465116277e-07,
+      "loss": 0.0015,
+      "step": 9775
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.8690697674418603e-07,
+      "loss": 0.0014,
+      "step": 9776
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.8689534883720932e-07,
+      "loss": 0.0009,
+      "step": 9777
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.8688372093023256e-07,
+      "loss": 0.0039,
+      "step": 9778
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.868720930232558e-07,
+      "loss": 0.0031,
+      "step": 9779
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.8686046511627905e-07,
+      "loss": 0.001,
+      "step": 9780
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.868488372093023e-07,
+      "loss": 0.0008,
+      "step": 9781
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.8683720930232558e-07,
+      "loss": 0.0009,
+      "step": 9782
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.8682558139534884e-07,
+      "loss": 0.0009,
+      "step": 9783
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.8681395348837208e-07,
+      "loss": 0.0019,
+      "step": 9784
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.8680232558139534e-07,
+      "loss": 0.0009,
+      "step": 9785
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.867906976744186e-07,
+      "loss": 0.0013,
+      "step": 9786
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.8677906976744186e-07,
+      "loss": 0.001,
+      "step": 9787
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.867674418604651e-07,
+      "loss": 0.0015,
+      "step": 9788
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.8675581395348836e-07,
+      "loss": 0.0009,
+      "step": 9789
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.8674418604651162e-07,
+      "loss": 0.001,
+      "step": 9790
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.8673255813953488e-07,
+      "loss": 0.0007,
+      "step": 9791
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.8672093023255812e-07,
+      "loss": 0.3983,
+      "step": 9792
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.8670930232558138e-07,
+      "loss": 0.001,
+      "step": 9793
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.8669767441860467e-07,
+      "loss": 0.0023,
+      "step": 9794
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.866860465116279e-07,
+      "loss": 0.0012,
+      "step": 9795
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.8667441860465116e-07,
+      "loss": 0.0011,
+      "step": 9796
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.866627906976744e-07,
+      "loss": 0.0013,
+      "step": 9797
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.8665116279069769e-07,
+      "loss": 0.0008,
+      "step": 9798
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.8663953488372092e-07,
+      "loss": 0.3062,
+      "step": 9799
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.8662790697674418e-07,
+      "loss": 0.0018,
+      "step": 9800
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.8661627906976742e-07,
+      "loss": 0.0012,
+      "step": 9801
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.866046511627907e-07,
+      "loss": 0.001,
+      "step": 9802
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.8659302325581394e-07,
+      "loss": 0.001,
+      "step": 9803
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.865813953488372e-07,
+      "loss": 0.0015,
+      "step": 9804
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.8656976744186044e-07,
+      "loss": 0.0013,
+      "step": 9805
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.865581395348837e-07,
+      "loss": 3.1361,
+      "step": 9806
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.86546511627907e-07,
+      "loss": 0.0011,
+      "step": 9807
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.8653488372093022e-07,
+      "loss": 0.0011,
+      "step": 9808
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.8652325581395349e-07,
+      "loss": 0.0011,
+      "step": 9809
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.8651162790697672e-07,
+      "loss": 0.0012,
+      "step": 9810
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.865e-07,
+      "loss": 0.0011,
+      "step": 9811
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.8648837209302324e-07,
+      "loss": 0.0012,
+      "step": 9812
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.864767441860465e-07,
+      "loss": 0.0009,
+      "step": 9813
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.8646511627906974e-07,
+      "loss": 0.0016,
+      "step": 9814
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.8645348837209303e-07,
+      "loss": 0.001,
+      "step": 9815
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.8644186046511626e-07,
+      "loss": 0.0024,
+      "step": 9816
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.8643023255813953e-07,
+      "loss": 0.0011,
+      "step": 9817
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.8641860465116276e-07,
+      "loss": 0.0013,
+      "step": 9818
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.8640697674418605e-07,
+      "loss": 0.0012,
+      "step": 9819
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.863953488372093e-07,
+      "loss": 0.0014,
+      "step": 9820
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.8638372093023255e-07,
+      "loss": 0.0017,
+      "step": 9821
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.8637209302325578e-07,
+      "loss": 0.0009,
+      "step": 9822
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.8636046511627907e-07,
+      "loss": 0.0014,
+      "step": 9823
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.9893617021276596,
+      "eval_loss": 0.15840478241443634,
+      "eval_runtime": 1.5537,
+      "eval_samples_per_second": 665.511,
+      "eval_steps_per_second": 10.942,
+      "step": 9823
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.8634883720930233e-07,
+      "loss": 0.001,
+      "step": 9824
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.8633720930232557e-07,
+      "loss": 0.0014,
+      "step": 9825
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.8632558139534883e-07,
+      "loss": 0.0034,
+      "step": 9826
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.863139534883721e-07,
+      "loss": 0.001,
+      "step": 9827
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.8630232558139535e-07,
+      "loss": 0.0011,
+      "step": 9828
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.862906976744186e-07,
+      "loss": 1.8116,
+      "step": 9829
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.8627906976744185e-07,
+      "loss": 0.0012,
+      "step": 9830
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.8626744186046514e-07,
+      "loss": 0.0014,
+      "step": 9831
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.8625581395348837e-07,
+      "loss": 0.0018,
+      "step": 9832
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.862441860465116e-07,
+      "loss": 0.0012,
+      "step": 9833
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.8623255813953487e-07,
+      "loss": 0.0014,
+      "step": 9834
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.862209302325581e-07,
+      "loss": 0.001,
+      "step": 9835
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.862093023255814e-07,
+      "loss": 0.0012,
+      "step": 9836
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.8619767441860465e-07,
+      "loss": 0.0013,
+      "step": 9837
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.861860465116279e-07,
+      "loss": 0.0015,
+      "step": 9838
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.8617441860465115e-07,
+      "loss": 0.0012,
+      "step": 9839
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.8616279069767441e-07,
+      "loss": 0.0012,
+      "step": 9840
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.8615116279069768e-07,
+      "loss": 0.001,
+      "step": 9841
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.861395348837209e-07,
+      "loss": 1.4919,
+      "step": 9842
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.8612790697674417e-07,
+      "loss": 0.0018,
+      "step": 9843
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.8611627906976743e-07,
+      "loss": 0.0015,
+      "step": 9844
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.861046511627907e-07,
+      "loss": 0.0011,
+      "step": 9845
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.8609302325581393e-07,
+      "loss": 0.0017,
+      "step": 9846
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.860813953488372e-07,
+      "loss": 0.0011,
+      "step": 9847
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.8606976744186048e-07,
+      "loss": 0.0009,
+      "step": 9848
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.8605813953488372e-07,
+      "loss": 0.0031,
+      "step": 9849
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.8604651162790698e-07,
+      "loss": 0.0087,
+      "step": 9850
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.860348837209302e-07,
+      "loss": 0.0013,
+      "step": 9851
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.860232558139535e-07,
+      "loss": 0.0012,
+      "step": 9852
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.8601162790697674e-07,
+      "loss": 0.0016,
+      "step": 9853
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.86e-07,
+      "loss": 0.0013,
+      "step": 9854
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.8598837209302323e-07,
+      "loss": 0.0008,
+      "step": 9855
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.8597674418604652e-07,
+      "loss": 0.0008,
+      "step": 9856
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.8596511627906976e-07,
+      "loss": 0.001,
+      "step": 9857
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.8595348837209302e-07,
+      "loss": 0.001,
+      "step": 9858
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.8594186046511625e-07,
+      "loss": 2.9336,
+      "step": 9859
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.8593023255813952e-07,
+      "loss": 0.0013,
+      "step": 9860
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.859186046511628e-07,
+      "loss": 0.0017,
+      "step": 9861
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.8590697674418604e-07,
+      "loss": 0.0011,
+      "step": 9862
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.8589534883720927e-07,
+      "loss": 0.001,
+      "step": 9863
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.8588372093023254e-07,
+      "loss": 0.0015,
+      "step": 9864
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.8587209302325582e-07,
+      "loss": 0.0015,
+      "step": 9865
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.8586046511627906e-07,
+      "loss": 0.0024,
+      "step": 9866
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.8584883720930232e-07,
+      "loss": 0.0015,
+      "step": 9867
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.8583720930232556e-07,
+      "loss": 0.0013,
+      "step": 9868
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.8582558139534884e-07,
+      "loss": 0.0019,
+      "step": 9869
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.8581395348837208e-07,
+      "loss": 0.0015,
+      "step": 9870
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.8580232558139534e-07,
+      "loss": 0.0029,
+      "step": 9871
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.8579069767441858e-07,
+      "loss": 0.0017,
+      "step": 9872
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.8577906976744186e-07,
+      "loss": 0.0009,
+      "step": 9873
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.857674418604651e-07,
+      "loss": 0.0016,
+      "step": 9874
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.8575581395348836e-07,
+      "loss": 0.2263,
+      "step": 9875
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.857441860465116e-07,
+      "loss": 0.001,
+      "step": 9876
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.8573255813953488e-07,
+      "loss": 0.0015,
+      "step": 9877
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.8572093023255815e-07,
+      "loss": 0.0011,
+      "step": 9878
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.8570930232558138e-07,
+      "loss": 0.0017,
+      "step": 9879
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.8569767441860464e-07,
+      "loss": 0.001,
+      "step": 9880
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.856860465116279e-07,
+      "loss": 0.001,
+      "step": 9881
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.8567441860465117e-07,
+      "loss": 0.0011,
+      "step": 9882
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.856627906976744e-07,
+      "loss": 0.001,
+      "step": 9883
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.8565116279069766e-07,
+      "loss": 0.0011,
+      "step": 9884
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.856395348837209e-07,
+      "loss": 0.0011,
+      "step": 9885
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.856279069767442e-07,
+      "loss": 0.0009,
+      "step": 9886
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.8561627906976742e-07,
+      "loss": 0.0033,
+      "step": 9887
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.8560465116279068e-07,
+      "loss": 1.7725,
+      "step": 9888
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.8559302325581392e-07,
+      "loss": 0.001,
+      "step": 9889
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.855813953488372e-07,
+      "loss": 0.0011,
+      "step": 9890
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.8556976744186047e-07,
+      "loss": 0.0017,
+      "step": 9891
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.855581395348837e-07,
+      "loss": 0.0009,
+      "step": 9892
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.8554651162790697e-07,
+      "loss": 0.0445,
+      "step": 9893
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.8553488372093023e-07,
+      "loss": 0.0011,
+      "step": 9894
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.855232558139535e-07,
+      "loss": 0.0012,
+      "step": 9895
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.8551162790697673e-07,
+      "loss": 0.0022,
+      "step": 9896
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.855e-07,
+      "loss": 0.0017,
+      "step": 9897
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.8548837209302325e-07,
+      "loss": 0.001,
+      "step": 9898
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.854767441860465e-07,
+      "loss": 0.0009,
+      "step": 9899
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.8546511627906975e-07,
+      "loss": 0.001,
+      "step": 9900
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.85453488372093e-07,
+      "loss": 0.0013,
+      "step": 9901
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.854418604651163e-07,
+      "loss": 0.001,
+      "step": 9902
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.8543023255813953e-07,
+      "loss": 0.0012,
+      "step": 9903
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.854186046511628e-07,
+      "loss": 0.0011,
+      "step": 9904
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.8540697674418603e-07,
+      "loss": 0.0012,
+      "step": 9905
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.8539534883720932e-07,
+      "loss": 0.0017,
+      "step": 9906
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.8538372093023255e-07,
+      "loss": 0.0011,
+      "step": 9907
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.853720930232558e-07,
+      "loss": 0.0012,
+      "step": 9908
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.8536046511627905e-07,
+      "loss": 0.001,
+      "step": 9909
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.8534883720930234e-07,
+      "loss": 0.0009,
+      "step": 9910
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.8533720930232557e-07,
+      "loss": 0.001,
+      "step": 9911
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.8532558139534883e-07,
+      "loss": 0.0011,
+      "step": 9912
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.8531395348837207e-07,
+      "loss": 0.0013,
+      "step": 9913
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.8530232558139533e-07,
+      "loss": 0.0009,
+      "step": 9914
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.8529069767441862e-07,
+      "loss": 0.0012,
+      "step": 9915
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.8527906976744185e-07,
+      "loss": 0.0055,
+      "step": 9916
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.852674418604651e-07,
+      "loss": 0.001,
+      "step": 9917
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.8525581395348835e-07,
+      "loss": 0.0015,
+      "step": 9918
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.8524418604651164e-07,
+      "loss": 0.001,
+      "step": 9919
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.8523255813953487e-07,
+      "loss": 0.0012,
+      "step": 9920
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.8522093023255814e-07,
+      "loss": 0.0007,
+      "step": 9921
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.8520930232558137e-07,
+      "loss": 0.0034,
+      "step": 9922
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.8519767441860466e-07,
+      "loss": 0.0307,
+      "step": 9923
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.851860465116279e-07,
+      "loss": 0.0011,
+      "step": 9924
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.8517441860465116e-07,
+      "loss": 0.1524,
+      "step": 9925
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.851627906976744e-07,
+      "loss": 0.0015,
+      "step": 9926
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.8515116279069768e-07,
+      "loss": 0.0099,
+      "step": 9927
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.8513953488372091e-07,
+      "loss": 0.0015,
+      "step": 9928
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.8512790697674418e-07,
+      "loss": 0.0019,
+      "step": 9929
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.851162790697674e-07,
+      "loss": 0.0013,
+      "step": 9930
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.851046511627907e-07,
+      "loss": 0.0016,
+      "step": 9931
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.8509302325581396e-07,
+      "loss": 0.0011,
+      "step": 9932
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.850813953488372e-07,
+      "loss": 0.001,
+      "step": 9933
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.8506976744186046e-07,
+      "loss": 0.001,
+      "step": 9934
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.8505813953488372e-07,
+      "loss": 0.0011,
+      "step": 9935
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.8504651162790698e-07,
+      "loss": 0.0018,
+      "step": 9936
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.8503488372093022e-07,
+      "loss": 0.0009,
+      "step": 9937
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.8502325581395348e-07,
+      "loss": 0.0012,
+      "step": 9938
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.8501162790697671e-07,
+      "loss": 0.0012,
+      "step": 9939
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.85e-07,
+      "loss": 0.0012,
+      "step": 9940
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.8498837209302324e-07,
+      "loss": 0.0009,
+      "step": 9941
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.849767441860465e-07,
+      "loss": 0.0049,
+      "step": 9942
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.8496511627906973e-07,
+      "loss": 0.0011,
+      "step": 9943
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.8495348837209302e-07,
+      "loss": 0.0014,
+      "step": 9944
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.8494186046511628e-07,
+      "loss": 0.0013,
+      "step": 9945
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.8493023255813952e-07,
+      "loss": 0.0015,
+      "step": 9946
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.8491860465116278e-07,
+      "loss": 0.0012,
+      "step": 9947
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.8490697674418604e-07,
+      "loss": 0.0018,
+      "step": 9948
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.848953488372093e-07,
+      "loss": 0.0012,
+      "step": 9949
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8488372093023254e-07,
+      "loss": 0.0013,
+      "step": 9950
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.848720930232558e-07,
+      "loss": 0.0034,
+      "step": 9951
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8486046511627906e-07,
+      "loss": 0.0014,
+      "step": 9952
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8484883720930233e-07,
+      "loss": 0.0012,
+      "step": 9953
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8483720930232556e-07,
+      "loss": 0.0011,
+      "step": 9954
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8482558139534882e-07,
+      "loss": 0.0014,
+      "step": 9955
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.848139534883721e-07,
+      "loss": 0.0022,
+      "step": 9956
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8480232558139535e-07,
+      "loss": 0.0013,
+      "step": 9957
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.847906976744186e-07,
+      "loss": 0.001,
+      "step": 9958
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8477906976744184e-07,
+      "loss": 0.0012,
+      "step": 9959
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8476744186046513e-07,
+      "loss": 0.0198,
+      "step": 9960
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8475581395348837e-07,
+      "loss": 0.001,
+      "step": 9961
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8474418604651163e-07,
+      "loss": 0.0019,
+      "step": 9962
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8473255813953486e-07,
+      "loss": 0.001,
+      "step": 9963
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8472093023255815e-07,
+      "loss": 0.0015,
+      "step": 9964
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8470930232558139e-07,
+      "loss": 0.0009,
+      "step": 9965
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8469767441860465e-07,
+      "loss": 0.0013,
+      "step": 9966
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8468604651162788e-07,
+      "loss": 0.0009,
+      "step": 9967
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8467441860465114e-07,
+      "loss": 0.0013,
+      "step": 9968
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8466279069767443e-07,
+      "loss": 0.001,
+      "step": 9969
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8465116279069767e-07,
+      "loss": 0.001,
+      "step": 9970
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.846395348837209e-07,
+      "loss": 0.0017,
+      "step": 9971
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.8462790697674417e-07,
+      "loss": 0.0012,
+      "step": 9972
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.8461627906976745e-07,
+      "loss": 0.0009,
+      "step": 9973
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.846046511627907e-07,
+      "loss": 0.001,
+      "step": 9974
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.8459302325581395e-07,
+      "loss": 0.0011,
+      "step": 9975
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.8458139534883719e-07,
+      "loss": 0.001,
+      "step": 9976
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.8456976744186047e-07,
+      "loss": 0.0011,
+      "step": 9977
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.845581395348837e-07,
+      "loss": 0.0014,
+      "step": 9978
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.8454651162790697e-07,
+      "loss": 0.0012,
+      "step": 9979
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.845348837209302e-07,
+      "loss": 0.0202,
+      "step": 9980
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.845232558139535e-07,
+      "loss": 0.1688,
+      "step": 9981
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.8451162790697673e-07,
+      "loss": 0.0019,
+      "step": 9982
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.845e-07,
+      "loss": 0.0011,
+      "step": 9983
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.8448837209302323e-07,
+      "loss": 0.4404,
+      "step": 9984
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.8447674418604651e-07,
+      "loss": 0.001,
+      "step": 9985
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.8446511627906978e-07,
+      "loss": 0.0013,
+      "step": 9986
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.84453488372093e-07,
+      "loss": 0.0013,
+      "step": 9987
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.8444186046511627e-07,
+      "loss": 0.001,
+      "step": 9988
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.8443023255813953e-07,
+      "loss": 0.0012,
+      "step": 9989
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.844186046511628e-07,
+      "loss": 1.0666,
+      "step": 9990
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.8440697674418603e-07,
+      "loss": 0.0028,
+      "step": 9991
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.843953488372093e-07,
+      "loss": 0.0011,
+      "step": 9992
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.8438372093023253e-07,
+      "loss": 0.0015,
+      "step": 9993
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.8437209302325582e-07,
+      "loss": 0.001,
+      "step": 9994
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.8436046511627905e-07,
+      "loss": 0.001,
+      "step": 9995
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.8434883720930231e-07,
+      "loss": 0.0009,
+      "step": 9996
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.8433720930232555e-07,
+      "loss": 0.0011,
+      "step": 9997
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.8432558139534884e-07,
+      "loss": 0.0012,
+      "step": 9998
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.843139534883721e-07,
+      "loss": 0.0009,
+      "step": 9999
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.8430232558139533e-07,
+      "loss": 0.0012,
+      "step": 10000
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.8429069767441857e-07,
+      "loss": 0.0008,
+      "step": 10001
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.8427906976744186e-07,
+      "loss": 0.0012,
+      "step": 10002
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.8426744186046512e-07,
+      "loss": 0.0011,
+      "step": 10003
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.8425581395348835e-07,
+      "loss": 0.0018,
+      "step": 10004
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.8424418604651162e-07,
+      "loss": 0.0012,
+      "step": 10005
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.8423255813953488e-07,
+      "loss": 0.0008,
+      "step": 10006
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.8422093023255814e-07,
+      "loss": 0.0333,
+      "step": 10007
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.8420930232558138e-07,
+      "loss": 0.0014,
+      "step": 10008
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.8419767441860464e-07,
+      "loss": 1.3219,
+      "step": 10009
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.8418604651162792e-07,
+      "loss": 0.0012,
+      "step": 10010
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.8417441860465116e-07,
+      "loss": 0.0011,
+      "step": 10011
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.841627906976744e-07,
+      "loss": 0.0011,
+      "step": 10012
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.8415116279069766e-07,
+      "loss": 0.0009,
+      "step": 10013
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.8413953488372095e-07,
+      "loss": 0.0012,
+      "step": 10014
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.8412790697674418e-07,
+      "loss": 0.0013,
+      "step": 10015
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.8411627906976744e-07,
+      "loss": 0.0008,
+      "step": 10016
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.8410465116279068e-07,
+      "loss": 0.001,
+      "step": 10017
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.8409302325581397e-07,
+      "loss": 0.2114,
+      "step": 10018
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.840813953488372e-07,
+      "loss": 0.0011,
+      "step": 10019
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.8406976744186046e-07,
+      "loss": 0.0032,
+      "step": 10020
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.840581395348837e-07,
+      "loss": 0.0013,
+      "step": 10021
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.8404651162790696e-07,
+      "loss": 0.0013,
+      "step": 10022
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.8403488372093022e-07,
+      "loss": 0.0009,
+      "step": 10023
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.8402325581395348e-07,
+      "loss": 0.0011,
+      "step": 10024
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.8401162790697672e-07,
+      "loss": 0.001,
+      "step": 10025
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.8399999999999998e-07,
+      "loss": 0.0019,
+      "step": 10026
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.8398837209302327e-07,
+      "loss": 0.0431,
+      "step": 10027
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.839767441860465e-07,
+      "loss": 0.0011,
+      "step": 10028
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.8396511627906977e-07,
+      "loss": 0.0009,
+      "step": 10029
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.83953488372093e-07,
+      "loss": 0.0009,
+      "step": 10030
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.839418604651163e-07,
+      "loss": 0.0011,
+      "step": 10031
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.8393023255813952e-07,
+      "loss": 0.0008,
+      "step": 10032
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.8391860465116279e-07,
+      "loss": 0.0009,
+      "step": 10033
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.8390697674418602e-07,
+      "loss": 0.0007,
+      "step": 10034
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.838953488372093e-07,
+      "loss": 0.0129,
+      "step": 10035
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.8388372093023254e-07,
+      "loss": 0.0012,
+      "step": 10036
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.838720930232558e-07,
+      "loss": 0.0019,
+      "step": 10037
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.8386046511627904e-07,
+      "loss": 0.0012,
+      "step": 10038
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.8384883720930233e-07,
+      "loss": 0.0011,
+      "step": 10039
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.838372093023256e-07,
+      "loss": 0.0025,
+      "step": 10040
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.8382558139534883e-07,
+      "loss": 0.0018,
+      "step": 10041
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.838139534883721e-07,
+      "loss": 0.0016,
+      "step": 10042
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.8380232558139535e-07,
+      "loss": 0.0011,
+      "step": 10043
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.837906976744186e-07,
+      "loss": 0.0018,
+      "step": 10044
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.8377906976744185e-07,
+      "loss": 0.0222,
+      "step": 10045
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.837674418604651e-07,
+      "loss": 0.0012,
+      "step": 10046
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.8375581395348834e-07,
+      "loss": 0.001,
+      "step": 10047
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.8374418604651163e-07,
+      "loss": 0.0014,
+      "step": 10048
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.8373255813953487e-07,
+      "loss": 0.0009,
+      "step": 10049
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.8372093023255813e-07,
+      "loss": 0.0026,
+      "step": 10050
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.8370930232558136e-07,
+      "loss": 0.0141,
+      "step": 10051
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.8369767441860465e-07,
+      "loss": 0.0013,
+      "step": 10052
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.8368604651162791e-07,
+      "loss": 0.0009,
+      "step": 10053
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.8367441860465115e-07,
+      "loss": 0.001,
+      "step": 10054
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.8366279069767438e-07,
+      "loss": 0.0009,
+      "step": 10055
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.8365116279069767e-07,
+      "loss": 0.001,
+      "step": 10056
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.8363953488372093e-07,
+      "loss": 1.0734,
+      "step": 10057
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.8362790697674417e-07,
+      "loss": 0.0019,
+      "step": 10058
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.8361627906976743e-07,
+      "loss": 0.0009,
+      "step": 10059
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.836046511627907e-07,
+      "loss": 0.0008,
+      "step": 10060
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.8359302325581395e-07,
+      "loss": 0.0028,
+      "step": 10061
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.835813953488372e-07,
+      "loss": 0.0012,
+      "step": 10062
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.8356976744186045e-07,
+      "loss": 0.0009,
+      "step": 10063
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.8355813953488374e-07,
+      "loss": 0.0012,
+      "step": 10064
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.8354651162790698e-07,
+      "loss": 0.0013,
+      "step": 10065
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.835348837209302e-07,
+      "loss": 0.0012,
+      "step": 10066
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.8352325581395347e-07,
+      "loss": 0.0012,
+      "step": 10067
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.8351162790697676e-07,
+      "loss": 0.0011,
+      "step": 10068
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.835e-07,
+      "loss": 0.0012,
+      "step": 10069
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.8348837209302326e-07,
+      "loss": 3.2288,
+      "step": 10070
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.834767441860465e-07,
+      "loss": 0.0012,
+      "step": 10071
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.8346511627906975e-07,
+      "loss": 0.0012,
+      "step": 10072
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.8345348837209302e-07,
+      "loss": 0.0011,
+      "step": 10073
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.8344186046511628e-07,
+      "loss": 0.001,
+      "step": 10074
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.834302325581395e-07,
+      "loss": 0.0024,
+      "step": 10075
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.8341860465116277e-07,
+      "loss": 1.7574,
+      "step": 10076
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.8340697674418604e-07,
+      "loss": 0.0153,
+      "step": 10077
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.833953488372093e-07,
+      "loss": 0.028,
+      "step": 10078
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.8338372093023253e-07,
+      "loss": 0.0011,
+      "step": 10079
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.833720930232558e-07,
+      "loss": 1.6117,
+      "step": 10080
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.8336046511627908e-07,
+      "loss": 0.0012,
+      "step": 10081
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.8334883720930232e-07,
+      "loss": 0.0054,
+      "step": 10082
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.8333720930232558e-07,
+      "loss": 0.0039,
+      "step": 10083
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.8332558139534882e-07,
+      "loss": 0.0008,
+      "step": 10084
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.833139534883721e-07,
+      "loss": 0.0011,
+      "step": 10085
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.8330232558139534e-07,
+      "loss": 0.0011,
+      "step": 10086
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.832906976744186e-07,
+      "loss": 0.001,
+      "step": 10087
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.8327906976744184e-07,
+      "loss": 0.001,
+      "step": 10088
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.8326744186046512e-07,
+      "loss": 0.0016,
+      "step": 10089
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.8325581395348836e-07,
+      "loss": 0.0009,
+      "step": 10090
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.8324418604651162e-07,
+      "loss": 0.0015,
+      "step": 10091
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.8323255813953486e-07,
+      "loss": 0.0012,
+      "step": 10092
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.8322093023255814e-07,
+      "loss": 0.0015,
+      "step": 10093
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.832093023255814e-07,
+      "loss": 0.001,
+      "step": 10094
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.8319767441860464e-07,
+      "loss": 0.0018,
+      "step": 10095
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.831860465116279e-07,
+      "loss": 0.0009,
+      "step": 10096
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.8317441860465116e-07,
+      "loss": 0.0007,
+      "step": 10097
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.8316279069767443e-07,
+      "loss": 0.001,
+      "step": 10098
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.8315116279069766e-07,
+      "loss": 0.001,
+      "step": 10099
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.8313953488372092e-07,
+      "loss": 0.0011,
+      "step": 10100
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.8312790697674416e-07,
+      "loss": 0.0013,
+      "step": 10101
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.8311627906976745e-07,
+      "loss": 0.0012,
+      "step": 10102
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.8310465116279068e-07,
+      "loss": 0.0011,
+      "step": 10103
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.8309302325581394e-07,
+      "loss": 0.0012,
+      "step": 10104
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.8308139534883718e-07,
+      "loss": 2.0063,
+      "step": 10105
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.8306976744186047e-07,
+      "loss": 0.0011,
+      "step": 10106
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.830581395348837e-07,
+      "loss": 0.0026,
+      "step": 10107
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.8304651162790696e-07,
+      "loss": 0.0041,
+      "step": 10108
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.830348837209302e-07,
+      "loss": 0.0009,
+      "step": 10109
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.830232558139535e-07,
+      "loss": 0.0014,
+      "step": 10110
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.8301162790697675e-07,
+      "loss": 0.0012,
+      "step": 10111
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.8299999999999998e-07,
+      "loss": 0.0013,
+      "step": 10112
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.8298837209302325e-07,
+      "loss": 0.0009,
+      "step": 10113
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.829767441860465e-07,
+      "loss": 0.0009,
+      "step": 10114
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.8296511627906977e-07,
+      "loss": 0.0009,
+      "step": 10115
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.82953488372093e-07,
+      "loss": 0.0011,
+      "step": 10116
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.8294186046511627e-07,
+      "loss": 0.0009,
+      "step": 10117
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.8293023255813953e-07,
+      "loss": 0.0019,
+      "step": 10118
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.829186046511628e-07,
+      "loss": 0.0007,
+      "step": 10119
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.8290697674418603e-07,
+      "loss": 0.0009,
+      "step": 10120
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.828953488372093e-07,
+      "loss": 0.001,
+      "step": 10121
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.8288372093023257e-07,
+      "loss": 0.0014,
+      "step": 10122
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.828720930232558e-07,
+      "loss": 0.0014,
+      "step": 10123
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.8286046511627907e-07,
+      "loss": 0.0012,
+      "step": 10124
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.828488372093023e-07,
+      "loss": 0.0015,
+      "step": 10125
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.8283720930232557e-07,
+      "loss": 0.0013,
+      "step": 10126
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.8282558139534883e-07,
+      "loss": 0.0013,
+      "step": 10127
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.828139534883721e-07,
+      "loss": 0.0012,
+      "step": 10128
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.8280232558139533e-07,
+      "loss": 0.0014,
+      "step": 10129
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.827906976744186e-07,
+      "loss": 0.001,
+      "step": 10130
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.8277906976744185e-07,
+      "loss": 0.0018,
+      "step": 10131
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.827674418604651e-07,
+      "loss": 0.0012,
+      "step": 10132
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.8275581395348835e-07,
+      "loss": 0.0009,
+      "step": 10133
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.827441860465116e-07,
+      "loss": 0.0011,
+      "step": 10134
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.827325581395349e-07,
+      "loss": 0.0016,
+      "step": 10135
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.8272093023255813e-07,
+      "loss": 0.0011,
+      "step": 10136
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.827093023255814e-07,
+      "loss": 0.0017,
+      "step": 10137
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.8269767441860463e-07,
+      "loss": 0.0013,
+      "step": 10138
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.8268604651162792e-07,
+      "loss": 0.001,
+      "step": 10139
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.8267441860465115e-07,
+      "loss": 0.001,
+      "step": 10140
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.8266279069767442e-07,
+      "loss": 0.0013,
+      "step": 10141
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.8265116279069765e-07,
+      "loss": 0.0008,
+      "step": 10142
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.8263953488372094e-07,
+      "loss": 0.002,
+      "step": 10143
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.8262790697674417e-07,
+      "loss": 0.0013,
+      "step": 10144
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.8261627906976744e-07,
+      "loss": 0.0041,
+      "step": 10145
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.8260465116279067e-07,
+      "loss": 0.001,
+      "step": 10146
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.8259302325581396e-07,
+      "loss": 0.0012,
+      "step": 10147
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.8258139534883722e-07,
+      "loss": 0.0012,
+      "step": 10148
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.8256976744186046e-07,
+      "loss": 0.0052,
+      "step": 10149
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.825581395348837e-07,
+      "loss": 0.0012,
+      "step": 10150
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.8254651162790698e-07,
+      "loss": 0.0009,
+      "step": 10151
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.8253488372093024e-07,
+      "loss": 0.001,
+      "step": 10152
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.8252325581395348e-07,
+      "loss": 0.0017,
+      "step": 10153
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.8251162790697674e-07,
+      "loss": 0.0012,
+      "step": 10154
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.8249999999999997e-07,
+      "loss": 0.0008,
+      "step": 10155
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.8248837209302326e-07,
+      "loss": 0.001,
+      "step": 10156
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.824767441860465e-07,
+      "loss": 0.0011,
+      "step": 10157
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.8246511627906976e-07,
+      "loss": 0.0015,
+      "step": 10158
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.82453488372093e-07,
+      "loss": 0.0012,
+      "step": 10159
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.8244186046511628e-07,
+      "loss": 0.0011,
+      "step": 10160
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.8243023255813952e-07,
+      "loss": 0.0008,
+      "step": 10161
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.8241860465116278e-07,
+      "loss": 0.001,
+      "step": 10162
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.8240697674418601e-07,
+      "loss": 0.0012,
+      "step": 10163
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.823953488372093e-07,
+      "loss": 0.0009,
+      "step": 10164
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.8238372093023256e-07,
+      "loss": 0.0009,
+      "step": 10165
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.823720930232558e-07,
+      "loss": 0.0012,
+      "step": 10166
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.8236046511627906e-07,
+      "loss": 0.2527,
+      "step": 10167
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.8234883720930232e-07,
+      "loss": 0.0016,
+      "step": 10168
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.8233720930232558e-07,
+      "loss": 0.0009,
+      "step": 10169
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.8232558139534882e-07,
+      "loss": 0.0013,
+      "step": 10170
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.8231395348837208e-07,
+      "loss": 0.0011,
+      "step": 10171
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.8230232558139534e-07,
+      "loss": 0.0016,
+      "step": 10172
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.822906976744186e-07,
+      "loss": 0.0008,
+      "step": 10173
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.8227906976744184e-07,
+      "loss": 0.0018,
+      "step": 10174
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.822674418604651e-07,
+      "loss": 0.001,
+      "step": 10175
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.822558139534884e-07,
+      "loss": 1.9058,
+      "step": 10176
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.8224418604651162e-07,
+      "loss": 0.0008,
+      "step": 10177
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.8223255813953489e-07,
+      "loss": 0.0025,
+      "step": 10178
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.8222093023255812e-07,
+      "loss": 0.001,
+      "step": 10179
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.8220930232558138e-07,
+      "loss": 0.0012,
+      "step": 10180
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.8219767441860465e-07,
+      "loss": 0.0014,
+      "step": 10181
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.821860465116279e-07,
+      "loss": 0.001,
+      "step": 10182
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.8217441860465114e-07,
+      "loss": 0.0009,
+      "step": 10183
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.821627906976744e-07,
+      "loss": 0.0293,
+      "step": 10184
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.8215116279069767e-07,
+      "loss": 0.0007,
+      "step": 10185
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.8213953488372093e-07,
+      "loss": 0.0064,
+      "step": 10186
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.8212790697674416e-07,
+      "loss": 0.0009,
+      "step": 10187
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.8211627906976742e-07,
+      "loss": 0.0011,
+      "step": 10188
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.821046511627907e-07,
+      "loss": 0.0011,
+      "step": 10189
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.8209302325581395e-07,
+      "loss": 2.9337,
+      "step": 10190
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.820813953488372e-07,
+      "loss": 1.5521,
+      "step": 10191
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.8206976744186044e-07,
+      "loss": 0.0009,
+      "step": 10192
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.8205813953488373e-07,
+      "loss": 0.013,
+      "step": 10193
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.8204651162790697e-07,
+      "loss": 0.0018,
+      "step": 10194
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.8203488372093023e-07,
+      "loss": 0.001,
+      "step": 10195
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.8202325581395347e-07,
+      "loss": 0.0011,
+      "step": 10196
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.8201162790697675e-07,
+      "loss": 0.0013,
+      "step": 10197
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.82e-07,
+      "loss": 0.0011,
+      "step": 10198
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.8198837209302325e-07,
+      "loss": 0.0009,
+      "step": 10199
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.8197674418604649e-07,
+      "loss": 0.001,
+      "step": 10200
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.8196511627906977e-07,
+      "loss": 0.0017,
+      "step": 10201
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.8195348837209304e-07,
+      "loss": 0.0013,
+      "step": 10202
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.8194186046511627e-07,
+      "loss": 0.0013,
+      "step": 10203
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.819302325581395e-07,
+      "loss": 2.4404,
+      "step": 10204
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.8191860465116277e-07,
+      "loss": 0.3301,
+      "step": 10205
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.8190697674418606e-07,
+      "loss": 0.7562,
+      "step": 10206
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.818953488372093e-07,
+      "loss": 0.0009,
+      "step": 10207
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.8188372093023255e-07,
+      "loss": 0.003,
+      "step": 10208
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.818720930232558e-07,
+      "loss": 0.001,
+      "step": 10209
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.8186046511627908e-07,
+      "loss": 0.0018,
+      "step": 10210
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.818488372093023e-07,
+      "loss": 0.0012,
+      "step": 10211
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.8183720930232557e-07,
+      "loss": 0.001,
+      "step": 10212
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.818255813953488e-07,
+      "loss": 0.0011,
+      "step": 10213
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.818139534883721e-07,
+      "loss": 0.0014,
+      "step": 10214
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.8180232558139533e-07,
+      "loss": 0.3943,
+      "step": 10215
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.817906976744186e-07,
+      "loss": 0.5058,
+      "step": 10216
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.8177906976744183e-07,
+      "loss": 0.0008,
+      "step": 10217
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.8176744186046512e-07,
+      "loss": 0.0007,
+      "step": 10218
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.8175581395348838e-07,
+      "loss": 0.0018,
+      "step": 10219
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.8174418604651161e-07,
+      "loss": 0.0013,
+      "step": 10220
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.8173255813953488e-07,
+      "loss": 0.0009,
+      "step": 10221
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.8172093023255814e-07,
+      "loss": 0.0013,
+      "step": 10222
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.817093023255814e-07,
+      "loss": 0.0014,
+      "step": 10223
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.8169767441860463e-07,
+      "loss": 0.001,
+      "step": 10224
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.816860465116279e-07,
+      "loss": 0.4808,
+      "step": 10225
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.8167441860465116e-07,
+      "loss": 0.0013,
+      "step": 10226
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.8166279069767442e-07,
+      "loss": 0.0009,
+      "step": 10227
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.8165116279069765e-07,
+      "loss": 0.001,
+      "step": 10228
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.8163953488372092e-07,
+      "loss": 0.0014,
+      "step": 10229
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.816279069767442e-07,
+      "loss": 0.0008,
+      "step": 10230
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.8161627906976744e-07,
+      "loss": 0.0012,
+      "step": 10231
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.816046511627907e-07,
+      "loss": 0.0013,
+      "step": 10232
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.8159302325581394e-07,
+      "loss": 0.0014,
+      "step": 10233
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.815813953488372e-07,
+      "loss": 0.0011,
+      "step": 10234
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8156976744186046e-07,
+      "loss": 0.0008,
+      "step": 10235
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8155813953488372e-07,
+      "loss": 0.0016,
+      "step": 10236
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8154651162790696e-07,
+      "loss": 0.001,
+      "step": 10237
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8153488372093022e-07,
+      "loss": 0.4253,
+      "step": 10238
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8152325581395348e-07,
+      "loss": 0.0173,
+      "step": 10239
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.8151162790697674e-07,
+      "loss": 0.0008,
+      "step": 10240
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.8149999999999998e-07,
+      "loss": 0.0011,
+      "step": 10241
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.8148837209302324e-07,
+      "loss": 1.3857,
+      "step": 10242
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.8147674418604653e-07,
+      "loss": 0.0009,
+      "step": 10243
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.8146511627906976e-07,
+      "loss": 0.0011,
+      "step": 10244
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.81453488372093e-07,
+      "loss": 0.0007,
+      "step": 10245
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.8144186046511626e-07,
+      "loss": 0.0014,
+      "step": 10246
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.8143023255813955e-07,
+      "loss": 0.0011,
+      "step": 10247
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.8141860465116278e-07,
+      "loss": 0.0012,
+      "step": 10248
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.8140697674418604e-07,
+      "loss": 0.0012,
+      "step": 10249
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.8139534883720928e-07,
+      "loss": 0.0014,
+      "step": 10250
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.8138372093023257e-07,
+      "loss": 0.0017,
+      "step": 10251
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.813720930232558e-07,
+      "loss": 0.0009,
+      "step": 10252
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.8136046511627907e-07,
+      "loss": 0.001,
+      "step": 10253
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.813488372093023e-07,
+      "loss": 0.0013,
+      "step": 10254
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.813372093023256e-07,
+      "loss": 0.0014,
+      "step": 10255
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.8132558139534882e-07,
+      "loss": 0.0016,
+      "step": 10256
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.8131395348837209e-07,
+      "loss": 0.0009,
+      "step": 10257
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.8130232558139532e-07,
+      "loss": 0.5775,
+      "step": 10258
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.8129069767441858e-07,
+      "loss": 0.0011,
+      "step": 10259
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.8127906976744187e-07,
+      "loss": 0.0013,
+      "step": 10260
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.812674418604651e-07,
+      "loss": 0.0015,
+      "step": 10261
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.8125581395348837e-07,
+      "loss": 0.001,
+      "step": 10262
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.812441860465116e-07,
+      "loss": 0.001,
+      "step": 10263
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.812325581395349e-07,
+      "loss": 0.0008,
+      "step": 10264
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.8122093023255813e-07,
+      "loss": 0.0015,
+      "step": 10265
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.812093023255814e-07,
+      "loss": 0.001,
+      "step": 10266
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.8119767441860462e-07,
+      "loss": 0.0009,
+      "step": 10267
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.811860465116279e-07,
+      "loss": 0.0009,
+      "step": 10268
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.8117441860465115e-07,
+      "loss": 0.0012,
+      "step": 10269
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.811627906976744e-07,
+      "loss": 0.0012,
+      "step": 10270
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.8115116279069764e-07,
+      "loss": 0.2979,
+      "step": 10271
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.8113953488372093e-07,
+      "loss": 0.0016,
+      "step": 10272
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.811279069767442e-07,
+      "loss": 0.0014,
+      "step": 10273
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.8111627906976743e-07,
+      "loss": 0.001,
+      "step": 10274
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.811046511627907e-07,
+      "loss": 0.001,
+      "step": 10275
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.8109302325581395e-07,
+      "loss": 0.001,
+      "step": 10276
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.8108139534883721e-07,
+      "loss": 0.001,
+      "step": 10277
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.8106976744186045e-07,
+      "loss": 0.001,
+      "step": 10278
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.810581395348837e-07,
+      "loss": 0.0012,
+      "step": 10279
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.8104651162790697e-07,
+      "loss": 0.0013,
+      "step": 10280
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8103488372093023e-07,
+      "loss": 0.0011,
+      "step": 10281
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8102325581395347e-07,
+      "loss": 0.0011,
+      "step": 10282
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8101162790697673e-07,
+      "loss": 0.0009,
+      "step": 10283
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8100000000000002e-07,
+      "loss": 0.0012,
+      "step": 10284
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.8098837209302325e-07,
+      "loss": 0.001,
+      "step": 10285
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.8097674418604652e-07,
+      "loss": 0.0009,
+      "step": 10286
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.8096511627906975e-07,
+      "loss": 0.0016,
+      "step": 10287
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.80953488372093e-07,
+      "loss": 0.0011,
+      "step": 10288
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.8094186046511627e-07,
+      "loss": 0.0009,
+      "step": 10289
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.8093023255813954e-07,
+      "loss": 0.0647,
+      "step": 10290
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.8091860465116277e-07,
+      "loss": 0.0014,
+      "step": 10291
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.8090697674418603e-07,
+      "loss": 0.0009,
+      "step": 10292
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.808953488372093e-07,
+      "loss": 0.0008,
+      "step": 10293
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.8088372093023256e-07,
+      "loss": 0.0013,
+      "step": 10294
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.808720930232558e-07,
+      "loss": 0.0012,
+      "step": 10295
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.8086046511627905e-07,
+      "loss": 0.0013,
+      "step": 10296
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.8084883720930234e-07,
+      "loss": 0.0011,
+      "step": 10297
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.8083720930232558e-07,
+      "loss": 0.0013,
+      "step": 10298
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.808255813953488e-07,
+      "loss": 0.001,
+      "step": 10299
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.8081395348837207e-07,
+      "loss": 0.0011,
+      "step": 10300
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.8080232558139536e-07,
+      "loss": 0.0011,
+      "step": 10301
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.807906976744186e-07,
+      "loss": 0.0008,
+      "step": 10302
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8077906976744186e-07,
+      "loss": 0.0009,
+      "step": 10303
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.807674418604651e-07,
+      "loss": 0.0013,
+      "step": 10304
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8075581395348838e-07,
+      "loss": 0.0011,
+      "step": 10305
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8074418604651162e-07,
+      "loss": 0.0011,
+      "step": 10306
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.8073255813953488e-07,
+      "loss": 0.0016,
+      "step": 10307
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.8072093023255812e-07,
+      "loss": 0.0008,
+      "step": 10308
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.807093023255814e-07,
+      "loss": 0.001,
+      "step": 10309
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.8069767441860464e-07,
+      "loss": 0.0028,
+      "step": 10310
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.806860465116279e-07,
+      "loss": 0.0012,
+      "step": 10311
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.8067441860465114e-07,
+      "loss": 0.0008,
+      "step": 10312
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.806627906976744e-07,
+      "loss": 0.0011,
+      "step": 10313
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.8065116279069769e-07,
+      "loss": 0.0014,
+      "step": 10314
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.8063953488372092e-07,
+      "loss": 0.0014,
+      "step": 10315
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.8062790697674418e-07,
+      "loss": 0.0011,
+      "step": 10316
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.8061627906976742e-07,
+      "loss": 0.0047,
+      "step": 10317
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.806046511627907e-07,
+      "loss": 0.0015,
+      "step": 10318
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.8059302325581394e-07,
+      "loss": 0.0009,
+      "step": 10319
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.805813953488372e-07,
+      "loss": 0.0061,
+      "step": 10320
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.8056976744186044e-07,
+      "loss": 0.0011,
+      "step": 10321
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.8055813953488373e-07,
+      "loss": 0.0018,
+      "step": 10322
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.8054651162790696e-07,
+      "loss": 0.0011,
+      "step": 10323
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.8053488372093022e-07,
+      "loss": 0.0009,
+      "step": 10324
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.8052325581395346e-07,
+      "loss": 0.0024,
+      "step": 10325
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.8051162790697675e-07,
+      "loss": 1.7537,
+      "step": 10326
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.805e-07,
+      "loss": 0.2607,
+      "step": 10327
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.8048837209302324e-07,
+      "loss": 0.0011,
+      "step": 10328
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.804767441860465e-07,
+      "loss": 0.0012,
+      "step": 10329
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.8046511627906977e-07,
+      "loss": 0.0008,
+      "step": 10330
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.8045348837209303e-07,
+      "loss": 0.0013,
+      "step": 10331
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.8044186046511626e-07,
+      "loss": 0.0009,
+      "step": 10332
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.8043023255813953e-07,
+      "loss": 0.0026,
+      "step": 10333
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.804186046511628e-07,
+      "loss": 0.0007,
+      "step": 10334
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.8040697674418605e-07,
+      "loss": 0.0011,
+      "step": 10335
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.8039534883720928e-07,
+      "loss": 0.0011,
+      "step": 10336
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.8038372093023255e-07,
+      "loss": 0.0011,
+      "step": 10337
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.8037209302325583e-07,
+      "loss": 0.0283,
+      "step": 10338
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.8036046511627907e-07,
+      "loss": 0.0011,
+      "step": 10339
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.8034883720930233e-07,
+      "loss": 0.001,
+      "step": 10340
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.9893617021276596,
+      "eval_loss": 0.15206366777420044,
+      "eval_runtime": 1.5556,
+      "eval_samples_per_second": 664.715,
+      "eval_steps_per_second": 10.929,
+      "step": 10340
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.8033720930232557e-07,
+      "loss": 0.0093,
+      "step": 10341
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.803255813953488e-07,
+      "loss": 0.0008,
+      "step": 10342
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.803139534883721e-07,
+      "loss": 0.0009,
+      "step": 10343
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.8030232558139535e-07,
+      "loss": 0.0012,
+      "step": 10344
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.8029069767441859e-07,
+      "loss": 0.0025,
+      "step": 10345
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.8027906976744185e-07,
+      "loss": 0.0054,
+      "step": 10346
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.802674418604651e-07,
+      "loss": 0.0013,
+      "step": 10347
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.8025581395348837e-07,
+      "loss": 0.0009,
+      "step": 10348
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.802441860465116e-07,
+      "loss": 0.0011,
+      "step": 10349
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.8023255813953487e-07,
+      "loss": 2.024,
+      "step": 10350
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.8022093023255816e-07,
+      "loss": 0.4132,
+      "step": 10351
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.802093023255814e-07,
+      "loss": 0.0013,
+      "step": 10352
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.8019767441860463e-07,
+      "loss": 0.0009,
+      "step": 10353
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.801860465116279e-07,
+      "loss": 0.0012,
+      "step": 10354
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.8017441860465118e-07,
+      "loss": 0.8531,
+      "step": 10355
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.801627906976744e-07,
+      "loss": 0.0016,
+      "step": 10356
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.8015116279069767e-07,
+      "loss": 0.0024,
+      "step": 10357
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.801395348837209e-07,
+      "loss": 0.0013,
+      "step": 10358
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.801279069767442e-07,
+      "loss": 0.0012,
+      "step": 10359
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.8011627906976743e-07,
+      "loss": 0.0008,
+      "step": 10360
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.801046511627907e-07,
+      "loss": 0.2775,
+      "step": 10361
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.8009302325581393e-07,
+      "loss": 0.0012,
+      "step": 10362
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.8008139534883722e-07,
+      "loss": 0.0015,
+      "step": 10363
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.8006976744186045e-07,
+      "loss": 0.0009,
+      "step": 10364
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.8005813953488372e-07,
+      "loss": 0.0008,
+      "step": 10365
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.8004651162790695e-07,
+      "loss": 0.0011,
+      "step": 10366
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.800348837209302e-07,
+      "loss": 0.0012,
+      "step": 10367
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.800232558139535e-07,
+      "loss": 0.0009,
+      "step": 10368
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.8001162790697674e-07,
+      "loss": 0.0013,
+      "step": 10369
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.8e-07,
+      "loss": 0.0011,
+      "step": 10370
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.7998837209302323e-07,
+      "loss": 0.0015,
+      "step": 10371
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.7997674418604652e-07,
+      "loss": 0.0009,
+      "step": 10372
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.7996511627906976e-07,
+      "loss": 0.0012,
+      "step": 10373
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.7995348837209302e-07,
+      "loss": 1.2234,
+      "step": 10374
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.7994186046511625e-07,
+      "loss": 0.7376,
+      "step": 10375
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.7993023255813954e-07,
+      "loss": 0.0011,
+      "step": 10376
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.7991860465116278e-07,
+      "loss": 0.0013,
+      "step": 10377
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.7990697674418604e-07,
+      "loss": 0.0009,
+      "step": 10378
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.7989534883720927e-07,
+      "loss": 0.0009,
+      "step": 10379
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.7988372093023256e-07,
+      "loss": 0.0009,
+      "step": 10380
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.7987209302325582e-07,
+      "loss": 0.3484,
+      "step": 10381
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.7986046511627906e-07,
+      "loss": 0.001,
+      "step": 10382
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.798488372093023e-07,
+      "loss": 0.0012,
+      "step": 10383
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.7983720930232558e-07,
+      "loss": 0.002,
+      "step": 10384
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.7982558139534884e-07,
+      "loss": 0.0287,
+      "step": 10385
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.7981395348837208e-07,
+      "loss": 0.0012,
+      "step": 10386
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.7980232558139534e-07,
+      "loss": 1.529,
+      "step": 10387
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.797906976744186e-07,
+      "loss": 0.0017,
+      "step": 10388
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.7977906976744186e-07,
+      "loss": 3.1383,
+      "step": 10389
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.797674418604651e-07,
+      "loss": 0.0014,
+      "step": 10390
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.7975581395348836e-07,
+      "loss": 0.001,
+      "step": 10391
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.797441860465116e-07,
+      "loss": 0.0012,
+      "step": 10392
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.7973255813953488e-07,
+      "loss": 0.0008,
+      "step": 10393
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.7972093023255812e-07,
+      "loss": 0.0013,
+      "step": 10394
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.7970930232558138e-07,
+      "loss": 0.0013,
+      "step": 10395
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.7969767441860462e-07,
+      "loss": 0.0087,
+      "step": 10396
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.796860465116279e-07,
+      "loss": 0.001,
+      "step": 10397
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.7967441860465117e-07,
+      "loss": 0.0007,
+      "step": 10398
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.796627906976744e-07,
+      "loss": 0.0018,
+      "step": 10399
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.7965116279069766e-07,
+      "loss": 0.0011,
+      "step": 10400
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.7963953488372092e-07,
+      "loss": 0.0013,
+      "step": 10401
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.7962790697674419e-07,
+      "loss": 0.0013,
+      "step": 10402
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.7961627906976742e-07,
+      "loss": 0.5969,
+      "step": 10403
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.7960465116279068e-07,
+      "loss": 0.001,
+      "step": 10404
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.7959302325581395e-07,
+      "loss": 0.0019,
+      "step": 10405
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.795813953488372e-07,
+      "loss": 0.0009,
+      "step": 10406
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.7956976744186044e-07,
+      "loss": 0.0009,
+      "step": 10407
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.795581395348837e-07,
+      "loss": 0.0009,
+      "step": 10408
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.79546511627907e-07,
+      "loss": 0.001,
+      "step": 10409
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.7953488372093023e-07,
+      "loss": 0.0006,
+      "step": 10410
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.795232558139535e-07,
+      "loss": 0.0012,
+      "step": 10411
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.7951162790697672e-07,
+      "loss": 0.0012,
+      "step": 10412
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.795e-07,
+      "loss": 0.6377,
+      "step": 10413
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.7948837209302325e-07,
+      "loss": 0.0011,
+      "step": 10414
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.794767441860465e-07,
+      "loss": 0.0011,
+      "step": 10415
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.7946511627906974e-07,
+      "loss": 0.0011,
+      "step": 10416
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.7945348837209303e-07,
+      "loss": 2.0635,
+      "step": 10417
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.7944186046511627e-07,
+      "loss": 0.0014,
+      "step": 10418
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.7943023255813953e-07,
+      "loss": 0.0091,
+      "step": 10419
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.7941860465116277e-07,
+      "loss": 2.0889,
+      "step": 10420
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.7940697674418603e-07,
+      "loss": 0.001,
+      "step": 10421
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.7939534883720931e-07,
+      "loss": 0.0014,
+      "step": 10422
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.7938372093023255e-07,
+      "loss": 0.002,
+      "step": 10423
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.793720930232558e-07,
+      "loss": 0.001,
+      "step": 10424
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.7936046511627905e-07,
+      "loss": 0.0032,
+      "step": 10425
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.7934883720930234e-07,
+      "loss": 0.0011,
+      "step": 10426
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.7933720930232557e-07,
+      "loss": 0.0011,
+      "step": 10427
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.7932558139534883e-07,
+      "loss": 0.0017,
+      "step": 10428
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.7931395348837207e-07,
+      "loss": 0.0031,
+      "step": 10429
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.7930232558139536e-07,
+      "loss": 0.0009,
+      "step": 10430
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.792906976744186e-07,
+      "loss": 0.0007,
+      "step": 10431
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.7927906976744185e-07,
+      "loss": 0.001,
+      "step": 10432
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.792674418604651e-07,
+      "loss": 0.001,
+      "step": 10433
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.7925581395348838e-07,
+      "loss": 0.0015,
+      "step": 10434
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.7924418604651164e-07,
+      "loss": 1.7799,
+      "step": 10435
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.7923255813953487e-07,
+      "loss": 0.0015,
+      "step": 10436
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.792209302325581e-07,
+      "loss": 0.0012,
+      "step": 10437
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.792093023255814e-07,
+      "loss": 0.001,
+      "step": 10438
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.7919767441860466e-07,
+      "loss": 0.0009,
+      "step": 10439
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.791860465116279e-07,
+      "loss": 0.003,
+      "step": 10440
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.7917441860465116e-07,
+      "loss": 0.0013,
+      "step": 10441
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.7916279069767442e-07,
+      "loss": 0.0013,
+      "step": 10442
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.7915116279069768e-07,
+      "loss": 0.0025,
+      "step": 10443
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.7913953488372091e-07,
+      "loss": 0.0014,
+      "step": 10444
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.7912790697674418e-07,
+      "loss": 0.0012,
+      "step": 10445
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.791162790697674e-07,
+      "loss": 0.0012,
+      "step": 10446
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.791046511627907e-07,
+      "loss": 0.0011,
+      "step": 10447
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.7909302325581393e-07,
+      "loss": 0.001,
+      "step": 10448
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.790813953488372e-07,
+      "loss": 0.0009,
+      "step": 10449
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.7906976744186043e-07,
+      "loss": 0.0013,
+      "step": 10450
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.7905813953488372e-07,
+      "loss": 0.0011,
+      "step": 10451
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.7904651162790698e-07,
+      "loss": 0.0012,
+      "step": 10452
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.7903488372093022e-07,
+      "loss": 0.0012,
+      "step": 10453
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.7902325581395348e-07,
+      "loss": 1.0859,
+      "step": 10454
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.7901162790697674e-07,
+      "loss": 0.0237,
+      "step": 10455
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.79e-07,
+      "loss": 0.0011,
+      "step": 10456
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.7898837209302324e-07,
+      "loss": 0.001,
+      "step": 10457
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.789767441860465e-07,
+      "loss": 0.001,
+      "step": 10458
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.7896511627906976e-07,
+      "loss": 0.0012,
+      "step": 10459
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.7895348837209302e-07,
+      "loss": 0.0021,
+      "step": 10460
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.7894186046511626e-07,
+      "loss": 0.0009,
+      "step": 10461
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.7893023255813952e-07,
+      "loss": 0.0013,
+      "step": 10462
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.789186046511628e-07,
+      "loss": 0.0015,
+      "step": 10463
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.7890697674418604e-07,
+      "loss": 0.0014,
+      "step": 10464
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.788953488372093e-07,
+      "loss": 0.0007,
+      "step": 10465
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.7888372093023254e-07,
+      "loss": 0.0053,
+      "step": 10466
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.7887209302325583e-07,
+      "loss": 0.001,
+      "step": 10467
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.7886046511627906e-07,
+      "loss": 0.0012,
+      "step": 10468
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.7884883720930232e-07,
+      "loss": 0.0012,
+      "step": 10469
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.7883720930232556e-07,
+      "loss": 0.0009,
+      "step": 10470
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.7882558139534885e-07,
+      "loss": 0.001,
+      "step": 10471
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.7881395348837208e-07,
+      "loss": 0.001,
+      "step": 10472
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.7880232558139534e-07,
+      "loss": 0.0008,
+      "step": 10473
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.7879069767441858e-07,
+      "loss": 0.0018,
+      "step": 10474
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.7877906976744184e-07,
+      "loss": 0.0015,
+      "step": 10475
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.7876744186046513e-07,
+      "loss": 0.0041,
+      "step": 10476
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.7875581395348837e-07,
+      "loss": 0.0012,
+      "step": 10477
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.7874418604651163e-07,
+      "loss": 0.0013,
+      "step": 10478
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.7873255813953486e-07,
+      "loss": 0.1405,
+      "step": 10479
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.7872093023255815e-07,
+      "loss": 0.0012,
+      "step": 10480
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.7870930232558139e-07,
+      "loss": 0.0012,
+      "step": 10481
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.7869767441860465e-07,
+      "loss": 0.001,
+      "step": 10482
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.7868604651162788e-07,
+      "loss": 0.0011,
+      "step": 10483
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.7867441860465117e-07,
+      "loss": 0.0008,
+      "step": 10484
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.786627906976744e-07,
+      "loss": 0.0008,
+      "step": 10485
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.7865116279069767e-07,
+      "loss": 0.0021,
+      "step": 10486
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.786395348837209e-07,
+      "loss": 0.0009,
+      "step": 10487
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.786279069767442e-07,
+      "loss": 0.0018,
+      "step": 10488
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.7861627906976743e-07,
+      "loss": 0.0012,
+      "step": 10489
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.786046511627907e-07,
+      "loss": 0.0016,
+      "step": 10490
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.7859302325581392e-07,
+      "loss": 0.0011,
+      "step": 10491
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.785813953488372e-07,
+      "loss": 0.0011,
+      "step": 10492
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.7856976744186047e-07,
+      "loss": 0.0011,
+      "step": 10493
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.785581395348837e-07,
+      "loss": 0.0015,
+      "step": 10494
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.7854651162790697e-07,
+      "loss": 0.001,
+      "step": 10495
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.7853488372093023e-07,
+      "loss": 0.0014,
+      "step": 10496
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.785232558139535e-07,
+      "loss": 0.0012,
+      "step": 10497
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.7851162790697673e-07,
+      "loss": 0.0012,
+      "step": 10498
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.785e-07,
+      "loss": 0.0018,
+      "step": 10499
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.7848837209302323e-07,
+      "loss": 0.0009,
+      "step": 10500
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.7847674418604651e-07,
+      "loss": 0.0013,
+      "step": 10501
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.7846511627906975e-07,
+      "loss": 0.0008,
+      "step": 10502
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.78453488372093e-07,
+      "loss": 0.1911,
+      "step": 10503
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.7844186046511625e-07,
+      "loss": 0.0017,
+      "step": 10504
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.7843023255813953e-07,
+      "loss": 0.0013,
+      "step": 10505
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.784186046511628e-07,
+      "loss": 0.0016,
+      "step": 10506
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.7840697674418603e-07,
+      "loss": 0.0009,
+      "step": 10507
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.783953488372093e-07,
+      "loss": 0.0018,
+      "step": 10508
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.7838372093023255e-07,
+      "loss": 0.0008,
+      "step": 10509
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.7837209302325582e-07,
+      "loss": 0.0013,
+      "step": 10510
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.7836046511627905e-07,
+      "loss": 0.0008,
+      "step": 10511
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.783488372093023e-07,
+      "loss": 0.0011,
+      "step": 10512
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.7833720930232557e-07,
+      "loss": 0.0011,
+      "step": 10513
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.7832558139534884e-07,
+      "loss": 0.0013,
+      "step": 10514
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.7831395348837207e-07,
+      "loss": 0.001,
+      "step": 10515
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.7830232558139533e-07,
+      "loss": 0.0012,
+      "step": 10516
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.7829069767441862e-07,
+      "loss": 0.0029,
+      "step": 10517
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.7827906976744186e-07,
+      "loss": 0.0008,
+      "step": 10518
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.7826744186046512e-07,
+      "loss": 0.0007,
+      "step": 10519
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.7825581395348835e-07,
+      "loss": 0.0016,
+      "step": 10520
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.7824418604651164e-07,
+      "loss": 0.0012,
+      "step": 10521
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.7823255813953488e-07,
+      "loss": 0.0016,
+      "step": 10522
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.7822093023255814e-07,
+      "loss": 0.001,
+      "step": 10523
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.7820930232558137e-07,
+      "loss": 0.2837,
+      "step": 10524
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.7819767441860464e-07,
+      "loss": 0.0008,
+      "step": 10525
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.781860465116279e-07,
+      "loss": 0.0009,
+      "step": 10526
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.7817441860465116e-07,
+      "loss": 0.001,
+      "step": 10527
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.781627906976744e-07,
+      "loss": 0.024,
+      "step": 10528
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.7815116279069766e-07,
+      "loss": 0.0009,
+      "step": 10529
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.7813953488372094e-07,
+      "loss": 0.0011,
+      "step": 10530
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.7812790697674418e-07,
+      "loss": 0.0008,
+      "step": 10531
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.7811627906976742e-07,
+      "loss": 0.001,
+      "step": 10532
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.7810465116279068e-07,
+      "loss": 0.0008,
+      "step": 10533
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.7809302325581396e-07,
+      "loss": 0.0015,
+      "step": 10534
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.780813953488372e-07,
+      "loss": 0.001,
+      "step": 10535
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.7806976744186046e-07,
+      "loss": 0.0009,
+      "step": 10536
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.780581395348837e-07,
+      "loss": 0.0009,
+      "step": 10537
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.7804651162790699e-07,
+      "loss": 0.0008,
+      "step": 10538
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.7803488372093022e-07,
+      "loss": 0.002,
+      "step": 10539
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.7802325581395348e-07,
+      "loss": 0.001,
+      "step": 10540
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.7801162790697672e-07,
+      "loss": 0.0013,
+      "step": 10541
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.78e-07,
+      "loss": 0.0031,
+      "step": 10542
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.7798837209302324e-07,
+      "loss": 0.004,
+      "step": 10543
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.779767441860465e-07,
+      "loss": 0.0014,
+      "step": 10544
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.7796511627906974e-07,
+      "loss": 0.001,
+      "step": 10545
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.7795348837209303e-07,
+      "loss": 0.0009,
+      "step": 10546
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.779418604651163e-07,
+      "loss": 0.0032,
+      "step": 10547
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.7793023255813952e-07,
+      "loss": 0.0009,
+      "step": 10548
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.7791860465116278e-07,
+      "loss": 0.0007,
+      "step": 10549
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.7790697674418605e-07,
+      "loss": 0.0014,
+      "step": 10550
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.778953488372093e-07,
+      "loss": 0.0011,
+      "step": 10551
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.7788372093023254e-07,
+      "loss": 0.0015,
+      "step": 10552
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.778720930232558e-07,
+      "loss": 0.0011,
+      "step": 10553
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.7786046511627904e-07,
+      "loss": 0.001,
+      "step": 10554
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.7784883720930233e-07,
+      "loss": 0.0079,
+      "step": 10555
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.7783720930232556e-07,
+      "loss": 0.0019,
+      "step": 10556
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.7782558139534883e-07,
+      "loss": 0.0009,
+      "step": 10557
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.7781395348837206e-07,
+      "loss": 0.0007,
+      "step": 10558
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.7780232558139535e-07,
+      "loss": 0.0011,
+      "step": 10559
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.777906976744186e-07,
+      "loss": 0.0012,
+      "step": 10560
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.7777906976744185e-07,
+      "loss": 0.0009,
+      "step": 10561
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.777674418604651e-07,
+      "loss": 0.0009,
+      "step": 10562
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.7775581395348837e-07,
+      "loss": 0.0012,
+      "step": 10563
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.7774418604651163e-07,
+      "loss": 0.001,
+      "step": 10564
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.7773255813953487e-07,
+      "loss": 0.0028,
+      "step": 10565
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.7772093023255813e-07,
+      "loss": 0.0012,
+      "step": 10566
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.777093023255814e-07,
+      "loss": 0.001,
+      "step": 10567
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.7769767441860465e-07,
+      "loss": 0.001,
+      "step": 10568
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.7768604651162789e-07,
+      "loss": 0.0008,
+      "step": 10569
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.7767441860465115e-07,
+      "loss": 0.0008,
+      "step": 10570
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.7766279069767444e-07,
+      "loss": 0.0017,
+      "step": 10571
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.7765116279069767e-07,
+      "loss": 0.0024,
+      "step": 10572
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.7763953488372093e-07,
+      "loss": 0.0011,
+      "step": 10573
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.7762790697674417e-07,
+      "loss": 0.0011,
+      "step": 10574
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.7761627906976746e-07,
+      "loss": 0.0013,
+      "step": 10575
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.776046511627907e-07,
+      "loss": 0.0012,
+      "step": 10576
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.7759302325581395e-07,
+      "loss": 0.001,
+      "step": 10577
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.775813953488372e-07,
+      "loss": 0.0017,
+      "step": 10578
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.7756976744186045e-07,
+      "loss": 0.0011,
+      "step": 10579
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.775581395348837e-07,
+      "loss": 0.0012,
+      "step": 10580
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.7754651162790697e-07,
+      "loss": 0.0071,
+      "step": 10581
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.775348837209302e-07,
+      "loss": 0.001,
+      "step": 10582
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.7752325581395347e-07,
+      "loss": 0.0012,
+      "step": 10583
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.7751162790697676e-07,
+      "loss": 0.0768,
+      "step": 10584
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.775e-07,
+      "loss": 0.001,
+      "step": 10585
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.7748837209302323e-07,
+      "loss": 0.0011,
+      "step": 10586
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.774767441860465e-07,
+      "loss": 0.0018,
+      "step": 10587
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.7746511627906978e-07,
+      "loss": 0.0012,
+      "step": 10588
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.7745348837209301e-07,
+      "loss": 0.0008,
+      "step": 10589
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.7744186046511628e-07,
+      "loss": 0.8451,
+      "step": 10590
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.774302325581395e-07,
+      "loss": 0.0012,
+      "step": 10591
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.774186046511628e-07,
+      "loss": 0.001,
+      "step": 10592
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.7740697674418604e-07,
+      "loss": 0.0011,
+      "step": 10593
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.773953488372093e-07,
+      "loss": 0.0014,
+      "step": 10594
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.7738372093023253e-07,
+      "loss": 0.0026,
+      "step": 10595
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.7737209302325582e-07,
+      "loss": 0.0009,
+      "step": 10596
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.7736046511627906e-07,
+      "loss": 1.1645,
+      "step": 10597
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.7734883720930232e-07,
+      "loss": 0.001,
+      "step": 10598
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.7733720930232555e-07,
+      "loss": 0.001,
+      "step": 10599
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.7732558139534884e-07,
+      "loss": 0.001,
+      "step": 10600
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.773139534883721e-07,
+      "loss": 0.001,
+      "step": 10601
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.7730232558139534e-07,
+      "loss": 0.0015,
+      "step": 10602
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.772906976744186e-07,
+      "loss": 0.0016,
+      "step": 10603
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.7727906976744186e-07,
+      "loss": 0.0009,
+      "step": 10604
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.7726744186046512e-07,
+      "loss": 0.0011,
+      "step": 10605
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.7725581395348836e-07,
+      "loss": 0.001,
+      "step": 10606
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.7724418604651162e-07,
+      "loss": 0.0007,
+      "step": 10607
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.7723255813953486e-07,
+      "loss": 0.0017,
+      "step": 10608
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.7722093023255814e-07,
+      "loss": 0.001,
+      "step": 10609
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.7720930232558138e-07,
+      "loss": 0.0012,
+      "step": 10610
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.7719767441860464e-07,
+      "loss": 0.0012,
+      "step": 10611
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.7718604651162788e-07,
+      "loss": 0.001,
+      "step": 10612
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.7717441860465116e-07,
+      "loss": 0.0009,
+      "step": 10613
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.7716279069767443e-07,
+      "loss": 0.0013,
+      "step": 10614
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.7715116279069766e-07,
+      "loss": 0.001,
+      "step": 10615
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.771395348837209e-07,
+      "loss": 0.0011,
+      "step": 10616
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.7712790697674418e-07,
+      "loss": 0.0007,
+      "step": 10617
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.7711627906976745e-07,
+      "loss": 0.0012,
+      "step": 10618
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.7710465116279068e-07,
+      "loss": 0.0009,
+      "step": 10619
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.7709302325581394e-07,
+      "loss": 0.0008,
+      "step": 10620
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.770813953488372e-07,
+      "loss": 0.0009,
+      "step": 10621
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.7706976744186047e-07,
+      "loss": 0.0009,
+      "step": 10622
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.770581395348837e-07,
+      "loss": 0.0011,
+      "step": 10623
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.7704651162790696e-07,
+      "loss": 0.0009,
+      "step": 10624
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.7703488372093025e-07,
+      "loss": 0.0009,
+      "step": 10625
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.7702325581395349e-07,
+      "loss": 0.0012,
+      "step": 10626
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.7701162790697672e-07,
+      "loss": 0.001,
+      "step": 10627
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.7699999999999998e-07,
+      "loss": 0.0063,
+      "step": 10628
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.7698837209302327e-07,
+      "loss": 0.0012,
+      "step": 10629
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.769767441860465e-07,
+      "loss": 0.0008,
+      "step": 10630
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.7696511627906977e-07,
+      "loss": 2.8478,
+      "step": 10631
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.76953488372093e-07,
+      "loss": 0.0008,
+      "step": 10632
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.7694186046511627e-07,
+      "loss": 0.0008,
+      "step": 10633
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.7693023255813953e-07,
+      "loss": 0.0011,
+      "step": 10634
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.769186046511628e-07,
+      "loss": 0.001,
+      "step": 10635
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.7690697674418602e-07,
+      "loss": 0.0013,
+      "step": 10636
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.7689534883720929e-07,
+      "loss": 0.001,
+      "step": 10637
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.7688372093023255e-07,
+      "loss": 0.001,
+      "step": 10638
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.768720930232558e-07,
+      "loss": 0.0023,
+      "step": 10639
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.7686046511627904e-07,
+      "loss": 0.0011,
+      "step": 10640
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.768488372093023e-07,
+      "loss": 0.0011,
+      "step": 10641
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.768372093023256e-07,
+      "loss": 0.0013,
+      "step": 10642
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.7682558139534883e-07,
+      "loss": 0.0015,
+      "step": 10643
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.768139534883721e-07,
+      "loss": 0.0013,
+      "step": 10644
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.7680232558139533e-07,
+      "loss": 0.0008,
+      "step": 10645
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.7679069767441861e-07,
+      "loss": 0.0012,
+      "step": 10646
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.7677906976744185e-07,
+      "loss": 0.0026,
+      "step": 10647
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.767674418604651e-07,
+      "loss": 0.0013,
+      "step": 10648
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.7675581395348835e-07,
+      "loss": 0.3103,
+      "step": 10649
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.7674418604651164e-07,
+      "loss": 0.0011,
+      "step": 10650
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.7673255813953487e-07,
+      "loss": 0.001,
+      "step": 10651
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.7672093023255813e-07,
+      "loss": 0.0016,
+      "step": 10652
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.7670930232558137e-07,
+      "loss": 0.0009,
+      "step": 10653
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.7669767441860466e-07,
+      "loss": 0.0012,
+      "step": 10654
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.7668604651162792e-07,
+      "loss": 0.0009,
+      "step": 10655
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.7667441860465115e-07,
+      "loss": 0.0009,
+      "step": 10656
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.7666279069767441e-07,
+      "loss": 0.0024,
+      "step": 10657
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7665116279069765e-07,
+      "loss": 0.001,
+      "step": 10658
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7663953488372094e-07,
+      "loss": 0.0009,
+      "step": 10659
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7662790697674417e-07,
+      "loss": 0.0036,
+      "step": 10660
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7661627906976743e-07,
+      "loss": 0.0008,
+      "step": 10661
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7660465116279067e-07,
+      "loss": 0.001,
+      "step": 10662
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7659302325581396e-07,
+      "loss": 0.0009,
+      "step": 10663
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.765813953488372e-07,
+      "loss": 0.0011,
+      "step": 10664
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.7656976744186046e-07,
+      "loss": 0.0013,
+      "step": 10665
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.765581395348837e-07,
+      "loss": 0.0008,
+      "step": 10666
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.7654651162790698e-07,
+      "loss": 0.001,
+      "step": 10667
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.7653488372093024e-07,
+      "loss": 0.0012,
+      "step": 10668
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.7652325581395348e-07,
+      "loss": 0.001,
+      "step": 10669
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.765116279069767e-07,
+      "loss": 0.068,
+      "step": 10670
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.765e-07,
+      "loss": 0.0082,
+      "step": 10671
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.7648837209302326e-07,
+      "loss": 0.0009,
+      "step": 10672
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.764767441860465e-07,
+      "loss": 0.001,
+      "step": 10673
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7646511627906976e-07,
+      "loss": 0.001,
+      "step": 10674
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7645348837209302e-07,
+      "loss": 0.0008,
+      "step": 10675
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7644186046511628e-07,
+      "loss": 0.001,
+      "step": 10676
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7643023255813952e-07,
+      "loss": 0.001,
+      "step": 10677
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7641860465116278e-07,
+      "loss": 0.0277,
+      "step": 10678
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.7640697674418607e-07,
+      "loss": 0.0011,
+      "step": 10679
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.763953488372093e-07,
+      "loss": 0.0011,
+      "step": 10680
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.7638372093023254e-07,
+      "loss": 0.0012,
+      "step": 10681
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.763720930232558e-07,
+      "loss": 0.001,
+      "step": 10682
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.7636046511627909e-07,
+      "loss": 0.0008,
+      "step": 10683
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7634883720930232e-07,
+      "loss": 0.001,
+      "step": 10684
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7633720930232558e-07,
+      "loss": 0.0015,
+      "step": 10685
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7632558139534882e-07,
+      "loss": 0.001,
+      "step": 10686
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7631395348837208e-07,
+      "loss": 0.0011,
+      "step": 10687
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7630232558139534e-07,
+      "loss": 0.0014,
+      "step": 10688
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.762906976744186e-07,
+      "loss": 0.0007,
+      "step": 10689
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7627906976744184e-07,
+      "loss": 0.0007,
+      "step": 10690
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.762674418604651e-07,
+      "loss": 0.001,
+      "step": 10691
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7625581395348836e-07,
+      "loss": 0.0017,
+      "step": 10692
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7624418604651162e-07,
+      "loss": 0.003,
+      "step": 10693
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.7623255813953486e-07,
+      "loss": 0.001,
+      "step": 10694
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.7622093023255812e-07,
+      "loss": 0.001,
+      "step": 10695
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.762093023255814e-07,
+      "loss": 0.001,
+      "step": 10696
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.7619767441860464e-07,
+      "loss": 0.0032,
+      "step": 10697
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.761860465116279e-07,
+      "loss": 0.0011,
+      "step": 10698
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.7617441860465114e-07,
+      "loss": 0.0012,
+      "step": 10699
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7616279069767443e-07,
+      "loss": 0.0011,
+      "step": 10700
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7615116279069766e-07,
+      "loss": 0.0009,
+      "step": 10701
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7613953488372093e-07,
+      "loss": 0.0016,
+      "step": 10702
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7612790697674416e-07,
+      "loss": 0.0014,
+      "step": 10703
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.7611627906976745e-07,
+      "loss": 0.0016,
+      "step": 10704
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7610465116279069e-07,
+      "loss": 0.0011,
+      "step": 10705
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7609302325581395e-07,
+      "loss": 0.0009,
+      "step": 10706
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7608139534883718e-07,
+      "loss": 0.0008,
+      "step": 10707
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7606976744186047e-07,
+      "loss": 0.0008,
+      "step": 10708
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.7605813953488373e-07,
+      "loss": 0.0006,
+      "step": 10709
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.7604651162790697e-07,
+      "loss": 0.0014,
+      "step": 10710
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.7603488372093023e-07,
+      "loss": 0.0009,
+      "step": 10711
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.7602325581395346e-07,
+      "loss": 0.001,
+      "step": 10712
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.7601162790697675e-07,
+      "loss": 0.001,
+      "step": 10713
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.76e-07,
+      "loss": 0.0008,
+      "step": 10714
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7598837209302325e-07,
+      "loss": 0.0012,
+      "step": 10715
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7597674418604648e-07,
+      "loss": 0.0014,
+      "step": 10716
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7596511627906977e-07,
+      "loss": 0.0012,
+      "step": 10717
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.75953488372093e-07,
+      "loss": 0.0012,
+      "step": 10718
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7594186046511627e-07,
+      "loss": 0.0008,
+      "step": 10719
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.759302325581395e-07,
+      "loss": 0.0014,
+      "step": 10720
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.759186046511628e-07,
+      "loss": 0.3004,
+      "step": 10721
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.7590697674418606e-07,
+      "loss": 0.0011,
+      "step": 10722
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.758953488372093e-07,
+      "loss": 0.0011,
+      "step": 10723
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.7588372093023253e-07,
+      "loss": 0.0117,
+      "step": 10724
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.7587209302325581e-07,
+      "loss": 0.0023,
+      "step": 10725
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.7586046511627908e-07,
+      "loss": 0.0015,
+      "step": 10726
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.758488372093023e-07,
+      "loss": 0.0008,
+      "step": 10727
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.7583720930232557e-07,
+      "loss": 0.0009,
+      "step": 10728
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.7582558139534883e-07,
+      "loss": 0.0018,
+      "step": 10729
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.758139534883721e-07,
+      "loss": 0.0012,
+      "step": 10730
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.7580232558139533e-07,
+      "loss": 0.001,
+      "step": 10731
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.757906976744186e-07,
+      "loss": 0.0016,
+      "step": 10732
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.7577906976744185e-07,
+      "loss": 0.0009,
+      "step": 10733
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.7576744186046512e-07,
+      "loss": 2.9986,
+      "step": 10734
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.7575581395348835e-07,
+      "loss": 0.0009,
+      "step": 10735
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.757441860465116e-07,
+      "loss": 0.0009,
+      "step": 10736
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.757325581395349e-07,
+      "loss": 0.0011,
+      "step": 10737
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.7572093023255814e-07,
+      "loss": 0.0012,
+      "step": 10738
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.757093023255814e-07,
+      "loss": 2.3783,
+      "step": 10739
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.7569767441860463e-07,
+      "loss": 0.0011,
+      "step": 10740
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.756860465116279e-07,
+      "loss": 0.0008,
+      "step": 10741
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.7567441860465116e-07,
+      "loss": 0.001,
+      "step": 10742
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.7566279069767442e-07,
+      "loss": 1.2206,
+      "step": 10743
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.7565116279069765e-07,
+      "loss": 0.001,
+      "step": 10744
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.7563953488372092e-07,
+      "loss": 0.0011,
+      "step": 10745
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7562790697674418e-07,
+      "loss": 0.0013,
+      "step": 10746
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7561627906976744e-07,
+      "loss": 0.0013,
+      "step": 10747
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7560465116279067e-07,
+      "loss": 0.0008,
+      "step": 10748
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7559302325581394e-07,
+      "loss": 0.0011,
+      "step": 10749
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7558139534883722e-07,
+      "loss": 0.0009,
+      "step": 10750
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.7556976744186046e-07,
+      "loss": 0.0012,
+      "step": 10751
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.7555813953488372e-07,
+      "loss": 0.0014,
+      "step": 10752
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.7554651162790696e-07,
+      "loss": 0.0012,
+      "step": 10753
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.7553488372093024e-07,
+      "loss": 0.0017,
+      "step": 10754
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.7552325581395348e-07,
+      "loss": 0.0011,
+      "step": 10755
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.7551162790697674e-07,
+      "loss": 0.0011,
+      "step": 10756
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.7549999999999998e-07,
+      "loss": 0.0009,
+      "step": 10757
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.7548837209302326e-07,
+      "loss": 0.0015,
+      "step": 10758
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.754767441860465e-07,
+      "loss": 0.0009,
+      "step": 10759
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.7546511627906976e-07,
+      "loss": 0.0013,
+      "step": 10760
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.75453488372093e-07,
+      "loss": 0.0035,
+      "step": 10761
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.7544186046511629e-07,
+      "loss": 0.0016,
+      "step": 10762
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.7543023255813955e-07,
+      "loss": 0.0009,
+      "step": 10763
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.7541860465116278e-07,
+      "loss": 0.0015,
+      "step": 10764
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.7540697674418602e-07,
+      "loss": 0.0012,
+      "step": 10765
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.7539534883720928e-07,
+      "loss": 0.0015,
+      "step": 10766
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.7538372093023257e-07,
+      "loss": 0.004,
+      "step": 10767
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.753720930232558e-07,
+      "loss": 0.0009,
+      "step": 10768
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.7536046511627906e-07,
+      "loss": 0.0018,
+      "step": 10769
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.753488372093023e-07,
+      "loss": 0.0009,
+      "step": 10770
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.753372093023256e-07,
+      "loss": 0.0009,
+      "step": 10771
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.7532558139534882e-07,
+      "loss": 0.0015,
+      "step": 10772
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.7531395348837208e-07,
+      "loss": 0.0047,
+      "step": 10773
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.7530232558139532e-07,
+      "loss": 0.0013,
+      "step": 10774
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.752906976744186e-07,
+      "loss": 0.0012,
+      "step": 10775
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.7527906976744184e-07,
+      "loss": 0.001,
+      "step": 10776
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.752674418604651e-07,
+      "loss": 0.0011,
+      "step": 10777
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.7525581395348834e-07,
+      "loss": 0.0008,
+      "step": 10778
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.7524418604651163e-07,
+      "loss": 0.0012,
+      "step": 10779
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.752325581395349e-07,
+      "loss": 0.0009,
+      "step": 10780
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.7522093023255813e-07,
+      "loss": 0.0012,
+      "step": 10781
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.752093023255814e-07,
+      "loss": 0.0008,
+      "step": 10782
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.7519767441860465e-07,
+      "loss": 0.0011,
+      "step": 10783
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.751860465116279e-07,
+      "loss": 0.0017,
+      "step": 10784
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.7517441860465115e-07,
+      "loss": 0.001,
+      "step": 10785
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.751627906976744e-07,
+      "loss": 0.0009,
+      "step": 10786
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.7515116279069767e-07,
+      "loss": 0.001,
+      "step": 10787
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.7513953488372093e-07,
+      "loss": 0.0009,
+      "step": 10788
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.7512790697674417e-07,
+      "loss": 0.0013,
+      "step": 10789
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.7511627906976743e-07,
+      "loss": 0.0009,
+      "step": 10790
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.7510465116279072e-07,
+      "loss": 0.001,
+      "step": 10791
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.7509302325581395e-07,
+      "loss": 0.0011,
+      "step": 10792
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.750813953488372e-07,
+      "loss": 2.221,
+      "step": 10793
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.7506976744186045e-07,
+      "loss": 0.0009,
+      "step": 10794
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.750581395348837e-07,
+      "loss": 0.0007,
+      "step": 10795
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.7504651162790697e-07,
+      "loss": 0.0013,
+      "step": 10796
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.7503488372093023e-07,
+      "loss": 0.0014,
+      "step": 10797
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.7502325581395347e-07,
+      "loss": 0.0012,
+      "step": 10798
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.7501162790697673e-07,
+      "loss": 0.0011,
+      "step": 10799
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.75e-07,
+      "loss": 0.0009,
+      "step": 10800
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.7498837209302325e-07,
+      "loss": 0.001,
+      "step": 10801
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.749767441860465e-07,
+      "loss": 0.0009,
+      "step": 10802
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.7496511627906975e-07,
+      "loss": 0.0009,
+      "step": 10803
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.7495348837209304e-07,
+      "loss": 0.0009,
+      "step": 10804
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.7494186046511627e-07,
+      "loss": 0.0011,
+      "step": 10805
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.7493023255813954e-07,
+      "loss": 0.0013,
+      "step": 10806
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.7491860465116277e-07,
+      "loss": 0.0013,
+      "step": 10807
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.7490697674418606e-07,
+      "loss": 0.0009,
+      "step": 10808
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.748953488372093e-07,
+      "loss": 0.0009,
+      "step": 10809
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.7488372093023256e-07,
+      "loss": 0.0011,
+      "step": 10810
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.748720930232558e-07,
+      "loss": 0.0014,
+      "step": 10811
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.7486046511627908e-07,
+      "loss": 0.0008,
+      "step": 10812
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.7484883720930231e-07,
+      "loss": 0.0008,
+      "step": 10813
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.7483720930232558e-07,
+      "loss": 0.0021,
+      "step": 10814
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.748255813953488e-07,
+      "loss": 0.0053,
+      "step": 10815
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.748139534883721e-07,
+      "loss": 0.0012,
+      "step": 10816
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.7480232558139536e-07,
+      "loss": 0.0009,
+      "step": 10817
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.747906976744186e-07,
+      "loss": 0.0008,
+      "step": 10818
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.7477906976744183e-07,
+      "loss": 0.0012,
+      "step": 10819
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.747674418604651e-07,
+      "loss": 0.001,
+      "step": 10820
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.7475581395348838e-07,
+      "loss": 0.0013,
+      "step": 10821
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.7474418604651162e-07,
+      "loss": 0.0008,
+      "step": 10822
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.7473255813953488e-07,
+      "loss": 0.2882,
+      "step": 10823
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.7472093023255811e-07,
+      "loss": 0.0009,
+      "step": 10824
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.747093023255814e-07,
+      "loss": 0.0077,
+      "step": 10825
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.7469767441860464e-07,
+      "loss": 0.0007,
+      "step": 10826
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.746860465116279e-07,
+      "loss": 0.0191,
+      "step": 10827
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.7467441860465113e-07,
+      "loss": 0.0011,
+      "step": 10828
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.7466279069767442e-07,
+      "loss": 0.0011,
+      "step": 10829
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.7465116279069766e-07,
+      "loss": 0.001,
+      "step": 10830
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.7463953488372092e-07,
+      "loss": 0.0013,
+      "step": 10831
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.7462790697674416e-07,
+      "loss": 0.0008,
+      "step": 10832
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.7461627906976744e-07,
+      "loss": 0.001,
+      "step": 10833
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.746046511627907e-07,
+      "loss": 0.0009,
+      "step": 10834
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.7459302325581394e-07,
+      "loss": 0.0012,
+      "step": 10835
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.745813953488372e-07,
+      "loss": 1.8959,
+      "step": 10836
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.7456976744186046e-07,
+      "loss": 0.0012,
+      "step": 10837
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.7455813953488373e-07,
+      "loss": 0.0011,
+      "step": 10838
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.7454651162790696e-07,
+      "loss": 0.001,
+      "step": 10839
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.7453488372093022e-07,
+      "loss": 0.0018,
+      "step": 10840
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.7452325581395348e-07,
+      "loss": 0.0065,
+      "step": 10841
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.7451162790697675e-07,
+      "loss": 0.0013,
+      "step": 10842
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.7449999999999998e-07,
+      "loss": 0.0013,
+      "step": 10843
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.7448837209302324e-07,
+      "loss": 0.0057,
+      "step": 10844
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.7447674418604648e-07,
+      "loss": 0.001,
+      "step": 10845
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.7446511627906977e-07,
+      "loss": 0.001,
+      "step": 10846
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.7445348837209303e-07,
+      "loss": 0.0009,
+      "step": 10847
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.7444186046511626e-07,
+      "loss": 0.0007,
+      "step": 10848
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.7443023255813952e-07,
+      "loss": 0.0008,
+      "step": 10849
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.7441860465116279e-07,
+      "loss": 0.2885,
+      "step": 10850
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.7440697674418605e-07,
+      "loss": 0.0042,
+      "step": 10851
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.7439534883720928e-07,
+      "loss": 0.001,
+      "step": 10852
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.7438372093023255e-07,
+      "loss": 0.0011,
+      "step": 10853
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.743720930232558e-07,
+      "loss": 0.0011,
+      "step": 10854
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.7436046511627907e-07,
+      "loss": 0.0603,
+      "step": 10855
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.743488372093023e-07,
+      "loss": 0.0009,
+      "step": 10856
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.7433720930232557e-07,
+      "loss": 0.0007,
+      "step": 10857
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.9893617021276596,
+      "eval_loss": 0.1551298201084137,
+      "eval_runtime": 1.5582,
+      "eval_samples_per_second": 663.571,
+      "eval_steps_per_second": 10.91,
+      "step": 10857
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.7432558139534885e-07,
+      "loss": 0.0008,
+      "step": 10858
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.743139534883721e-07,
+      "loss": 0.001,
+      "step": 10859
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.7430232558139535e-07,
+      "loss": 0.0012,
+      "step": 10860
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.7429069767441859e-07,
+      "loss": 0.0009,
+      "step": 10861
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.7427906976744187e-07,
+      "loss": 0.0008,
+      "step": 10862
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.742674418604651e-07,
+      "loss": 0.0009,
+      "step": 10863
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.7425581395348837e-07,
+      "loss": 0.0013,
+      "step": 10864
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.742441860465116e-07,
+      "loss": 0.0008,
+      "step": 10865
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.742325581395349e-07,
+      "loss": 0.0011,
+      "step": 10866
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.7422093023255813e-07,
+      "loss": 0.0007,
+      "step": 10867
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.742093023255814e-07,
+      "loss": 0.0013,
+      "step": 10868
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.7419767441860463e-07,
+      "loss": 0.0007,
+      "step": 10869
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.7418604651162791e-07,
+      "loss": 0.0012,
+      "step": 10870
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.7417441860465115e-07,
+      "loss": 0.0069,
+      "step": 10871
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.741627906976744e-07,
+      "loss": 0.0009,
+      "step": 10872
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.7415116279069765e-07,
+      "loss": 0.0009,
+      "step": 10873
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.741395348837209e-07,
+      "loss": 0.0012,
+      "step": 10874
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.741279069767442e-07,
+      "loss": 0.0008,
+      "step": 10875
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.7411627906976743e-07,
+      "loss": 0.0012,
+      "step": 10876
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.741046511627907e-07,
+      "loss": 0.0007,
+      "step": 10877
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.7409302325581393e-07,
+      "loss": 0.0054,
+      "step": 10878
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.7408139534883722e-07,
+      "loss": 0.0008,
+      "step": 10879
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.7406976744186045e-07,
+      "loss": 0.0012,
+      "step": 10880
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.7405813953488371e-07,
+      "loss": 0.0009,
+      "step": 10881
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.7404651162790695e-07,
+      "loss": 0.001,
+      "step": 10882
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.7403488372093024e-07,
+      "loss": 0.0012,
+      "step": 10883
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.7402325581395347e-07,
+      "loss": 0.0011,
+      "step": 10884
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.7401162790697673e-07,
+      "loss": 0.0022,
+      "step": 10885
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.7399999999999997e-07,
+      "loss": 0.001,
+      "step": 10886
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.7398837209302326e-07,
+      "loss": 0.0012,
+      "step": 10887
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.7397674418604652e-07,
+      "loss": 0.0008,
+      "step": 10888
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.7396511627906976e-07,
+      "loss": 0.0014,
+      "step": 10889
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.7395348837209302e-07,
+      "loss": 0.001,
+      "step": 10890
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.7394186046511628e-07,
+      "loss": 0.0018,
+      "step": 10891
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.7393023255813954e-07,
+      "loss": 0.0015,
+      "step": 10892
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.7391860465116278e-07,
+      "loss": 0.001,
+      "step": 10893
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.7390697674418604e-07,
+      "loss": 0.0009,
+      "step": 10894
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.738953488372093e-07,
+      "loss": 0.001,
+      "step": 10895
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.7388372093023256e-07,
+      "loss": 0.0014,
+      "step": 10896
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.738720930232558e-07,
+      "loss": 0.0008,
+      "step": 10897
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.7386046511627906e-07,
+      "loss": 0.001,
+      "step": 10898
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.738488372093023e-07,
+      "loss": 0.0013,
+      "step": 10899
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.7383720930232558e-07,
+      "loss": 0.0013,
+      "step": 10900
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.7382558139534884e-07,
+      "loss": 0.0012,
+      "step": 10901
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.7381395348837208e-07,
+      "loss": 0.0014,
+      "step": 10902
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.738023255813953e-07,
+      "loss": 0.0009,
+      "step": 10903
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.737906976744186e-07,
+      "loss": 0.0011,
+      "step": 10904
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.7377906976744186e-07,
+      "loss": 0.001,
+      "step": 10905
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.737674418604651e-07,
+      "loss": 0.0043,
+      "step": 10906
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.7375581395348836e-07,
+      "loss": 0.0019,
+      "step": 10907
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.7374418604651162e-07,
+      "loss": 0.0007,
+      "step": 10908
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.7373255813953488e-07,
+      "loss": 0.0011,
+      "step": 10909
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.7372093023255812e-07,
+      "loss": 0.8156,
+      "step": 10910
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.7370930232558138e-07,
+      "loss": 0.0009,
+      "step": 10911
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.7369767441860467e-07,
+      "loss": 0.0009,
+      "step": 10912
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.736860465116279e-07,
+      "loss": 0.0011,
+      "step": 10913
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.7367441860465114e-07,
+      "loss": 0.0008,
+      "step": 10914
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.736627906976744e-07,
+      "loss": 0.0009,
+      "step": 10915
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.736511627906977e-07,
+      "loss": 0.0007,
+      "step": 10916
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.7363953488372092e-07,
+      "loss": 0.0016,
+      "step": 10917
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.7362790697674419e-07,
+      "loss": 0.0009,
+      "step": 10918
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.7361627906976742e-07,
+      "loss": 0.001,
+      "step": 10919
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.736046511627907e-07,
+      "loss": 0.001,
+      "step": 10920
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.7359302325581394e-07,
+      "loss": 0.0017,
+      "step": 10921
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.735813953488372e-07,
+      "loss": 0.0009,
+      "step": 10922
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.7356976744186044e-07,
+      "loss": 0.0011,
+      "step": 10923
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.7355813953488373e-07,
+      "loss": 0.0009,
+      "step": 10924
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.7354651162790696e-07,
+      "loss": 0.0007,
+      "step": 10925
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.7353488372093023e-07,
+      "loss": 0.0007,
+      "step": 10926
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.7352325581395346e-07,
+      "loss": 0.0009,
+      "step": 10927
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.7351162790697672e-07,
+      "loss": 0.0012,
+      "step": 10928
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.735e-07,
+      "loss": 0.001,
+      "step": 10929
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.7348837209302325e-07,
+      "loss": 0.0012,
+      "step": 10930
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.734767441860465e-07,
+      "loss": 2.0924,
+      "step": 10931
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.7346511627906974e-07,
+      "loss": 0.0011,
+      "step": 10932
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.7345348837209303e-07,
+      "loss": 0.0011,
+      "step": 10933
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.7344186046511627e-07,
+      "loss": 0.0011,
+      "step": 10934
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.7343023255813953e-07,
+      "loss": 0.0008,
+      "step": 10935
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.7341860465116276e-07,
+      "loss": 0.0011,
+      "step": 10936
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.7340697674418605e-07,
+      "loss": 0.001,
+      "step": 10937
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.733953488372093e-07,
+      "loss": 0.0008,
+      "step": 10938
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.7338372093023255e-07,
+      "loss": 0.0012,
+      "step": 10939
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.7337209302325578e-07,
+      "loss": 0.001,
+      "step": 10940
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.7336046511627907e-07,
+      "loss": 0.0014,
+      "step": 10941
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.7334883720930233e-07,
+      "loss": 0.0011,
+      "step": 10942
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.7333720930232557e-07,
+      "loss": 0.001,
+      "step": 10943
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.7332558139534883e-07,
+      "loss": 0.0011,
+      "step": 10944
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.733139534883721e-07,
+      "loss": 0.001,
+      "step": 10945
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.7330232558139535e-07,
+      "loss": 0.0012,
+      "step": 10946
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.732906976744186e-07,
+      "loss": 0.001,
+      "step": 10947
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.7327906976744185e-07,
+      "loss": 0.0013,
+      "step": 10948
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.7326744186046511e-07,
+      "loss": 0.1841,
+      "step": 10949
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.7325581395348838e-07,
+      "loss": 0.0009,
+      "step": 10950
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.732441860465116e-07,
+      "loss": 0.0011,
+      "step": 10951
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.7323255813953487e-07,
+      "loss": 0.001,
+      "step": 10952
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.732209302325581e-07,
+      "loss": 0.001,
+      "step": 10953
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.732093023255814e-07,
+      "loss": 0.0019,
+      "step": 10954
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.7319767441860466e-07,
+      "loss": 0.0009,
+      "step": 10955
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.731860465116279e-07,
+      "loss": 0.0699,
+      "step": 10956
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.7317441860465113e-07,
+      "loss": 0.0008,
+      "step": 10957
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.7316279069767442e-07,
+      "loss": 0.0008,
+      "step": 10958
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.7315116279069768e-07,
+      "loss": 0.7543,
+      "step": 10959
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.731395348837209e-07,
+      "loss": 0.0007,
+      "step": 10960
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.7312790697674417e-07,
+      "loss": 0.001,
+      "step": 10961
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.7311627906976744e-07,
+      "loss": 0.0008,
+      "step": 10962
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.731046511627907e-07,
+      "loss": 0.0008,
+      "step": 10963
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.7309302325581393e-07,
+      "loss": 0.3491,
+      "step": 10964
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.730813953488372e-07,
+      "loss": 0.0008,
+      "step": 10965
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.7306976744186048e-07,
+      "loss": 0.0008,
+      "step": 10966
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.7305813953488372e-07,
+      "loss": 0.0011,
+      "step": 10967
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.7304651162790695e-07,
+      "loss": 0.0008,
+      "step": 10968
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.7303488372093022e-07,
+      "loss": 0.001,
+      "step": 10969
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.730232558139535e-07,
+      "loss": 0.0008,
+      "step": 10970
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.7301162790697674e-07,
+      "loss": 0.0011,
+      "step": 10971
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.73e-07,
+      "loss": 0.0053,
+      "step": 10972
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.7298837209302324e-07,
+      "loss": 0.0026,
+      "step": 10973
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.7297674418604652e-07,
+      "loss": 0.0495,
+      "step": 10974
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.7296511627906976e-07,
+      "loss": 0.0007,
+      "step": 10975
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.7295348837209302e-07,
+      "loss": 0.0011,
+      "step": 10976
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.7294186046511626e-07,
+      "loss": 0.0008,
+      "step": 10977
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.7293023255813952e-07,
+      "loss": 0.0009,
+      "step": 10978
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.7291860465116278e-07,
+      "loss": 0.001,
+      "step": 10979
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.7290697674418604e-07,
+      "loss": 0.0013,
+      "step": 10980
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.7289534883720928e-07,
+      "loss": 1.0057,
+      "step": 10981
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.7288372093023254e-07,
+      "loss": 0.0008,
+      "step": 10982
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.7287209302325583e-07,
+      "loss": 0.2616,
+      "step": 10983
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.7286046511627906e-07,
+      "loss": 0.0021,
+      "step": 10984
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.7284883720930232e-07,
+      "loss": 0.001,
+      "step": 10985
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.7283720930232556e-07,
+      "loss": 0.0007,
+      "step": 10986
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.7282558139534885e-07,
+      "loss": 0.001,
+      "step": 10987
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.7281395348837208e-07,
+      "loss": 0.0013,
+      "step": 10988
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.7280232558139534e-07,
+      "loss": 0.0159,
+      "step": 10989
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.7279069767441858e-07,
+      "loss": 0.0016,
+      "step": 10990
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.7277906976744187e-07,
+      "loss": 0.001,
+      "step": 10991
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.727674418604651e-07,
+      "loss": 0.0012,
+      "step": 10992
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.7275581395348836e-07,
+      "loss": 0.001,
+      "step": 10993
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.727441860465116e-07,
+      "loss": 0.0011,
+      "step": 10994
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.727325581395349e-07,
+      "loss": 0.0045,
+      "step": 10995
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.7272093023255815e-07,
+      "loss": 0.0011,
+      "step": 10996
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.7270930232558138e-07,
+      "loss": 0.0014,
+      "step": 10997
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.7269767441860462e-07,
+      "loss": 0.001,
+      "step": 10998
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.726860465116279e-07,
+      "loss": 0.0014,
+      "step": 10999
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.7267441860465117e-07,
+      "loss": 0.0011,
+      "step": 11000
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.726627906976744e-07,
+      "loss": 0.0011,
+      "step": 11001
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.7265116279069767e-07,
+      "loss": 0.0011,
+      "step": 11002
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.7263953488372093e-07,
+      "loss": 0.0011,
+      "step": 11003
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.726279069767442e-07,
+      "loss": 0.001,
+      "step": 11004
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.7261627906976743e-07,
+      "loss": 0.0008,
+      "step": 11005
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.726046511627907e-07,
+      "loss": 0.003,
+      "step": 11006
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.7259302325581392e-07,
+      "loss": 0.0013,
+      "step": 11007
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.725813953488372e-07,
+      "loss": 0.0009,
+      "step": 11008
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.7256976744186045e-07,
+      "loss": 0.0008,
+      "step": 11009
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.725581395348837e-07,
+      "loss": 0.0009,
+      "step": 11010
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.7254651162790694e-07,
+      "loss": 0.0008,
+      "step": 11011
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.7253488372093023e-07,
+      "loss": 0.0008,
+      "step": 11012
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.725232558139535e-07,
+      "loss": 0.0013,
+      "step": 11013
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.7251162790697673e-07,
+      "loss": 0.0012,
+      "step": 11014
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.725e-07,
+      "loss": 1.9808,
+      "step": 11015
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.7248837209302325e-07,
+      "loss": 0.0009,
+      "step": 11016
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.724767441860465e-07,
+      "loss": 0.1654,
+      "step": 11017
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.7246511627906975e-07,
+      "loss": 0.0016,
+      "step": 11018
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.72453488372093e-07,
+      "loss": 0.0008,
+      "step": 11019
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.7244186046511627e-07,
+      "loss": 0.0013,
+      "step": 11020
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.7243023255813953e-07,
+      "loss": 0.0009,
+      "step": 11021
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.7241860465116277e-07,
+      "loss": 0.0008,
+      "step": 11022
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.7240697674418603e-07,
+      "loss": 0.4326,
+      "step": 11023
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.7239534883720932e-07,
+      "loss": 0.0006,
+      "step": 11024
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.7238372093023255e-07,
+      "loss": 0.0013,
+      "step": 11025
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.7237209302325582e-07,
+      "loss": 0.0017,
+      "step": 11026
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.7236046511627905e-07,
+      "loss": 0.0008,
+      "step": 11027
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.7234883720930234e-07,
+      "loss": 0.001,
+      "step": 11028
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.7233720930232557e-07,
+      "loss": 0.0012,
+      "step": 11029
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.7232558139534884e-07,
+      "loss": 0.0011,
+      "step": 11030
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.7231395348837207e-07,
+      "loss": 0.0192,
+      "step": 11031
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.7230232558139533e-07,
+      "loss": 0.0159,
+      "step": 11032
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.722906976744186e-07,
+      "loss": 0.0008,
+      "step": 11033
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.7227906976744186e-07,
+      "loss": 0.0009,
+      "step": 11034
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.722674418604651e-07,
+      "loss": 0.0011,
+      "step": 11035
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.7225581395348835e-07,
+      "loss": 0.0009,
+      "step": 11036
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.7224418604651164e-07,
+      "loss": 0.0008,
+      "step": 11037
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.7223255813953488e-07,
+      "loss": 0.0013,
+      "step": 11038
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.7222093023255814e-07,
+      "loss": 0.0028,
+      "step": 11039
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.7220930232558137e-07,
+      "loss": 2.3404,
+      "step": 11040
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.7219767441860466e-07,
+      "loss": 0.0009,
+      "step": 11041
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.721860465116279e-07,
+      "loss": 0.0011,
+      "step": 11042
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.7217441860465116e-07,
+      "loss": 0.001,
+      "step": 11043
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.721627906976744e-07,
+      "loss": 0.001,
+      "step": 11044
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.7215116279069768e-07,
+      "loss": 0.0009,
+      "step": 11045
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.7213953488372092e-07,
+      "loss": 0.0011,
+      "step": 11046
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.7212790697674418e-07,
+      "loss": 0.0012,
+      "step": 11047
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.7211627906976741e-07,
+      "loss": 0.001,
+      "step": 11048
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.721046511627907e-07,
+      "loss": 0.001,
+      "step": 11049
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.7209302325581396e-07,
+      "loss": 0.0018,
+      "step": 11050
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.720813953488372e-07,
+      "loss": 0.001,
+      "step": 11051
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.7206976744186043e-07,
+      "loss": 0.0014,
+      "step": 11052
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.7205813953488372e-07,
+      "loss": 0.2342,
+      "step": 11053
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.7204651162790698e-07,
+      "loss": 1.2928,
+      "step": 11054
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.7203488372093022e-07,
+      "loss": 0.0008,
+      "step": 11055
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.7202325581395348e-07,
+      "loss": 0.0015,
+      "step": 11056
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.7201162790697674e-07,
+      "loss": 0.0013,
+      "step": 11057
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.72e-07,
+      "loss": 0.0016,
+      "step": 11058
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.7198837209302324e-07,
+      "loss": 0.0009,
+      "step": 11059
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.719767441860465e-07,
+      "loss": 0.022,
+      "step": 11060
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.7196511627906974e-07,
+      "loss": 1.8767,
+      "step": 11061
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.7195348837209303e-07,
+      "loss": 0.0011,
+      "step": 11062
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.7194186046511626e-07,
+      "loss": 0.0008,
+      "step": 11063
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.7193023255813952e-07,
+      "loss": 0.0009,
+      "step": 11064
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.7191860465116276e-07,
+      "loss": 0.0017,
+      "step": 11065
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.7190697674418605e-07,
+      "loss": 0.0008,
+      "step": 11066
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.718953488372093e-07,
+      "loss": 0.001,
+      "step": 11067
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.7188372093023254e-07,
+      "loss": 0.0007,
+      "step": 11068
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.718720930232558e-07,
+      "loss": 0.0016,
+      "step": 11069
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.7186046511627907e-07,
+      "loss": 0.0012,
+      "step": 11070
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.7184883720930233e-07,
+      "loss": 0.001,
+      "step": 11071
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.7183720930232556e-07,
+      "loss": 0.001,
+      "step": 11072
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.7182558139534882e-07,
+      "loss": 0.001,
+      "step": 11073
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.7181395348837209e-07,
+      "loss": 0.0009,
+      "step": 11074
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.7180232558139535e-07,
+      "loss": 0.0007,
+      "step": 11075
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.7179069767441858e-07,
+      "loss": 0.0008,
+      "step": 11076
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.7177906976744185e-07,
+      "loss": 0.0009,
+      "step": 11077
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.7176744186046513e-07,
+      "loss": 0.0008,
+      "step": 11078
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.7175581395348837e-07,
+      "loss": 0.0008,
+      "step": 11079
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.7174418604651163e-07,
+      "loss": 0.0008,
+      "step": 11080
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.7173255813953487e-07,
+      "loss": 0.0008,
+      "step": 11081
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.7172093023255815e-07,
+      "loss": 0.0009,
+      "step": 11082
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.717093023255814e-07,
+      "loss": 0.0008,
+      "step": 11083
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.7169767441860465e-07,
+      "loss": 0.0007,
+      "step": 11084
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.7168604651162789e-07,
+      "loss": 0.001,
+      "step": 11085
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.7167441860465115e-07,
+      "loss": 0.0008,
+      "step": 11086
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.716627906976744e-07,
+      "loss": 0.001,
+      "step": 11087
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.7165116279069767e-07,
+      "loss": 0.0008,
+      "step": 11088
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.716395348837209e-07,
+      "loss": 0.0012,
+      "step": 11089
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.7162790697674417e-07,
+      "loss": 0.0016,
+      "step": 11090
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.7161627906976746e-07,
+      "loss": 0.001,
+      "step": 11091
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.716046511627907e-07,
+      "loss": 0.0011,
+      "step": 11092
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.7159302325581395e-07,
+      "loss": 0.0012,
+      "step": 11093
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.715813953488372e-07,
+      "loss": 0.0009,
+      "step": 11094
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.7156976744186048e-07,
+      "loss": 0.0012,
+      "step": 11095
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.715581395348837e-07,
+      "loss": 0.001,
+      "step": 11096
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.7154651162790697e-07,
+      "loss": 0.0009,
+      "step": 11097
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.715348837209302e-07,
+      "loss": 0.0014,
+      "step": 11098
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.715232558139535e-07,
+      "loss": 0.0016,
+      "step": 11099
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.7151162790697673e-07,
+      "loss": 0.0009,
+      "step": 11100
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.715e-07,
+      "loss": 0.0009,
+      "step": 11101
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.7148837209302323e-07,
+      "loss": 0.0014,
+      "step": 11102
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.7147674418604652e-07,
+      "loss": 0.0011,
+      "step": 11103
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.7146511627906978e-07,
+      "loss": 0.0008,
+      "step": 11104
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.7145348837209301e-07,
+      "loss": 0.0007,
+      "step": 11105
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.7144186046511625e-07,
+      "loss": 0.0009,
+      "step": 11106
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.7143023255813954e-07,
+      "loss": 0.001,
+      "step": 11107
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.714186046511628e-07,
+      "loss": 0.0011,
+      "step": 11108
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.7140697674418603e-07,
+      "loss": 0.001,
+      "step": 11109
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.713953488372093e-07,
+      "loss": 0.0008,
+      "step": 11110
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.7138372093023253e-07,
+      "loss": 0.0009,
+      "step": 11111
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.7137209302325582e-07,
+      "loss": 0.0011,
+      "step": 11112
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.7136046511627905e-07,
+      "loss": 0.001,
+      "step": 11113
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.7134883720930232e-07,
+      "loss": 0.0007,
+      "step": 11114
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.7133720930232555e-07,
+      "loss": 0.0009,
+      "step": 11115
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.7132558139534884e-07,
+      "loss": 0.001,
+      "step": 11116
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.7131395348837208e-07,
+      "loss": 0.001,
+      "step": 11117
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.7130232558139534e-07,
+      "loss": 0.0013,
+      "step": 11118
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.7129069767441857e-07,
+      "loss": 0.0012,
+      "step": 11119
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.7127906976744186e-07,
+      "loss": 0.001,
+      "step": 11120
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.7126744186046512e-07,
+      "loss": 0.0008,
+      "step": 11121
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.7125581395348836e-07,
+      "loss": 0.0008,
+      "step": 11122
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.7124418604651162e-07,
+      "loss": 0.0017,
+      "step": 11123
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.7123255813953488e-07,
+      "loss": 0.0006,
+      "step": 11124
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.7122093023255814e-07,
+      "loss": 0.001,
+      "step": 11125
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.7120930232558138e-07,
+      "loss": 0.0047,
+      "step": 11126
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.7119767441860464e-07,
+      "loss": 0.002,
+      "step": 11127
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.711860465116279e-07,
+      "loss": 0.0007,
+      "step": 11128
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.7117441860465116e-07,
+      "loss": 0.001,
+      "step": 11129
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.711627906976744e-07,
+      "loss": 0.001,
+      "step": 11130
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.7115116279069766e-07,
+      "loss": 0.0016,
+      "step": 11131
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.7113953488372095e-07,
+      "loss": 0.0011,
+      "step": 11132
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.7112790697674418e-07,
+      "loss": 0.001,
+      "step": 11133
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.7111627906976745e-07,
+      "loss": 0.001,
+      "step": 11134
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.7110465116279068e-07,
+      "loss": 0.0019,
+      "step": 11135
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.7109302325581397e-07,
+      "loss": 0.0011,
+      "step": 11136
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.710813953488372e-07,
+      "loss": 0.001,
+      "step": 11137
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.7106976744186047e-07,
+      "loss": 0.0012,
+      "step": 11138
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.710581395348837e-07,
+      "loss": 0.0011,
+      "step": 11139
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.7104651162790696e-07,
+      "loss": 0.4121,
+      "step": 11140
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.7103488372093022e-07,
+      "loss": 0.0009,
+      "step": 11141
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.7102325581395349e-07,
+      "loss": 0.001,
+      "step": 11142
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.7101162790697672e-07,
+      "loss": 0.0008,
+      "step": 11143
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.7099999999999998e-07,
+      "loss": 0.0009,
+      "step": 11144
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.7098837209302327e-07,
+      "loss": 0.0008,
+      "step": 11145
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.709767441860465e-07,
+      "loss": 0.0009,
+      "step": 11146
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.7096511627906974e-07,
+      "loss": 0.001,
+      "step": 11147
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.70953488372093e-07,
+      "loss": 0.0023,
+      "step": 11148
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.709418604651163e-07,
+      "loss": 0.0012,
+      "step": 11149
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.7093023255813953e-07,
+      "loss": 0.0013,
+      "step": 11150
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.709186046511628e-07,
+      "loss": 0.0014,
+      "step": 11151
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.7090697674418602e-07,
+      "loss": 0.0008,
+      "step": 11152
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.708953488372093e-07,
+      "loss": 0.0012,
+      "step": 11153
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.7088372093023255e-07,
+      "loss": 0.001,
+      "step": 11154
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.708720930232558e-07,
+      "loss": 0.0009,
+      "step": 11155
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.7086046511627904e-07,
+      "loss": 0.0009,
+      "step": 11156
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.7084883720930233e-07,
+      "loss": 0.0028,
+      "step": 11157
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.7083720930232557e-07,
+      "loss": 0.0012,
+      "step": 11158
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.7082558139534883e-07,
+      "loss": 0.0017,
+      "step": 11159
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.7081395348837206e-07,
+      "loss": 0.0011,
+      "step": 11160
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.7080232558139535e-07,
+      "loss": 0.0008,
+      "step": 11161
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.7079069767441861e-07,
+      "loss": 0.0008,
+      "step": 11162
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.7077906976744185e-07,
+      "loss": 1.293,
+      "step": 11163
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.707674418604651e-07,
+      "loss": 0.001,
+      "step": 11164
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.7075581395348835e-07,
+      "loss": 0.0013,
+      "step": 11165
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.7074418604651163e-07,
+      "loss": 0.0012,
+      "step": 11166
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.7073255813953487e-07,
+      "loss": 0.0008,
+      "step": 11167
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.7072093023255813e-07,
+      "loss": 0.0009,
+      "step": 11168
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.7070930232558137e-07,
+      "loss": 0.0009,
+      "step": 11169
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.7069767441860465e-07,
+      "loss": 0.0011,
+      "step": 11170
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.706860465116279e-07,
+      "loss": 0.0009,
+      "step": 11171
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.7067441860465115e-07,
+      "loss": 0.0011,
+      "step": 11172
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.706627906976744e-07,
+      "loss": 0.0012,
+      "step": 11173
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.7065116279069768e-07,
+      "loss": 0.0015,
+      "step": 11174
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.7063953488372094e-07,
+      "loss": 0.001,
+      "step": 11175
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.7062790697674417e-07,
+      "loss": 0.0008,
+      "step": 11176
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.7061627906976743e-07,
+      "loss": 0.001,
+      "step": 11177
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.706046511627907e-07,
+      "loss": 0.0008,
+      "step": 11178
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.7059302325581396e-07,
+      "loss": 0.0008,
+      "step": 11179
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.705813953488372e-07,
+      "loss": 0.0009,
+      "step": 11180
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.7056976744186045e-07,
+      "loss": 0.001,
+      "step": 11181
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.7055813953488372e-07,
+      "loss": 0.0009,
+      "step": 11182
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.7054651162790698e-07,
+      "loss": 0.0012,
+      "step": 11183
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.705348837209302e-07,
+      "loss": 0.0014,
+      "step": 11184
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.7052325581395347e-07,
+      "loss": 0.001,
+      "step": 11185
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.7051162790697676e-07,
+      "loss": 0.0007,
+      "step": 11186
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.705e-07,
+      "loss": 0.0007,
+      "step": 11187
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.7048837209302326e-07,
+      "loss": 0.0011,
+      "step": 11188
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.704767441860465e-07,
+      "loss": 0.001,
+      "step": 11189
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.7046511627906978e-07,
+      "loss": 0.0008,
+      "step": 11190
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.7045348837209302e-07,
+      "loss": 0.0011,
+      "step": 11191
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.7044186046511628e-07,
+      "loss": 0.0013,
+      "step": 11192
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.7043023255813952e-07,
+      "loss": 0.001,
+      "step": 11193
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.7041860465116278e-07,
+      "loss": 0.0008,
+      "step": 11194
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.7040697674418604e-07,
+      "loss": 0.0008,
+      "step": 11195
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.703953488372093e-07,
+      "loss": 0.0012,
+      "step": 11196
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.7038372093023254e-07,
+      "loss": 0.0008,
+      "step": 11197
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.703720930232558e-07,
+      "loss": 0.0052,
+      "step": 11198
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.7036046511627909e-07,
+      "loss": 0.0008,
+      "step": 11199
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.7034883720930232e-07,
+      "loss": 0.0011,
+      "step": 11200
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.7033720930232556e-07,
+      "loss": 0.0012,
+      "step": 11201
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.7032558139534882e-07,
+      "loss": 0.0011,
+      "step": 11202
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.703139534883721e-07,
+      "loss": 0.0012,
+      "step": 11203
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.7030232558139534e-07,
+      "loss": 0.0033,
+      "step": 11204
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.702906976744186e-07,
+      "loss": 0.0008,
+      "step": 11205
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.7027906976744184e-07,
+      "loss": 0.001,
+      "step": 11206
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.7026744186046513e-07,
+      "loss": 0.0009,
+      "step": 11207
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.7025581395348836e-07,
+      "loss": 0.4889,
+      "step": 11208
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.7024418604651162e-07,
+      "loss": 0.0007,
+      "step": 11209
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.7023255813953486e-07,
+      "loss": 0.0009,
+      "step": 11210
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.7022093023255815e-07,
+      "loss": 0.0017,
+      "step": 11211
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.7020930232558138e-07,
+      "loss": 0.0009,
+      "step": 11212
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.7019767441860464e-07,
+      "loss": 0.001,
+      "step": 11213
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.7018604651162788e-07,
+      "loss": 0.0009,
+      "step": 11214
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.7017441860465117e-07,
+      "loss": 0.0012,
+      "step": 11215
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.7016279069767443e-07,
+      "loss": 0.0008,
+      "step": 11216
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.7015116279069766e-07,
+      "loss": 0.0009,
+      "step": 11217
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.7013953488372093e-07,
+      "loss": 0.0008,
+      "step": 11218
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.7012790697674416e-07,
+      "loss": 0.0013,
+      "step": 11219
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.7011627906976745e-07,
+      "loss": 0.0018,
+      "step": 11220
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.7010465116279068e-07,
+      "loss": 0.0008,
+      "step": 11221
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.7009302325581395e-07,
+      "loss": 0.0006,
+      "step": 11222
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.7008139534883718e-07,
+      "loss": 0.0009,
+      "step": 11223
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.7006976744186047e-07,
+      "loss": 0.0007,
+      "step": 11224
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.700581395348837e-07,
+      "loss": 0.0008,
+      "step": 11225
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.7004651162790697e-07,
+      "loss": 0.0007,
+      "step": 11226
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.700348837209302e-07,
+      "loss": 0.0006,
+      "step": 11227
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.700232558139535e-07,
+      "loss": 0.0012,
+      "step": 11228
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.7001162790697675e-07,
+      "loss": 0.0008,
+      "step": 11229
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.7e-07,
+      "loss": 0.0009,
+      "step": 11230
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.6998837209302325e-07,
+      "loss": 0.0008,
+      "step": 11231
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.699767441860465e-07,
+      "loss": 2.0918,
+      "step": 11232
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.6996511627906977e-07,
+      "loss": 0.001,
+      "step": 11233
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.69953488372093e-07,
+      "loss": 0.002,
+      "step": 11234
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.6994186046511627e-07,
+      "loss": 1.9234,
+      "step": 11235
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.6993023255813953e-07,
+      "loss": 0.0012,
+      "step": 11236
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.699186046511628e-07,
+      "loss": 0.0008,
+      "step": 11237
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.6990697674418603e-07,
+      "loss": 0.001,
+      "step": 11238
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.698953488372093e-07,
+      "loss": 0.0011,
+      "step": 11239
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.6988372093023258e-07,
+      "loss": 0.0015,
+      "step": 11240
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.698720930232558e-07,
+      "loss": 0.0011,
+      "step": 11241
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.6986046511627907e-07,
+      "loss": 0.0014,
+      "step": 11242
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.698488372093023e-07,
+      "loss": 0.0474,
+      "step": 11243
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.698372093023256e-07,
+      "loss": 0.001,
+      "step": 11244
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.6982558139534883e-07,
+      "loss": 0.0008,
+      "step": 11245
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.698139534883721e-07,
+      "loss": 0.001,
+      "step": 11246
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.6980232558139533e-07,
+      "loss": 0.0295,
+      "step": 11247
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.697906976744186e-07,
+      "loss": 0.0008,
+      "step": 11248
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.6977906976744185e-07,
+      "loss": 0.0014,
+      "step": 11249
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.6976744186046512e-07,
+      "loss": 0.0011,
+      "step": 11250
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.6975581395348835e-07,
+      "loss": 0.0014,
+      "step": 11251
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.697441860465116e-07,
+      "loss": 0.001,
+      "step": 11252
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.6973255813953487e-07,
+      "loss": 0.001,
+      "step": 11253
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.6972093023255814e-07,
+      "loss": 0.0008,
+      "step": 11254
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.6970930232558137e-07,
+      "loss": 0.0012,
+      "step": 11255
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.6969767441860463e-07,
+      "loss": 0.0009,
+      "step": 11256
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.6968604651162792e-07,
+      "loss": 0.0009,
+      "step": 11257
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.6967441860465116e-07,
+      "loss": 0.0152,
+      "step": 11258
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.6966279069767442e-07,
+      "loss": 0.001,
+      "step": 11259
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.6965116279069765e-07,
+      "loss": 0.0011,
+      "step": 11260
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.6963953488372094e-07,
+      "loss": 0.001,
+      "step": 11261
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.6962790697674418e-07,
+      "loss": 0.0009,
+      "step": 11262
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.6961627906976744e-07,
+      "loss": 0.0523,
+      "step": 11263
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.6960465116279067e-07,
+      "loss": 0.0007,
+      "step": 11264
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.6959302325581396e-07,
+      "loss": 3.6064,
+      "step": 11265
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.695813953488372e-07,
+      "loss": 0.0007,
+      "step": 11266
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.6956976744186046e-07,
+      "loss": 0.0008,
+      "step": 11267
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.695581395348837e-07,
+      "loss": 0.0007,
+      "step": 11268
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.6954651162790698e-07,
+      "loss": 0.0019,
+      "step": 11269
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.6953488372093024e-07,
+      "loss": 0.0008,
+      "step": 11270
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.6952325581395348e-07,
+      "loss": 0.0113,
+      "step": 11271
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.6951162790697674e-07,
+      "loss": 0.0009,
+      "step": 11272
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.6949999999999998e-07,
+      "loss": 0.0009,
+      "step": 11273
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.6948837209302326e-07,
+      "loss": 0.0023,
+      "step": 11274
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.694767441860465e-07,
+      "loss": 0.0008,
+      "step": 11275
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.6946511627906976e-07,
+      "loss": 0.0009,
+      "step": 11276
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.69453488372093e-07,
+      "loss": 0.0013,
+      "step": 11277
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.6944186046511628e-07,
+      "loss": 0.0007,
+      "step": 11278
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.6943023255813952e-07,
+      "loss": 0.0047,
+      "step": 11279
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.6941860465116278e-07,
+      "loss": 0.0022,
+      "step": 11280
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.6940697674418602e-07,
+      "loss": 0.0017,
+      "step": 11281
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.693953488372093e-07,
+      "loss": 0.0013,
+      "step": 11282
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.6938372093023257e-07,
+      "loss": 0.0009,
+      "step": 11283
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.693720930232558e-07,
+      "loss": 0.0008,
+      "step": 11284
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.6936046511627904e-07,
+      "loss": 0.0011,
+      "step": 11285
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.6934883720930233e-07,
+      "loss": 0.0009,
+      "step": 11286
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.693372093023256e-07,
+      "loss": 0.0009,
+      "step": 11287
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.6932558139534882e-07,
+      "loss": 0.0013,
+      "step": 11288
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.6931395348837208e-07,
+      "loss": 0.0009,
+      "step": 11289
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.6930232558139535e-07,
+      "loss": 2.4098,
+      "step": 11290
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.692906976744186e-07,
+      "loss": 0.0012,
+      "step": 11291
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.6927906976744184e-07,
+      "loss": 0.0016,
+      "step": 11292
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.692674418604651e-07,
+      "loss": 2.9305,
+      "step": 11293
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.692558139534884e-07,
+      "loss": 0.0011,
+      "step": 11294
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.6924418604651163e-07,
+      "loss": 0.0013,
+      "step": 11295
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.6923255813953486e-07,
+      "loss": 0.0017,
+      "step": 11296
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.6922093023255812e-07,
+      "loss": 0.001,
+      "step": 11297
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.6920930232558136e-07,
+      "loss": 0.0011,
+      "step": 11298
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.6919767441860465e-07,
+      "loss": 0.0007,
+      "step": 11299
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.691860465116279e-07,
+      "loss": 0.001,
+      "step": 11300
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.6917441860465115e-07,
+      "loss": 0.0009,
+      "step": 11301
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.691627906976744e-07,
+      "loss": 0.0008,
+      "step": 11302
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.6915116279069767e-07,
+      "loss": 0.0012,
+      "step": 11303
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.6913953488372093e-07,
+      "loss": 0.0008,
+      "step": 11304
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.6912790697674417e-07,
+      "loss": 0.0008,
+      "step": 11305
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.6911627906976743e-07,
+      "loss": 0.0014,
+      "step": 11306
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.691046511627907e-07,
+      "loss": 0.0008,
+      "step": 11307
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.6909302325581395e-07,
+      "loss": 0.1726,
+      "step": 11308
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.6908139534883719e-07,
+      "loss": 0.0015,
+      "step": 11309
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.6906976744186045e-07,
+      "loss": 0.0014,
+      "step": 11310
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.6905813953488374e-07,
+      "loss": 0.001,
+      "step": 11311
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.6904651162790697e-07,
+      "loss": 0.0429,
+      "step": 11312
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.6903488372093023e-07,
+      "loss": 0.0008,
+      "step": 11313
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.6902325581395347e-07,
+      "loss": 0.0008,
+      "step": 11314
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.6901162790697676e-07,
+      "loss": 0.0011,
+      "step": 11315
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.69e-07,
+      "loss": 0.0006,
+      "step": 11316
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.6898837209302325e-07,
+      "loss": 0.0008,
+      "step": 11317
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.689767441860465e-07,
+      "loss": 0.1489,
+      "step": 11318
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.6896511627906978e-07,
+      "loss": 0.0015,
+      "step": 11319
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.68953488372093e-07,
+      "loss": 0.0007,
+      "step": 11320
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.6894186046511627e-07,
+      "loss": 0.0008,
+      "step": 11321
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.689302325581395e-07,
+      "loss": 0.0008,
+      "step": 11322
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.689186046511628e-07,
+      "loss": 0.0016,
+      "step": 11323
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.6890697674418606e-07,
+      "loss": 0.0015,
+      "step": 11324
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.688953488372093e-07,
+      "loss": 0.0011,
+      "step": 11325
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.6888372093023256e-07,
+      "loss": 0.001,
+      "step": 11326
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.688720930232558e-07,
+      "loss": 0.0008,
+      "step": 11327
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.6886046511627908e-07,
+      "loss": 0.0011,
+      "step": 11328
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.6884883720930231e-07,
+      "loss": 0.0018,
+      "step": 11329
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.6883720930232558e-07,
+      "loss": 0.0009,
+      "step": 11330
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.688255813953488e-07,
+      "loss": 0.0008,
+      "step": 11331
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.688139534883721e-07,
+      "loss": 0.0012,
+      "step": 11332
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.6880232558139533e-07,
+      "loss": 0.0256,
+      "step": 11333
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.687906976744186e-07,
+      "loss": 1.9724,
+      "step": 11334
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.6877906976744183e-07,
+      "loss": 0.001,
+      "step": 11335
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.6876744186046512e-07,
+      "loss": 0.0011,
+      "step": 11336
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.6875581395348838e-07,
+      "loss": 1.6284,
+      "step": 11337
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.6874418604651162e-07,
+      "loss": 0.0013,
+      "step": 11338
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.6873255813953485e-07,
+      "loss": 0.0009,
+      "step": 11339
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.6872093023255814e-07,
+      "loss": 0.0007,
+      "step": 11340
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.687093023255814e-07,
+      "loss": 0.001,
+      "step": 11341
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.6869767441860464e-07,
+      "loss": 0.0007,
+      "step": 11342
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.686860465116279e-07,
+      "loss": 0.0012,
+      "step": 11343
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.6867441860465116e-07,
+      "loss": 0.2755,
+      "step": 11344
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.6866279069767442e-07,
+      "loss": 0.001,
+      "step": 11345
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.6865116279069766e-07,
+      "loss": 0.0009,
+      "step": 11346
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.6863953488372092e-07,
+      "loss": 0.0008,
+      "step": 11347
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.686279069767442e-07,
+      "loss": 0.0372,
+      "step": 11348
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.6861627906976744e-07,
+      "loss": 0.0011,
+      "step": 11349
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.6860465116279068e-07,
+      "loss": 0.0015,
+      "step": 11350
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.6859302325581394e-07,
+      "loss": 0.0022,
+      "step": 11351
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.6858139534883717e-07,
+      "loss": 0.0008,
+      "step": 11352
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.6856976744186046e-07,
+      "loss": 0.0009,
+      "step": 11353
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.6855813953488372e-07,
+      "loss": 1.7826,
+      "step": 11354
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.6854651162790696e-07,
+      "loss": 0.2136,
+      "step": 11355
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.6853488372093022e-07,
+      "loss": 0.0011,
+      "step": 11356
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.6852325581395348e-07,
+      "loss": 0.0013,
+      "step": 11357
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.6851162790697674e-07,
+      "loss": 0.0011,
+      "step": 11358
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.6849999999999998e-07,
+      "loss": 0.0007,
+      "step": 11359
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.6848837209302324e-07,
+      "loss": 1.6663,
+      "step": 11360
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.684767441860465e-07,
+      "loss": 0.0008,
+      "step": 11361
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.6846511627906977e-07,
+      "loss": 0.0009,
+      "step": 11362
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.68453488372093e-07,
+      "loss": 0.002,
+      "step": 11363
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.6844186046511626e-07,
+      "loss": 0.0009,
+      "step": 11364
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.6843023255813955e-07,
+      "loss": 0.0048,
+      "step": 11365
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.6841860465116279e-07,
+      "loss": 0.0009,
+      "step": 11366
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.6840697674418605e-07,
+      "loss": 0.0007,
+      "step": 11367
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.6839534883720928e-07,
+      "loss": 0.001,
+      "step": 11368
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.6838372093023257e-07,
+      "loss": 0.001,
+      "step": 11369
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.683720930232558e-07,
+      "loss": 0.001,
+      "step": 11370
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.6836046511627907e-07,
+      "loss": 0.0011,
+      "step": 11371
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.683488372093023e-07,
+      "loss": 0.0017,
+      "step": 11372
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.683372093023256e-07,
+      "loss": 0.0008,
+      "step": 11373
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.6832558139534883e-07,
+      "loss": 0.0009,
+      "step": 11374
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.14615371823310852,
+      "eval_runtime": 1.5568,
+      "eval_samples_per_second": 664.178,
+      "eval_steps_per_second": 10.92,
+      "step": 11374
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.683139534883721e-07,
+      "loss": 0.0011,
+      "step": 11375
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.6830232558139532e-07,
+      "loss": 0.0009,
+      "step": 11376
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.682906976744186e-07,
+      "loss": 0.001,
+      "step": 11377
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.6827906976744187e-07,
+      "loss": 0.7288,
+      "step": 11378
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.682674418604651e-07,
+      "loss": 0.001,
+      "step": 11379
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.6825581395348834e-07,
+      "loss": 0.0012,
+      "step": 11380
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.682441860465116e-07,
+      "loss": 0.0008,
+      "step": 11381
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.682325581395349e-07,
+      "loss": 0.001,
+      "step": 11382
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.6822093023255813e-07,
+      "loss": 0.0031,
+      "step": 11383
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.682093023255814e-07,
+      "loss": 0.001,
+      "step": 11384
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.6819767441860463e-07,
+      "loss": 0.0007,
+      "step": 11385
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.6818604651162791e-07,
+      "loss": 0.0008,
+      "step": 11386
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.6817441860465115e-07,
+      "loss": 0.001,
+      "step": 11387
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.681627906976744e-07,
+      "loss": 0.0009,
+      "step": 11388
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.6815116279069765e-07,
+      "loss": 0.0008,
+      "step": 11389
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.6813953488372093e-07,
+      "loss": 0.001,
+      "step": 11390
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.6812790697674417e-07,
+      "loss": 0.0005,
+      "step": 11391
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.6811627906976743e-07,
+      "loss": 0.2886,
+      "step": 11392
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.6810465116279067e-07,
+      "loss": 0.0014,
+      "step": 11393
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.6809302325581395e-07,
+      "loss": 0.0011,
+      "step": 11394
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.6808139534883722e-07,
+      "loss": 0.0016,
+      "step": 11395
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.6806976744186045e-07,
+      "loss": 0.0008,
+      "step": 11396
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.6805813953488371e-07,
+      "loss": 0.0008,
+      "step": 11397
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.6804651162790698e-07,
+      "loss": 0.0009,
+      "step": 11398
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.6803488372093024e-07,
+      "loss": 0.001,
+      "step": 11399
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.6802325581395347e-07,
+      "loss": 0.0009,
+      "step": 11400
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.6801162790697673e-07,
+      "loss": 0.0016,
+      "step": 11401
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.68e-07,
+      "loss": 0.0015,
+      "step": 11402
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.6798837209302326e-07,
+      "loss": 0.0011,
+      "step": 11403
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.679767441860465e-07,
+      "loss": 0.0009,
+      "step": 11404
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.6796511627906975e-07,
+      "loss": 0.0006,
+      "step": 11405
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.67953488372093e-07,
+      "loss": 0.0012,
+      "step": 11406
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.6794186046511628e-07,
+      "loss": 0.0008,
+      "step": 11407
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.6793023255813954e-07,
+      "loss": 0.0008,
+      "step": 11408
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.6791860465116277e-07,
+      "loss": 0.0009,
+      "step": 11409
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.6790697674418604e-07,
+      "loss": 0.0011,
+      "step": 11410
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.678953488372093e-07,
+      "loss": 0.0008,
+      "step": 11411
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.6788372093023256e-07,
+      "loss": 0.0014,
+      "step": 11412
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.678720930232558e-07,
+      "loss": 0.0009,
+      "step": 11413
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.6786046511627906e-07,
+      "loss": 0.0008,
+      "step": 11414
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.6784883720930232e-07,
+      "loss": 0.0007,
+      "step": 11415
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.6783720930232558e-07,
+      "loss": 0.001,
+      "step": 11416
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.6782558139534882e-07,
+      "loss": 0.0009,
+      "step": 11417
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.6781395348837208e-07,
+      "loss": 0.1759,
+      "step": 11418
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.6780232558139537e-07,
+      "loss": 0.0007,
+      "step": 11419
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.677906976744186e-07,
+      "loss": 0.0009,
+      "step": 11420
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.6777906976744186e-07,
+      "loss": 0.0024,
+      "step": 11421
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.677674418604651e-07,
+      "loss": 0.0009,
+      "step": 11422
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.6775581395348839e-07,
+      "loss": 0.0009,
+      "step": 11423
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.6774418604651162e-07,
+      "loss": 0.002,
+      "step": 11424
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.6773255813953488e-07,
+      "loss": 0.0008,
+      "step": 11425
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.6772093023255812e-07,
+      "loss": 0.0011,
+      "step": 11426
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.677093023255814e-07,
+      "loss": 0.0125,
+      "step": 11427
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.6769767441860464e-07,
+      "loss": 0.001,
+      "step": 11428
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.676860465116279e-07,
+      "loss": 0.0008,
+      "step": 11429
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.6767441860465114e-07,
+      "loss": 0.0016,
+      "step": 11430
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.676627906976744e-07,
+      "loss": 0.0009,
+      "step": 11431
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.676511627906977e-07,
+      "loss": 0.001,
+      "step": 11432
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.6763953488372092e-07,
+      "loss": 1.6141,
+      "step": 11433
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.6762790697674416e-07,
+      "loss": 0.0029,
+      "step": 11434
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.6761627906976742e-07,
+      "loss": 0.0035,
+      "step": 11435
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.676046511627907e-07,
+      "loss": 0.0007,
+      "step": 11436
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.6759302325581394e-07,
+      "loss": 0.0011,
+      "step": 11437
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.675813953488372e-07,
+      "loss": 0.0036,
+      "step": 11438
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.6756976744186044e-07,
+      "loss": 0.0009,
+      "step": 11439
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.6755813953488373e-07,
+      "loss": 0.0008,
+      "step": 11440
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.6754651162790696e-07,
+      "loss": 0.0009,
+      "step": 11441
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.6753488372093023e-07,
+      "loss": 0.001,
+      "step": 11442
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.6752325581395346e-07,
+      "loss": 0.0008,
+      "step": 11443
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.6751162790697675e-07,
+      "loss": 0.0008,
+      "step": 11444
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.6749999999999998e-07,
+      "loss": 0.0009,
+      "step": 11445
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.6748837209302325e-07,
+      "loss": 0.0008,
+      "step": 11446
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.6747674418604648e-07,
+      "loss": 0.0009,
+      "step": 11447
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.6746511627906977e-07,
+      "loss": 0.0011,
+      "step": 11448
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.6745348837209303e-07,
+      "loss": 0.0011,
+      "step": 11449
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.6744186046511627e-07,
+      "loss": 0.0008,
+      "step": 11450
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.6743023255813953e-07,
+      "loss": 0.001,
+      "step": 11451
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.674186046511628e-07,
+      "loss": 0.0844,
+      "step": 11452
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.6740697674418605e-07,
+      "loss": 0.0012,
+      "step": 11453
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.673953488372093e-07,
+      "loss": 0.0017,
+      "step": 11454
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.6738372093023255e-07,
+      "loss": 0.0007,
+      "step": 11455
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.673720930232558e-07,
+      "loss": 0.0008,
+      "step": 11456
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.6736046511627907e-07,
+      "loss": 0.0008,
+      "step": 11457
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.673488372093023e-07,
+      "loss": 0.0009,
+      "step": 11458
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.6733720930232557e-07,
+      "loss": 0.0015,
+      "step": 11459
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.673255813953488e-07,
+      "loss": 0.0011,
+      "step": 11460
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.673139534883721e-07,
+      "loss": 1.9585,
+      "step": 11461
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.6730232558139535e-07,
+      "loss": 2.606,
+      "step": 11462
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.672906976744186e-07,
+      "loss": 0.4626,
+      "step": 11463
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.6727906976744185e-07,
+      "loss": 2.2852,
+      "step": 11464
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.672674418604651e-07,
+      "loss": 0.0007,
+      "step": 11465
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.6725581395348837e-07,
+      "loss": 0.0009,
+      "step": 11466
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.672441860465116e-07,
+      "loss": 0.0311,
+      "step": 11467
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.6723255813953487e-07,
+      "loss": 0.001,
+      "step": 11468
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.6722093023255813e-07,
+      "loss": 0.0022,
+      "step": 11469
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.672093023255814e-07,
+      "loss": 0.0008,
+      "step": 11470
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.6719767441860463e-07,
+      "loss": 0.001,
+      "step": 11471
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.671860465116279e-07,
+      "loss": 0.0008,
+      "step": 11472
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.6717441860465118e-07,
+      "loss": 0.0009,
+      "step": 11473
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.6716279069767442e-07,
+      "loss": 0.0011,
+      "step": 11474
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.6715116279069768e-07,
+      "loss": 0.0009,
+      "step": 11475
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.671395348837209e-07,
+      "loss": 0.002,
+      "step": 11476
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.671279069767442e-07,
+      "loss": 0.0008,
+      "step": 11477
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.6711627906976744e-07,
+      "loss": 0.0008,
+      "step": 11478
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.671046511627907e-07,
+      "loss": 0.0011,
+      "step": 11479
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.6709302325581393e-07,
+      "loss": 0.001,
+      "step": 11480
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.6708139534883722e-07,
+      "loss": 0.2273,
+      "step": 11481
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.6706976744186046e-07,
+      "loss": 0.0013,
+      "step": 11482
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.6705813953488372e-07,
+      "loss": 0.0013,
+      "step": 11483
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.6704651162790695e-07,
+      "loss": 0.0011,
+      "step": 11484
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.6703488372093021e-07,
+      "loss": 0.0008,
+      "step": 11485
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.670232558139535e-07,
+      "loss": 0.344,
+      "step": 11486
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.6701162790697674e-07,
+      "loss": 0.001,
+      "step": 11487
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.6699999999999997e-07,
+      "loss": 0.0012,
+      "step": 11488
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.6698837209302324e-07,
+      "loss": 0.001,
+      "step": 11489
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.6697674418604652e-07,
+      "loss": 0.0008,
+      "step": 11490
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.6696511627906976e-07,
+      "loss": 0.001,
+      "step": 11491
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.6695348837209302e-07,
+      "loss": 0.0011,
+      "step": 11492
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.6694186046511626e-07,
+      "loss": 0.0009,
+      "step": 11493
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.6693023255813954e-07,
+      "loss": 0.001,
+      "step": 11494
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.6691860465116278e-07,
+      "loss": 0.2092,
+      "step": 11495
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.6690697674418604e-07,
+      "loss": 0.001,
+      "step": 11496
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.6689534883720928e-07,
+      "loss": 2.2283,
+      "step": 11497
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.6688372093023256e-07,
+      "loss": 0.0014,
+      "step": 11498
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.668720930232558e-07,
+      "loss": 0.0009,
+      "step": 11499
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.6686046511627906e-07,
+      "loss": 0.0007,
+      "step": 11500
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.668488372093023e-07,
+      "loss": 0.0009,
+      "step": 11501
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.6683720930232558e-07,
+      "loss": 0.0011,
+      "step": 11502
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.6682558139534885e-07,
+      "loss": 0.0015,
+      "step": 11503
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.6681395348837208e-07,
+      "loss": 0.0008,
+      "step": 11504
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.6680232558139534e-07,
+      "loss": 0.0009,
+      "step": 11505
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.667906976744186e-07,
+      "loss": 0.0012,
+      "step": 11506
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.6677906976744187e-07,
+      "loss": 1.2458,
+      "step": 11507
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.667674418604651e-07,
+      "loss": 0.0083,
+      "step": 11508
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.6675581395348836e-07,
+      "loss": 0.0015,
+      "step": 11509
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.6674418604651163e-07,
+      "loss": 0.0009,
+      "step": 11510
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.667325581395349e-07,
+      "loss": 0.0009,
+      "step": 11511
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.6672093023255812e-07,
+      "loss": 0.0009,
+      "step": 11512
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.6670930232558138e-07,
+      "loss": 0.001,
+      "step": 11513
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.6669767441860462e-07,
+      "loss": 0.0085,
+      "step": 11514
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.666860465116279e-07,
+      "loss": 0.0302,
+      "step": 11515
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.6667441860465117e-07,
+      "loss": 0.0008,
+      "step": 11516
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.666627906976744e-07,
+      "loss": 0.0014,
+      "step": 11517
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.6665116279069764e-07,
+      "loss": 0.0008,
+      "step": 11518
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.6663953488372093e-07,
+      "loss": 0.001,
+      "step": 11519
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.666279069767442e-07,
+      "loss": 0.0008,
+      "step": 11520
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.6661627906976742e-07,
+      "loss": 0.0008,
+      "step": 11521
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.6660465116279069e-07,
+      "loss": 0.0008,
+      "step": 11522
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.6659302325581395e-07,
+      "loss": 0.0009,
+      "step": 11523
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.665813953488372e-07,
+      "loss": 0.001,
+      "step": 11524
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.6656976744186044e-07,
+      "loss": 0.0008,
+      "step": 11525
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.665581395348837e-07,
+      "loss": 0.0008,
+      "step": 11526
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.66546511627907e-07,
+      "loss": 0.0016,
+      "step": 11527
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.6653488372093023e-07,
+      "loss": 0.0012,
+      "step": 11528
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.6652325581395347e-07,
+      "loss": 0.001,
+      "step": 11529
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.6651162790697673e-07,
+      "loss": 0.001,
+      "step": 11530
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.6650000000000002e-07,
+      "loss": 0.0011,
+      "step": 11531
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.6648837209302325e-07,
+      "loss": 0.0011,
+      "step": 11532
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.664767441860465e-07,
+      "loss": 0.0007,
+      "step": 11533
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.6646511627906975e-07,
+      "loss": 0.0006,
+      "step": 11534
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.6645348837209304e-07,
+      "loss": 0.0019,
+      "step": 11535
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.6644186046511627e-07,
+      "loss": 0.001,
+      "step": 11536
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.6643023255813953e-07,
+      "loss": 2.9496,
+      "step": 11537
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.6641860465116277e-07,
+      "loss": 0.1892,
+      "step": 11538
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.6640697674418603e-07,
+      "loss": 0.0018,
+      "step": 11539
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.663953488372093e-07,
+      "loss": 0.0012,
+      "step": 11540
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.6638372093023255e-07,
+      "loss": 0.6717,
+      "step": 11541
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.663720930232558e-07,
+      "loss": 0.0008,
+      "step": 11542
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.6636046511627905e-07,
+      "loss": 0.0008,
+      "step": 11543
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.6634883720930234e-07,
+      "loss": 0.0014,
+      "step": 11544
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.6633720930232557e-07,
+      "loss": 0.0009,
+      "step": 11545
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.6632558139534884e-07,
+      "loss": 0.0009,
+      "step": 11546
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.6631395348837207e-07,
+      "loss": 0.0012,
+      "step": 11547
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.6630232558139536e-07,
+      "loss": 0.001,
+      "step": 11548
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.662906976744186e-07,
+      "loss": 0.0011,
+      "step": 11549
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.6627906976744186e-07,
+      "loss": 0.0011,
+      "step": 11550
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.662674418604651e-07,
+      "loss": 0.0006,
+      "step": 11551
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.6625581395348838e-07,
+      "loss": 0.0007,
+      "step": 11552
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.6624418604651161e-07,
+      "loss": 0.0012,
+      "step": 11553
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.6623255813953488e-07,
+      "loss": 0.0009,
+      "step": 11554
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.662209302325581e-07,
+      "loss": 0.3701,
+      "step": 11555
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.662093023255814e-07,
+      "loss": 0.0009,
+      "step": 11556
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.6619767441860466e-07,
+      "loss": 0.0011,
+      "step": 11557
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.661860465116279e-07,
+      "loss": 0.0011,
+      "step": 11558
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.6617441860465116e-07,
+      "loss": 0.001,
+      "step": 11559
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.6616279069767442e-07,
+      "loss": 0.0037,
+      "step": 11560
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.6615116279069768e-07,
+      "loss": 0.0011,
+      "step": 11561
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.6613953488372092e-07,
+      "loss": 0.0014,
+      "step": 11562
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.6612790697674418e-07,
+      "loss": 0.0011,
+      "step": 11563
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.6611627906976744e-07,
+      "loss": 0.0008,
+      "step": 11564
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.661046511627907e-07,
+      "loss": 0.0008,
+      "step": 11565
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.6609302325581394e-07,
+      "loss": 0.0007,
+      "step": 11566
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.660813953488372e-07,
+      "loss": 0.0014,
+      "step": 11567
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.6606976744186043e-07,
+      "loss": 0.0009,
+      "step": 11568
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.6605813953488372e-07,
+      "loss": 0.0013,
+      "step": 11569
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.6604651162790698e-07,
+      "loss": 0.0007,
+      "step": 11570
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.6603488372093022e-07,
+      "loss": 0.0008,
+      "step": 11571
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.6602325581395345e-07,
+      "loss": 0.001,
+      "step": 11572
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.6601162790697674e-07,
+      "loss": 0.0751,
+      "step": 11573
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.66e-07,
+      "loss": 0.0015,
+      "step": 11574
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.6598837209302324e-07,
+      "loss": 0.0015,
+      "step": 11575
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.659767441860465e-07,
+      "loss": 0.0011,
+      "step": 11576
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.6596511627906976e-07,
+      "loss": 0.0007,
+      "step": 11577
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.6595348837209302e-07,
+      "loss": 0.0009,
+      "step": 11578
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.6594186046511626e-07,
+      "loss": 0.0009,
+      "step": 11579
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.6593023255813952e-07,
+      "loss": 0.0012,
+      "step": 11580
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.659186046511628e-07,
+      "loss": 0.0078,
+      "step": 11581
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.6590697674418604e-07,
+      "loss": 0.001,
+      "step": 11582
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.6589534883720928e-07,
+      "loss": 0.0006,
+      "step": 11583
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.6588372093023254e-07,
+      "loss": 0.0008,
+      "step": 11584
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.6587209302325583e-07,
+      "loss": 0.0007,
+      "step": 11585
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.6586046511627907e-07,
+      "loss": 0.0027,
+      "step": 11586
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.6584883720930233e-07,
+      "loss": 0.001,
+      "step": 11587
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.6583720930232556e-07,
+      "loss": 0.0008,
+      "step": 11588
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.6582558139534885e-07,
+      "loss": 0.0014,
+      "step": 11589
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.6581395348837209e-07,
+      "loss": 0.001,
+      "step": 11590
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.6580232558139535e-07,
+      "loss": 0.0009,
+      "step": 11591
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.6579069767441858e-07,
+      "loss": 0.0008,
+      "step": 11592
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.6577906976744184e-07,
+      "loss": 0.0011,
+      "step": 11593
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.657674418604651e-07,
+      "loss": 0.0016,
+      "step": 11594
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.6575581395348837e-07,
+      "loss": 0.0009,
+      "step": 11595
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.657441860465116e-07,
+      "loss": 0.0012,
+      "step": 11596
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.6573255813953486e-07,
+      "loss": 0.0008,
+      "step": 11597
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.6572093023255815e-07,
+      "loss": 0.0124,
+      "step": 11598
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.657093023255814e-07,
+      "loss": 0.0009,
+      "step": 11599
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.6569767441860465e-07,
+      "loss": 0.0143,
+      "step": 11600
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.6568604651162789e-07,
+      "loss": 0.0008,
+      "step": 11601
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.6567441860465117e-07,
+      "loss": 1.6588,
+      "step": 11602
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.656627906976744e-07,
+      "loss": 0.0008,
+      "step": 11603
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.6565116279069767e-07,
+      "loss": 0.0012,
+      "step": 11604
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.656395348837209e-07,
+      "loss": 0.0007,
+      "step": 11605
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.656279069767442e-07,
+      "loss": 0.0007,
+      "step": 11606
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.6561627906976743e-07,
+      "loss": 0.0013,
+      "step": 11607
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.656046511627907e-07,
+      "loss": 0.0009,
+      "step": 11608
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.6559302325581393e-07,
+      "loss": 0.0043,
+      "step": 11609
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.6558139534883721e-07,
+      "loss": 2.8713,
+      "step": 11610
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.6556976744186048e-07,
+      "loss": 0.0009,
+      "step": 11611
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.655581395348837e-07,
+      "loss": 0.0207,
+      "step": 11612
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.6554651162790697e-07,
+      "loss": 0.0007,
+      "step": 11613
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.6553488372093023e-07,
+      "loss": 0.0009,
+      "step": 11614
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.655232558139535e-07,
+      "loss": 0.0014,
+      "step": 11615
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.6551162790697673e-07,
+      "loss": 0.0007,
+      "step": 11616
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.655e-07,
+      "loss": 0.0007,
+      "step": 11617
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.6548837209302323e-07,
+      "loss": 0.0011,
+      "step": 11618
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.6547674418604652e-07,
+      "loss": 0.001,
+      "step": 11619
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.6546511627906975e-07,
+      "loss": 0.0009,
+      "step": 11620
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.6545348837209301e-07,
+      "loss": 0.0015,
+      "step": 11621
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.6544186046511625e-07,
+      "loss": 0.0059,
+      "step": 11622
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.6543023255813954e-07,
+      "loss": 0.0012,
+      "step": 11623
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.6541860465116277e-07,
+      "loss": 0.0009,
+      "step": 11624
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.6540697674418603e-07,
+      "loss": 0.0011,
+      "step": 11625
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.6539534883720927e-07,
+      "loss": 0.0008,
+      "step": 11626
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.6538372093023256e-07,
+      "loss": 0.003,
+      "step": 11627
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.6537209302325582e-07,
+      "loss": 0.0011,
+      "step": 11628
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.6536046511627905e-07,
+      "loss": 0.001,
+      "step": 11629
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.6534883720930232e-07,
+      "loss": 0.0008,
+      "step": 11630
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.6533720930232558e-07,
+      "loss": 0.0027,
+      "step": 11631
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.6532558139534884e-07,
+      "loss": 0.0015,
+      "step": 11632
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.6531395348837207e-07,
+      "loss": 0.0009,
+      "step": 11633
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.6530232558139534e-07,
+      "loss": 0.0019,
+      "step": 11634
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.652906976744186e-07,
+      "loss": 0.0007,
+      "step": 11635
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.6527906976744186e-07,
+      "loss": 0.0008,
+      "step": 11636
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.652674418604651e-07,
+      "loss": 0.0017,
+      "step": 11637
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.6525581395348836e-07,
+      "loss": 0.001,
+      "step": 11638
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.6524418604651164e-07,
+      "loss": 0.001,
+      "step": 11639
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.6523255813953488e-07,
+      "loss": 0.0014,
+      "step": 11640
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.6522093023255814e-07,
+      "loss": 0.0007,
+      "step": 11641
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.6520930232558138e-07,
+      "loss": 0.0008,
+      "step": 11642
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.6519767441860467e-07,
+      "loss": 0.0009,
+      "step": 11643
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.651860465116279e-07,
+      "loss": 0.0009,
+      "step": 11644
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.6517441860465116e-07,
+      "loss": 0.0009,
+      "step": 11645
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.651627906976744e-07,
+      "loss": 0.0007,
+      "step": 11646
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.6515116279069766e-07,
+      "loss": 0.0012,
+      "step": 11647
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.6513953488372092e-07,
+      "loss": 0.0006,
+      "step": 11648
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.6512790697674418e-07,
+      "loss": 0.0006,
+      "step": 11649
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.6511627906976742e-07,
+      "loss": 0.0006,
+      "step": 11650
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.6510465116279068e-07,
+      "loss": 0.0008,
+      "step": 11651
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.6509302325581397e-07,
+      "loss": 0.0006,
+      "step": 11652
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.650813953488372e-07,
+      "loss": 0.0007,
+      "step": 11653
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.6506976744186046e-07,
+      "loss": 0.0022,
+      "step": 11654
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.650581395348837e-07,
+      "loss": 0.0041,
+      "step": 11655
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 1.65046511627907e-07,
+      "loss": 0.0009,
+      "step": 11656
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 1.6503488372093022e-07,
+      "loss": 0.0007,
+      "step": 11657
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 1.6502325581395348e-07,
+      "loss": 1.3531,
+      "step": 11658
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 1.6501162790697672e-07,
+      "loss": 0.0008,
+      "step": 11659
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 1.65e-07,
+      "loss": 0.0011,
+      "step": 11660
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.6498837209302324e-07,
+      "loss": 0.0009,
+      "step": 11661
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.649767441860465e-07,
+      "loss": 0.001,
+      "step": 11662
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.6496511627906974e-07,
+      "loss": 0.0009,
+      "step": 11663
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.6495348837209303e-07,
+      "loss": 0.0008,
+      "step": 11664
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.649418604651163e-07,
+      "loss": 0.0007,
+      "step": 11665
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.6493023255813953e-07,
+      "loss": 0.0008,
+      "step": 11666
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 1.6491860465116276e-07,
+      "loss": 0.001,
+      "step": 11667
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 1.6490697674418605e-07,
+      "loss": 0.001,
+      "step": 11668
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 1.648953488372093e-07,
+      "loss": 0.0013,
+      "step": 11669
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 1.6488372093023255e-07,
+      "loss": 0.0021,
+      "step": 11670
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 1.648720930232558e-07,
+      "loss": 0.0007,
+      "step": 11671
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.6486046511627904e-07,
+      "loss": 0.0009,
+      "step": 11672
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.6484883720930233e-07,
+      "loss": 0.0006,
+      "step": 11673
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.6483720930232557e-07,
+      "loss": 0.0011,
+      "step": 11674
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.6482558139534883e-07,
+      "loss": 0.0169,
+      "step": 11675
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.6481395348837206e-07,
+      "loss": 0.0011,
+      "step": 11676
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 1.6480232558139535e-07,
+      "loss": 0.0007,
+      "step": 11677
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 1.647906976744186e-07,
+      "loss": 0.0009,
+      "step": 11678
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 1.6477906976744185e-07,
+      "loss": 0.001,
+      "step": 11679
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 1.6476744186046508e-07,
+      "loss": 0.0006,
+      "step": 11680
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 1.6475581395348837e-07,
+      "loss": 0.0011,
+      "step": 11681
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 1.6474418604651163e-07,
+      "loss": 0.0008,
+      "step": 11682
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 1.6473255813953487e-07,
+      "loss": 0.001,
+      "step": 11683
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 1.6472093023255813e-07,
+      "loss": 0.0009,
+      "step": 11684
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 1.647093023255814e-07,
+      "loss": 0.0008,
+      "step": 11685
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 1.6469767441860465e-07,
+      "loss": 0.0009,
+      "step": 11686
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 1.646860465116279e-07,
+      "loss": 0.001,
+      "step": 11687
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 1.6467441860465115e-07,
+      "loss": 0.0009,
+      "step": 11688
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 1.646627906976744e-07,
+      "loss": 0.2751,
+      "step": 11689
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 1.6465116279069767e-07,
+      "loss": 0.0008,
+      "step": 11690
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 1.646395348837209e-07,
+      "loss": 0.0008,
+      "step": 11691
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.6462790697674417e-07,
+      "loss": 0.0008,
+      "step": 11692
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.6461627906976746e-07,
+      "loss": 0.0009,
+      "step": 11693
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.646046511627907e-07,
+      "loss": 0.0007,
+      "step": 11694
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.6459302325581396e-07,
+      "loss": 0.0007,
+      "step": 11695
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.645813953488372e-07,
+      "loss": 0.0015,
+      "step": 11696
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 1.6456976744186048e-07,
+      "loss": 0.0023,
+      "step": 11697
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 1.6455813953488372e-07,
+      "loss": 0.0009,
+      "step": 11698
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 1.6454651162790698e-07,
+      "loss": 0.0013,
+      "step": 11699
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 1.645348837209302e-07,
+      "loss": 0.0036,
+      "step": 11700
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 1.6452325581395347e-07,
+      "loss": 0.0008,
+      "step": 11701
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 1.6451162790697674e-07,
+      "loss": 0.001,
+      "step": 11702
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 1.645e-07,
+      "loss": 0.0014,
+      "step": 11703
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 1.6448837209302323e-07,
+      "loss": 0.0009,
+      "step": 11704
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 1.644767441860465e-07,
+      "loss": 0.0007,
+      "step": 11705
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 1.6446511627906978e-07,
+      "loss": 0.0008,
+      "step": 11706
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 1.6445348837209302e-07,
+      "loss": 0.0008,
+      "step": 11707
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.6444186046511628e-07,
+      "loss": 0.0009,
+      "step": 11708
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.6443023255813951e-07,
+      "loss": 0.0008,
+      "step": 11709
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.644186046511628e-07,
+      "loss": 0.0008,
+      "step": 11710
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.6440697674418604e-07,
+      "loss": 0.0012,
+      "step": 11711
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.643953488372093e-07,
+      "loss": 0.0009,
+      "step": 11712
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 1.6438372093023254e-07,
+      "loss": 0.5755,
+      "step": 11713
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 1.6437209302325582e-07,
+      "loss": 0.0012,
+      "step": 11714
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 1.6436046511627906e-07,
+      "loss": 0.8415,
+      "step": 11715
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 1.6434883720930232e-07,
+      "loss": 0.0209,
+      "step": 11716
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 1.6433720930232556e-07,
+      "loss": 0.0013,
+      "step": 11717
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.6432558139534884e-07,
+      "loss": 0.0014,
+      "step": 11718
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.643139534883721e-07,
+      "loss": 0.0009,
+      "step": 11719
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.6430232558139534e-07,
+      "loss": 0.001,
+      "step": 11720
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.6429069767441858e-07,
+      "loss": 0.0018,
+      "step": 11721
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.6427906976744186e-07,
+      "loss": 0.0013,
+      "step": 11722
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 1.6426744186046513e-07,
+      "loss": 0.0025,
+      "step": 11723
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 1.6425581395348836e-07,
+      "loss": 0.0209,
+      "step": 11724
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 1.6424418604651162e-07,
+      "loss": 0.0009,
+      "step": 11725
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 1.6423255813953486e-07,
+      "loss": 0.0012,
+      "step": 11726
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 1.6422093023255815e-07,
+      "loss": 0.0007,
+      "step": 11727
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 1.6420930232558138e-07,
+      "loss": 0.0014,
+      "step": 11728
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.6419767441860464e-07,
+      "loss": 0.001,
+      "step": 11729
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.6418604651162788e-07,
+      "loss": 0.0007,
+      "step": 11730
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.6417441860465117e-07,
+      "loss": 0.0012,
+      "step": 11731
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.641627906976744e-07,
+      "loss": 0.0009,
+      "step": 11732
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.6415116279069766e-07,
+      "loss": 0.0007,
+      "step": 11733
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 1.641395348837209e-07,
+      "loss": 0.0011,
+      "step": 11734
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 1.641279069767442e-07,
+      "loss": 0.0009,
+      "step": 11735
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 1.6411627906976745e-07,
+      "loss": 0.0007,
+      "step": 11736
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 1.6410465116279068e-07,
+      "loss": 0.0011,
+      "step": 11737
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 1.6409302325581395e-07,
+      "loss": 0.0009,
+      "step": 11738
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 1.640813953488372e-07,
+      "loss": 0.0032,
+      "step": 11739
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 1.6406976744186047e-07,
+      "loss": 0.0008,
+      "step": 11740
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 1.640581395348837e-07,
+      "loss": 0.0008,
+      "step": 11741
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 1.6404651162790697e-07,
+      "loss": 0.0007,
+      "step": 11742
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 1.6403488372093023e-07,
+      "loss": 0.0009,
+      "step": 11743
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.640232558139535e-07,
+      "loss": 0.0008,
+      "step": 11744
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.6401162790697672e-07,
+      "loss": 0.0007,
+      "step": 11745
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.6399999999999999e-07,
+      "loss": 0.001,
+      "step": 11746
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.6398837209302327e-07,
+      "loss": 0.0008,
+      "step": 11747
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.639767441860465e-07,
+      "loss": 0.1568,
+      "step": 11748
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 1.6396511627906977e-07,
+      "loss": 0.0009,
+      "step": 11749
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 1.63953488372093e-07,
+      "loss": 0.001,
+      "step": 11750
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 1.6394186046511627e-07,
+      "loss": 0.0014,
+      "step": 11751
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 1.6393023255813953e-07,
+      "loss": 0.001,
+      "step": 11752
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 1.639186046511628e-07,
+      "loss": 0.0007,
+      "step": 11753
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.6390697674418603e-07,
+      "loss": 0.0012,
+      "step": 11754
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.638953488372093e-07,
+      "loss": 0.0008,
+      "step": 11755
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.6388372093023255e-07,
+      "loss": 0.0007,
+      "step": 11756
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.638720930232558e-07,
+      "loss": 0.001,
+      "step": 11757
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.6386046511627905e-07,
+      "loss": 0.0024,
+      "step": 11758
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.638488372093023e-07,
+      "loss": 0.0009,
+      "step": 11759
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 1.638372093023256e-07,
+      "loss": 0.0013,
+      "step": 11760
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 1.6382558139534883e-07,
+      "loss": 0.002,
+      "step": 11761
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 1.6381395348837207e-07,
+      "loss": 0.001,
+      "step": 11762
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 1.6380232558139533e-07,
+      "loss": 0.0011,
+      "step": 11763
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 1.6379069767441862e-07,
+      "loss": 0.0011,
+      "step": 11764
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.6377906976744185e-07,
+      "loss": 0.0009,
+      "step": 11765
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.6376744186046511e-07,
+      "loss": 0.0007,
+      "step": 11766
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.6375581395348835e-07,
+      "loss": 0.0009,
+      "step": 11767
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.6374418604651164e-07,
+      "loss": 1.5405,
+      "step": 11768
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.6373255813953487e-07,
+      "loss": 0.0008,
+      "step": 11769
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.6372093023255813e-07,
+      "loss": 0.0009,
+      "step": 11770
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.6370930232558137e-07,
+      "loss": 0.0016,
+      "step": 11771
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.6369767441860466e-07,
+      "loss": 0.001,
+      "step": 11772
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.636860465116279e-07,
+      "loss": 0.0009,
+      "step": 11773
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 1.6367441860465116e-07,
+      "loss": 0.0005,
+      "step": 11774
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 1.636627906976744e-07,
+      "loss": 0.0022,
+      "step": 11775
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 1.6365116279069768e-07,
+      "loss": 0.001,
+      "step": 11776
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 1.6363953488372094e-07,
+      "loss": 0.0013,
+      "step": 11777
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 1.6362790697674418e-07,
+      "loss": 0.0029,
+      "step": 11778
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 1.6361627906976744e-07,
+      "loss": 0.0009,
+      "step": 11779
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.6360465116279067e-07,
+      "loss": 0.0006,
+      "step": 11780
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.6359302325581396e-07,
+      "loss": 0.0057,
+      "step": 11781
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.635813953488372e-07,
+      "loss": 0.0007,
+      "step": 11782
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.6356976744186046e-07,
+      "loss": 0.0009,
+      "step": 11783
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.635581395348837e-07,
+      "loss": 0.0008,
+      "step": 11784
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.6354651162790698e-07,
+      "loss": 0.0012,
+      "step": 11785
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 1.6353488372093022e-07,
+      "loss": 0.001,
+      "step": 11786
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 1.6352325581395348e-07,
+      "loss": 0.0008,
+      "step": 11787
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 1.6351162790697671e-07,
+      "loss": 0.0018,
+      "step": 11788
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 1.635e-07,
+      "loss": 0.001,
+      "step": 11789
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 1.6348837209302326e-07,
+      "loss": 0.0011,
+      "step": 11790
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.634767441860465e-07,
+      "loss": 0.001,
+      "step": 11791
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.6346511627906976e-07,
+      "loss": 0.0008,
+      "step": 11792
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.6345348837209302e-07,
+      "loss": 1.9132,
+      "step": 11793
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.6344186046511628e-07,
+      "loss": 0.001,
+      "step": 11794
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.6343023255813952e-07,
+      "loss": 0.0009,
+      "step": 11795
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 1.6341860465116278e-07,
+      "loss": 0.0007,
+      "step": 11796
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 1.6340697674418604e-07,
+      "loss": 0.0006,
+      "step": 11797
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 1.633953488372093e-07,
+      "loss": 0.0011,
+      "step": 11798
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 1.6338372093023254e-07,
+      "loss": 0.0008,
+      "step": 11799
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 1.633720930232558e-07,
+      "loss": 0.0008,
+      "step": 11800
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 1.633604651162791e-07,
+      "loss": 0.0016,
+      "step": 11801
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 1.6334883720930232e-07,
+      "loss": 0.0009,
+      "step": 11802
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 1.6333720930232559e-07,
+      "loss": 0.0013,
+      "step": 11803
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 1.6332558139534882e-07,
+      "loss": 0.0006,
+      "step": 11804
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 1.6331395348837206e-07,
+      "loss": 0.0022,
+      "step": 11805
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.6330232558139534e-07,
+      "loss": 0.0019,
+      "step": 11806
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.632906976744186e-07,
+      "loss": 0.0009,
+      "step": 11807
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.6327906976744184e-07,
+      "loss": 0.0007,
+      "step": 11808
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.632674418604651e-07,
+      "loss": 0.0285,
+      "step": 11809
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 1.6325581395348837e-07,
+      "loss": 0.0007,
+      "step": 11810
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.6324418604651163e-07,
+      "loss": 0.0016,
+      "step": 11811
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.6323255813953486e-07,
+      "loss": 0.0011,
+      "step": 11812
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.6322093023255812e-07,
+      "loss": 0.8299,
+      "step": 11813
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.632093023255814e-07,
+      "loss": 0.0008,
+      "step": 11814
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.6319767441860465e-07,
+      "loss": 0.0006,
+      "step": 11815
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.6318604651162788e-07,
+      "loss": 0.0008,
+      "step": 11816
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 1.6317441860465114e-07,
+      "loss": 0.0009,
+      "step": 11817
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 1.6316279069767443e-07,
+      "loss": 0.0008,
+      "step": 11818
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 1.6315116279069767e-07,
+      "loss": 0.0007,
+      "step": 11819
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 1.6313953488372093e-07,
+      "loss": 0.0012,
+      "step": 11820
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 1.6312790697674416e-07,
+      "loss": 0.0376,
+      "step": 11821
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 1.6311627906976745e-07,
+      "loss": 0.0018,
+      "step": 11822
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 1.631046511627907e-07,
+      "loss": 0.0008,
+      "step": 11823
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 1.6309302325581395e-07,
+      "loss": 0.0008,
+      "step": 11824
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 1.6308139534883718e-07,
+      "loss": 0.0008,
+      "step": 11825
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 1.6306976744186047e-07,
+      "loss": 3.3283,
+      "step": 11826
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 1.630581395348837e-07,
+      "loss": 0.0007,
+      "step": 11827
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 1.6304651162790697e-07,
+      "loss": 0.0057,
+      "step": 11828
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 1.630348837209302e-07,
+      "loss": 0.0009,
+      "step": 11829
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 1.630232558139535e-07,
+      "loss": 0.001,
+      "step": 11830
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 1.6301162790697676e-07,
+      "loss": 0.0007,
+      "step": 11831
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 1.63e-07,
+      "loss": 0.0006,
+      "step": 11832
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 1.6298837209302325e-07,
+      "loss": 0.001,
+      "step": 11833
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 1.629767441860465e-07,
+      "loss": 0.0007,
+      "step": 11834
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 1.6296511627906978e-07,
+      "loss": 0.0016,
+      "step": 11835
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 1.62953488372093e-07,
+      "loss": 0.0011,
+      "step": 11836
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.6294186046511627e-07,
+      "loss": 0.0006,
+      "step": 11837
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.629302325581395e-07,
+      "loss": 0.0009,
+      "step": 11838
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.629186046511628e-07,
+      "loss": 0.0007,
+      "step": 11839
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.6290697674418603e-07,
+      "loss": 0.0008,
+      "step": 11840
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.628953488372093e-07,
+      "loss": 0.0009,
+      "step": 11841
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.6288372093023253e-07,
+      "loss": 0.0011,
+      "step": 11842
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.6287209302325582e-07,
+      "loss": 0.001,
+      "step": 11843
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.6286046511627908e-07,
+      "loss": 0.0012,
+      "step": 11844
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.6284883720930231e-07,
+      "loss": 0.0017,
+      "step": 11845
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.6283720930232558e-07,
+      "loss": 0.001,
+      "step": 11846
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 1.6282558139534884e-07,
+      "loss": 0.0009,
+      "step": 11847
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 1.628139534883721e-07,
+      "loss": 0.0009,
+      "step": 11848
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 1.6280232558139533e-07,
+      "loss": 0.0041,
+      "step": 11849
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 1.627906976744186e-07,
+      "loss": 0.001,
+      "step": 11850
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 1.6277906976744186e-07,
+      "loss": 0.0014,
+      "step": 11851
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 1.6276744186046512e-07,
+      "loss": 0.0006,
+      "step": 11852
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 1.6275581395348835e-07,
+      "loss": 0.001,
+      "step": 11853
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 1.6274418604651162e-07,
+      "loss": 0.0007,
+      "step": 11854
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 1.627325581395349e-07,
+      "loss": 0.001,
+      "step": 11855
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 1.6272093023255814e-07,
+      "loss": 0.0017,
+      "step": 11856
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 1.627093023255814e-07,
+      "loss": 0.0009,
+      "step": 11857
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 1.6269767441860464e-07,
+      "loss": 0.0009,
+      "step": 11858
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 1.6268604651162787e-07,
+      "loss": 0.0005,
+      "step": 11859
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 1.6267441860465116e-07,
+      "loss": 0.001,
+      "step": 11860
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 1.6266279069767442e-07,
+      "loss": 0.0012,
+      "step": 11861
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 1.6265116279069766e-07,
+      "loss": 0.0007,
+      "step": 11862
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 1.6263953488372092e-07,
+      "loss": 0.1472,
+      "step": 11863
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 1.6262790697674418e-07,
+      "loss": 0.0013,
+      "step": 11864
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 1.6261627906976744e-07,
+      "loss": 0.001,
+      "step": 11865
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 1.6260465116279068e-07,
+      "loss": 0.0007,
+      "step": 11866
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 1.6259302325581394e-07,
+      "loss": 0.0007,
+      "step": 11867
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.6258139534883723e-07,
+      "loss": 0.0007,
+      "step": 11868
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.6256976744186046e-07,
+      "loss": 0.0009,
+      "step": 11869
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.625581395348837e-07,
+      "loss": 0.0008,
+      "step": 11870
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.6254651162790696e-07,
+      "loss": 0.0007,
+      "step": 11871
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.6253488372093025e-07,
+      "loss": 0.0008,
+      "step": 11872
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 1.6252325581395348e-07,
+      "loss": 0.0009,
+      "step": 11873
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 1.6251162790697674e-07,
+      "loss": 0.0009,
+      "step": 11874
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 1.6249999999999998e-07,
+      "loss": 0.0011,
+      "step": 11875
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 1.6248837209302327e-07,
+      "loss": 0.0008,
+      "step": 11876
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 1.624767441860465e-07,
+      "loss": 0.0016,
+      "step": 11877
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 1.6246511627906976e-07,
+      "loss": 0.0015,
+      "step": 11878
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.62453488372093e-07,
+      "loss": 0.0007,
+      "step": 11879
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.624418604651163e-07,
+      "loss": 0.001,
+      "step": 11880
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.6243023255813952e-07,
+      "loss": 0.0011,
+      "step": 11881
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.6241860465116278e-07,
+      "loss": 0.4322,
+      "step": 11882
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.6240697674418602e-07,
+      "loss": 0.0008,
+      "step": 11883
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 1.6239534883720928e-07,
+      "loss": 0.0007,
+      "step": 11884
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 1.6238372093023257e-07,
+      "loss": 0.0018,
+      "step": 11885
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 1.623720930232558e-07,
+      "loss": 0.0008,
+      "step": 11886
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 1.6236046511627907e-07,
+      "loss": 0.001,
+      "step": 11887
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 1.623488372093023e-07,
+      "loss": 0.0008,
+      "step": 11888
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 1.623372093023256e-07,
+      "loss": 0.0007,
+      "step": 11889
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 1.6232558139534883e-07,
+      "loss": 0.0006,
+      "step": 11890
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 1.623139534883721e-07,
+      "loss": 0.001,
+      "step": 11891
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.9893617021276596,
+      "eval_loss": 0.15252994000911713,
+      "eval_runtime": 1.5526,
+      "eval_samples_per_second": 665.971,
+      "eval_steps_per_second": 10.949,
+      "step": 11891
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 1.6230232558139532e-07,
+      "loss": 0.0011,
+      "step": 11892
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 1.622906976744186e-07,
+      "loss": 0.0008,
+      "step": 11893
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.6227906976744185e-07,
+      "loss": 0.5242,
+      "step": 11894
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.622674418604651e-07,
+      "loss": 0.0016,
+      "step": 11895
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.6225581395348834e-07,
+      "loss": 0.0008,
+      "step": 11896
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.6224418604651163e-07,
+      "loss": 0.0012,
+      "step": 11897
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.622325581395349e-07,
+      "loss": 0.0035,
+      "step": 11898
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 1.6222093023255813e-07,
+      "loss": 0.0021,
+      "step": 11899
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 1.6220930232558136e-07,
+      "loss": 0.0028,
+      "step": 11900
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 1.6219767441860465e-07,
+      "loss": 0.001,
+      "step": 11901
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 1.6218604651162791e-07,
+      "loss": 0.0007,
+      "step": 11902
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 1.6217441860465115e-07,
+      "loss": 0.0027,
+      "step": 11903
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.621627906976744e-07,
+      "loss": 0.1124,
+      "step": 11904
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.6215116279069767e-07,
+      "loss": 0.0009,
+      "step": 11905
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.6213953488372093e-07,
+      "loss": 0.0007,
+      "step": 11906
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.6212790697674417e-07,
+      "loss": 0.0008,
+      "step": 11907
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.6211627906976743e-07,
+      "loss": 0.0011,
+      "step": 11908
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.6210465116279072e-07,
+      "loss": 0.001,
+      "step": 11909
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 1.6209302325581395e-07,
+      "loss": 0.0007,
+      "step": 11910
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 1.620813953488372e-07,
+      "loss": 0.0008,
+      "step": 11911
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 1.6206976744186045e-07,
+      "loss": 0.0008,
+      "step": 11912
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 1.6205813953488369e-07,
+      "loss": 0.0008,
+      "step": 11913
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 1.6204651162790697e-07,
+      "loss": 0.0011,
+      "step": 11914
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.6203488372093024e-07,
+      "loss": 0.0008,
+      "step": 11915
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.6202325581395347e-07,
+      "loss": 0.0011,
+      "step": 11916
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.6201162790697673e-07,
+      "loss": 0.0009,
+      "step": 11917
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.62e-07,
+      "loss": 0.0009,
+      "step": 11918
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.6198837209302326e-07,
+      "loss": 0.0009,
+      "step": 11919
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 1.619767441860465e-07,
+      "loss": 0.0012,
+      "step": 11920
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 1.6196511627906975e-07,
+      "loss": 0.0032,
+      "step": 11921
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 1.6195348837209302e-07,
+      "loss": 3.38,
+      "step": 11922
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 1.6194186046511628e-07,
+      "loss": 0.0009,
+      "step": 11923
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 1.619302325581395e-07,
+      "loss": 0.0092,
+      "step": 11924
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 1.6191860465116277e-07,
+      "loss": 0.0016,
+      "step": 11925
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 1.6190697674418606e-07,
+      "loss": 0.0009,
+      "step": 11926
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 1.618953488372093e-07,
+      "loss": 0.0008,
+      "step": 11927
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 1.6188372093023256e-07,
+      "loss": 0.0009,
+      "step": 11928
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 1.618720930232558e-07,
+      "loss": 0.001,
+      "step": 11929
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 1.6186046511627908e-07,
+      "loss": 0.0009,
+      "step": 11930
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 1.6184883720930232e-07,
+      "loss": 0.0009,
+      "step": 11931
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 1.6183720930232558e-07,
+      "loss": 1.8095,
+      "step": 11932
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 1.6182558139534881e-07,
+      "loss": 0.0009,
+      "step": 11933
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 1.618139534883721e-07,
+      "loss": 0.0009,
+      "step": 11934
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.6180232558139534e-07,
+      "loss": 0.0017,
+      "step": 11935
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.617906976744186e-07,
+      "loss": 0.0011,
+      "step": 11936
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.6177906976744183e-07,
+      "loss": 0.0007,
+      "step": 11937
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.617674418604651e-07,
+      "loss": 0.0667,
+      "step": 11938
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.6175581395348838e-07,
+      "loss": 0.0008,
+      "step": 11939
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.6174418604651162e-07,
+      "loss": 0.0008,
+      "step": 11940
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.6173255813953488e-07,
+      "loss": 0.0012,
+      "step": 11941
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.6172093023255812e-07,
+      "loss": 0.001,
+      "step": 11942
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.617093023255814e-07,
+      "loss": 0.0009,
+      "step": 11943
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.6169767441860464e-07,
+      "loss": 0.0008,
+      "step": 11944
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.616860465116279e-07,
+      "loss": 0.0012,
+      "step": 11945
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 1.6167441860465114e-07,
+      "loss": 0.0009,
+      "step": 11946
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 1.6166279069767443e-07,
+      "loss": 0.0009,
+      "step": 11947
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 1.6165116279069766e-07,
+      "loss": 0.3866,
+      "step": 11948
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 1.6163953488372092e-07,
+      "loss": 3.257,
+      "step": 11949
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 1.6162790697674416e-07,
+      "loss": 0.001,
+      "step": 11950
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 1.6161627906976745e-07,
+      "loss": 0.0007,
+      "step": 11951
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 1.616046511627907e-07,
+      "loss": 0.0007,
+      "step": 11952
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 1.6159302325581394e-07,
+      "loss": 0.001,
+      "step": 11953
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 1.6158139534883718e-07,
+      "loss": 0.0008,
+      "step": 11954
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 1.6156976744186047e-07,
+      "loss": 2.1203,
+      "step": 11955
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 1.6155813953488373e-07,
+      "loss": 0.0005,
+      "step": 11956
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 1.6154651162790696e-07,
+      "loss": 0.0008,
+      "step": 11957
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 1.6153488372093023e-07,
+      "loss": 0.0009,
+      "step": 11958
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 1.6152325581395349e-07,
+      "loss": 0.0009,
+      "step": 11959
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 1.6151162790697675e-07,
+      "loss": 0.0008,
+      "step": 11960
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 1.6149999999999998e-07,
+      "loss": 0.0009,
+      "step": 11961
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 1.6148837209302325e-07,
+      "loss": 0.0007,
+      "step": 11962
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 1.6147674418604653e-07,
+      "loss": 0.0009,
+      "step": 11963
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 1.6146511627906977e-07,
+      "loss": 0.0007,
+      "step": 11964
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 1.61453488372093e-07,
+      "loss": 0.0007,
+      "step": 11965
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.6144186046511627e-07,
+      "loss": 0.0008,
+      "step": 11966
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.614302325581395e-07,
+      "loss": 0.001,
+      "step": 11967
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.614186046511628e-07,
+      "loss": 0.0009,
+      "step": 11968
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.6140697674418605e-07,
+      "loss": 0.0011,
+      "step": 11969
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.6139534883720929e-07,
+      "loss": 0.0008,
+      "step": 11970
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 1.6138372093023255e-07,
+      "loss": 0.0009,
+      "step": 11971
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.613720930232558e-07,
+      "loss": 0.0012,
+      "step": 11972
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.6136046511627907e-07,
+      "loss": 0.0008,
+      "step": 11973
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.613488372093023e-07,
+      "loss": 0.001,
+      "step": 11974
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.6133720930232557e-07,
+      "loss": 0.0009,
+      "step": 11975
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.6132558139534883e-07,
+      "loss": 0.0008,
+      "step": 11976
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 1.613139534883721e-07,
+      "loss": 0.0009,
+      "step": 11977
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 1.6130232558139533e-07,
+      "loss": 0.0007,
+      "step": 11978
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 1.612906976744186e-07,
+      "loss": 0.003,
+      "step": 11979
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 1.6127906976744188e-07,
+      "loss": 0.0007,
+      "step": 11980
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 1.612674418604651e-07,
+      "loss": 0.001,
+      "step": 11981
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 1.6125581395348837e-07,
+      "loss": 0.0008,
+      "step": 11982
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 1.612441860465116e-07,
+      "loss": 0.0008,
+      "step": 11983
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 1.612325581395349e-07,
+      "loss": 0.0011,
+      "step": 11984
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 1.6122093023255813e-07,
+      "loss": 0.0014,
+      "step": 11985
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 1.612093023255814e-07,
+      "loss": 0.001,
+      "step": 11986
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 1.6119767441860463e-07,
+      "loss": 0.0008,
+      "step": 11987
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 1.6118604651162792e-07,
+      "loss": 0.0009,
+      "step": 11988
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 1.6117441860465115e-07,
+      "loss": 0.0008,
+      "step": 11989
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 1.6116279069767441e-07,
+      "loss": 0.001,
+      "step": 11990
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 1.6115116279069765e-07,
+      "loss": 0.001,
+      "step": 11991
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 1.611395348837209e-07,
+      "loss": 2.1413,
+      "step": 11992
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 1.611279069767442e-07,
+      "loss": 0.0008,
+      "step": 11993
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 1.6111627906976743e-07,
+      "loss": 0.002,
+      "step": 11994
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 1.611046511627907e-07,
+      "loss": 0.0008,
+      "step": 11995
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 1.6109302325581393e-07,
+      "loss": 0.001,
+      "step": 11996
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.6108139534883722e-07,
+      "loss": 0.0008,
+      "step": 11997
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.6106976744186046e-07,
+      "loss": 0.0008,
+      "step": 11998
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.6105813953488372e-07,
+      "loss": 0.0008,
+      "step": 11999
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.6104651162790695e-07,
+      "loss": 0.0013,
+      "step": 12000
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.6103488372093024e-07,
+      "loss": 0.0009,
+      "step": 12001
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.6102325581395348e-07,
+      "loss": 0.4361,
+      "step": 12002
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.6101162790697674e-07,
+      "loss": 0.0007,
+      "step": 12003
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.6099999999999997e-07,
+      "loss": 0.0011,
+      "step": 12004
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.6098837209302326e-07,
+      "loss": 0.0007,
+      "step": 12005
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.609767441860465e-07,
+      "loss": 0.0012,
+      "step": 12006
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.6096511627906976e-07,
+      "loss": 0.0007,
+      "step": 12007
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.60953488372093e-07,
+      "loss": 0.001,
+      "step": 12008
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.6094186046511628e-07,
+      "loss": 0.0008,
+      "step": 12009
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.6093023255813954e-07,
+      "loss": 0.0007,
+      "step": 12010
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.6091860465116278e-07,
+      "loss": 0.0031,
+      "step": 12011
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.6090697674418604e-07,
+      "loss": 2.5007,
+      "step": 12012
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 1.608953488372093e-07,
+      "loss": 0.0013,
+      "step": 12013
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 1.6088372093023256e-07,
+      "loss": 0.1243,
+      "step": 12014
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 1.608720930232558e-07,
+      "loss": 0.0012,
+      "step": 12015
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 1.6086046511627906e-07,
+      "loss": 0.0008,
+      "step": 12016
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 1.6084883720930232e-07,
+      "loss": 0.0005,
+      "step": 12017
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.6083720930232558e-07,
+      "loss": 0.0009,
+      "step": 12018
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.6082558139534882e-07,
+      "loss": 0.0008,
+      "step": 12019
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.6081395348837208e-07,
+      "loss": 0.0008,
+      "step": 12020
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.6080232558139532e-07,
+      "loss": 0.001,
+      "step": 12021
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.607906976744186e-07,
+      "loss": 0.1832,
+      "step": 12022
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 1.6077906976744187e-07,
+      "loss": 0.0015,
+      "step": 12023
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 1.607674418604651e-07,
+      "loss": 0.0014,
+      "step": 12024
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 1.6075581395348836e-07,
+      "loss": 0.001,
+      "step": 12025
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 1.6074418604651162e-07,
+      "loss": 0.383,
+      "step": 12026
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 1.6073255813953489e-07,
+      "loss": 0.0008,
+      "step": 12027
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 1.6072093023255812e-07,
+      "loss": 0.0008,
+      "step": 12028
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.6070930232558138e-07,
+      "loss": 0.0008,
+      "step": 12029
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.6069767441860464e-07,
+      "loss": 0.259,
+      "step": 12030
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.606860465116279e-07,
+      "loss": 0.0011,
+      "step": 12031
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.6067441860465114e-07,
+      "loss": 0.0008,
+      "step": 12032
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.606627906976744e-07,
+      "loss": 0.0008,
+      "step": 12033
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 1.606511627906977e-07,
+      "loss": 0.0009,
+      "step": 12034
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 1.6063953488372093e-07,
+      "loss": 0.0008,
+      "step": 12035
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 1.606279069767442e-07,
+      "loss": 0.0008,
+      "step": 12036
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 1.6061627906976742e-07,
+      "loss": 0.0006,
+      "step": 12037
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 1.606046511627907e-07,
+      "loss": 0.0007,
+      "step": 12038
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.6059302325581395e-07,
+      "loss": 0.0009,
+      "step": 12039
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.605813953488372e-07,
+      "loss": 0.0011,
+      "step": 12040
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.6056976744186044e-07,
+      "loss": 0.0025,
+      "step": 12041
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.6055813953488373e-07,
+      "loss": 0.001,
+      "step": 12042
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 1.6054651162790697e-07,
+      "loss": 0.0007,
+      "step": 12043
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 1.6053488372093023e-07,
+      "loss": 0.0009,
+      "step": 12044
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 1.6052325581395346e-07,
+      "loss": 0.002,
+      "step": 12045
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 1.6051162790697673e-07,
+      "loss": 0.001,
+      "step": 12046
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 1.6050000000000001e-07,
+      "loss": 0.0008,
+      "step": 12047
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 1.6048837209302325e-07,
+      "loss": 0.0007,
+      "step": 12048
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 1.6047674418604648e-07,
+      "loss": 0.0014,
+      "step": 12049
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 1.6046511627906975e-07,
+      "loss": 0.0008,
+      "step": 12050
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 1.6045348837209303e-07,
+      "loss": 0.1317,
+      "step": 12051
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 1.6044186046511627e-07,
+      "loss": 0.001,
+      "step": 12052
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 1.6043023255813953e-07,
+      "loss": 0.0018,
+      "step": 12053
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.6041860465116277e-07,
+      "loss": 0.001,
+      "step": 12054
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.6040697674418606e-07,
+      "loss": 0.0008,
+      "step": 12055
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.603953488372093e-07,
+      "loss": 0.0007,
+      "step": 12056
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.6038372093023255e-07,
+      "loss": 0.0007,
+      "step": 12057
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.603720930232558e-07,
+      "loss": 0.0008,
+      "step": 12058
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.6036046511627908e-07,
+      "loss": 0.0017,
+      "step": 12059
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 1.603488372093023e-07,
+      "loss": 0.0011,
+      "step": 12060
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 1.6033720930232557e-07,
+      "loss": 0.0016,
+      "step": 12061
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 1.603255813953488e-07,
+      "loss": 0.0008,
+      "step": 12062
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 1.603139534883721e-07,
+      "loss": 0.0008,
+      "step": 12063
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 1.6030232558139536e-07,
+      "loss": 0.0016,
+      "step": 12064
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.602906976744186e-07,
+      "loss": 0.0009,
+      "step": 12065
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.6027906976744185e-07,
+      "loss": 0.0009,
+      "step": 12066
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.6026744186046512e-07,
+      "loss": 0.0013,
+      "step": 12067
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.6025581395348838e-07,
+      "loss": 0.0025,
+      "step": 12068
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.6024418604651161e-07,
+      "loss": 0.0008,
+      "step": 12069
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 1.6023255813953487e-07,
+      "loss": 0.0012,
+      "step": 12070
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 1.602209302325581e-07,
+      "loss": 2.56,
+      "step": 12071
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 1.602093023255814e-07,
+      "loss": 0.0009,
+      "step": 12072
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 1.6019767441860463e-07,
+      "loss": 0.0008,
+      "step": 12073
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 1.601860465116279e-07,
+      "loss": 0.0007,
+      "step": 12074
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 1.6017441860465113e-07,
+      "loss": 0.0007,
+      "step": 12075
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 1.6016279069767442e-07,
+      "loss": 0.001,
+      "step": 12076
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 1.6015116279069768e-07,
+      "loss": 0.0022,
+      "step": 12077
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 1.6013953488372092e-07,
+      "loss": 0.0008,
+      "step": 12078
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 1.6012790697674418e-07,
+      "loss": 0.0006,
+      "step": 12079
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 1.6011627906976744e-07,
+      "loss": 0.0008,
+      "step": 12080
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 1.601046511627907e-07,
+      "loss": 0.0007,
+      "step": 12081
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 1.6009302325581394e-07,
+      "loss": 0.0006,
+      "step": 12082
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 1.600813953488372e-07,
+      "loss": 0.001,
+      "step": 12083
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 1.6006976744186046e-07,
+      "loss": 0.0006,
+      "step": 12084
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.6005813953488372e-07,
+      "loss": 0.0006,
+      "step": 12085
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.6004651162790696e-07,
+      "loss": 0.0009,
+      "step": 12086
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.6003488372093022e-07,
+      "loss": 0.0011,
+      "step": 12087
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.600232558139535e-07,
+      "loss": 0.0011,
+      "step": 12088
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.6001162790697674e-07,
+      "loss": 0.0006,
+      "step": 12089
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.6e-07,
+      "loss": 0.0007,
+      "step": 12090
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.5998837209302324e-07,
+      "loss": 0.0009,
+      "step": 12091
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.5997674418604653e-07,
+      "loss": 0.0007,
+      "step": 12092
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.5996511627906976e-07,
+      "loss": 0.001,
+      "step": 12093
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.5995348837209302e-07,
+      "loss": 0.0011,
+      "step": 12094
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.5994186046511626e-07,
+      "loss": 0.0007,
+      "step": 12095
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 1.5993023255813955e-07,
+      "loss": 0.0013,
+      "step": 12096
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 1.5991860465116278e-07,
+      "loss": 0.0009,
+      "step": 12097
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 1.5990697674418604e-07,
+      "loss": 0.0009,
+      "step": 12098
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 1.5989534883720928e-07,
+      "loss": 0.0009,
+      "step": 12099
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 1.5988372093023254e-07,
+      "loss": 0.0008,
+      "step": 12100
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.5987209302325583e-07,
+      "loss": 0.0007,
+      "step": 12101
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.5986046511627906e-07,
+      "loss": 0.0013,
+      "step": 12102
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.598488372093023e-07,
+      "loss": 0.0011,
+      "step": 12103
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.5983720930232556e-07,
+      "loss": 0.0011,
+      "step": 12104
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.5982558139534885e-07,
+      "loss": 0.0008,
+      "step": 12105
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 1.5981395348837208e-07,
+      "loss": 0.0007,
+      "step": 12106
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 1.5980232558139535e-07,
+      "loss": 0.001,
+      "step": 12107
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 1.5979069767441858e-07,
+      "loss": 0.0009,
+      "step": 12108
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 1.5977906976744187e-07,
+      "loss": 0.0031,
+      "step": 12109
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 1.597674418604651e-07,
+      "loss": 0.0009,
+      "step": 12110
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 1.5975581395348837e-07,
+      "loss": 0.0011,
+      "step": 12111
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 1.597441860465116e-07,
+      "loss": 0.5656,
+      "step": 12112
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 1.597325581395349e-07,
+      "loss": 0.0007,
+      "step": 12113
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 1.5972093023255813e-07,
+      "loss": 0.1254,
+      "step": 12114
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 1.597093023255814e-07,
+      "loss": 0.0009,
+      "step": 12115
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.5969767441860462e-07,
+      "loss": 0.0007,
+      "step": 12116
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.596860465116279e-07,
+      "loss": 0.0008,
+      "step": 12117
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.5967441860465117e-07,
+      "loss": 0.0009,
+      "step": 12118
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.596627906976744e-07,
+      "loss": 0.0012,
+      "step": 12119
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.5965116279069767e-07,
+      "loss": 0.0006,
+      "step": 12120
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 1.5963953488372093e-07,
+      "loss": 0.0009,
+      "step": 12121
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 1.596279069767442e-07,
+      "loss": 0.0009,
+      "step": 12122
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 1.5961627906976743e-07,
+      "loss": 0.0019,
+      "step": 12123
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 1.596046511627907e-07,
+      "loss": 0.0011,
+      "step": 12124
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 1.5959302325581393e-07,
+      "loss": 0.0009,
+      "step": 12125
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 1.595813953488372e-07,
+      "loss": 0.0007,
+      "step": 12126
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 1.5956976744186045e-07,
+      "loss": 0.001,
+      "step": 12127
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 1.595581395348837e-07,
+      "loss": 0.0016,
+      "step": 12128
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 1.5954651162790695e-07,
+      "loss": 0.0009,
+      "step": 12129
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 1.5953488372093023e-07,
+      "loss": 0.0007,
+      "step": 12130
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 1.595232558139535e-07,
+      "loss": 0.0011,
+      "step": 12131
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 1.5951162790697673e-07,
+      "loss": 0.001,
+      "step": 12132
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 1.5949999999999997e-07,
+      "loss": 0.0006,
+      "step": 12133
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 1.5948837209302325e-07,
+      "loss": 0.0009,
+      "step": 12134
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 1.5947674418604652e-07,
+      "loss": 0.0009,
+      "step": 12135
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 1.5946511627906975e-07,
+      "loss": 0.0011,
+      "step": 12136
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.59453488372093e-07,
+      "loss": 0.0009,
+      "step": 12137
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.5944186046511627e-07,
+      "loss": 0.001,
+      "step": 12138
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.5943023255813954e-07,
+      "loss": 0.0009,
+      "step": 12139
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.5941860465116277e-07,
+      "loss": 0.001,
+      "step": 12140
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.5940697674418603e-07,
+      "loss": 0.0015,
+      "step": 12141
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 1.5939534883720932e-07,
+      "loss": 0.0013,
+      "step": 12142
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 1.5938372093023256e-07,
+      "loss": 0.0006,
+      "step": 12143
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 1.593720930232558e-07,
+      "loss": 0.0025,
+      "step": 12144
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 1.5936046511627905e-07,
+      "loss": 0.0007,
+      "step": 12145
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 1.5934883720930234e-07,
+      "loss": 0.0008,
+      "step": 12146
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.5933720930232558e-07,
+      "loss": 0.0011,
+      "step": 12147
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.5932558139534884e-07,
+      "loss": 0.0007,
+      "step": 12148
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.5931395348837207e-07,
+      "loss": 0.001,
+      "step": 12149
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.5930232558139536e-07,
+      "loss": 0.0011,
+      "step": 12150
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.592906976744186e-07,
+      "loss": 0.0008,
+      "step": 12151
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.5927906976744186e-07,
+      "loss": 0.0006,
+      "step": 12152
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 1.592674418604651e-07,
+      "loss": 0.0111,
+      "step": 12153
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 1.5925581395348836e-07,
+      "loss": 0.0007,
+      "step": 12154
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 1.5924418604651162e-07,
+      "loss": 2.7148,
+      "step": 12155
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 1.5923255813953488e-07,
+      "loss": 0.001,
+      "step": 12156
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 1.5922093023255811e-07,
+      "loss": 0.0009,
+      "step": 12157
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 1.5920930232558138e-07,
+      "loss": 0.0015,
+      "step": 12158
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 1.5919767441860466e-07,
+      "loss": 0.0008,
+      "step": 12159
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 1.591860465116279e-07,
+      "loss": 0.0008,
+      "step": 12160
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 1.5917441860465116e-07,
+      "loss": 0.001,
+      "step": 12161
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 1.591627906976744e-07,
+      "loss": 0.0008,
+      "step": 12162
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 1.5915116279069768e-07,
+      "loss": 0.0016,
+      "step": 12163
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 1.5913953488372092e-07,
+      "loss": 0.0011,
+      "step": 12164
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 1.5912790697674418e-07,
+      "loss": 0.0008,
+      "step": 12165
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 1.5911627906976742e-07,
+      "loss": 0.0011,
+      "step": 12166
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 1.591046511627907e-07,
+      "loss": 0.0041,
+      "step": 12167
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 1.5909302325581394e-07,
+      "loss": 0.0008,
+      "step": 12168
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 1.590813953488372e-07,
+      "loss": 0.0011,
+      "step": 12169
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 1.5906976744186044e-07,
+      "loss": 0.001,
+      "step": 12170
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 1.5905813953488373e-07,
+      "loss": 0.0007,
+      "step": 12171
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 1.59046511627907e-07,
+      "loss": 0.1939,
+      "step": 12172
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 1.5903488372093022e-07,
+      "loss": 0.0006,
+      "step": 12173
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 1.5902325581395348e-07,
+      "loss": 0.0015,
+      "step": 12174
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 1.5901162790697675e-07,
+      "loss": 0.0009,
+      "step": 12175
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 1.59e-07,
+      "loss": 1.1301,
+      "step": 12176
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 1.5898837209302324e-07,
+      "loss": 0.0006,
+      "step": 12177
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.589767441860465e-07,
+      "loss": 0.001,
+      "step": 12178
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.5896511627906974e-07,
+      "loss": 0.0008,
+      "step": 12179
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.5895348837209303e-07,
+      "loss": 0.0007,
+      "step": 12180
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.5894186046511626e-07,
+      "loss": 0.0006,
+      "step": 12181
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.5893023255813952e-07,
+      "loss": 0.0013,
+      "step": 12182
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.5891860465116276e-07,
+      "loss": 0.0008,
+      "step": 12183
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 1.5890697674418605e-07,
+      "loss": 0.0014,
+      "step": 12184
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 1.588953488372093e-07,
+      "loss": 0.0043,
+      "step": 12185
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 1.5888372093023255e-07,
+      "loss": 0.0007,
+      "step": 12186
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 1.5887209302325578e-07,
+      "loss": 0.0006,
+      "step": 12187
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 1.5886046511627907e-07,
+      "loss": 0.0009,
+      "step": 12188
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 1.5884883720930233e-07,
+      "loss": 0.0007,
+      "step": 12189
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 1.5883720930232557e-07,
+      "loss": 0.0008,
+      "step": 12190
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 1.5882558139534883e-07,
+      "loss": 0.0006,
+      "step": 12191
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 1.588139534883721e-07,
+      "loss": 0.0008,
+      "step": 12192
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 1.5880232558139535e-07,
+      "loss": 0.0008,
+      "step": 12193
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 1.5879069767441859e-07,
+      "loss": 0.0007,
+      "step": 12194
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 1.5877906976744185e-07,
+      "loss": 0.0012,
+      "step": 12195
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 1.5876744186046514e-07,
+      "loss": 0.0009,
+      "step": 12196
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 1.5875581395348837e-07,
+      "loss": 0.0007,
+      "step": 12197
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 1.587441860465116e-07,
+      "loss": 0.0007,
+      "step": 12198
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 1.5873255813953487e-07,
+      "loss": 0.0008,
+      "step": 12199
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 1.5872093023255816e-07,
+      "loss": 1.112,
+      "step": 12200
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 1.587093023255814e-07,
+      "loss": 0.0008,
+      "step": 12201
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 1.5869767441860465e-07,
+      "loss": 0.0009,
+      "step": 12202
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 1.586860465116279e-07,
+      "loss": 0.0011,
+      "step": 12203
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.5867441860465115e-07,
+      "loss": 0.0009,
+      "step": 12204
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.586627906976744e-07,
+      "loss": 0.3233,
+      "step": 12205
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.5865116279069767e-07,
+      "loss": 0.0006,
+      "step": 12206
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.586395348837209e-07,
+      "loss": 0.0008,
+      "step": 12207
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.5862790697674417e-07,
+      "loss": 0.0012,
+      "step": 12208
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 1.5861627906976743e-07,
+      "loss": 0.9914,
+      "step": 12209
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 1.586046511627907e-07,
+      "loss": 0.0008,
+      "step": 12210
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 1.5859302325581393e-07,
+      "loss": 0.3791,
+      "step": 12211
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 1.585813953488372e-07,
+      "loss": 0.0009,
+      "step": 12212
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 1.5856976744186048e-07,
+      "loss": 0.001,
+      "step": 12213
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 1.5855813953488371e-07,
+      "loss": 0.0022,
+      "step": 12214
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.5854651162790698e-07,
+      "loss": 0.0005,
+      "step": 12215
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.585348837209302e-07,
+      "loss": 0.0007,
+      "step": 12216
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.585232558139535e-07,
+      "loss": 0.0006,
+      "step": 12217
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.5851162790697673e-07,
+      "loss": 0.015,
+      "step": 12218
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.585e-07,
+      "loss": 0.0024,
+      "step": 12219
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 1.5848837209302323e-07,
+      "loss": 0.0013,
+      "step": 12220
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 1.5847674418604652e-07,
+      "loss": 0.0013,
+      "step": 12221
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 1.5846511627906976e-07,
+      "loss": 0.0017,
+      "step": 12222
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 1.5845348837209302e-07,
+      "loss": 0.0011,
+      "step": 12223
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 1.5844186046511625e-07,
+      "loss": 0.0006,
+      "step": 12224
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.5843023255813954e-07,
+      "loss": 0.0008,
+      "step": 12225
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.584186046511628e-07,
+      "loss": 0.0008,
+      "step": 12226
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.5840697674418604e-07,
+      "loss": 0.0009,
+      "step": 12227
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.583953488372093e-07,
+      "loss": 0.0011,
+      "step": 12228
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.5838372093023256e-07,
+      "loss": 0.0007,
+      "step": 12229
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 1.5837209302325582e-07,
+      "loss": 0.0007,
+      "step": 12230
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 1.5836046511627906e-07,
+      "loss": 0.0009,
+      "step": 12231
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 1.5834883720930232e-07,
+      "loss": 0.0008,
+      "step": 12232
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 1.5833720930232555e-07,
+      "loss": 0.0007,
+      "step": 12233
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 1.5832558139534884e-07,
+      "loss": 0.0006,
+      "step": 12234
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 1.5831395348837208e-07,
+      "loss": 0.0008,
+      "step": 12235
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 1.5830232558139534e-07,
+      "loss": 0.0006,
+      "step": 12236
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 1.5829069767441857e-07,
+      "loss": 0.0008,
+      "step": 12237
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 1.5827906976744186e-07,
+      "loss": 0.0012,
+      "step": 12238
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 1.5826744186046512e-07,
+      "loss": 0.0612,
+      "step": 12239
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.5825581395348836e-07,
+      "loss": 0.0014,
+      "step": 12240
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.582441860465116e-07,
+      "loss": 0.0009,
+      "step": 12241
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.5823255813953488e-07,
+      "loss": 0.001,
+      "step": 12242
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.5822093023255815e-07,
+      "loss": 0.0011,
+      "step": 12243
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.5820930232558138e-07,
+      "loss": 0.0009,
+      "step": 12244
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.5819767441860464e-07,
+      "loss": 0.0006,
+      "step": 12245
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 1.581860465116279e-07,
+      "loss": 0.0014,
+      "step": 12246
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 1.5817441860465117e-07,
+      "loss": 0.0007,
+      "step": 12247
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 1.581627906976744e-07,
+      "loss": 0.0008,
+      "step": 12248
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 1.5815116279069766e-07,
+      "loss": 0.0013,
+      "step": 12249
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 1.5813953488372092e-07,
+      "loss": 0.0007,
+      "step": 12250
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.5812790697674419e-07,
+      "loss": 0.0009,
+      "step": 12251
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.5811627906976742e-07,
+      "loss": 0.0009,
+      "step": 12252
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.5810465116279068e-07,
+      "loss": 0.0007,
+      "step": 12253
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.5809302325581397e-07,
+      "loss": 0.001,
+      "step": 12254
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.580813953488372e-07,
+      "loss": 0.0008,
+      "step": 12255
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 1.5806976744186047e-07,
+      "loss": 0.0007,
+      "step": 12256
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 1.580581395348837e-07,
+      "loss": 0.0009,
+      "step": 12257
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 1.5804651162790697e-07,
+      "loss": 0.001,
+      "step": 12258
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 1.5803488372093023e-07,
+      "loss": 0.0015,
+      "step": 12259
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 1.580232558139535e-07,
+      "loss": 0.0007,
+      "step": 12260
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.5801162790697672e-07,
+      "loss": 0.0006,
+      "step": 12261
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.5799999999999999e-07,
+      "loss": 0.0007,
+      "step": 12262
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.5798837209302325e-07,
+      "loss": 0.0009,
+      "step": 12263
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.579767441860465e-07,
+      "loss": 0.0006,
+      "step": 12264
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.5796511627906974e-07,
+      "loss": 0.0008,
+      "step": 12265
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 1.57953488372093e-07,
+      "loss": 0.001,
+      "step": 12266
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 1.579418604651163e-07,
+      "loss": 0.0011,
+      "step": 12267
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 1.5793023255813953e-07,
+      "loss": 0.0093,
+      "step": 12268
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 1.579186046511628e-07,
+      "loss": 0.0008,
+      "step": 12269
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 1.5790697674418603e-07,
+      "loss": 0.0006,
+      "step": 12270
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.5789534883720931e-07,
+      "loss": 0.0011,
+      "step": 12271
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.5788372093023255e-07,
+      "loss": 0.0009,
+      "step": 12272
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.578720930232558e-07,
+      "loss": 0.0008,
+      "step": 12273
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.5786046511627905e-07,
+      "loss": 0.0007,
+      "step": 12274
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.5784883720930233e-07,
+      "loss": 0.0011,
+      "step": 12275
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.5783720930232557e-07,
+      "loss": 0.0008,
+      "step": 12276
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 1.5782558139534883e-07,
+      "loss": 1.0737,
+      "step": 12277
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 1.5781395348837207e-07,
+      "loss": 0.0006,
+      "step": 12278
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 1.5780232558139536e-07,
+      "loss": 0.0007,
+      "step": 12279
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 1.5779069767441862e-07,
+      "loss": 0.0008,
+      "step": 12280
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 1.5777906976744185e-07,
+      "loss": 0.0009,
+      "step": 12281
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 1.577674418604651e-07,
+      "loss": 0.0011,
+      "step": 12282
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 1.5775581395348838e-07,
+      "loss": 0.001,
+      "step": 12283
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 1.5774418604651164e-07,
+      "loss": 0.001,
+      "step": 12284
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 1.5773255813953487e-07,
+      "loss": 0.0009,
+      "step": 12285
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 1.5772093023255813e-07,
+      "loss": 0.0008,
+      "step": 12286
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.5770930232558137e-07,
+      "loss": 0.0007,
+      "step": 12287
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.5769767441860466e-07,
+      "loss": 0.0006,
+      "step": 12288
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.576860465116279e-07,
+      "loss": 0.0006,
+      "step": 12289
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.5767441860465115e-07,
+      "loss": 0.0006,
+      "step": 12290
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.576627906976744e-07,
+      "loss": 0.001,
+      "step": 12291
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 1.5765116279069768e-07,
+      "loss": 0.0008,
+      "step": 12292
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 1.576395348837209e-07,
+      "loss": 0.0009,
+      "step": 12293
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 1.5762790697674417e-07,
+      "loss": 0.0027,
+      "step": 12294
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 1.576162790697674e-07,
+      "loss": 0.2909,
+      "step": 12295
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 1.576046511627907e-07,
+      "loss": 0.0012,
+      "step": 12296
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.5759302325581396e-07,
+      "loss": 0.0007,
+      "step": 12297
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.575813953488372e-07,
+      "loss": 3.0665,
+      "step": 12298
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.5756976744186046e-07,
+      "loss": 0.001,
+      "step": 12299
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.5755813953488372e-07,
+      "loss": 0.0007,
+      "step": 12300
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.5754651162790698e-07,
+      "loss": 0.0008,
+      "step": 12301
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.5753488372093022e-07,
+      "loss": 0.0007,
+      "step": 12302
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.5752325581395348e-07,
+      "loss": 0.0008,
+      "step": 12303
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.5751162790697674e-07,
+      "loss": 0.0009,
+      "step": 12304
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.575e-07,
+      "loss": 0.002,
+      "step": 12305
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.5748837209302324e-07,
+      "loss": 0.0008,
+      "step": 12306
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 1.574767441860465e-07,
+      "loss": 0.0014,
+      "step": 12307
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 1.5746511627906979e-07,
+      "loss": 0.0006,
+      "step": 12308
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 1.5745348837209302e-07,
+      "loss": 0.0011,
+      "step": 12309
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 1.5744186046511628e-07,
+      "loss": 0.195,
+      "step": 12310
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 1.5743023255813952e-07,
+      "loss": 0.0008,
+      "step": 12311
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 1.5741860465116278e-07,
+      "loss": 0.0008,
+      "step": 12312
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 1.5740697674418604e-07,
+      "loss": 0.001,
+      "step": 12313
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 1.573953488372093e-07,
+      "loss": 0.0015,
+      "step": 12314
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 1.5738372093023254e-07,
+      "loss": 0.0008,
+      "step": 12315
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 1.573720930232558e-07,
+      "loss": 0.0007,
+      "step": 12316
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 1.5736046511627906e-07,
+      "loss": 1.8051,
+      "step": 12317
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 1.5734883720930232e-07,
+      "loss": 0.0011,
+      "step": 12318
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 1.5733720930232556e-07,
+      "loss": 0.0007,
+      "step": 12319
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 1.5732558139534882e-07,
+      "loss": 0.0007,
+      "step": 12320
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 1.573139534883721e-07,
+      "loss": 0.0009,
+      "step": 12321
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 1.5730232558139534e-07,
+      "loss": 0.0017,
+      "step": 12322
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 1.572906976744186e-07,
+      "loss": 0.0008,
+      "step": 12323
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 1.5727906976744184e-07,
+      "loss": 0.0011,
+      "step": 12324
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 1.5726744186046513e-07,
+      "loss": 0.0035,
+      "step": 12325
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 1.5725581395348836e-07,
+      "loss": 0.0015,
+      "step": 12326
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 1.5724418604651163e-07,
+      "loss": 0.0129,
+      "step": 12327
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.5723255813953486e-07,
+      "loss": 0.0014,
+      "step": 12328
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.5722093023255815e-07,
+      "loss": 0.0008,
+      "step": 12329
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.5720930232558138e-07,
+      "loss": 0.0008,
+      "step": 12330
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.5719767441860465e-07,
+      "loss": 0.001,
+      "step": 12331
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.5718604651162788e-07,
+      "loss": 0.0016,
+      "step": 12332
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.5717441860465117e-07,
+      "loss": 0.0008,
+      "step": 12333
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 1.5716279069767443e-07,
+      "loss": 0.0009,
+      "step": 12334
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 1.5715116279069767e-07,
+      "loss": 0.0007,
+      "step": 12335
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 1.571395348837209e-07,
+      "loss": 0.0018,
+      "step": 12336
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 1.5712790697674416e-07,
+      "loss": 0.0006,
+      "step": 12337
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 1.5711627906976745e-07,
+      "loss": 0.0009,
+      "step": 12338
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.571046511627907e-07,
+      "loss": 0.0006,
+      "step": 12339
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.5709302325581395e-07,
+      "loss": 0.0007,
+      "step": 12340
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.5708139534883718e-07,
+      "loss": 0.001,
+      "step": 12341
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.5706976744186047e-07,
+      "loss": 0.0009,
+      "step": 12342
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 1.570581395348837e-07,
+      "loss": 0.0008,
+      "step": 12343
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 1.5704651162790697e-07,
+      "loss": 0.0009,
+      "step": 12344
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 1.570348837209302e-07,
+      "loss": 0.0013,
+      "step": 12345
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 1.570232558139535e-07,
+      "loss": 0.0007,
+      "step": 12346
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 1.5701162790697673e-07,
+      "loss": 0.0035,
+      "step": 12347
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 1.57e-07,
+      "loss": 0.0008,
+      "step": 12348
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 1.5698837209302322e-07,
+      "loss": 0.0008,
+      "step": 12349
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 1.569767441860465e-07,
+      "loss": 0.0118,
+      "step": 12350
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 1.5696511627906977e-07,
+      "loss": 0.0011,
+      "step": 12351
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 1.56953488372093e-07,
+      "loss": 0.001,
+      "step": 12352
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 1.5694186046511627e-07,
+      "loss": 0.0975,
+      "step": 12353
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 1.5693023255813953e-07,
+      "loss": 0.0007,
+      "step": 12354
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 1.569186046511628e-07,
+      "loss": 0.0009,
+      "step": 12355
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 1.5690697674418603e-07,
+      "loss": 0.001,
+      "step": 12356
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 1.568953488372093e-07,
+      "loss": 0.0007,
+      "step": 12357
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 1.5688372093023255e-07,
+      "loss": 0.0008,
+      "step": 12358
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 1.5687209302325582e-07,
+      "loss": 0.0009,
+      "step": 12359
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 1.5686046511627905e-07,
+      "loss": 0.0008,
+      "step": 12360
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 1.568488372093023e-07,
+      "loss": 0.0008,
+      "step": 12361
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 1.568372093023256e-07,
+      "loss": 0.0014,
+      "step": 12362
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 1.5682558139534884e-07,
+      "loss": 0.0132,
+      "step": 12363
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 1.568139534883721e-07,
+      "loss": 0.0007,
+      "step": 12364
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.5680232558139533e-07,
+      "loss": 0.0011,
+      "step": 12365
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.567906976744186e-07,
+      "loss": 0.0008,
+      "step": 12366
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.5677906976744186e-07,
+      "loss": 0.0012,
+      "step": 12367
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.5676744186046512e-07,
+      "loss": 0.0006,
+      "step": 12368
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.5675581395348835e-07,
+      "loss": 0.0006,
+      "step": 12369
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 1.5674418604651162e-07,
+      "loss": 0.001,
+      "step": 12370
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 1.5673255813953488e-07,
+      "loss": 0.0009,
+      "step": 12371
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 1.5672093023255814e-07,
+      "loss": 0.0007,
+      "step": 12372
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 1.5670930232558137e-07,
+      "loss": 0.0008,
+      "step": 12373
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 1.5669767441860464e-07,
+      "loss": 0.0009,
+      "step": 12374
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.5668604651162792e-07,
+      "loss": 0.0006,
+      "step": 12375
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.5667441860465116e-07,
+      "loss": 2.674,
+      "step": 12376
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.5666279069767442e-07,
+      "loss": 0.0007,
+      "step": 12377
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.5665116279069766e-07,
+      "loss": 0.0006,
+      "step": 12378
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.5663953488372094e-07,
+      "loss": 0.0008,
+      "step": 12379
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 1.5662790697674418e-07,
+      "loss": 0.0007,
+      "step": 12380
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 1.5661627906976744e-07,
+      "loss": 0.0007,
+      "step": 12381
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 1.5660465116279068e-07,
+      "loss": 0.0007,
+      "step": 12382
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 1.5659302325581396e-07,
+      "loss": 0.0009,
+      "step": 12383
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 1.565813953488372e-07,
+      "loss": 0.0011,
+      "step": 12384
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.5656976744186046e-07,
+      "loss": 0.0014,
+      "step": 12385
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.565581395348837e-07,
+      "loss": 0.0009,
+      "step": 12386
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.5654651162790698e-07,
+      "loss": 0.0006,
+      "step": 12387
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.5653488372093022e-07,
+      "loss": 0.0009,
+      "step": 12388
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.5652325581395348e-07,
+      "loss": 0.001,
+      "step": 12389
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.5651162790697672e-07,
+      "loss": 0.0015,
+      "step": 12390
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.5649999999999998e-07,
+      "loss": 0.0007,
+      "step": 12391
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.5648837209302327e-07,
+      "loss": 0.0007,
+      "step": 12392
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.564767441860465e-07,
+      "loss": 0.0006,
+      "step": 12393
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.5646511627906976e-07,
+      "loss": 0.0007,
+      "step": 12394
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.56453488372093e-07,
+      "loss": 0.001,
+      "step": 12395
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 1.564418604651163e-07,
+      "loss": 0.0053,
+      "step": 12396
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 1.5643023255813952e-07,
+      "loss": 0.0007,
+      "step": 12397
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 1.5641860465116278e-07,
+      "loss": 0.0008,
+      "step": 12398
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 1.5640697674418602e-07,
+      "loss": 0.0012,
+      "step": 12399
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 1.563953488372093e-07,
+      "loss": 0.0008,
+      "step": 12400
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.5638372093023254e-07,
+      "loss": 0.0005,
+      "step": 12401
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.563720930232558e-07,
+      "loss": 0.0013,
+      "step": 12402
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.5636046511627904e-07,
+      "loss": 0.0008,
+      "step": 12403
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.5634883720930233e-07,
+      "loss": 0.0012,
+      "step": 12404
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.563372093023256e-07,
+      "loss": 0.001,
+      "step": 12405
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 1.5632558139534882e-07,
+      "loss": 0.0008,
+      "step": 12406
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 1.5631395348837209e-07,
+      "loss": 0.0012,
+      "step": 12407
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 1.5630232558139535e-07,
+      "loss": 0.0007,
+      "step": 12408
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.15161991119384766,
+      "eval_runtime": 1.5555,
+      "eval_samples_per_second": 664.742,
+      "eval_steps_per_second": 10.929,
+      "step": 12408
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 1.562906976744186e-07,
+      "loss": 1.9547,
+      "step": 12409
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 1.5627906976744185e-07,
+      "loss": 0.0007,
+      "step": 12410
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.562674418604651e-07,
+      "loss": 0.0009,
+      "step": 12411
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.5625581395348837e-07,
+      "loss": 0.0009,
+      "step": 12412
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.5624418604651163e-07,
+      "loss": 0.0007,
+      "step": 12413
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.5623255813953487e-07,
+      "loss": 0.0008,
+      "step": 12414
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.5622093023255813e-07,
+      "loss": 0.001,
+      "step": 12415
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 1.5620930232558142e-07,
+      "loss": 0.0007,
+      "step": 12416
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 1.5619767441860465e-07,
+      "loss": 0.0008,
+      "step": 12417
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 1.561860465116279e-07,
+      "loss": 0.0009,
+      "step": 12418
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 1.5617441860465115e-07,
+      "loss": 0.0006,
+      "step": 12419
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 1.5616279069767438e-07,
+      "loss": 0.0011,
+      "step": 12420
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.5615116279069767e-07,
+      "loss": 0.0007,
+      "step": 12421
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.5613953488372093e-07,
+      "loss": 0.0009,
+      "step": 12422
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.5612790697674417e-07,
+      "loss": 0.0027,
+      "step": 12423
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.5611627906976743e-07,
+      "loss": 0.0013,
+      "step": 12424
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.561046511627907e-07,
+      "loss": 0.0011,
+      "step": 12425
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.5609302325581395e-07,
+      "loss": 0.0009,
+      "step": 12426
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 1.560813953488372e-07,
+      "loss": 0.0009,
+      "step": 12427
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 1.5606976744186045e-07,
+      "loss": 0.0007,
+      "step": 12428
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 1.5605813953488374e-07,
+      "loss": 0.0086,
+      "step": 12429
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 1.5604651162790697e-07,
+      "loss": 0.0005,
+      "step": 12430
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 1.560348837209302e-07,
+      "loss": 0.0006,
+      "step": 12431
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 1.5602325581395347e-07,
+      "loss": 0.0009,
+      "step": 12432
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 1.5601162790697676e-07,
+      "loss": 0.001,
+      "step": 12433
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 1.56e-07,
+      "loss": 0.0011,
+      "step": 12434
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 1.5598837209302326e-07,
+      "loss": 0.0009,
+      "step": 12435
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 1.559767441860465e-07,
+      "loss": 0.0008,
+      "step": 12436
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.5596511627906978e-07,
+      "loss": 0.0008,
+      "step": 12437
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.5595348837209301e-07,
+      "loss": 0.0015,
+      "step": 12438
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.5594186046511628e-07,
+      "loss": 0.0009,
+      "step": 12439
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.559302325581395e-07,
+      "loss": 0.0009,
+      "step": 12440
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 1.559186046511628e-07,
+      "loss": 0.0008,
+      "step": 12441
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 1.5590697674418603e-07,
+      "loss": 0.0007,
+      "step": 12442
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 1.558953488372093e-07,
+      "loss": 0.0006,
+      "step": 12443
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 1.5588372093023253e-07,
+      "loss": 0.0007,
+      "step": 12444
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 1.558720930232558e-07,
+      "loss": 0.0094,
+      "step": 12445
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 1.5586046511627908e-07,
+      "loss": 0.0007,
+      "step": 12446
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.5584883720930232e-07,
+      "loss": 0.0006,
+      "step": 12447
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.5583720930232558e-07,
+      "loss": 0.0009,
+      "step": 12448
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.5582558139534881e-07,
+      "loss": 0.0011,
+      "step": 12449
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.558139534883721e-07,
+      "loss": 0.0006,
+      "step": 12450
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.5580232558139534e-07,
+      "loss": 0.0007,
+      "step": 12451
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 1.557906976744186e-07,
+      "loss": 0.0012,
+      "step": 12452
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 1.5577906976744183e-07,
+      "loss": 0.0007,
+      "step": 12453
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 1.5576744186046512e-07,
+      "loss": 0.0009,
+      "step": 12454
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 1.5575581395348836e-07,
+      "loss": 0.0006,
+      "step": 12455
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 1.5574418604651162e-07,
+      "loss": 0.001,
+      "step": 12456
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 1.5573255813953485e-07,
+      "loss": 0.0005,
+      "step": 12457
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 1.5572093023255814e-07,
+      "loss": 0.0006,
+      "step": 12458
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 1.557093023255814e-07,
+      "loss": 0.0014,
+      "step": 12459
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 1.5569767441860464e-07,
+      "loss": 0.0025,
+      "step": 12460
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 1.556860465116279e-07,
+      "loss": 0.0051,
+      "step": 12461
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 1.5567441860465116e-07,
+      "loss": 0.001,
+      "step": 12462
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 1.5566279069767442e-07,
+      "loss": 0.0029,
+      "step": 12463
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 1.5565116279069766e-07,
+      "loss": 0.0006,
+      "step": 12464
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 1.5563953488372092e-07,
+      "loss": 0.02,
+      "step": 12465
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 1.5562790697674418e-07,
+      "loss": 0.0006,
+      "step": 12466
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 1.5561627906976745e-07,
+      "loss": 0.0119,
+      "step": 12467
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 1.5560465116279068e-07,
+      "loss": 0.0009,
+      "step": 12468
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 1.5559302325581394e-07,
+      "loss": 0.0009,
+      "step": 12469
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 1.5558139534883723e-07,
+      "loss": 0.0009,
+      "step": 12470
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 1.5556976744186047e-07,
+      "loss": 0.0039,
+      "step": 12471
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 1.5555813953488373e-07,
+      "loss": 0.0008,
+      "step": 12472
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.5554651162790696e-07,
+      "loss": 0.0008,
+      "step": 12473
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.555348837209302e-07,
+      "loss": 0.0008,
+      "step": 12474
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.5552325581395349e-07,
+      "loss": 0.0008,
+      "step": 12475
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.5551162790697675e-07,
+      "loss": 0.0013,
+      "step": 12476
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 1.5549999999999998e-07,
+      "loss": 0.0007,
+      "step": 12477
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 1.5548837209302324e-07,
+      "loss": 0.0009,
+      "step": 12478
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 1.554767441860465e-07,
+      "loss": 0.0008,
+      "step": 12479
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 1.5546511627906977e-07,
+      "loss": 0.226,
+      "step": 12480
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 1.55453488372093e-07,
+      "loss": 0.0007,
+      "step": 12481
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 1.5544186046511626e-07,
+      "loss": 0.0007,
+      "step": 12482
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 1.5543023255813955e-07,
+      "loss": 0.0009,
+      "step": 12483
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 1.554186046511628e-07,
+      "loss": 0.0008,
+      "step": 12484
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 1.5540697674418602e-07,
+      "loss": 0.0011,
+      "step": 12485
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 1.5539534883720929e-07,
+      "loss": 0.0008,
+      "step": 12486
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 1.5538372093023257e-07,
+      "loss": 0.0005,
+      "step": 12487
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 1.553720930232558e-07,
+      "loss": 0.0008,
+      "step": 12488
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.5536046511627907e-07,
+      "loss": 0.0007,
+      "step": 12489
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.553488372093023e-07,
+      "loss": 0.0007,
+      "step": 12490
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.553372093023256e-07,
+      "loss": 0.0007,
+      "step": 12491
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.5532558139534883e-07,
+      "loss": 0.0009,
+      "step": 12492
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.553139534883721e-07,
+      "loss": 0.0007,
+      "step": 12493
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 1.5530232558139533e-07,
+      "loss": 0.0005,
+      "step": 12494
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 1.5529069767441861e-07,
+      "loss": 0.0008,
+      "step": 12495
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 1.5527906976744185e-07,
+      "loss": 0.0067,
+      "step": 12496
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 1.552674418604651e-07,
+      "loss": 0.0009,
+      "step": 12497
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 1.5525581395348835e-07,
+      "loss": 0.0009,
+      "step": 12498
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 1.552441860465116e-07,
+      "loss": 0.0012,
+      "step": 12499
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 1.552325581395349e-07,
+      "loss": 0.001,
+      "step": 12500
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 1.5522093023255813e-07,
+      "loss": 0.0012,
+      "step": 12501
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 1.552093023255814e-07,
+      "loss": 0.0009,
+      "step": 12502
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 1.5519767441860463e-07,
+      "loss": 0.0007,
+      "step": 12503
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.5518604651162792e-07,
+      "loss": 0.001,
+      "step": 12504
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.5517441860465115e-07,
+      "loss": 0.0006,
+      "step": 12505
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.5516279069767441e-07,
+      "loss": 0.0018,
+      "step": 12506
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.5515116279069765e-07,
+      "loss": 0.0008,
+      "step": 12507
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.5513953488372094e-07,
+      "loss": 0.0011,
+      "step": 12508
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 1.5512790697674417e-07,
+      "loss": 0.0008,
+      "step": 12509
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 1.5511627906976743e-07,
+      "loss": 0.1798,
+      "step": 12510
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 1.5510465116279067e-07,
+      "loss": 0.0011,
+      "step": 12511
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 1.5509302325581396e-07,
+      "loss": 0.0007,
+      "step": 12512
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 1.5508139534883722e-07,
+      "loss": 0.0008,
+      "step": 12513
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.5506976744186045e-07,
+      "loss": 0.0007,
+      "step": 12514
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.550581395348837e-07,
+      "loss": 0.0006,
+      "step": 12515
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.5504651162790698e-07,
+      "loss": 0.001,
+      "step": 12516
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.5503488372093024e-07,
+      "loss": 0.0008,
+      "step": 12517
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.5502325581395347e-07,
+      "loss": 0.0007,
+      "step": 12518
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.5501162790697674e-07,
+      "loss": 0.0008,
+      "step": 12519
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 1.55e-07,
+      "loss": 0.0007,
+      "step": 12520
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 1.5498837209302326e-07,
+      "loss": 0.0009,
+      "step": 12521
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 1.549767441860465e-07,
+      "loss": 0.0006,
+      "step": 12522
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 1.5496511627906976e-07,
+      "loss": 0.0007,
+      "step": 12523
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 1.54953488372093e-07,
+      "loss": 0.0008,
+      "step": 12524
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.5494186046511628e-07,
+      "loss": 0.0007,
+      "step": 12525
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.5493023255813952e-07,
+      "loss": 0.0007,
+      "step": 12526
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.5491860465116278e-07,
+      "loss": 0.0011,
+      "step": 12527
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.54906976744186e-07,
+      "loss": 0.0009,
+      "step": 12528
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.548953488372093e-07,
+      "loss": 0.0008,
+      "step": 12529
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 1.5488372093023256e-07,
+      "loss": 0.004,
+      "step": 12530
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 1.548720930232558e-07,
+      "loss": 0.0008,
+      "step": 12531
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 1.5486046511627906e-07,
+      "loss": 0.0006,
+      "step": 12532
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 1.5484883720930232e-07,
+      "loss": 0.0006,
+      "step": 12533
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 1.5483720930232558e-07,
+      "loss": 0.001,
+      "step": 12534
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 1.5482558139534882e-07,
+      "loss": 0.0018,
+      "step": 12535
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 1.5481395348837208e-07,
+      "loss": 0.0008,
+      "step": 12536
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 1.5480232558139534e-07,
+      "loss": 0.0008,
+      "step": 12537
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 1.547906976744186e-07,
+      "loss": 0.0008,
+      "step": 12538
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 1.5477906976744184e-07,
+      "loss": 0.0012,
+      "step": 12539
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.547674418604651e-07,
+      "loss": 0.0009,
+      "step": 12540
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.547558139534884e-07,
+      "loss": 0.0007,
+      "step": 12541
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.5474418604651162e-07,
+      "loss": 0.0009,
+      "step": 12542
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.5473255813953489e-07,
+      "loss": 0.0009,
+      "step": 12543
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.5472093023255812e-07,
+      "loss": 0.0014,
+      "step": 12544
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.547093023255814e-07,
+      "loss": 0.0007,
+      "step": 12545
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 1.5469767441860464e-07,
+      "loss": 0.0009,
+      "step": 12546
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 1.546860465116279e-07,
+      "loss": 0.7989,
+      "step": 12547
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 1.5467441860465114e-07,
+      "loss": 0.0008,
+      "step": 12548
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 1.5466279069767443e-07,
+      "loss": 0.0008,
+      "step": 12549
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 1.5465116279069766e-07,
+      "loss": 0.0008,
+      "step": 12550
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 1.5463953488372093e-07,
+      "loss": 0.012,
+      "step": 12551
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 1.5462790697674416e-07,
+      "loss": 0.0007,
+      "step": 12552
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 1.5461627906976742e-07,
+      "loss": 0.0026,
+      "step": 12553
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 1.546046511627907e-07,
+      "loss": 0.0006,
+      "step": 12554
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 1.5459302325581395e-07,
+      "loss": 0.0008,
+      "step": 12555
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 1.545813953488372e-07,
+      "loss": 0.0016,
+      "step": 12556
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 1.5456976744186044e-07,
+      "loss": 0.0007,
+      "step": 12557
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 1.5455813953488373e-07,
+      "loss": 0.0006,
+      "step": 12558
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 1.5454651162790697e-07,
+      "loss": 0.0006,
+      "step": 12559
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 1.5453488372093023e-07,
+      "loss": 0.1355,
+      "step": 12560
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 1.5452325581395346e-07,
+      "loss": 0.0007,
+      "step": 12561
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 1.5451162790697675e-07,
+      "loss": 0.0009,
+      "step": 12562
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 1.545e-07,
+      "loss": 0.0007,
+      "step": 12563
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 1.5448837209302325e-07,
+      "loss": 0.0007,
+      "step": 12564
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 1.5447674418604648e-07,
+      "loss": 0.4015,
+      "step": 12565
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 1.5446511627906977e-07,
+      "loss": 0.0006,
+      "step": 12566
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 1.5445348837209303e-07,
+      "loss": 0.0008,
+      "step": 12567
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 1.5444186046511627e-07,
+      "loss": 0.0009,
+      "step": 12568
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 1.544302325581395e-07,
+      "loss": 0.0008,
+      "step": 12569
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 1.544186046511628e-07,
+      "loss": 0.0006,
+      "step": 12570
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.5440697674418605e-07,
+      "loss": 0.0012,
+      "step": 12571
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.543953488372093e-07,
+      "loss": 0.0008,
+      "step": 12572
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.5438372093023255e-07,
+      "loss": 0.0007,
+      "step": 12573
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.543720930232558e-07,
+      "loss": 0.0007,
+      "step": 12574
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.5436046511627907e-07,
+      "loss": 0.0006,
+      "step": 12575
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.543488372093023e-07,
+      "loss": 0.0005,
+      "step": 12576
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 1.5433720930232557e-07,
+      "loss": 0.0009,
+      "step": 12577
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 1.543255813953488e-07,
+      "loss": 0.0006,
+      "step": 12578
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 1.543139534883721e-07,
+      "loss": 0.0007,
+      "step": 12579
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 1.5430232558139533e-07,
+      "loss": 1.7687,
+      "step": 12580
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 1.542906976744186e-07,
+      "loss": 0.0005,
+      "step": 12581
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 1.5427906976744183e-07,
+      "loss": 0.0007,
+      "step": 12582
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 1.5426744186046512e-07,
+      "loss": 0.0006,
+      "step": 12583
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 1.5425581395348838e-07,
+      "loss": 0.0008,
+      "step": 12584
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 1.542441860465116e-07,
+      "loss": 0.0007,
+      "step": 12585
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 1.5423255813953487e-07,
+      "loss": 0.0008,
+      "step": 12586
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 1.5422093023255814e-07,
+      "loss": 0.0014,
+      "step": 12587
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 1.542093023255814e-07,
+      "loss": 0.0007,
+      "step": 12588
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 1.5419767441860463e-07,
+      "loss": 0.0016,
+      "step": 12589
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 1.541860465116279e-07,
+      "loss": 0.0007,
+      "step": 12590
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 1.5417441860465116e-07,
+      "loss": 0.0007,
+      "step": 12591
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 1.5416279069767442e-07,
+      "loss": 0.0265,
+      "step": 12592
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 1.5415116279069765e-07,
+      "loss": 0.001,
+      "step": 12593
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 1.5413953488372091e-07,
+      "loss": 0.0009,
+      "step": 12594
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 1.541279069767442e-07,
+      "loss": 0.0015,
+      "step": 12595
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 1.5411627906976744e-07,
+      "loss": 0.0008,
+      "step": 12596
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 1.541046511627907e-07,
+      "loss": 0.0012,
+      "step": 12597
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 1.5409302325581394e-07,
+      "loss": 0.1383,
+      "step": 12598
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 1.5408139534883722e-07,
+      "loss": 0.001,
+      "step": 12599
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 1.5406976744186046e-07,
+      "loss": 0.0012,
+      "step": 12600
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 1.5405813953488372e-07,
+      "loss": 0.0006,
+      "step": 12601
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 1.5404651162790696e-07,
+      "loss": 0.0009,
+      "step": 12602
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 1.5403488372093024e-07,
+      "loss": 0.0007,
+      "step": 12603
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 1.5402325581395348e-07,
+      "loss": 0.0009,
+      "step": 12604
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 1.5401162790697674e-07,
+      "loss": 0.0007,
+      "step": 12605
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 1.5399999999999998e-07,
+      "loss": 0.0007,
+      "step": 12606
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 1.5398837209302324e-07,
+      "loss": 0.0012,
+      "step": 12607
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.5397674418604653e-07,
+      "loss": 0.0009,
+      "step": 12608
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.5396511627906976e-07,
+      "loss": 0.001,
+      "step": 12609
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.5395348837209302e-07,
+      "loss": 0.0007,
+      "step": 12610
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.5394186046511626e-07,
+      "loss": 0.0008,
+      "step": 12611
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.5393023255813955e-07,
+      "loss": 0.0005,
+      "step": 12612
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 1.5391860465116278e-07,
+      "loss": 0.0024,
+      "step": 12613
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 1.5390697674418604e-07,
+      "loss": 0.0007,
+      "step": 12614
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 1.5389534883720928e-07,
+      "loss": 0.0875,
+      "step": 12615
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 1.5388372093023257e-07,
+      "loss": 0.001,
+      "step": 12616
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 1.538720930232558e-07,
+      "loss": 0.0006,
+      "step": 12617
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 1.5386046511627906e-07,
+      "loss": 0.0015,
+      "step": 12618
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 1.538488372093023e-07,
+      "loss": 0.0006,
+      "step": 12619
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 1.538372093023256e-07,
+      "loss": 0.0007,
+      "step": 12620
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 1.5382558139534885e-07,
+      "loss": 0.0008,
+      "step": 12621
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 1.5381395348837208e-07,
+      "loss": 0.0006,
+      "step": 12622
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 1.5380232558139532e-07,
+      "loss": 0.0009,
+      "step": 12623
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 1.537906976744186e-07,
+      "loss": 0.001,
+      "step": 12624
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 1.5377906976744187e-07,
+      "loss": 0.2443,
+      "step": 12625
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 1.537674418604651e-07,
+      "loss": 0.0006,
+      "step": 12626
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 1.5375581395348837e-07,
+      "loss": 0.0008,
+      "step": 12627
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 1.5374418604651163e-07,
+      "loss": 0.0008,
+      "step": 12628
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 1.537325581395349e-07,
+      "loss": 0.0008,
+      "step": 12629
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 1.5372093023255812e-07,
+      "loss": 0.0007,
+      "step": 12630
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 1.5370930232558139e-07,
+      "loss": 0.0008,
+      "step": 12631
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 1.5369767441860462e-07,
+      "loss": 0.0006,
+      "step": 12632
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 1.536860465116279e-07,
+      "loss": 0.0007,
+      "step": 12633
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 1.5367441860465115e-07,
+      "loss": 0.001,
+      "step": 12634
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 1.536627906976744e-07,
+      "loss": 0.0007,
+      "step": 12635
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 1.5365116279069764e-07,
+      "loss": 0.0008,
+      "step": 12636
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 1.5363953488372093e-07,
+      "loss": 0.0008,
+      "step": 12637
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 1.536279069767442e-07,
+      "loss": 0.0007,
+      "step": 12638
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.5361627906976743e-07,
+      "loss": 0.0008,
+      "step": 12639
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.536046511627907e-07,
+      "loss": 0.0011,
+      "step": 12640
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.5359302325581395e-07,
+      "loss": 0.0007,
+      "step": 12641
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.535813953488372e-07,
+      "loss": 0.0008,
+      "step": 12642
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.5356976744186045e-07,
+      "loss": 0.0031,
+      "step": 12643
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 1.535581395348837e-07,
+      "loss": 0.0009,
+      "step": 12644
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 1.5354651162790697e-07,
+      "loss": 0.0006,
+      "step": 12645
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 1.5353488372093023e-07,
+      "loss": 0.0006,
+      "step": 12646
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 1.5352325581395347e-07,
+      "loss": 0.0077,
+      "step": 12647
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 1.5351162790697673e-07,
+      "loss": 0.0007,
+      "step": 12648
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 1.5350000000000002e-07,
+      "loss": 0.0008,
+      "step": 12649
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 1.5348837209302325e-07,
+      "loss": 0.0011,
+      "step": 12650
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 1.5347674418604651e-07,
+      "loss": 0.0007,
+      "step": 12651
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 1.5346511627906975e-07,
+      "loss": 0.001,
+      "step": 12652
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 1.5345348837209304e-07,
+      "loss": 0.0011,
+      "step": 12653
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 1.5344186046511627e-07,
+      "loss": 0.0009,
+      "step": 12654
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 1.5343023255813954e-07,
+      "loss": 0.0006,
+      "step": 12655
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 1.5341860465116277e-07,
+      "loss": 0.0007,
+      "step": 12656
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 1.5340697674418603e-07,
+      "loss": 0.0008,
+      "step": 12657
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 1.533953488372093e-07,
+      "loss": 0.0006,
+      "step": 12658
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 1.5338372093023256e-07,
+      "loss": 0.0008,
+      "step": 12659
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 1.533720930232558e-07,
+      "loss": 0.0008,
+      "step": 12660
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 1.5336046511627905e-07,
+      "loss": 0.0005,
+      "step": 12661
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 1.5334883720930234e-07,
+      "loss": 0.0008,
+      "step": 12662
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 1.5333720930232558e-07,
+      "loss": 0.0006,
+      "step": 12663
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.533255813953488e-07,
+      "loss": 0.0007,
+      "step": 12664
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.5331395348837207e-07,
+      "loss": 0.0011,
+      "step": 12665
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.5330232558139536e-07,
+      "loss": 0.0008,
+      "step": 12666
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.532906976744186e-07,
+      "loss": 0.0007,
+      "step": 12667
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.5327906976744186e-07,
+      "loss": 0.0034,
+      "step": 12668
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.532674418604651e-07,
+      "loss": 0.001,
+      "step": 12669
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 1.5325581395348838e-07,
+      "loss": 0.0007,
+      "step": 12670
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 1.5324418604651162e-07,
+      "loss": 0.0007,
+      "step": 12671
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 1.5323255813953488e-07,
+      "loss": 0.0007,
+      "step": 12672
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 1.5322093023255811e-07,
+      "loss": 0.0016,
+      "step": 12673
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 1.532093023255814e-07,
+      "loss": 0.0007,
+      "step": 12674
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.5319767441860464e-07,
+      "loss": 0.0008,
+      "step": 12675
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.531860465116279e-07,
+      "loss": 0.0008,
+      "step": 12676
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.5317441860465113e-07,
+      "loss": 0.0012,
+      "step": 12677
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.5316279069767442e-07,
+      "loss": 0.0007,
+      "step": 12678
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.5315116279069768e-07,
+      "loss": 0.0008,
+      "step": 12679
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 1.5313953488372092e-07,
+      "loss": 0.0008,
+      "step": 12680
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 1.5312790697674418e-07,
+      "loss": 0.0009,
+      "step": 12681
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 1.5311627906976744e-07,
+      "loss": 0.0008,
+      "step": 12682
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 1.531046511627907e-07,
+      "loss": 0.0006,
+      "step": 12683
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 1.5309302325581394e-07,
+      "loss": 0.0008,
+      "step": 12684
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 1.530813953488372e-07,
+      "loss": 0.0009,
+      "step": 12685
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 1.5306976744186044e-07,
+      "loss": 0.0009,
+      "step": 12686
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 1.5305813953488372e-07,
+      "loss": 0.0008,
+      "step": 12687
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 1.5304651162790696e-07,
+      "loss": 0.0009,
+      "step": 12688
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 1.5303488372093022e-07,
+      "loss": 0.0009,
+      "step": 12689
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 1.5302325581395346e-07,
+      "loss": 2.0627,
+      "step": 12690
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 1.5301162790697675e-07,
+      "loss": 0.155,
+      "step": 12691
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 1.53e-07,
+      "loss": 0.0007,
+      "step": 12692
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 1.5298837209302324e-07,
+      "loss": 0.0009,
+      "step": 12693
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 1.529767441860465e-07,
+      "loss": 0.0006,
+      "step": 12694
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 1.5296511627906977e-07,
+      "loss": 0.0007,
+      "step": 12695
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 1.5295348837209303e-07,
+      "loss": 0.0013,
+      "step": 12696
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 1.5294186046511626e-07,
+      "loss": 0.7598,
+      "step": 12697
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 1.5293023255813952e-07,
+      "loss": 0.0009,
+      "step": 12698
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 1.5291860465116279e-07,
+      "loss": 0.0009,
+      "step": 12699
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 1.5290697674418605e-07,
+      "loss": 0.0007,
+      "step": 12700
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 1.5289534883720928e-07,
+      "loss": 0.0008,
+      "step": 12701
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 1.5288372093023254e-07,
+      "loss": 2.1857,
+      "step": 12702
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 1.5287209302325583e-07,
+      "loss": 0.0009,
+      "step": 12703
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 1.5286046511627907e-07,
+      "loss": 0.0006,
+      "step": 12704
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 1.5284883720930233e-07,
+      "loss": 0.0005,
+      "step": 12705
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 1.5283720930232556e-07,
+      "loss": 0.0011,
+      "step": 12706
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 1.5282558139534885e-07,
+      "loss": 0.0018,
+      "step": 12707
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 1.528139534883721e-07,
+      "loss": 0.0009,
+      "step": 12708
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 1.5280232558139535e-07,
+      "loss": 0.0009,
+      "step": 12709
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 1.5279069767441859e-07,
+      "loss": 0.0007,
+      "step": 12710
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 1.5277906976744185e-07,
+      "loss": 0.0006,
+      "step": 12711
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 1.527674418604651e-07,
+      "loss": 0.0009,
+      "step": 12712
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 1.5275581395348837e-07,
+      "loss": 0.0005,
+      "step": 12713
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 1.527441860465116e-07,
+      "loss": 2.9203,
+      "step": 12714
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 1.5273255813953487e-07,
+      "loss": 0.1425,
+      "step": 12715
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 1.5272093023255816e-07,
+      "loss": 0.0009,
+      "step": 12716
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 1.527093023255814e-07,
+      "loss": 0.0006,
+      "step": 12717
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 1.5269767441860463e-07,
+      "loss": 0.0008,
+      "step": 12718
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 1.526860465116279e-07,
+      "loss": 0.0005,
+      "step": 12719
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 1.5267441860465118e-07,
+      "loss": 0.0006,
+      "step": 12720
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 1.526627906976744e-07,
+      "loss": 0.0015,
+      "step": 12721
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 1.5265116279069767e-07,
+      "loss": 0.0018,
+      "step": 12722
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 1.526395348837209e-07,
+      "loss": 0.0041,
+      "step": 12723
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 1.526279069767442e-07,
+      "loss": 0.0009,
+      "step": 12724
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 1.5261627906976743e-07,
+      "loss": 0.0008,
+      "step": 12725
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 1.526046511627907e-07,
+      "loss": 0.0008,
+      "step": 12726
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 1.5259302325581393e-07,
+      "loss": 0.0006,
+      "step": 12727
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 1.5258139534883722e-07,
+      "loss": 0.0011,
+      "step": 12728
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 1.5256976744186045e-07,
+      "loss": 0.0696,
+      "step": 12729
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 1.5255813953488371e-07,
+      "loss": 0.0008,
+      "step": 12730
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 1.5254651162790695e-07,
+      "loss": 0.0009,
+      "step": 12731
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 1.5253488372093024e-07,
+      "loss": 0.0007,
+      "step": 12732
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 1.525232558139535e-07,
+      "loss": 0.0008,
+      "step": 12733
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 1.5251162790697673e-07,
+      "loss": 0.0006,
+      "step": 12734
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 1.525e-07,
+      "loss": 0.0006,
+      "step": 12735
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 1.5248837209302326e-07,
+      "loss": 0.0009,
+      "step": 12736
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 1.5247674418604652e-07,
+      "loss": 0.0009,
+      "step": 12737
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 1.5246511627906975e-07,
+      "loss": 0.001,
+      "step": 12738
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 1.5245348837209302e-07,
+      "loss": 0.0007,
+      "step": 12739
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 1.5244186046511625e-07,
+      "loss": 0.0008,
+      "step": 12740
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 1.5243023255813954e-07,
+      "loss": 0.0007,
+      "step": 12741
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 1.5241860465116277e-07,
+      "loss": 0.0009,
+      "step": 12742
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 1.5240697674418604e-07,
+      "loss": 0.001,
+      "step": 12743
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 1.5239534883720927e-07,
+      "loss": 0.001,
+      "step": 12744
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 1.5238372093023256e-07,
+      "loss": 0.0007,
+      "step": 12745
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 1.5237209302325582e-07,
+      "loss": 2.9939,
+      "step": 12746
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 1.5236046511627906e-07,
+      "loss": 0.001,
+      "step": 12747
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 1.5234883720930232e-07,
+      "loss": 0.0006,
+      "step": 12748
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 1.5233720930232558e-07,
+      "loss": 0.0007,
+      "step": 12749
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 1.5232558139534884e-07,
+      "loss": 0.0012,
+      "step": 12750
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 1.5231395348837208e-07,
+      "loss": 0.0007,
+      "step": 12751
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 1.5230232558139534e-07,
+      "loss": 0.0015,
+      "step": 12752
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 1.522906976744186e-07,
+      "loss": 0.0006,
+      "step": 12753
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 1.5227906976744186e-07,
+      "loss": 0.0009,
+      "step": 12754
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 1.522674418604651e-07,
+      "loss": 0.0007,
+      "step": 12755
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 1.5225581395348836e-07,
+      "loss": 0.0005,
+      "step": 12756
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.5224418604651165e-07,
+      "loss": 0.0009,
+      "step": 12757
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.5223255813953488e-07,
+      "loss": 0.0008,
+      "step": 12758
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.5222093023255814e-07,
+      "loss": 0.0011,
+      "step": 12759
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.5220930232558138e-07,
+      "loss": 0.0016,
+      "step": 12760
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.5219767441860467e-07,
+      "loss": 0.0008,
+      "step": 12761
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 1.521860465116279e-07,
+      "loss": 0.0011,
+      "step": 12762
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 1.5217441860465116e-07,
+      "loss": 0.0008,
+      "step": 12763
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 1.521627906976744e-07,
+      "loss": 0.0006,
+      "step": 12764
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 1.5215116279069766e-07,
+      "loss": 0.0013,
+      "step": 12765
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 1.5213953488372092e-07,
+      "loss": 0.0007,
+      "step": 12766
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 1.5212790697674419e-07,
+      "loss": 0.0011,
+      "step": 12767
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 1.5211627906976742e-07,
+      "loss": 0.001,
+      "step": 12768
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 1.5210465116279068e-07,
+      "loss": 0.0006,
+      "step": 12769
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 1.5209302325581394e-07,
+      "loss": 0.0021,
+      "step": 12770
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 1.520813953488372e-07,
+      "loss": 0.001,
+      "step": 12771
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 1.5206976744186044e-07,
+      "loss": 0.0009,
+      "step": 12772
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 1.520581395348837e-07,
+      "loss": 0.0013,
+      "step": 12773
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 1.52046511627907e-07,
+      "loss": 0.0008,
+      "step": 12774
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 1.5203488372093023e-07,
+      "loss": 0.0008,
+      "step": 12775
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 1.520232558139535e-07,
+      "loss": 0.0006,
+      "step": 12776
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 1.5201162790697672e-07,
+      "loss": 0.0009,
+      "step": 12777
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 1.52e-07,
+      "loss": 0.0401,
+      "step": 12778
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 1.5198837209302325e-07,
+      "loss": 0.0008,
+      "step": 12779
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 1.519767441860465e-07,
+      "loss": 0.0007,
+      "step": 12780
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 1.5196511627906974e-07,
+      "loss": 2.3372,
+      "step": 12781
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 1.5195348837209303e-07,
+      "loss": 0.0006,
+      "step": 12782
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 1.5194186046511627e-07,
+      "loss": 0.0006,
+      "step": 12783
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 1.5193023255813953e-07,
+      "loss": 0.0006,
+      "step": 12784
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 1.5191860465116276e-07,
+      "loss": 2.2953,
+      "step": 12785
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 1.5190697674418605e-07,
+      "loss": 0.0009,
+      "step": 12786
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 1.5189534883720931e-07,
+      "loss": 0.0007,
+      "step": 12787
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 1.5188372093023255e-07,
+      "loss": 0.0006,
+      "step": 12788
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 1.518720930232558e-07,
+      "loss": 0.001,
+      "step": 12789
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 1.5186046511627907e-07,
+      "loss": 0.0006,
+      "step": 12790
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 1.5184883720930233e-07,
+      "loss": 0.0008,
+      "step": 12791
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 1.5183720930232557e-07,
+      "loss": 0.0006,
+      "step": 12792
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 1.5182558139534883e-07,
+      "loss": 0.0005,
+      "step": 12793
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 1.5181395348837207e-07,
+      "loss": 0.0006,
+      "step": 12794
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 1.5180232558139535e-07,
+      "loss": 0.0029,
+      "step": 12795
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 1.517906976744186e-07,
+      "loss": 0.0008,
+      "step": 12796
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 1.5177906976744185e-07,
+      "loss": 0.001,
+      "step": 12797
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 1.5176744186046509e-07,
+      "loss": 0.0007,
+      "step": 12798
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 1.5175581395348837e-07,
+      "loss": 0.0009,
+      "step": 12799
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 1.5174418604651164e-07,
+      "loss": 0.0009,
+      "step": 12800
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 1.5173255813953487e-07,
+      "loss": 0.0048,
+      "step": 12801
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 1.517209302325581e-07,
+      "loss": 0.0007,
+      "step": 12802
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 1.517093023255814e-07,
+      "loss": 0.0007,
+      "step": 12803
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 1.5169767441860466e-07,
+      "loss": 0.6226,
+      "step": 12804
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 1.516860465116279e-07,
+      "loss": 0.0007,
+      "step": 12805
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 1.5167441860465115e-07,
+      "loss": 0.0009,
+      "step": 12806
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 1.5166279069767442e-07,
+      "loss": 0.0013,
+      "step": 12807
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 1.5165116279069768e-07,
+      "loss": 0.2405,
+      "step": 12808
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 1.516395348837209e-07,
+      "loss": 0.0008,
+      "step": 12809
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 1.5162790697674417e-07,
+      "loss": 0.0007,
+      "step": 12810
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 1.5161627906976746e-07,
+      "loss": 0.0008,
+      "step": 12811
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 1.516046511627907e-07,
+      "loss": 0.0006,
+      "step": 12812
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 1.5159302325581393e-07,
+      "loss": 0.0033,
+      "step": 12813
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 1.515813953488372e-07,
+      "loss": 0.0007,
+      "step": 12814
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 1.5156976744186048e-07,
+      "loss": 0.0009,
+      "step": 12815
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 1.5155813953488372e-07,
+      "loss": 0.0012,
+      "step": 12816
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 1.5154651162790698e-07,
+      "loss": 0.0008,
+      "step": 12817
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 1.5153488372093021e-07,
+      "loss": 0.0009,
+      "step": 12818
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 1.5152325581395348e-07,
+      "loss": 0.0007,
+      "step": 12819
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 1.5151162790697674e-07,
+      "loss": 0.0006,
+      "step": 12820
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 1.515e-07,
+      "loss": 3.2329,
+      "step": 12821
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 1.5148837209302324e-07,
+      "loss": 0.3516,
+      "step": 12822
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 1.514767441860465e-07,
+      "loss": 0.0008,
+      "step": 12823
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 1.5146511627906976e-07,
+      "loss": 0.0059,
+      "step": 12824
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 1.5145348837209302e-07,
+      "loss": 0.0007,
+      "step": 12825
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 1.5144186046511626e-07,
+      "loss": 0.001,
+      "step": 12826
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 1.5143023255813952e-07,
+      "loss": 0.001,
+      "step": 12827
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 1.514186046511628e-07,
+      "loss": 0.0006,
+      "step": 12828
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 1.5140697674418604e-07,
+      "loss": 0.0005,
+      "step": 12829
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 1.513953488372093e-07,
+      "loss": 0.0006,
+      "step": 12830
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 1.5138372093023254e-07,
+      "loss": 0.4239,
+      "step": 12831
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 1.5137209302325583e-07,
+      "loss": 0.0011,
+      "step": 12832
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 1.5136046511627906e-07,
+      "loss": 0.001,
+      "step": 12833
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 1.5134883720930232e-07,
+      "loss": 0.0035,
+      "step": 12834
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 1.5133720930232556e-07,
+      "loss": 0.0009,
+      "step": 12835
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 1.5132558139534885e-07,
+      "loss": 0.001,
+      "step": 12836
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 1.5131395348837208e-07,
+      "loss": 0.0007,
+      "step": 12837
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 1.5130232558139534e-07,
+      "loss": 0.0017,
+      "step": 12838
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 1.5129069767441858e-07,
+      "loss": 0.0006,
+      "step": 12839
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.5127906976744187e-07,
+      "loss": 0.0009,
+      "step": 12840
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.5126744186046513e-07,
+      "loss": 0.0007,
+      "step": 12841
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.5125581395348836e-07,
+      "loss": 0.0008,
+      "step": 12842
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.5124418604651163e-07,
+      "loss": 0.0008,
+      "step": 12843
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 1.5123255813953486e-07,
+      "loss": 0.0009,
+      "step": 12844
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 1.5122093023255815e-07,
+      "loss": 0.0005,
+      "step": 12845
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 1.5120930232558138e-07,
+      "loss": 0.0007,
+      "step": 12846
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 1.5119767441860465e-07,
+      "loss": 0.0008,
+      "step": 12847
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 1.5118604651162788e-07,
+      "loss": 0.0006,
+      "step": 12848
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 1.5117441860465117e-07,
+      "loss": 3.5328,
+      "step": 12849
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 1.511627906976744e-07,
+      "loss": 0.0007,
+      "step": 12850
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 1.5115116279069767e-07,
+      "loss": 0.0006,
+      "step": 12851
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 1.511395348837209e-07,
+      "loss": 0.0011,
+      "step": 12852
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 1.511279069767442e-07,
+      "loss": 0.0011,
+      "step": 12853
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 1.5111627906976745e-07,
+      "loss": 0.0008,
+      "step": 12854
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 1.5110465116279069e-07,
+      "loss": 0.0009,
+      "step": 12855
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 1.5109302325581392e-07,
+      "loss": 0.0008,
+      "step": 12856
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 1.510813953488372e-07,
+      "loss": 0.0011,
+      "step": 12857
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 1.5106976744186047e-07,
+      "loss": 0.0009,
+      "step": 12858
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 1.510581395348837e-07,
+      "loss": 0.0007,
+      "step": 12859
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 1.5104651162790697e-07,
+      "loss": 0.453,
+      "step": 12860
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 1.5103488372093023e-07,
+      "loss": 0.0007,
+      "step": 12861
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 1.510232558139535e-07,
+      "loss": 0.0007,
+      "step": 12862
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 1.5101162790697673e-07,
+      "loss": 0.0007,
+      "step": 12863
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 1.51e-07,
+      "loss": 0.0013,
+      "step": 12864
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 1.5098837209302328e-07,
+      "loss": 1.3854,
+      "step": 12865
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 1.509767441860465e-07,
+      "loss": 0.0008,
+      "step": 12866
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 1.5096511627906975e-07,
+      "loss": 0.0006,
+      "step": 12867
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 1.50953488372093e-07,
+      "loss": 0.0035,
+      "step": 12868
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 1.509418604651163e-07,
+      "loss": 0.0007,
+      "step": 12869
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 1.5093023255813953e-07,
+      "loss": 0.0009,
+      "step": 12870
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 1.509186046511628e-07,
+      "loss": 0.0009,
+      "step": 12871
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 1.5090697674418603e-07,
+      "loss": 0.0007,
+      "step": 12872
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 1.508953488372093e-07,
+      "loss": 0.0009,
+      "step": 12873
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 1.5088372093023255e-07,
+      "loss": 0.0006,
+      "step": 12874
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 1.5087209302325581e-07,
+      "loss": 0.0011,
+      "step": 12875
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 1.5086046511627905e-07,
+      "loss": 0.0007,
+      "step": 12876
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 1.508488372093023e-07,
+      "loss": 0.0007,
+      "step": 12877
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 1.5083720930232557e-07,
+      "loss": 0.0013,
+      "step": 12878
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 1.5082558139534884e-07,
+      "loss": 0.0207,
+      "step": 12879
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 1.5081395348837207e-07,
+      "loss": 0.0016,
+      "step": 12880
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 1.5080232558139533e-07,
+      "loss": 0.0009,
+      "step": 12881
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 1.5079069767441862e-07,
+      "loss": 0.001,
+      "step": 12882
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 1.5077906976744186e-07,
+      "loss": 0.0008,
+      "step": 12883
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 1.5076744186046512e-07,
+      "loss": 0.0011,
+      "step": 12884
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 1.5075581395348835e-07,
+      "loss": 0.0008,
+      "step": 12885
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 1.5074418604651164e-07,
+      "loss": 0.0009,
+      "step": 12886
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 1.5073255813953488e-07,
+      "loss": 0.0011,
+      "step": 12887
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 1.5072093023255814e-07,
+      "loss": 0.0012,
+      "step": 12888
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 1.5070930232558137e-07,
+      "loss": 0.0008,
+      "step": 12889
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 1.5069767441860466e-07,
+      "loss": 0.0007,
+      "step": 12890
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 1.506860465116279e-07,
+      "loss": 0.0007,
+      "step": 12891
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 1.5067441860465116e-07,
+      "loss": 0.0009,
+      "step": 12892
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 1.506627906976744e-07,
+      "loss": 0.0007,
+      "step": 12893
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 1.5065116279069768e-07,
+      "loss": 0.0007,
+      "step": 12894
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 1.5063953488372094e-07,
+      "loss": 0.0006,
+      "step": 12895
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 1.5062790697674418e-07,
+      "loss": 0.0006,
+      "step": 12896
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 1.5061627906976741e-07,
+      "loss": 0.0006,
+      "step": 12897
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 1.5060465116279068e-07,
+      "loss": 0.0007,
+      "step": 12898
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 1.5059302325581396e-07,
+      "loss": 0.002,
+      "step": 12899
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 1.505813953488372e-07,
+      "loss": 0.0007,
+      "step": 12900
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 1.5056976744186046e-07,
+      "loss": 0.001,
+      "step": 12901
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 1.505581395348837e-07,
+      "loss": 0.0006,
+      "step": 12902
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 1.5054651162790698e-07,
+      "loss": 0.0005,
+      "step": 12903
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 1.5053488372093022e-07,
+      "loss": 0.9909,
+      "step": 12904
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 1.5052325581395348e-07,
+      "loss": 0.0008,
+      "step": 12905
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 1.5051162790697672e-07,
+      "loss": 0.0009,
+      "step": 12906
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.505e-07,
+      "loss": 0.0007,
+      "step": 12907
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.5048837209302324e-07,
+      "loss": 0.0007,
+      "step": 12908
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.504767441860465e-07,
+      "loss": 0.0008,
+      "step": 12909
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.5046511627906974e-07,
+      "loss": 0.0007,
+      "step": 12910
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.5045348837209302e-07,
+      "loss": 0.0008,
+      "step": 12911
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 1.5044186046511629e-07,
+      "loss": 0.0006,
+      "step": 12912
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 1.5043023255813952e-07,
+      "loss": 0.0008,
+      "step": 12913
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 1.5041860465116278e-07,
+      "loss": 0.0006,
+      "step": 12914
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 1.5040697674418605e-07,
+      "loss": 0.0006,
+      "step": 12915
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 1.503953488372093e-07,
+      "loss": 0.0006,
+      "step": 12916
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 1.5038372093023254e-07,
+      "loss": 0.0007,
+      "step": 12917
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 1.503720930232558e-07,
+      "loss": 0.0011,
+      "step": 12918
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 1.5036046511627907e-07,
+      "loss": 0.0006,
+      "step": 12919
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 1.5034883720930233e-07,
+      "loss": 0.0018,
+      "step": 12920
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 1.5033720930232556e-07,
+      "loss": 0.0014,
+      "step": 12921
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 1.5032558139534882e-07,
+      "loss": 0.0006,
+      "step": 12922
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 1.503139534883721e-07,
+      "loss": 0.0008,
+      "step": 12923
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 1.5030232558139535e-07,
+      "loss": 0.0076,
+      "step": 12924
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 1.502906976744186e-07,
+      "loss": 0.0007,
+      "step": 12925
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.1471852958202362,
+      "eval_runtime": 1.566,
+      "eval_samples_per_second": 660.261,
+      "eval_steps_per_second": 10.855,
+      "step": 12925
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 1.5027906976744184e-07,
+      "loss": 0.0008,
+      "step": 12926
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 1.502674418604651e-07,
+      "loss": 0.0007,
+      "step": 12927
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 1.5025581395348837e-07,
+      "loss": 0.0007,
+      "step": 12928
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 1.5024418604651163e-07,
+      "loss": 0.0007,
+      "step": 12929
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 1.5023255813953486e-07,
+      "loss": 0.001,
+      "step": 12930
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 1.5022093023255813e-07,
+      "loss": 0.0007,
+      "step": 12931
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 1.502093023255814e-07,
+      "loss": 0.0008,
+      "step": 12932
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 1.5019767441860465e-07,
+      "loss": 0.0008,
+      "step": 12933
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 1.5018604651162789e-07,
+      "loss": 0.0009,
+      "step": 12934
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 1.5017441860465115e-07,
+      "loss": 0.0013,
+      "step": 12935
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 1.5016279069767444e-07,
+      "loss": 0.0011,
+      "step": 12936
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 1.5015116279069767e-07,
+      "loss": 0.0011,
+      "step": 12937
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.5013953488372093e-07,
+      "loss": 0.0009,
+      "step": 12938
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.5012790697674417e-07,
+      "loss": 0.0031,
+      "step": 12939
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.5011627906976746e-07,
+      "loss": 0.0009,
+      "step": 12940
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.501046511627907e-07,
+      "loss": 0.0008,
+      "step": 12941
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.5009302325581395e-07,
+      "loss": 0.0011,
+      "step": 12942
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 1.500813953488372e-07,
+      "loss": 0.0013,
+      "step": 12943
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 1.5006976744186048e-07,
+      "loss": 0.0014,
+      "step": 12944
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 1.500581395348837e-07,
+      "loss": 0.0009,
+      "step": 12945
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 1.5004651162790697e-07,
+      "loss": 0.0006,
+      "step": 12946
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 1.500348837209302e-07,
+      "loss": 0.0005,
+      "step": 12947
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 1.500232558139535e-07,
+      "loss": 0.0012,
+      "step": 12948
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 1.5001162790697676e-07,
+      "loss": 0.0007,
+      "step": 12949
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 1.5e-07,
+      "loss": 0.001,
+      "step": 12950
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 1.4998837209302323e-07,
+      "loss": 0.001,
+      "step": 12951
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 1.4997674418604652e-07,
+      "loss": 0.0082,
+      "step": 12952
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 1.4996511627906975e-07,
+      "loss": 0.0011,
+      "step": 12953
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.4995348837209301e-07,
+      "loss": 0.0009,
+      "step": 12954
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.4994186046511628e-07,
+      "loss": 0.0007,
+      "step": 12955
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.4993023255813954e-07,
+      "loss": 0.0007,
+      "step": 12956
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.4991860465116277e-07,
+      "loss": 0.0008,
+      "step": 12957
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 1.4990697674418603e-07,
+      "loss": 0.0006,
+      "step": 12958
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.498953488372093e-07,
+      "loss": 0.0008,
+      "step": 12959
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.4988372093023256e-07,
+      "loss": 0.0018,
+      "step": 12960
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.498720930232558e-07,
+      "loss": 1.481,
+      "step": 12961
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.4986046511627905e-07,
+      "loss": 0.0008,
+      "step": 12962
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.4984883720930232e-07,
+      "loss": 0.0007,
+      "step": 12963
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 1.4983720930232558e-07,
+      "loss": 0.0007,
+      "step": 12964
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 1.4982558139534884e-07,
+      "loss": 0.0008,
+      "step": 12965
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 1.498139534883721e-07,
+      "loss": 0.0006,
+      "step": 12966
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 1.4980232558139534e-07,
+      "loss": 0.0051,
+      "step": 12967
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 1.497906976744186e-07,
+      "loss": 0.0127,
+      "step": 12968
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.4977906976744186e-07,
+      "loss": 0.001,
+      "step": 12969
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.4976744186046512e-07,
+      "loss": 0.0013,
+      "step": 12970
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.4975581395348836e-07,
+      "loss": 0.0009,
+      "step": 12971
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.4974418604651162e-07,
+      "loss": 0.006,
+      "step": 12972
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.4973255813953488e-07,
+      "loss": 0.0018,
+      "step": 12973
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 1.4972093023255814e-07,
+      "loss": 0.0006,
+      "step": 12974
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.4970930232558138e-07,
+      "loss": 0.0007,
+      "step": 12975
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.4969767441860467e-07,
+      "loss": 0.0005,
+      "step": 12976
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.496860465116279e-07,
+      "loss": 0.0852,
+      "step": 12977
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.4967441860465114e-07,
+      "loss": 0.0008,
+      "step": 12978
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 1.4966279069767442e-07,
+      "loss": 0.0008,
+      "step": 12979
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 1.4965116279069766e-07,
+      "loss": 0.0005,
+      "step": 12980
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 1.4963953488372092e-07,
+      "loss": 0.0007,
+      "step": 12981
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 1.4962790697674418e-07,
+      "loss": 0.0008,
+      "step": 12982
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 1.4961627906976744e-07,
+      "loss": 0.0011,
+      "step": 12983
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 1.4960465116279068e-07,
+      "loss": 0.0007,
+      "step": 12984
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 1.4959302325581394e-07,
+      "loss": 0.0008,
+      "step": 12985
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 1.495813953488372e-07,
+      "loss": 0.0006,
+      "step": 12986
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 1.4956976744186046e-07,
+      "loss": 0.0007,
+      "step": 12987
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 1.495581395348837e-07,
+      "loss": 0.0014,
+      "step": 12988
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 1.4954651162790696e-07,
+      "loss": 0.0008,
+      "step": 12989
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 1.4953488372093022e-07,
+      "loss": 0.0008,
+      "step": 12990
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 1.4952325581395349e-07,
+      "loss": 0.0008,
+      "step": 12991
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 1.4951162790697675e-07,
+      "loss": 0.001,
+      "step": 12992
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 1.495e-07,
+      "loss": 0.0007,
+      "step": 12993
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 1.4948837209302324e-07,
+      "loss": 0.0012,
+      "step": 12994
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 1.494767441860465e-07,
+      "loss": 0.0008,
+      "step": 12995
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 1.4946511627906977e-07,
+      "loss": 0.0011,
+      "step": 12996
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 1.4945348837209303e-07,
+      "loss": 0.0009,
+      "step": 12997
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 1.4944186046511626e-07,
+      "loss": 0.0012,
+      "step": 12998
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 1.4943023255813953e-07,
+      "loss": 0.0008,
+      "step": 12999
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.494186046511628e-07,
+      "loss": 0.0008,
+      "step": 13000
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.4940697674418605e-07,
+      "loss": 0.0008,
+      "step": 13001
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.4939534883720928e-07,
+      "loss": 0.0008,
+      "step": 13002
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.4938372093023257e-07,
+      "loss": 0.0008,
+      "step": 13003
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.493720930232558e-07,
+      "loss": 0.0008,
+      "step": 13004
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 1.4936046511627904e-07,
+      "loss": 0.001,
+      "step": 13005
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.4934883720930233e-07,
+      "loss": 0.367,
+      "step": 13006
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.4933720930232557e-07,
+      "loss": 0.0007,
+      "step": 13007
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.4932558139534883e-07,
+      "loss": 0.0006,
+      "step": 13008
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.493139534883721e-07,
+      "loss": 0.0007,
+      "step": 13009
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 1.4930232558139535e-07,
+      "loss": 2.0987,
+      "step": 13010
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.492906976744186e-07,
+      "loss": 0.0008,
+      "step": 13011
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.4927906976744185e-07,
+      "loss": 0.0007,
+      "step": 13012
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.492674418604651e-07,
+      "loss": 0.1135,
+      "step": 13013
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.4925581395348837e-07,
+      "loss": 0.0026,
+      "step": 13014
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 1.492441860465116e-07,
+      "loss": 0.0008,
+      "step": 13015
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 1.4923255813953487e-07,
+      "loss": 0.0006,
+      "step": 13016
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 1.4922093023255813e-07,
+      "loss": 0.0007,
+      "step": 13017
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 1.492093023255814e-07,
+      "loss": 0.0007,
+      "step": 13018
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 1.4919767441860463e-07,
+      "loss": 0.0007,
+      "step": 13019
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 1.4918604651162792e-07,
+      "loss": 0.0019,
+      "step": 13020
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 1.4917441860465115e-07,
+      "loss": 0.0007,
+      "step": 13021
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 1.491627906976744e-07,
+      "loss": 0.0007,
+      "step": 13022
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 1.4915116279069767e-07,
+      "loss": 0.0034,
+      "step": 13023
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 1.4913953488372094e-07,
+      "loss": 0.0006,
+      "step": 13024
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 1.4912790697674417e-07,
+      "loss": 0.0011,
+      "step": 13025
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.4911627906976743e-07,
+      "loss": 0.0007,
+      "step": 13026
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.491046511627907e-07,
+      "loss": 0.0009,
+      "step": 13027
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.4909302325581396e-07,
+      "loss": 0.0007,
+      "step": 13028
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.490813953488372e-07,
+      "loss": 0.0006,
+      "step": 13029
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.4906976744186045e-07,
+      "loss": 0.0005,
+      "step": 13030
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.4905813953488372e-07,
+      "loss": 0.0029,
+      "step": 13031
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.4904651162790695e-07,
+      "loss": 0.0008,
+      "step": 13032
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.4903488372093024e-07,
+      "loss": 0.001,
+      "step": 13033
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.4902325581395347e-07,
+      "loss": 0.0007,
+      "step": 13034
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.4901162790697674e-07,
+      "loss": 0.0009,
+      "step": 13035
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 1.49e-07,
+      "loss": 0.0007,
+      "step": 13036
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 1.4898837209302326e-07,
+      "loss": 0.0007,
+      "step": 13037
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 1.489767441860465e-07,
+      "loss": 0.0007,
+      "step": 13038
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 1.4896511627906976e-07,
+      "loss": 0.0012,
+      "step": 13039
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 1.4895348837209302e-07,
+      "loss": 0.0005,
+      "step": 13040
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 1.4894186046511628e-07,
+      "loss": 0.0008,
+      "step": 13041
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.4893023255813951e-07,
+      "loss": 0.0006,
+      "step": 13042
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.4891860465116278e-07,
+      "loss": 0.0006,
+      "step": 13043
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.4890697674418604e-07,
+      "loss": 0.0005,
+      "step": 13044
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.488953488372093e-07,
+      "loss": 0.0007,
+      "step": 13045
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 1.4888372093023254e-07,
+      "loss": 0.0009,
+      "step": 13046
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 1.4887209302325582e-07,
+      "loss": 0.0008,
+      "step": 13047
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 1.4886046511627906e-07,
+      "loss": 0.0007,
+      "step": 13048
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 1.4884883720930232e-07,
+      "loss": 0.1218,
+      "step": 13049
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 1.4883720930232558e-07,
+      "loss": 0.0006,
+      "step": 13050
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 1.4882558139534884e-07,
+      "loss": 0.0007,
+      "step": 13051
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 1.4881395348837208e-07,
+      "loss": 0.0015,
+      "step": 13052
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 1.4880232558139534e-07,
+      "loss": 0.0021,
+      "step": 13053
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 1.487906976744186e-07,
+      "loss": 0.0007,
+      "step": 13054
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 1.4877906976744186e-07,
+      "loss": 0.0006,
+      "step": 13055
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 1.487674418604651e-07,
+      "loss": 0.0005,
+      "step": 13056
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.4875581395348836e-07,
+      "loss": 0.0008,
+      "step": 13057
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.4874418604651162e-07,
+      "loss": 0.0009,
+      "step": 13058
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.4873255813953486e-07,
+      "loss": 0.4599,
+      "step": 13059
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.4872093023255815e-07,
+      "loss": 0.0007,
+      "step": 13060
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.4870930232558138e-07,
+      "loss": 0.0009,
+      "step": 13061
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 1.4869767441860464e-07,
+      "loss": 0.0011,
+      "step": 13062
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 1.486860465116279e-07,
+      "loss": 2.6045,
+      "step": 13063
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 1.4867441860465117e-07,
+      "loss": 0.0007,
+      "step": 13064
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 1.486627906976744e-07,
+      "loss": 0.0096,
+      "step": 13065
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 1.4865116279069766e-07,
+      "loss": 0.0006,
+      "step": 13066
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 1.4863953488372093e-07,
+      "loss": 0.0012,
+      "step": 13067
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 1.486279069767442e-07,
+      "loss": 0.0008,
+      "step": 13068
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 1.4861627906976742e-07,
+      "loss": 0.0005,
+      "step": 13069
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 1.4860465116279068e-07,
+      "loss": 0.0007,
+      "step": 13070
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 1.4859302325581395e-07,
+      "loss": 2.2413,
+      "step": 13071
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 1.485813953488372e-07,
+      "loss": 0.001,
+      "step": 13072
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.4856976744186044e-07,
+      "loss": 0.0008,
+      "step": 13073
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.4855813953488373e-07,
+      "loss": 0.0008,
+      "step": 13074
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.4854651162790697e-07,
+      "loss": 1.6172,
+      "step": 13075
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.4853488372093023e-07,
+      "loss": 0.0006,
+      "step": 13076
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 1.485232558139535e-07,
+      "loss": 0.0006,
+      "step": 13077
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 1.4851162790697675e-07,
+      "loss": 0.0008,
+      "step": 13078
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 1.4849999999999999e-07,
+      "loss": 0.0008,
+      "step": 13079
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 1.4848837209302325e-07,
+      "loss": 0.0006,
+      "step": 13080
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 1.484767441860465e-07,
+      "loss": 0.0025,
+      "step": 13081
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 1.4846511627906977e-07,
+      "loss": 0.0008,
+      "step": 13082
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 1.48453488372093e-07,
+      "loss": 0.0007,
+      "step": 13083
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 1.4844186046511627e-07,
+      "loss": 0.0006,
+      "step": 13084
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 1.4843023255813953e-07,
+      "loss": 0.0005,
+      "step": 13085
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 1.4841860465116277e-07,
+      "loss": 0.0018,
+      "step": 13086
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 1.4840697674418605e-07,
+      "loss": 0.0006,
+      "step": 13087
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.483953488372093e-07,
+      "loss": 0.0014,
+      "step": 13088
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.4838372093023255e-07,
+      "loss": 0.8544,
+      "step": 13089
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.483720930232558e-07,
+      "loss": 0.0006,
+      "step": 13090
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.4836046511627907e-07,
+      "loss": 0.0009,
+      "step": 13091
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.483488372093023e-07,
+      "loss": 0.0005,
+      "step": 13092
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 1.4833720930232557e-07,
+      "loss": 0.0008,
+      "step": 13093
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 1.4832558139534883e-07,
+      "loss": 0.0007,
+      "step": 13094
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 1.483139534883721e-07,
+      "loss": 0.0011,
+      "step": 13095
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 1.4830232558139533e-07,
+      "loss": 0.0006,
+      "step": 13096
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 1.482906976744186e-07,
+      "loss": 0.0007,
+      "step": 13097
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 1.4827906976744185e-07,
+      "loss": 0.0011,
+      "step": 13098
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 1.4826744186046511e-07,
+      "loss": 0.0009,
+      "step": 13099
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 1.4825581395348835e-07,
+      "loss": 0.0008,
+      "step": 13100
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 1.4824418604651164e-07,
+      "loss": 0.0007,
+      "step": 13101
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 1.4823255813953487e-07,
+      "loss": 0.0005,
+      "step": 13102
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 1.4822093023255814e-07,
+      "loss": 0.0006,
+      "step": 13103
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.482093023255814e-07,
+      "loss": 0.0014,
+      "step": 13104
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.4819767441860466e-07,
+      "loss": 0.0007,
+      "step": 13105
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.481860465116279e-07,
+      "loss": 0.001,
+      "step": 13106
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.4817441860465116e-07,
+      "loss": 0.0006,
+      "step": 13107
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 1.4816279069767442e-07,
+      "loss": 0.0012,
+      "step": 13108
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 1.4815116279069768e-07,
+      "loss": 3.2721,
+      "step": 13109
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 1.4813953488372091e-07,
+      "loss": 0.0008,
+      "step": 13110
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 1.4812790697674418e-07,
+      "loss": 0.0051,
+      "step": 13111
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 1.4811627906976744e-07,
+      "loss": 0.0009,
+      "step": 13112
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 1.4810465116279067e-07,
+      "loss": 0.0008,
+      "step": 13113
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 1.4809302325581396e-07,
+      "loss": 0.001,
+      "step": 13114
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 1.480813953488372e-07,
+      "loss": 0.0008,
+      "step": 13115
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 1.4806976744186046e-07,
+      "loss": 0.3613,
+      "step": 13116
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 1.4805813953488372e-07,
+      "loss": 0.0008,
+      "step": 13117
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 1.4804651162790698e-07,
+      "loss": 0.0046,
+      "step": 13118
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.4803488372093022e-07,
+      "loss": 0.0009,
+      "step": 13119
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.4802325581395348e-07,
+      "loss": 0.0008,
+      "step": 13120
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.4801162790697674e-07,
+      "loss": 0.0006,
+      "step": 13121
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.48e-07,
+      "loss": 0.0007,
+      "step": 13122
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.4798837209302324e-07,
+      "loss": 0.0007,
+      "step": 13123
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 1.479767441860465e-07,
+      "loss": 0.0008,
+      "step": 13124
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 1.4796511627906976e-07,
+      "loss": 0.0006,
+      "step": 13125
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 1.4795348837209302e-07,
+      "loss": 0.0007,
+      "step": 13126
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 1.4794186046511626e-07,
+      "loss": 0.0006,
+      "step": 13127
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 1.4793023255813955e-07,
+      "loss": 0.0008,
+      "step": 13128
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 1.4791860465116278e-07,
+      "loss": 0.0008,
+      "step": 13129
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.4790697674418604e-07,
+      "loss": 0.0012,
+      "step": 13130
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.478953488372093e-07,
+      "loss": 0.0006,
+      "step": 13131
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.4788372093023257e-07,
+      "loss": 0.0007,
+      "step": 13132
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.478720930232558e-07,
+      "loss": 0.0006,
+      "step": 13133
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 1.4786046511627906e-07,
+      "loss": 0.0008,
+      "step": 13134
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 1.4784883720930232e-07,
+      "loss": 0.0007,
+      "step": 13135
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 1.4783720930232559e-07,
+      "loss": 0.0006,
+      "step": 13136
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 1.4782558139534882e-07,
+      "loss": 0.0005,
+      "step": 13137
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 1.4781395348837208e-07,
+      "loss": 0.0007,
+      "step": 13138
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 1.4780232558139534e-07,
+      "loss": 0.0008,
+      "step": 13139
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.4779069767441858e-07,
+      "loss": 0.002,
+      "step": 13140
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.4777906976744184e-07,
+      "loss": 0.0006,
+      "step": 13141
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.477674418604651e-07,
+      "loss": 0.0007,
+      "step": 13142
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.4775581395348837e-07,
+      "loss": 0.0006,
+      "step": 13143
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 1.4774418604651163e-07,
+      "loss": 0.0007,
+      "step": 13144
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 1.477325581395349e-07,
+      "loss": 0.0022,
+      "step": 13145
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 1.4772093023255812e-07,
+      "loss": 0.0008,
+      "step": 13146
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 1.4770930232558139e-07,
+      "loss": 0.0008,
+      "step": 13147
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 1.4769767441860465e-07,
+      "loss": 0.0009,
+      "step": 13148
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 1.476860465116279e-07,
+      "loss": 0.0007,
+      "step": 13149
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.4767441860465114e-07,
+      "loss": 0.0007,
+      "step": 13150
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.476627906976744e-07,
+      "loss": 0.0006,
+      "step": 13151
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.4765116279069767e-07,
+      "loss": 0.0007,
+      "step": 13152
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.4763953488372093e-07,
+      "loss": 0.0013,
+      "step": 13153
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.4762790697674416e-07,
+      "loss": 0.0007,
+      "step": 13154
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 1.4761627906976745e-07,
+      "loss": 0.0008,
+      "step": 13155
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 1.476046511627907e-07,
+      "loss": 0.0013,
+      "step": 13156
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 1.4759302325581395e-07,
+      "loss": 0.0007,
+      "step": 13157
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 1.475813953488372e-07,
+      "loss": 0.0006,
+      "step": 13158
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 1.4756976744186047e-07,
+      "loss": 0.0009,
+      "step": 13159
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 1.475581395348837e-07,
+      "loss": 0.0012,
+      "step": 13160
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 1.4754651162790697e-07,
+      "loss": 0.0009,
+      "step": 13161
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 1.4753488372093023e-07,
+      "loss": 0.5604,
+      "step": 13162
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 1.475232558139535e-07,
+      "loss": 0.0008,
+      "step": 13163
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 1.4751162790697673e-07,
+      "loss": 0.0007,
+      "step": 13164
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 1.475e-07,
+      "loss": 0.0007,
+      "step": 13165
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 1.4748837209302325e-07,
+      "loss": 0.0006,
+      "step": 13166
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 1.474767441860465e-07,
+      "loss": 0.0008,
+      "step": 13167
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 1.4746511627906975e-07,
+      "loss": 0.0009,
+      "step": 13168
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 1.47453488372093e-07,
+      "loss": 0.0008,
+      "step": 13169
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 1.4744186046511627e-07,
+      "loss": 0.0017,
+      "step": 13170
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.4743023255813953e-07,
+      "loss": 0.0008,
+      "step": 13171
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.474186046511628e-07,
+      "loss": 0.0009,
+      "step": 13172
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.4740697674418603e-07,
+      "loss": 0.0008,
+      "step": 13173
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.473953488372093e-07,
+      "loss": 0.0012,
+      "step": 13174
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 1.4738372093023255e-07,
+      "loss": 0.0054,
+      "step": 13175
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 1.4737209302325582e-07,
+      "loss": 0.0008,
+      "step": 13176
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 1.4736046511627905e-07,
+      "loss": 0.0008,
+      "step": 13177
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 1.4734883720930231e-07,
+      "loss": 0.0007,
+      "step": 13178
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 1.4733720930232558e-07,
+      "loss": 0.0007,
+      "step": 13179
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 1.4732558139534884e-07,
+      "loss": 0.0006,
+      "step": 13180
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.4731395348837207e-07,
+      "loss": 0.0007,
+      "step": 13181
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.4730232558139536e-07,
+      "loss": 0.001,
+      "step": 13182
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.472906976744186e-07,
+      "loss": 0.0007,
+      "step": 13183
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.4727906976744186e-07,
+      "loss": 0.001,
+      "step": 13184
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.4726744186046512e-07,
+      "loss": 0.0007,
+      "step": 13185
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 1.4725581395348838e-07,
+      "loss": 0.0008,
+      "step": 13186
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.4724418604651162e-07,
+      "loss": 0.0008,
+      "step": 13187
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.4723255813953488e-07,
+      "loss": 0.0006,
+      "step": 13188
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.4722093023255814e-07,
+      "loss": 0.0006,
+      "step": 13189
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.4720930232558137e-07,
+      "loss": 0.0008,
+      "step": 13190
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 1.4719767441860464e-07,
+      "loss": 0.0007,
+      "step": 13191
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 1.471860465116279e-07,
+      "loss": 0.0006,
+      "step": 13192
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 1.4717441860465116e-07,
+      "loss": 0.0006,
+      "step": 13193
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 1.471627906976744e-07,
+      "loss": 0.0007,
+      "step": 13194
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 1.4715116279069766e-07,
+      "loss": 0.0008,
+      "step": 13195
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 1.4713953488372092e-07,
+      "loss": 0.0009,
+      "step": 13196
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 1.4712790697674418e-07,
+      "loss": 0.0008,
+      "step": 13197
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 1.4711627906976744e-07,
+      "loss": 0.0009,
+      "step": 13198
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 1.471046511627907e-07,
+      "loss": 0.0007,
+      "step": 13199
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 1.4709302325581394e-07,
+      "loss": 0.0006,
+      "step": 13200
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 1.470813953488372e-07,
+      "loss": 0.001,
+      "step": 13201
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 1.4706976744186046e-07,
+      "loss": 0.0008,
+      "step": 13202
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 1.4705813953488372e-07,
+      "loss": 0.0229,
+      "step": 13203
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 1.4704651162790696e-07,
+      "loss": 0.0009,
+      "step": 13204
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 1.4703488372093022e-07,
+      "loss": 0.0007,
+      "step": 13205
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 1.4702325581395348e-07,
+      "loss": 0.002,
+      "step": 13206
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.4701162790697674e-07,
+      "loss": 0.0005,
+      "step": 13207
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.4699999999999998e-07,
+      "loss": 0.001,
+      "step": 13208
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.4698837209302327e-07,
+      "loss": 0.0007,
+      "step": 13209
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.469767441860465e-07,
+      "loss": 0.1235,
+      "step": 13210
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 1.4696511627906976e-07,
+      "loss": 0.0008,
+      "step": 13211
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.4695348837209303e-07,
+      "loss": 0.0013,
+      "step": 13212
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.469418604651163e-07,
+      "loss": 0.0006,
+      "step": 13213
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.4693023255813952e-07,
+      "loss": 0.0005,
+      "step": 13214
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.4691860465116279e-07,
+      "loss": 0.0008,
+      "step": 13215
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.4690697674418605e-07,
+      "loss": 0.002,
+      "step": 13216
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 1.4689534883720928e-07,
+      "loss": 0.0008,
+      "step": 13217
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 1.4688372093023254e-07,
+      "loss": 0.0011,
+      "step": 13218
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 1.468720930232558e-07,
+      "loss": 0.0009,
+      "step": 13219
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 1.4686046511627907e-07,
+      "loss": 0.0007,
+      "step": 13220
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 1.468488372093023e-07,
+      "loss": 0.0006,
+      "step": 13221
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 1.4683720930232556e-07,
+      "loss": 0.0011,
+      "step": 13222
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 1.4682558139534883e-07,
+      "loss": 0.0008,
+      "step": 13223
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 1.468139534883721e-07,
+      "loss": 0.0007,
+      "step": 13224
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 1.4680232558139535e-07,
+      "loss": 0.0006,
+      "step": 13225
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 1.467906976744186e-07,
+      "loss": 1.4588,
+      "step": 13226
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 1.4677906976744185e-07,
+      "loss": 0.0009,
+      "step": 13227
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 1.467674418604651e-07,
+      "loss": 0.0006,
+      "step": 13228
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 1.4675581395348837e-07,
+      "loss": 0.0006,
+      "step": 13229
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 1.4674418604651163e-07,
+      "loss": 0.0005,
+      "step": 13230
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 1.4673255813953487e-07,
+      "loss": 0.0005,
+      "step": 13231
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 1.4672093023255813e-07,
+      "loss": 0.0006,
+      "step": 13232
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 1.467093023255814e-07,
+      "loss": 0.0007,
+      "step": 13233
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 1.4669767441860465e-07,
+      "loss": 0.0008,
+      "step": 13234
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 1.466860465116279e-07,
+      "loss": 0.0006,
+      "step": 13235
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 1.4667441860465118e-07,
+      "loss": 0.0007,
+      "step": 13236
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 1.466627906976744e-07,
+      "loss": 0.0058,
+      "step": 13237
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.4665116279069767e-07,
+      "loss": 0.0008,
+      "step": 13238
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.4663953488372093e-07,
+      "loss": 0.0007,
+      "step": 13239
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.466279069767442e-07,
+      "loss": 0.0008,
+      "step": 13240
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.4661627906976743e-07,
+      "loss": 0.0007,
+      "step": 13241
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 1.466046511627907e-07,
+      "loss": 0.0007,
+      "step": 13242
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.4659302325581395e-07,
+      "loss": 0.0067,
+      "step": 13243
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.465813953488372e-07,
+      "loss": 0.0007,
+      "step": 13244
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.4656976744186045e-07,
+      "loss": 0.0031,
+      "step": 13245
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.465581395348837e-07,
+      "loss": 0.0006,
+      "step": 13246
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.4654651162790697e-07,
+      "loss": 2.6886,
+      "step": 13247
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 1.465348837209302e-07,
+      "loss": 0.0009,
+      "step": 13248
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.4652325581395347e-07,
+      "loss": 0.0008,
+      "step": 13249
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.4651162790697673e-07,
+      "loss": 0.0007,
+      "step": 13250
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.465e-07,
+      "loss": 0.0006,
+      "step": 13251
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.4648837209302326e-07,
+      "loss": 0.0006,
+      "step": 13252
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 1.4647674418604652e-07,
+      "loss": 0.0006,
+      "step": 13253
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 1.4646511627906975e-07,
+      "loss": 0.0008,
+      "step": 13254
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 1.4645348837209302e-07,
+      "loss": 0.0013,
+      "step": 13255
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 1.4644186046511628e-07,
+      "loss": 0.0007,
+      "step": 13256
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 1.4643023255813954e-07,
+      "loss": 0.0131,
+      "step": 13257
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 1.4641860465116277e-07,
+      "loss": 0.0006,
+      "step": 13258
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.4640697674418604e-07,
+      "loss": 0.0006,
+      "step": 13259
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.463953488372093e-07,
+      "loss": 0.0007,
+      "step": 13260
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.4638372093023256e-07,
+      "loss": 0.0015,
+      "step": 13261
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.463720930232558e-07,
+      "loss": 0.1088,
+      "step": 13262
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 1.4636046511627908e-07,
+      "loss": 0.0009,
+      "step": 13263
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 1.4634883720930232e-07,
+      "loss": 0.0005,
+      "step": 13264
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 1.4633720930232558e-07,
+      "loss": 0.0006,
+      "step": 13265
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 1.4632558139534884e-07,
+      "loss": 0.0008,
+      "step": 13266
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 1.463139534883721e-07,
+      "loss": 0.0013,
+      "step": 13267
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 1.4630232558139534e-07,
+      "loss": 1.1094,
+      "step": 13268
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 1.462906976744186e-07,
+      "loss": 0.0016,
+      "step": 13269
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 1.4627906976744186e-07,
+      "loss": 0.0009,
+      "step": 13270
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 1.462674418604651e-07,
+      "loss": 0.0008,
+      "step": 13271
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 1.4625581395348836e-07,
+      "loss": 0.0008,
+      "step": 13272
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 1.4624418604651162e-07,
+      "loss": 0.0006,
+      "step": 13273
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.4623255813953488e-07,
+      "loss": 0.2265,
+      "step": 13274
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.4622093023255812e-07,
+      "loss": 0.0009,
+      "step": 13275
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.4620930232558138e-07,
+      "loss": 0.0007,
+      "step": 13276
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.4619767441860464e-07,
+      "loss": 0.0007,
+      "step": 13277
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.461860465116279e-07,
+      "loss": 0.0007,
+      "step": 13278
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 1.4617441860465114e-07,
+      "loss": 0.0007,
+      "step": 13279
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 1.4616279069767443e-07,
+      "loss": 0.0015,
+      "step": 13280
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 1.4615116279069766e-07,
+      "loss": 0.0013,
+      "step": 13281
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 1.4613953488372092e-07,
+      "loss": 0.0017,
+      "step": 13282
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 1.4612790697674418e-07,
+      "loss": 0.0007,
+      "step": 13283
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 1.4611627906976745e-07,
+      "loss": 0.0009,
+      "step": 13284
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 1.4610465116279068e-07,
+      "loss": 0.0008,
+      "step": 13285
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 1.4609302325581394e-07,
+      "loss": 0.0009,
+      "step": 13286
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 1.460813953488372e-07,
+      "loss": 0.0007,
+      "step": 13287
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 1.4606976744186047e-07,
+      "loss": 0.001,
+      "step": 13288
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 1.460581395348837e-07,
+      "loss": 0.0008,
+      "step": 13289
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 1.4604651162790696e-07,
+      "loss": 0.0006,
+      "step": 13290
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 1.4603488372093023e-07,
+      "loss": 0.0007,
+      "step": 13291
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 1.460232558139535e-07,
+      "loss": 0.0009,
+      "step": 13292
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 1.4601162790697675e-07,
+      "loss": 0.0029,
+      "step": 13293
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 1.46e-07,
+      "loss": 0.0007,
+      "step": 13294
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 1.4598837209302325e-07,
+      "loss": 0.0008,
+      "step": 13295
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 1.459767441860465e-07,
+      "loss": 0.0007,
+      "step": 13296
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 1.4596511627906977e-07,
+      "loss": 0.0012,
+      "step": 13297
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 1.45953488372093e-07,
+      "loss": 0.001,
+      "step": 13298
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 1.4594186046511627e-07,
+      "loss": 0.0007,
+      "step": 13299
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 1.4593023255813953e-07,
+      "loss": 0.0007,
+      "step": 13300
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 1.459186046511628e-07,
+      "loss": 0.0007,
+      "step": 13301
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 1.4590697674418602e-07,
+      "loss": 0.0009,
+      "step": 13302
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 1.4589534883720929e-07,
+      "loss": 0.0006,
+      "step": 13303
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 1.4588372093023255e-07,
+      "loss": 0.0007,
+      "step": 13304
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.458720930232558e-07,
+      "loss": 0.0006,
+      "step": 13305
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.4586046511627904e-07,
+      "loss": 0.001,
+      "step": 13306
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.4584883720930233e-07,
+      "loss": 0.0218,
+      "step": 13307
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.4583720930232557e-07,
+      "loss": 0.0013,
+      "step": 13308
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.4582558139534883e-07,
+      "loss": 0.0006,
+      "step": 13309
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 1.458139534883721e-07,
+      "loss": 0.0007,
+      "step": 13310
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 1.4580232558139535e-07,
+      "loss": 0.0011,
+      "step": 13311
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 1.457906976744186e-07,
+      "loss": 0.0793,
+      "step": 13312
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 1.4577906976744185e-07,
+      "loss": 0.0008,
+      "step": 13313
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 1.457674418604651e-07,
+      "loss": 0.0008,
+      "step": 13314
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 1.4575581395348837e-07,
+      "loss": 0.0006,
+      "step": 13315
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 1.457441860465116e-07,
+      "loss": 0.0495,
+      "step": 13316
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 1.4573255813953487e-07,
+      "loss": 0.0007,
+      "step": 13317
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 1.4572093023255813e-07,
+      "loss": 0.0006,
+      "step": 13318
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 1.457093023255814e-07,
+      "loss": 0.0009,
+      "step": 13319
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 1.4569767441860466e-07,
+      "loss": 0.0024,
+      "step": 13320
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 1.4568604651162792e-07,
+      "loss": 0.0031,
+      "step": 13321
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 1.4567441860465115e-07,
+      "loss": 0.001,
+      "step": 13322
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 1.4566279069767441e-07,
+      "loss": 0.0007,
+      "step": 13323
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 1.4565116279069768e-07,
+      "loss": 0.0008,
+      "step": 13324
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 1.456395348837209e-07,
+      "loss": 0.0007,
+      "step": 13325
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 1.4562790697674417e-07,
+      "loss": 0.0009,
+      "step": 13326
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 1.4561627906976744e-07,
+      "loss": 0.0044,
+      "step": 13327
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 1.456046511627907e-07,
+      "loss": 0.0009,
+      "step": 13328
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 1.4559302325581393e-07,
+      "loss": 0.0013,
+      "step": 13329
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 1.455813953488372e-07,
+      "loss": 0.0009,
+      "step": 13330
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.4556976744186046e-07,
+      "loss": 0.0009,
+      "step": 13331
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.4555813953488372e-07,
+      "loss": 0.0008,
+      "step": 13332
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.4554651162790695e-07,
+      "loss": 0.0008,
+      "step": 13333
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.4553488372093024e-07,
+      "loss": 0.0006,
+      "step": 13334
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.4552325581395348e-07,
+      "loss": 0.0006,
+      "step": 13335
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 1.4551162790697674e-07,
+      "loss": 0.0007,
+      "step": 13336
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.455e-07,
+      "loss": 0.0008,
+      "step": 13337
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.4548837209302326e-07,
+      "loss": 0.1151,
+      "step": 13338
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.454767441860465e-07,
+      "loss": 0.0008,
+      "step": 13339
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.4546511627906976e-07,
+      "loss": 0.0015,
+      "step": 13340
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 1.4545348837209302e-07,
+      "loss": 0.0008,
+      "step": 13341
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.4544186046511628e-07,
+      "loss": 2.5433,
+      "step": 13342
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.4543023255813952e-07,
+      "loss": 0.0008,
+      "step": 13343
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.4541860465116278e-07,
+      "loss": 0.1113,
+      "step": 13344
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.4540697674418604e-07,
+      "loss": 0.0044,
+      "step": 13345
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 1.453953488372093e-07,
+      "loss": 0.0048,
+      "step": 13346
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 1.4538372093023256e-07,
+      "loss": 0.836,
+      "step": 13347
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 1.4537209302325583e-07,
+      "loss": 0.0008,
+      "step": 13348
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 1.4536046511627906e-07,
+      "loss": 0.0008,
+      "step": 13349
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 1.4534883720930232e-07,
+      "loss": 0.0009,
+      "step": 13350
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 1.4533720930232558e-07,
+      "loss": 0.0008,
+      "step": 13351
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 1.4532558139534882e-07,
+      "loss": 0.0986,
+      "step": 13352
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 1.4531395348837208e-07,
+      "loss": 0.0007,
+      "step": 13353
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 1.4530232558139534e-07,
+      "loss": 0.0014,
+      "step": 13354
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 1.452906976744186e-07,
+      "loss": 0.0008,
+      "step": 13355
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 1.4527906976744184e-07,
+      "loss": 0.0008,
+      "step": 13356
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 1.452674418604651e-07,
+      "loss": 0.0007,
+      "step": 13357
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 1.4525581395348836e-07,
+      "loss": 0.0007,
+      "step": 13358
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 1.4524418604651162e-07,
+      "loss": 0.0006,
+      "step": 13359
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 1.4523255813953486e-07,
+      "loss": 0.0006,
+      "step": 13360
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 1.4522093023255815e-07,
+      "loss": 0.0007,
+      "step": 13361
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.4520930232558138e-07,
+      "loss": 0.0009,
+      "step": 13362
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.4519767441860464e-07,
+      "loss": 0.0007,
+      "step": 13363
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.451860465116279e-07,
+      "loss": 0.0005,
+      "step": 13364
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.4517441860465117e-07,
+      "loss": 0.0008,
+      "step": 13365
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.451627906976744e-07,
+      "loss": 0.0013,
+      "step": 13366
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 1.4515116279069767e-07,
+      "loss": 0.0008,
+      "step": 13367
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.4513953488372093e-07,
+      "loss": 2.4613,
+      "step": 13368
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.451279069767442e-07,
+      "loss": 0.001,
+      "step": 13369
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.4511627906976742e-07,
+      "loss": 0.001,
+      "step": 13370
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.4510465116279069e-07,
+      "loss": 0.0007,
+      "step": 13371
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 1.4509302325581395e-07,
+      "loss": 0.0009,
+      "step": 13372
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.450813953488372e-07,
+      "loss": 0.0009,
+      "step": 13373
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.4506976744186047e-07,
+      "loss": 0.0008,
+      "step": 13374
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.4505813953488373e-07,
+      "loss": 0.0013,
+      "step": 13375
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.4504651162790697e-07,
+      "loss": 0.0007,
+      "step": 13376
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 1.4503488372093023e-07,
+      "loss": 0.0009,
+      "step": 13377
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 1.450232558139535e-07,
+      "loss": 0.0009,
+      "step": 13378
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 1.4501162790697673e-07,
+      "loss": 0.0007,
+      "step": 13379
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 1.45e-07,
+      "loss": 0.0007,
+      "step": 13380
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 1.4498837209302325e-07,
+      "loss": 0.0032,
+      "step": 13381
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 1.449767441860465e-07,
+      "loss": 2.8738,
+      "step": 13382
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 1.4496511627906975e-07,
+      "loss": 0.0008,
+      "step": 13383
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 1.44953488372093e-07,
+      "loss": 0.001,
+      "step": 13384
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 1.4494186046511627e-07,
+      "loss": 0.0007,
+      "step": 13385
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 1.4493023255813953e-07,
+      "loss": 0.0008,
+      "step": 13386
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 1.4491860465116277e-07,
+      "loss": 0.0006,
+      "step": 13387
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 1.4490697674418606e-07,
+      "loss": 0.0008,
+      "step": 13388
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 1.448953488372093e-07,
+      "loss": 0.0007,
+      "step": 13389
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 1.4488372093023255e-07,
+      "loss": 0.0006,
+      "step": 13390
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 1.4487209302325581e-07,
+      "loss": 0.7314,
+      "step": 13391
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 1.4486046511627908e-07,
+      "loss": 0.001,
+      "step": 13392
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.448488372093023e-07,
+      "loss": 0.0009,
+      "step": 13393
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.4483720930232557e-07,
+      "loss": 0.0007,
+      "step": 13394
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.4482558139534883e-07,
+      "loss": 0.0008,
+      "step": 13395
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.448139534883721e-07,
+      "loss": 0.001,
+      "step": 13396
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.4480232558139533e-07,
+      "loss": 0.0009,
+      "step": 13397
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 1.447906976744186e-07,
+      "loss": 0.0009,
+      "step": 13398
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 1.4477906976744185e-07,
+      "loss": 0.0008,
+      "step": 13399
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 1.4476744186046512e-07,
+      "loss": 0.0012,
+      "step": 13400
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 1.4475581395348835e-07,
+      "loss": 0.0006,
+      "step": 13401
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 1.4474418604651164e-07,
+      "loss": 0.0008,
+      "step": 13402
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 1.4473255813953488e-07,
+      "loss": 0.0039,
+      "step": 13403
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.4472093023255814e-07,
+      "loss": 0.0006,
+      "step": 13404
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.447093023255814e-07,
+      "loss": 0.0006,
+      "step": 13405
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.4469767441860463e-07,
+      "loss": 0.0007,
+      "step": 13406
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.446860465116279e-07,
+      "loss": 0.0007,
+      "step": 13407
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 1.4467441860465116e-07,
+      "loss": 0.0006,
+      "step": 13408
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 1.4466279069767442e-07,
+      "loss": 0.0008,
+      "step": 13409
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 1.4465116279069765e-07,
+      "loss": 0.0006,
+      "step": 13410
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 1.4463953488372092e-07,
+      "loss": 0.0006,
+      "step": 13411
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 1.4462790697674418e-07,
+      "loss": 0.0008,
+      "step": 13412
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 1.4461627906976744e-07,
+      "loss": 0.0007,
+      "step": 13413
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 1.4460465116279067e-07,
+      "loss": 0.0007,
+      "step": 13414
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 1.4459302325581396e-07,
+      "loss": 0.0008,
+      "step": 13415
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 1.445813953488372e-07,
+      "loss": 0.0008,
+      "step": 13416
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 1.4456976744186046e-07,
+      "loss": 0.0032,
+      "step": 13417
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 1.4455813953488372e-07,
+      "loss": 0.0008,
+      "step": 13418
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 1.4454651162790698e-07,
+      "loss": 0.0009,
+      "step": 13419
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 1.4453488372093022e-07,
+      "loss": 0.0008,
+      "step": 13420
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 1.4452325581395348e-07,
+      "loss": 0.0007,
+      "step": 13421
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 1.4451162790697674e-07,
+      "loss": 0.0008,
+      "step": 13422
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 1.445e-07,
+      "loss": 0.0005,
+      "step": 13423
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.4448837209302324e-07,
+      "loss": 0.0007,
+      "step": 13424
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.444767441860465e-07,
+      "loss": 0.0011,
+      "step": 13425
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.4446511627906976e-07,
+      "loss": 0.0007,
+      "step": 13426
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.4445348837209302e-07,
+      "loss": 0.0006,
+      "step": 13427
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.4444186046511626e-07,
+      "loss": 0.0005,
+      "step": 13428
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 1.4443023255813955e-07,
+      "loss": 0.0009,
+      "step": 13429
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 1.4441860465116278e-07,
+      "loss": 0.0008,
+      "step": 13430
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 1.4440697674418604e-07,
+      "loss": 0.0011,
+      "step": 13431
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 1.443953488372093e-07,
+      "loss": 0.0008,
+      "step": 13432
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 1.4438372093023254e-07,
+      "loss": 0.0008,
+      "step": 13433
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 1.443720930232558e-07,
+      "loss": 0.0008,
+      "step": 13434
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 1.4436046511627906e-07,
+      "loss": 0.0006,
+      "step": 13435
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 1.4434883720930233e-07,
+      "loss": 0.0006,
+      "step": 13436
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 1.4433720930232556e-07,
+      "loss": 0.0654,
+      "step": 13437
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 1.4432558139534882e-07,
+      "loss": 0.0008,
+      "step": 13438
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 1.4431395348837209e-07,
+      "loss": 0.0006,
+      "step": 13439
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 1.4430232558139535e-07,
+      "loss": 0.0008,
+      "step": 13440
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 1.4429069767441858e-07,
+      "loss": 0.0007,
+      "step": 13441
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 1.4427906976744187e-07,
+      "loss": 0.0005,
+      "step": 13442
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.14857108891010284,
+      "eval_runtime": 1.5632,
+      "eval_samples_per_second": 661.449,
+      "eval_steps_per_second": 10.875,
+      "step": 13442
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 1.442674418604651e-07,
+      "loss": 0.0021,
+      "step": 13443
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 1.4425581395348837e-07,
+      "loss": 0.0006,
+      "step": 13444
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 1.4424418604651163e-07,
+      "loss": 0.0009,
+      "step": 13445
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 1.442325581395349e-07,
+      "loss": 0.0008,
+      "step": 13446
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 1.4422093023255813e-07,
+      "loss": 0.0005,
+      "step": 13447
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 1.442093023255814e-07,
+      "loss": 0.0007,
+      "step": 13448
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 1.4419767441860465e-07,
+      "loss": 0.0007,
+      "step": 13449
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 1.441860465116279e-07,
+      "loss": 0.0007,
+      "step": 13450
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 1.4417441860465115e-07,
+      "loss": 0.0012,
+      "step": 13451
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 1.441627906976744e-07,
+      "loss": 0.0009,
+      "step": 13452
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 1.4415116279069767e-07,
+      "loss": 0.0007,
+      "step": 13453
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 1.4413953488372093e-07,
+      "loss": 0.0011,
+      "step": 13454
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.4412790697674417e-07,
+      "loss": 0.0006,
+      "step": 13455
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.4411627906976745e-07,
+      "loss": 0.0007,
+      "step": 13456
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.441046511627907e-07,
+      "loss": 0.001,
+      "step": 13457
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.4409302325581395e-07,
+      "loss": 0.0006,
+      "step": 13458
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.4408139534883721e-07,
+      "loss": 0.0009,
+      "step": 13459
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 1.4406976744186045e-07,
+      "loss": 0.0006,
+      "step": 13460
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 1.440581395348837e-07,
+      "loss": 0.0009,
+      "step": 13461
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 1.4404651162790697e-07,
+      "loss": 0.0008,
+      "step": 13462
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 1.4403488372093023e-07,
+      "loss": 0.0006,
+      "step": 13463
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 1.4402325581395347e-07,
+      "loss": 0.0006,
+      "step": 13464
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 1.4401162790697673e-07,
+      "loss": 0.0006,
+      "step": 13465
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 1.44e-07,
+      "loss": 0.0008,
+      "step": 13466
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 1.4398837209302325e-07,
+      "loss": 0.0008,
+      "step": 13467
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 1.439767441860465e-07,
+      "loss": 0.0015,
+      "step": 13468
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 1.4396511627906978e-07,
+      "loss": 0.4661,
+      "step": 13469
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 1.43953488372093e-07,
+      "loss": 0.0007,
+      "step": 13470
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.4394186046511627e-07,
+      "loss": 0.0006,
+      "step": 13471
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.4393023255813954e-07,
+      "loss": 0.0007,
+      "step": 13472
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.439186046511628e-07,
+      "loss": 0.0011,
+      "step": 13473
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.4390697674418603e-07,
+      "loss": 0.0006,
+      "step": 13474
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 1.438953488372093e-07,
+      "loss": 0.0008,
+      "step": 13475
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 1.4388372093023256e-07,
+      "loss": 0.0006,
+      "step": 13476
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 1.4387209302325582e-07,
+      "loss": 0.0005,
+      "step": 13477
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 1.4386046511627905e-07,
+      "loss": 0.0006,
+      "step": 13478
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 1.4384883720930232e-07,
+      "loss": 0.2545,
+      "step": 13479
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 1.4383720930232558e-07,
+      "loss": 0.0007,
+      "step": 13480
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 1.4382558139534884e-07,
+      "loss": 0.0006,
+      "step": 13481
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 1.4381395348837207e-07,
+      "loss": 0.0007,
+      "step": 13482
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 1.4380232558139536e-07,
+      "loss": 0.0007,
+      "step": 13483
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 1.437906976744186e-07,
+      "loss": 0.0016,
+      "step": 13484
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 1.4377906976744186e-07,
+      "loss": 0.0005,
+      "step": 13485
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.4376744186046512e-07,
+      "loss": 0.0008,
+      "step": 13486
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.4375581395348836e-07,
+      "loss": 0.0008,
+      "step": 13487
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.4374418604651162e-07,
+      "loss": 0.0007,
+      "step": 13488
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.4373255813953488e-07,
+      "loss": 0.0012,
+      "step": 13489
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.4372093023255814e-07,
+      "loss": 0.0006,
+      "step": 13490
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 1.4370930232558138e-07,
+      "loss": 0.0005,
+      "step": 13491
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 1.4369767441860464e-07,
+      "loss": 0.1123,
+      "step": 13492
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 1.436860465116279e-07,
+      "loss": 0.001,
+      "step": 13493
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 1.4367441860465116e-07,
+      "loss": 0.0008,
+      "step": 13494
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 1.436627906976744e-07,
+      "loss": 0.0006,
+      "step": 13495
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 1.4365116279069768e-07,
+      "loss": 0.0005,
+      "step": 13496
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 1.4363953488372092e-07,
+      "loss": 0.0007,
+      "step": 13497
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 1.4362790697674418e-07,
+      "loss": 0.0006,
+      "step": 13498
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 1.4361627906976744e-07,
+      "loss": 0.0006,
+      "step": 13499
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 1.436046511627907e-07,
+      "loss": 0.0011,
+      "step": 13500
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 1.4359302325581394e-07,
+      "loss": 0.0009,
+      "step": 13501
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 1.435813953488372e-07,
+      "loss": 0.0007,
+      "step": 13502
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 1.4356976744186046e-07,
+      "loss": 0.0005,
+      "step": 13503
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 1.4355813953488373e-07,
+      "loss": 3.0339,
+      "step": 13504
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 1.4354651162790696e-07,
+      "loss": 0.0008,
+      "step": 13505
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 1.4353488372093022e-07,
+      "loss": 0.2191,
+      "step": 13506
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.4352325581395348e-07,
+      "loss": 0.0008,
+      "step": 13507
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.4351162790697675e-07,
+      "loss": 0.0008,
+      "step": 13508
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.4349999999999998e-07,
+      "loss": 0.0006,
+      "step": 13509
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.4348837209302324e-07,
+      "loss": 0.001,
+      "step": 13510
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 1.434767441860465e-07,
+      "loss": 0.3252,
+      "step": 13511
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 1.4346511627906977e-07,
+      "loss": 0.0005,
+      "step": 13512
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 1.4345348837209303e-07,
+      "loss": 0.0007,
+      "step": 13513
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 1.4344186046511626e-07,
+      "loss": 0.0006,
+      "step": 13514
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 1.4343023255813953e-07,
+      "loss": 0.0005,
+      "step": 13515
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 1.434186046511628e-07,
+      "loss": 0.0007,
+      "step": 13516
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.4340697674418605e-07,
+      "loss": 0.0008,
+      "step": 13517
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.4339534883720928e-07,
+      "loss": 0.0006,
+      "step": 13518
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.4338372093023255e-07,
+      "loss": 0.0005,
+      "step": 13519
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.433720930232558e-07,
+      "loss": 0.0007,
+      "step": 13520
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.4336046511627907e-07,
+      "loss": 0.0007,
+      "step": 13521
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 1.433488372093023e-07,
+      "loss": 0.5891,
+      "step": 13522
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 1.4333720930232557e-07,
+      "loss": 0.0005,
+      "step": 13523
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 1.4332558139534883e-07,
+      "loss": 0.001,
+      "step": 13524
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 1.433139534883721e-07,
+      "loss": 0.0007,
+      "step": 13525
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 1.4330232558139535e-07,
+      "loss": 0.0022,
+      "step": 13526
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 1.432906976744186e-07,
+      "loss": 0.0006,
+      "step": 13527
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.4327906976744185e-07,
+      "loss": 0.0007,
+      "step": 13528
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.432674418604651e-07,
+      "loss": 0.0008,
+      "step": 13529
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.4325581395348837e-07,
+      "loss": 0.0006,
+      "step": 13530
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.4324418604651163e-07,
+      "loss": 0.0007,
+      "step": 13531
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 1.4323255813953487e-07,
+      "loss": 0.7446,
+      "step": 13532
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 1.4322093023255813e-07,
+      "loss": 0.0005,
+      "step": 13533
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 1.432093023255814e-07,
+      "loss": 0.0006,
+      "step": 13534
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 1.4319767441860465e-07,
+      "loss": 0.0007,
+      "step": 13535
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 1.431860465116279e-07,
+      "loss": 0.0008,
+      "step": 13536
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 1.4317441860465115e-07,
+      "loss": 0.0013,
+      "step": 13537
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.431627906976744e-07,
+      "loss": 0.001,
+      "step": 13538
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.4315116279069765e-07,
+      "loss": 0.0007,
+      "step": 13539
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.4313953488372094e-07,
+      "loss": 0.0007,
+      "step": 13540
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.4312790697674417e-07,
+      "loss": 0.2299,
+      "step": 13541
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 1.4311627906976743e-07,
+      "loss": 0.0008,
+      "step": 13542
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 1.431046511627907e-07,
+      "loss": 0.0006,
+      "step": 13543
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 1.4309302325581396e-07,
+      "loss": 0.0006,
+      "step": 13544
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 1.430813953488372e-07,
+      "loss": 0.0007,
+      "step": 13545
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 1.4306976744186045e-07,
+      "loss": 0.001,
+      "step": 13546
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 1.4305813953488371e-07,
+      "loss": 0.0007,
+      "step": 13547
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.4304651162790698e-07,
+      "loss": 0.0013,
+      "step": 13548
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.430348837209302e-07,
+      "loss": 0.0014,
+      "step": 13549
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.4302325581395347e-07,
+      "loss": 0.0007,
+      "step": 13550
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.4301162790697673e-07,
+      "loss": 0.0006,
+      "step": 13551
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.43e-07,
+      "loss": 0.0006,
+      "step": 13552
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 1.4298837209302326e-07,
+      "loss": 0.0009,
+      "step": 13553
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 1.4297674418604652e-07,
+      "loss": 0.0008,
+      "step": 13554
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 1.4296511627906976e-07,
+      "loss": 0.0008,
+      "step": 13555
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 1.4295348837209302e-07,
+      "loss": 0.0012,
+      "step": 13556
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 1.4294186046511628e-07,
+      "loss": 0.0006,
+      "step": 13557
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 1.4293023255813954e-07,
+      "loss": 0.001,
+      "step": 13558
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 1.4291860465116278e-07,
+      "loss": 0.0009,
+      "step": 13559
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 1.4290697674418604e-07,
+      "loss": 0.0007,
+      "step": 13560
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 1.428953488372093e-07,
+      "loss": 0.0006,
+      "step": 13561
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 1.4288372093023256e-07,
+      "loss": 0.0007,
+      "step": 13562
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 1.428720930232558e-07,
+      "loss": 0.0006,
+      "step": 13563
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 1.4286046511627906e-07,
+      "loss": 0.0007,
+      "step": 13564
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 1.4284883720930232e-07,
+      "loss": 0.0006,
+      "step": 13565
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 1.4283720930232555e-07,
+      "loss": 0.0007,
+      "step": 13566
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 1.4282558139534884e-07,
+      "loss": 0.0006,
+      "step": 13567
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 1.4281395348837208e-07,
+      "loss": 0.0906,
+      "step": 13568
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 1.4280232558139534e-07,
+      "loss": 0.0007,
+      "step": 13569
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 1.427906976744186e-07,
+      "loss": 0.0006,
+      "step": 13570
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 1.4277906976744186e-07,
+      "loss": 0.0006,
+      "step": 13571
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 1.427674418604651e-07,
+      "loss": 0.0007,
+      "step": 13572
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 1.4275581395348836e-07,
+      "loss": 0.0007,
+      "step": 13573
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.4274418604651162e-07,
+      "loss": 0.0009,
+      "step": 13574
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.4273255813953488e-07,
+      "loss": 0.0005,
+      "step": 13575
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.4272093023255812e-07,
+      "loss": 0.0007,
+      "step": 13576
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.4270930232558138e-07,
+      "loss": 0.0013,
+      "step": 13577
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.4269767441860464e-07,
+      "loss": 0.0007,
+      "step": 13578
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 1.426860465116279e-07,
+      "loss": 0.0007,
+      "step": 13579
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 1.4267441860465117e-07,
+      "loss": 0.0005,
+      "step": 13580
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 1.4266279069767443e-07,
+      "loss": 0.0007,
+      "step": 13581
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 1.4265116279069766e-07,
+      "loss": 0.0008,
+      "step": 13582
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 1.4263953488372092e-07,
+      "loss": 0.0006,
+      "step": 13583
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 1.4262790697674419e-07,
+      "loss": 0.0011,
+      "step": 13584
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 1.4261627906976745e-07,
+      "loss": 0.0008,
+      "step": 13585
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 1.4260465116279068e-07,
+      "loss": 0.0007,
+      "step": 13586
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 1.4259302325581394e-07,
+      "loss": 0.0007,
+      "step": 13587
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 1.425813953488372e-07,
+      "loss": 0.0005,
+      "step": 13588
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 1.4256976744186047e-07,
+      "loss": 0.0006,
+      "step": 13589
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.425581395348837e-07,
+      "loss": 0.0008,
+      "step": 13590
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.4254651162790697e-07,
+      "loss": 0.0009,
+      "step": 13591
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.4253488372093023e-07,
+      "loss": 0.0006,
+      "step": 13592
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.4252325581395346e-07,
+      "loss": 0.0007,
+      "step": 13593
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 1.4251162790697675e-07,
+      "loss": 0.0006,
+      "step": 13594
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 1.4249999999999999e-07,
+      "loss": 0.0008,
+      "step": 13595
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 1.4248837209302325e-07,
+      "loss": 0.0007,
+      "step": 13596
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 1.424767441860465e-07,
+      "loss": 0.0006,
+      "step": 13597
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 1.4246511627906977e-07,
+      "loss": 2.0553,
+      "step": 13598
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 1.42453488372093e-07,
+      "loss": 0.0008,
+      "step": 13599
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 1.4244186046511627e-07,
+      "loss": 0.0005,
+      "step": 13600
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 1.4243023255813953e-07,
+      "loss": 0.0008,
+      "step": 13601
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 1.424186046511628e-07,
+      "loss": 0.0006,
+      "step": 13602
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 1.4240697674418603e-07,
+      "loss": 0.0007,
+      "step": 13603
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 1.423953488372093e-07,
+      "loss": 0.0007,
+      "step": 13604
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.4238372093023255e-07,
+      "loss": 0.0007,
+      "step": 13605
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.423720930232558e-07,
+      "loss": 0.0007,
+      "step": 13606
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.4236046511627907e-07,
+      "loss": 0.0012,
+      "step": 13607
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.4234883720930233e-07,
+      "loss": 0.0006,
+      "step": 13608
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.4233720930232557e-07,
+      "loss": 0.0028,
+      "step": 13609
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 1.4232558139534883e-07,
+      "loss": 0.1223,
+      "step": 13610
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 1.423139534883721e-07,
+      "loss": 0.0006,
+      "step": 13611
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 1.4230232558139536e-07,
+      "loss": 0.0007,
+      "step": 13612
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 1.422906976744186e-07,
+      "loss": 0.0009,
+      "step": 13613
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 1.4227906976744185e-07,
+      "loss": 3.0966,
+      "step": 13614
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 1.4226744186046511e-07,
+      "loss": 0.0042,
+      "step": 13615
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 1.4225581395348838e-07,
+      "loss": 0.0006,
+      "step": 13616
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 1.422441860465116e-07,
+      "loss": 1.6957,
+      "step": 13617
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 1.4223255813953487e-07,
+      "loss": 0.0008,
+      "step": 13618
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 1.4222093023255813e-07,
+      "loss": 0.001,
+      "step": 13619
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 1.4220930232558137e-07,
+      "loss": 0.0008,
+      "step": 13620
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 1.4219767441860466e-07,
+      "loss": 0.0008,
+      "step": 13621
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 1.421860465116279e-07,
+      "loss": 0.001,
+      "step": 13622
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 1.4217441860465115e-07,
+      "loss": 0.0006,
+      "step": 13623
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 1.4216279069767442e-07,
+      "loss": 0.0006,
+      "step": 13624
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 1.4215116279069768e-07,
+      "loss": 0.0023,
+      "step": 13625
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 1.4213953488372091e-07,
+      "loss": 0.0008,
+      "step": 13626
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 1.4212790697674418e-07,
+      "loss": 0.0008,
+      "step": 13627
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 1.4211627906976744e-07,
+      "loss": 0.0008,
+      "step": 13628
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 1.421046511627907e-07,
+      "loss": 0.0008,
+      "step": 13629
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 1.4209302325581393e-07,
+      "loss": 0.0008,
+      "step": 13630
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 1.420813953488372e-07,
+      "loss": 0.0007,
+      "step": 13631
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 1.4206976744186046e-07,
+      "loss": 0.0008,
+      "step": 13632
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 1.4205813953488372e-07,
+      "loss": 0.0006,
+      "step": 13633
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 1.4204651162790698e-07,
+      "loss": 0.0008,
+      "step": 13634
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 1.4203488372093024e-07,
+      "loss": 0.0008,
+      "step": 13635
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.4202325581395348e-07,
+      "loss": 0.0011,
+      "step": 13636
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.4201162790697674e-07,
+      "loss": 0.0007,
+      "step": 13637
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.42e-07,
+      "loss": 0.0006,
+      "step": 13638
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.4198837209302326e-07,
+      "loss": 0.0005,
+      "step": 13639
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.419767441860465e-07,
+      "loss": 0.0006,
+      "step": 13640
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 1.4196511627906976e-07,
+      "loss": 0.0006,
+      "step": 13641
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.4195348837209302e-07,
+      "loss": 0.0009,
+      "step": 13642
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.4194186046511626e-07,
+      "loss": 0.0006,
+      "step": 13643
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.4193023255813952e-07,
+      "loss": 0.0007,
+      "step": 13644
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.4191860465116278e-07,
+      "loss": 0.0008,
+      "step": 13645
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 1.4190697674418604e-07,
+      "loss": 0.0009,
+      "step": 13646
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.4189534883720928e-07,
+      "loss": 0.001,
+      "step": 13647
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.4188372093023257e-07,
+      "loss": 0.0008,
+      "step": 13648
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.418720930232558e-07,
+      "loss": 0.0015,
+      "step": 13649
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.4186046511627906e-07,
+      "loss": 0.0006,
+      "step": 13650
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 1.4184883720930232e-07,
+      "loss": 0.0007,
+      "step": 13651
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 1.4183720930232559e-07,
+      "loss": 0.0027,
+      "step": 13652
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 1.4182558139534882e-07,
+      "loss": 0.0093,
+      "step": 13653
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 1.4181395348837208e-07,
+      "loss": 0.0007,
+      "step": 13654
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 1.4180232558139534e-07,
+      "loss": 0.0008,
+      "step": 13655
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 1.417906976744186e-07,
+      "loss": 0.0007,
+      "step": 13656
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 1.4177906976744184e-07,
+      "loss": 0.0008,
+      "step": 13657
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 1.417674418604651e-07,
+      "loss": 0.0046,
+      "step": 13658
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 1.4175581395348836e-07,
+      "loss": 0.0779,
+      "step": 13659
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 1.4174418604651163e-07,
+      "loss": 0.0006,
+      "step": 13660
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 1.4173255813953486e-07,
+      "loss": 0.0006,
+      "step": 13661
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 1.4172093023255815e-07,
+      "loss": 0.0006,
+      "step": 13662
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 1.4170930232558138e-07,
+      "loss": 0.0007,
+      "step": 13663
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 1.4169767441860465e-07,
+      "loss": 0.0091,
+      "step": 13664
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 1.416860465116279e-07,
+      "loss": 0.0007,
+      "step": 13665
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 1.4167441860465117e-07,
+      "loss": 0.0006,
+      "step": 13666
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.416627906976744e-07,
+      "loss": 0.0006,
+      "step": 13667
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.4165116279069767e-07,
+      "loss": 0.0013,
+      "step": 13668
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.4163953488372093e-07,
+      "loss": 0.0007,
+      "step": 13669
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.4162790697674416e-07,
+      "loss": 0.0006,
+      "step": 13670
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.4161627906976743e-07,
+      "loss": 0.0007,
+      "step": 13671
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 1.416046511627907e-07,
+      "loss": 0.0008,
+      "step": 13672
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.4159302325581395e-07,
+      "loss": 0.0013,
+      "step": 13673
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.4158139534883718e-07,
+      "loss": 0.0006,
+      "step": 13674
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.4156976744186047e-07,
+      "loss": 0.0006,
+      "step": 13675
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.415581395348837e-07,
+      "loss": 0.0007,
+      "step": 13676
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 1.4154651162790697e-07,
+      "loss": 0.0007,
+      "step": 13677
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 1.4153488372093023e-07,
+      "loss": 0.0006,
+      "step": 13678
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 1.415232558139535e-07,
+      "loss": 0.002,
+      "step": 13679
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 1.4151162790697673e-07,
+      "loss": 0.0006,
+      "step": 13680
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 1.415e-07,
+      "loss": 0.0008,
+      "step": 13681
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 1.4148837209302325e-07,
+      "loss": 0.0014,
+      "step": 13682
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 1.4147674418604651e-07,
+      "loss": 0.0008,
+      "step": 13683
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 1.4146511627906975e-07,
+      "loss": 0.0005,
+      "step": 13684
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 1.41453488372093e-07,
+      "loss": 0.0009,
+      "step": 13685
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 1.4144186046511627e-07,
+      "loss": 0.0005,
+      "step": 13686
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 1.4143023255813953e-07,
+      "loss": 0.0634,
+      "step": 13687
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 1.4141860465116277e-07,
+      "loss": 0.0019,
+      "step": 13688
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 1.4140697674418606e-07,
+      "loss": 0.009,
+      "step": 13689
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 1.413953488372093e-07,
+      "loss": 0.0009,
+      "step": 13690
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 1.4138372093023255e-07,
+      "loss": 0.4781,
+      "step": 13691
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 1.4137209302325582e-07,
+      "loss": 0.0011,
+      "step": 13692
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 1.4136046511627908e-07,
+      "loss": 0.0006,
+      "step": 13693
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 1.413488372093023e-07,
+      "loss": 0.0011,
+      "step": 13694
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 1.4133720930232557e-07,
+      "loss": 0.0005,
+      "step": 13695
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 1.4132558139534884e-07,
+      "loss": 0.0007,
+      "step": 13696
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 1.4131395348837207e-07,
+      "loss": 0.0016,
+      "step": 13697
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.4130232558139533e-07,
+      "loss": 0.0009,
+      "step": 13698
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.412906976744186e-07,
+      "loss": 0.0009,
+      "step": 13699
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.4127906976744186e-07,
+      "loss": 2.8302,
+      "step": 13700
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.412674418604651e-07,
+      "loss": 0.0007,
+      "step": 13701
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.4125581395348838e-07,
+      "loss": 0.0008,
+      "step": 13702
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 1.4124418604651162e-07,
+      "loss": 0.0006,
+      "step": 13703
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.4123255813953488e-07,
+      "loss": 0.0007,
+      "step": 13704
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.4122093023255814e-07,
+      "loss": 0.0024,
+      "step": 13705
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.412093023255814e-07,
+      "loss": 0.0007,
+      "step": 13706
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.4119767441860464e-07,
+      "loss": 0.0067,
+      "step": 13707
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 1.411860465116279e-07,
+      "loss": 0.0007,
+      "step": 13708
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 1.4117441860465116e-07,
+      "loss": 0.0007,
+      "step": 13709
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 1.4116279069767442e-07,
+      "loss": 0.0005,
+      "step": 13710
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 1.4115116279069766e-07,
+      "loss": 0.0006,
+      "step": 13711
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 1.4113953488372092e-07,
+      "loss": 0.0007,
+      "step": 13712
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 1.4112790697674418e-07,
+      "loss": 0.0011,
+      "step": 13713
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 1.4111627906976744e-07,
+      "loss": 0.0005,
+      "step": 13714
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 1.4110465116279068e-07,
+      "loss": 0.0006,
+      "step": 13715
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 1.4109302325581396e-07,
+      "loss": 0.0008,
+      "step": 13716
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 1.410813953488372e-07,
+      "loss": 0.0008,
+      "step": 13717
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 1.4106976744186046e-07,
+      "loss": 0.0009,
+      "step": 13718
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 1.4105813953488372e-07,
+      "loss": 0.0006,
+      "step": 13719
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 1.4104651162790698e-07,
+      "loss": 0.0005,
+      "step": 13720
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 1.4103488372093022e-07,
+      "loss": 0.0007,
+      "step": 13721
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 1.4102325581395348e-07,
+      "loss": 0.0007,
+      "step": 13722
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 1.4101162790697674e-07,
+      "loss": 0.0008,
+      "step": 13723
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 1.4099999999999998e-07,
+      "loss": 0.0007,
+      "step": 13724
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 1.4098837209302324e-07,
+      "loss": 0.0005,
+      "step": 13725
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 1.409767441860465e-07,
+      "loss": 0.001,
+      "step": 13726
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 1.4096511627906976e-07,
+      "loss": 0.0006,
+      "step": 13727
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 1.40953488372093e-07,
+      "loss": 0.0006,
+      "step": 13728
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.409418604651163e-07,
+      "loss": 0.0006,
+      "step": 13729
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.4093023255813952e-07,
+      "loss": 0.0007,
+      "step": 13730
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.4091860465116278e-07,
+      "loss": 0.0007,
+      "step": 13731
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.4090697674418605e-07,
+      "loss": 1.6948,
+      "step": 13732
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.408953488372093e-07,
+      "loss": 0.0006,
+      "step": 13733
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 1.4088372093023254e-07,
+      "loss": 0.0007,
+      "step": 13734
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 1.408720930232558e-07,
+      "loss": 0.0038,
+      "step": 13735
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 1.4086046511627907e-07,
+      "loss": 0.0006,
+      "step": 13736
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 1.4084883720930233e-07,
+      "loss": 0.0006,
+      "step": 13737
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 1.4083720930232556e-07,
+      "loss": 0.0006,
+      "step": 13738
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 1.4082558139534883e-07,
+      "loss": 0.0005,
+      "step": 13739
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.408139534883721e-07,
+      "loss": 0.0006,
+      "step": 13740
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.4080232558139535e-07,
+      "loss": 0.0006,
+      "step": 13741
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.4079069767441858e-07,
+      "loss": 0.0008,
+      "step": 13742
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.4077906976744187e-07,
+      "loss": 0.0013,
+      "step": 13743
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 1.407674418604651e-07,
+      "loss": 0.0006,
+      "step": 13744
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 1.4075581395348837e-07,
+      "loss": 0.0006,
+      "step": 13745
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 1.4074418604651163e-07,
+      "loss": 0.0059,
+      "step": 13746
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 1.407325581395349e-07,
+      "loss": 0.0011,
+      "step": 13747
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 1.4072093023255813e-07,
+      "loss": 0.0005,
+      "step": 13748
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 1.407093023255814e-07,
+      "loss": 0.0006,
+      "step": 13749
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 1.4069767441860465e-07,
+      "loss": 0.0006,
+      "step": 13750
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 1.4068604651162789e-07,
+      "loss": 0.0006,
+      "step": 13751
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 1.4067441860465115e-07,
+      "loss": 0.0007,
+      "step": 13752
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 1.406627906976744e-07,
+      "loss": 0.0007,
+      "step": 13753
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 1.4065116279069767e-07,
+      "loss": 2.4756,
+      "step": 13754
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.406395348837209e-07,
+      "loss": 0.0008,
+      "step": 13755
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.406279069767442e-07,
+      "loss": 0.0008,
+      "step": 13756
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.4061627906976743e-07,
+      "loss": 0.0005,
+      "step": 13757
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.406046511627907e-07,
+      "loss": 0.0009,
+      "step": 13758
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 1.4059302325581395e-07,
+      "loss": 0.0012,
+      "step": 13759
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.4058139534883722e-07,
+      "loss": 0.0009,
+      "step": 13760
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.4056976744186045e-07,
+      "loss": 0.0009,
+      "step": 13761
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.405581395348837e-07,
+      "loss": 0.0008,
+      "step": 13762
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.4054651162790697e-07,
+      "loss": 0.0006,
+      "step": 13763
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.4053488372093024e-07,
+      "loss": 0.0008,
+      "step": 13764
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 1.4052325581395347e-07,
+      "loss": 0.0007,
+      "step": 13765
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.4051162790697673e-07,
+      "loss": 0.0006,
+      "step": 13766
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.405e-07,
+      "loss": 0.0006,
+      "step": 13767
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.4048837209302326e-07,
+      "loss": 0.0006,
+      "step": 13768
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.404767441860465e-07,
+      "loss": 0.0007,
+      "step": 13769
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 1.4046511627906978e-07,
+      "loss": 0.0008,
+      "step": 13770
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.4045348837209301e-07,
+      "loss": 0.0006,
+      "step": 13771
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.4044186046511628e-07,
+      "loss": 0.0006,
+      "step": 13772
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.4043023255813954e-07,
+      "loss": 0.0007,
+      "step": 13773
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.404186046511628e-07,
+      "loss": 0.0078,
+      "step": 13774
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 1.4040697674418603e-07,
+      "loss": 0.001,
+      "step": 13775
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 1.403953488372093e-07,
+      "loss": 0.0007,
+      "step": 13776
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 1.4038372093023256e-07,
+      "loss": 0.0007,
+      "step": 13777
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 1.403720930232558e-07,
+      "loss": 0.0009,
+      "step": 13778
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 1.4036046511627906e-07,
+      "loss": 0.0007,
+      "step": 13779
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 1.4034883720930232e-07,
+      "loss": 0.0009,
+      "step": 13780
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 1.4033720930232558e-07,
+      "loss": 0.0005,
+      "step": 13781
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 1.4032558139534881e-07,
+      "loss": 0.0007,
+      "step": 13782
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 1.4031395348837208e-07,
+      "loss": 0.001,
+      "step": 13783
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 1.4030232558139534e-07,
+      "loss": 0.0009,
+      "step": 13784
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 1.402906976744186e-07,
+      "loss": 0.0007,
+      "step": 13785
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 1.4027906976744186e-07,
+      "loss": 0.0007,
+      "step": 13786
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 1.4026744186046512e-07,
+      "loss": 0.0016,
+      "step": 13787
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 1.4025581395348836e-07,
+      "loss": 0.1073,
+      "step": 13788
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 1.4024418604651162e-07,
+      "loss": 0.0007,
+      "step": 13789
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 1.4023255813953488e-07,
+      "loss": 0.0008,
+      "step": 13790
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.4022093023255814e-07,
+      "loss": 0.0008,
+      "step": 13791
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.4020930232558138e-07,
+      "loss": 0.0007,
+      "step": 13792
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.4019767441860464e-07,
+      "loss": 0.0025,
+      "step": 13793
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.401860465116279e-07,
+      "loss": 0.0006,
+      "step": 13794
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.4017441860465116e-07,
+      "loss": 0.0008,
+      "step": 13795
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 1.401627906976744e-07,
+      "loss": 0.0006,
+      "step": 13796
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 1.4015116279069769e-07,
+      "loss": 0.0005,
+      "step": 13797
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 1.4013953488372092e-07,
+      "loss": 0.0007,
+      "step": 13798
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 1.4012790697674418e-07,
+      "loss": 0.0006,
+      "step": 13799
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 1.4011627906976745e-07,
+      "loss": 0.0007,
+      "step": 13800
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 1.401046511627907e-07,
+      "loss": 0.0009,
+      "step": 13801
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 1.4009302325581394e-07,
+      "loss": 0.0014,
+      "step": 13802
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 1.400813953488372e-07,
+      "loss": 0.0006,
+      "step": 13803
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 1.4006976744186047e-07,
+      "loss": 0.0163,
+      "step": 13804
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 1.400581395348837e-07,
+      "loss": 0.0006,
+      "step": 13805
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 1.4004651162790696e-07,
+      "loss": 0.0007,
+      "step": 13806
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.4003488372093022e-07,
+      "loss": 0.0005,
+      "step": 13807
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.4002325581395349e-07,
+      "loss": 0.0005,
+      "step": 13808
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.4001162790697672e-07,
+      "loss": 0.0017,
+      "step": 13809
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.3999999999999998e-07,
+      "loss": 0.0007,
+      "step": 13810
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 1.3998837209302324e-07,
+      "loss": 0.0009,
+      "step": 13811
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 1.399767441860465e-07,
+      "loss": 0.0015,
+      "step": 13812
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 1.3996511627906977e-07,
+      "loss": 0.0007,
+      "step": 13813
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 1.3995348837209303e-07,
+      "loss": 0.0005,
+      "step": 13814
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 1.3994186046511627e-07,
+      "loss": 0.0006,
+      "step": 13815
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 1.3993023255813953e-07,
+      "loss": 0.0008,
+      "step": 13816
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 1.399186046511628e-07,
+      "loss": 0.0006,
+      "step": 13817
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 1.3990697674418605e-07,
+      "loss": 0.0006,
+      "step": 13818
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 1.3989534883720929e-07,
+      "loss": 0.0007,
+      "step": 13819
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 1.3988372093023255e-07,
+      "loss": 0.2185,
+      "step": 13820
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 1.398720930232558e-07,
+      "loss": 2.3844,
+      "step": 13821
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.3986046511627907e-07,
+      "loss": 0.0018,
+      "step": 13822
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.398488372093023e-07,
+      "loss": 0.0006,
+      "step": 13823
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.398372093023256e-07,
+      "loss": 0.001,
+      "step": 13824
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.3982558139534883e-07,
+      "loss": 0.0036,
+      "step": 13825
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.398139534883721e-07,
+      "loss": 0.0006,
+      "step": 13826
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 1.3980232558139535e-07,
+      "loss": 0.0009,
+      "step": 13827
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 1.3979069767441861e-07,
+      "loss": 0.0008,
+      "step": 13828
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 1.3977906976744185e-07,
+      "loss": 0.0007,
+      "step": 13829
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 1.397674418604651e-07,
+      "loss": 0.0016,
+      "step": 13830
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 1.3975581395348837e-07,
+      "loss": 0.3555,
+      "step": 13831
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 1.397441860465116e-07,
+      "loss": 0.0012,
+      "step": 13832
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 1.3973255813953487e-07,
+      "loss": 0.0006,
+      "step": 13833
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 1.3972093023255813e-07,
+      "loss": 0.0008,
+      "step": 13834
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 1.397093023255814e-07,
+      "loss": 2.069,
+      "step": 13835
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 1.3969767441860463e-07,
+      "loss": 0.0096,
+      "step": 13836
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 1.396860465116279e-07,
+      "loss": 0.0006,
+      "step": 13837
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.3967441860465115e-07,
+      "loss": 0.0017,
+      "step": 13838
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.3966279069767441e-07,
+      "loss": 0.0009,
+      "step": 13839
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.3965116279069768e-07,
+      "loss": 0.0005,
+      "step": 13840
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.3963953488372094e-07,
+      "loss": 0.0007,
+      "step": 13841
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 1.3962790697674417e-07,
+      "loss": 0.0005,
+      "step": 13842
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.3961627906976743e-07,
+      "loss": 0.0006,
+      "step": 13843
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.396046511627907e-07,
+      "loss": 0.0006,
+      "step": 13844
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.3959302325581396e-07,
+      "loss": 0.0015,
+      "step": 13845
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.395813953488372e-07,
+      "loss": 0.0037,
+      "step": 13846
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.3956976744186045e-07,
+      "loss": 0.0009,
+      "step": 13847
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.3955813953488372e-07,
+      "loss": 0.0007,
+      "step": 13848
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.3954651162790698e-07,
+      "loss": 0.0007,
+      "step": 13849
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.3953488372093021e-07,
+      "loss": 0.0005,
+      "step": 13850
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.395232558139535e-07,
+      "loss": 0.0008,
+      "step": 13851
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.3951162790697674e-07,
+      "loss": 0.0006,
+      "step": 13852
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 1.395e-07,
+      "loss": 0.0006,
+      "step": 13853
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 1.3948837209302326e-07,
+      "loss": 0.0007,
+      "step": 13854
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 1.3947674418604652e-07,
+      "loss": 0.0011,
+      "step": 13855
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 1.3946511627906976e-07,
+      "loss": 0.0011,
+      "step": 13856
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 1.3945348837209302e-07,
+      "loss": 0.0007,
+      "step": 13857
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 1.3944186046511628e-07,
+      "loss": 0.0006,
+      "step": 13858
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 1.3943023255813952e-07,
+      "loss": 0.0007,
+      "step": 13859
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 1.3941860465116278e-07,
+      "loss": 0.0006,
+      "step": 13860
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 1.3940697674418604e-07,
+      "loss": 0.0006,
+      "step": 13861
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 1.393953488372093e-07,
+      "loss": 0.0008,
+      "step": 13862
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 1.3938372093023254e-07,
+      "loss": 0.0011,
+      "step": 13863
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 1.393720930232558e-07,
+      "loss": 0.103,
+      "step": 13864
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 1.3936046511627906e-07,
+      "loss": 2.2948,
+      "step": 13865
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 1.3934883720930232e-07,
+      "loss": 0.0005,
+      "step": 13866
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 1.3933720930232558e-07,
+      "loss": 0.0007,
+      "step": 13867
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 1.3932558139534884e-07,
+      "loss": 0.0011,
+      "step": 13868
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 1.3931395348837208e-07,
+      "loss": 0.0007,
+      "step": 13869
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 1.3930232558139534e-07,
+      "loss": 0.0009,
+      "step": 13870
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 1.392906976744186e-07,
+      "loss": 0.0006,
+      "step": 13871
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 1.3927906976744187e-07,
+      "loss": 0.0011,
+      "step": 13872
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 1.392674418604651e-07,
+      "loss": 0.0005,
+      "step": 13873
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.3925581395348836e-07,
+      "loss": 0.0006,
+      "step": 13874
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.3924418604651162e-07,
+      "loss": 0.0005,
+      "step": 13875
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.3923255813953489e-07,
+      "loss": 0.0011,
+      "step": 13876
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.3922093023255812e-07,
+      "loss": 0.0007,
+      "step": 13877
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 1.392093023255814e-07,
+      "loss": 0.0006,
+      "step": 13878
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.3919767441860464e-07,
+      "loss": 0.0007,
+      "step": 13879
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.391860465116279e-07,
+      "loss": 0.0007,
+      "step": 13880
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.3917441860465117e-07,
+      "loss": 0.0006,
+      "step": 13881
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.3916279069767443e-07,
+      "loss": 0.0004,
+      "step": 13882
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.3915116279069766e-07,
+      "loss": 0.0191,
+      "step": 13883
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 1.3913953488372093e-07,
+      "loss": 0.0007,
+      "step": 13884
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.391279069767442e-07,
+      "loss": 0.0007,
+      "step": 13885
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.3911627906976742e-07,
+      "loss": 0.0007,
+      "step": 13886
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.3910465116279068e-07,
+      "loss": 0.0007,
+      "step": 13887
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.3909302325581395e-07,
+      "loss": 0.0008,
+      "step": 13888
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 1.390813953488372e-07,
+      "loss": 0.0014,
+      "step": 13889
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 1.3906976744186044e-07,
+      "loss": 0.0005,
+      "step": 13890
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 1.390581395348837e-07,
+      "loss": 0.0014,
+      "step": 13891
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 1.3904651162790697e-07,
+      "loss": 0.0006,
+      "step": 13892
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 1.3903488372093023e-07,
+      "loss": 0.0046,
+      "step": 13893
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 1.390232558139535e-07,
+      "loss": 0.0007,
+      "step": 13894
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 1.3901162790697675e-07,
+      "loss": 0.0006,
+      "step": 13895
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 1.39e-07,
+      "loss": 0.0006,
+      "step": 13896
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 1.3898837209302325e-07,
+      "loss": 0.0009,
+      "step": 13897
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 1.389767441860465e-07,
+      "loss": 0.0007,
+      "step": 13898
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 1.3896511627906977e-07,
+      "loss": 0.0008,
+      "step": 13899
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 1.38953488372093e-07,
+      "loss": 0.0005,
+      "step": 13900
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 1.3894186046511627e-07,
+      "loss": 0.0006,
+      "step": 13901
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 1.3893023255813953e-07,
+      "loss": 0.0005,
+      "step": 13902
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 1.389186046511628e-07,
+      "loss": 0.0005,
+      "step": 13903
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 1.3890697674418603e-07,
+      "loss": 0.0007,
+      "step": 13904
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.388953488372093e-07,
+      "loss": 0.0006,
+      "step": 13905
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.3888372093023255e-07,
+      "loss": 0.0006,
+      "step": 13906
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.3887209302325581e-07,
+      "loss": 0.0006,
+      "step": 13907
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.3886046511627907e-07,
+      "loss": 0.0007,
+      "step": 13908
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 1.3884883720930234e-07,
+      "loss": 0.0006,
+      "step": 13909
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.3883720930232557e-07,
+      "loss": 0.0033,
+      "step": 13910
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.3882558139534883e-07,
+      "loss": 0.0007,
+      "step": 13911
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.388139534883721e-07,
+      "loss": 0.0007,
+      "step": 13912
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.3880232558139533e-07,
+      "loss": 0.0007,
+      "step": 13913
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.387906976744186e-07,
+      "loss": 0.0004,
+      "step": 13914
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 1.3877906976744185e-07,
+      "loss": 0.311,
+      "step": 13915
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 1.3876744186046512e-07,
+      "loss": 0.0013,
+      "step": 13916
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 1.3875581395348835e-07,
+      "loss": 0.0096,
+      "step": 13917
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 1.387441860465116e-07,
+      "loss": 0.0008,
+      "step": 13918
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 1.3873255813953487e-07,
+      "loss": 0.0011,
+      "step": 13919
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 1.3872093023255814e-07,
+      "loss": 0.0006,
+      "step": 13920
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 1.3870930232558137e-07,
+      "loss": 0.0007,
+      "step": 13921
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 1.3869767441860466e-07,
+      "loss": 0.0764,
+      "step": 13922
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 1.386860465116279e-07,
+      "loss": 0.0007,
+      "step": 13923
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 1.3867441860465116e-07,
+      "loss": 0.0068,
+      "step": 13924
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 1.3866279069767442e-07,
+      "loss": 0.0008,
+      "step": 13925
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.3865116279069768e-07,
+      "loss": 0.001,
+      "step": 13926
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.3863953488372092e-07,
+      "loss": 1.4515,
+      "step": 13927
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.3862790697674418e-07,
+      "loss": 0.0006,
+      "step": 13928
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.3861627906976744e-07,
+      "loss": 0.001,
+      "step": 13929
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 1.386046511627907e-07,
+      "loss": 0.0007,
+      "step": 13930
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 1.3859302325581394e-07,
+      "loss": 0.0009,
+      "step": 13931
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 1.385813953488372e-07,
+      "loss": 0.0776,
+      "step": 13932
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 1.3856976744186046e-07,
+      "loss": 0.0005,
+      "step": 13933
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 1.3855813953488372e-07,
+      "loss": 0.0005,
+      "step": 13934
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 1.3854651162790698e-07,
+      "loss": 0.0039,
+      "step": 13935
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.3853488372093024e-07,
+      "loss": 0.0005,
+      "step": 13936
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.3852325581395348e-07,
+      "loss": 0.0006,
+      "step": 13937
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.3851162790697674e-07,
+      "loss": 0.0008,
+      "step": 13938
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.385e-07,
+      "loss": 0.0006,
+      "step": 13939
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 1.3848837209302324e-07,
+      "loss": 0.0006,
+      "step": 13940
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.384767441860465e-07,
+      "loss": 0.0012,
+      "step": 13941
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.3846511627906976e-07,
+      "loss": 0.0014,
+      "step": 13942
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.3845348837209302e-07,
+      "loss": 0.0006,
+      "step": 13943
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.3844186046511626e-07,
+      "loss": 0.0065,
+      "step": 13944
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.3843023255813952e-07,
+      "loss": 0.0008,
+      "step": 13945
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 1.3841860465116278e-07,
+      "loss": 0.0006,
+      "step": 13946
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 1.3840697674418604e-07,
+      "loss": 0.0007,
+      "step": 13947
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 1.3839534883720928e-07,
+      "loss": 0.0007,
+      "step": 13948
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 1.3838372093023257e-07,
+      "loss": 0.0007,
+      "step": 13949
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 1.383720930232558e-07,
+      "loss": 0.0007,
+      "step": 13950
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 1.3836046511627906e-07,
+      "loss": 0.0007,
+      "step": 13951
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 1.3834883720930233e-07,
+      "loss": 0.0007,
+      "step": 13952
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 1.383372093023256e-07,
+      "loss": 0.0006,
+      "step": 13953
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 1.3832558139534882e-07,
+      "loss": 0.0006,
+      "step": 13954
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 1.3831395348837208e-07,
+      "loss": 0.1843,
+      "step": 13955
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 1.3830232558139535e-07,
+      "loss": 0.0015,
+      "step": 13956
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 1.382906976744186e-07,
+      "loss": 0.0017,
+      "step": 13957
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 1.3827906976744184e-07,
+      "loss": 0.0006,
+      "step": 13958
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 1.382674418604651e-07,
+      "loss": 0.0006,
+      "step": 13959
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.1476353406906128,
+      "eval_runtime": 1.5637,
+      "eval_samples_per_second": 661.244,
+      "eval_steps_per_second": 10.872,
+      "step": 13959
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 1.3825581395348837e-07,
+      "loss": 0.0009,
+      "step": 13960
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 1.3824418604651163e-07,
+      "loss": 0.0006,
+      "step": 13961
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 1.382325581395349e-07,
+      "loss": 0.0014,
+      "step": 13962
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 1.3822093023255812e-07,
+      "loss": 0.0006,
+      "step": 13963
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 1.3820930232558139e-07,
+      "loss": 0.0012,
+      "step": 13964
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 1.3819767441860465e-07,
+      "loss": 0.0006,
+      "step": 13965
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 1.381860465116279e-07,
+      "loss": 0.0008,
+      "step": 13966
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 1.3817441860465115e-07,
+      "loss": 0.0006,
+      "step": 13967
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 1.381627906976744e-07,
+      "loss": 0.0005,
+      "step": 13968
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 1.3815116279069767e-07,
+      "loss": 0.0016,
+      "step": 13969
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 1.3813953488372093e-07,
+      "loss": 0.0007,
+      "step": 13970
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 1.3812790697674417e-07,
+      "loss": 0.0005,
+      "step": 13971
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.3811627906976743e-07,
+      "loss": 0.0007,
+      "step": 13972
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.381046511627907e-07,
+      "loss": 0.0007,
+      "step": 13973
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.3809302325581395e-07,
+      "loss": 0.001,
+      "step": 13974
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.3808139534883719e-07,
+      "loss": 0.0008,
+      "step": 13975
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.3806976744186047e-07,
+      "loss": 0.0007,
+      "step": 13976
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 1.380581395348837e-07,
+      "loss": 0.0005,
+      "step": 13977
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 1.3804651162790697e-07,
+      "loss": 0.0006,
+      "step": 13978
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 1.3803488372093023e-07,
+      "loss": 0.0008,
+      "step": 13979
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 1.380232558139535e-07,
+      "loss": 0.001,
+      "step": 13980
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 1.3801162790697673e-07,
+      "loss": 0.0009,
+      "step": 13981
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 1.38e-07,
+      "loss": 0.0005,
+      "step": 13982
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 1.3798837209302325e-07,
+      "loss": 0.0006,
+      "step": 13983
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 1.3797674418604652e-07,
+      "loss": 0.0006,
+      "step": 13984
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 1.3796511627906975e-07,
+      "loss": 0.0006,
+      "step": 13985
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 1.37953488372093e-07,
+      "loss": 0.001,
+      "step": 13986
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 1.3794186046511627e-07,
+      "loss": 0.0007,
+      "step": 13987
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.3793023255813954e-07,
+      "loss": 0.0005,
+      "step": 13988
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.379186046511628e-07,
+      "loss": 0.0009,
+      "step": 13989
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.3790697674418603e-07,
+      "loss": 0.0007,
+      "step": 13990
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.378953488372093e-07,
+      "loss": 0.0006,
+      "step": 13991
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 1.3788372093023256e-07,
+      "loss": 0.0008,
+      "step": 13992
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 1.3787209302325582e-07,
+      "loss": 0.0004,
+      "step": 13993
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 1.3786046511627905e-07,
+      "loss": 0.0007,
+      "step": 13994
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 1.3784883720930231e-07,
+      "loss": 0.0007,
+      "step": 13995
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 1.3783720930232558e-07,
+      "loss": 0.0016,
+      "step": 13996
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 1.3782558139534884e-07,
+      "loss": 0.0009,
+      "step": 13997
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 1.3781395348837207e-07,
+      "loss": 0.001,
+      "step": 13998
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 1.3780232558139533e-07,
+      "loss": 0.0006,
+      "step": 13999
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 1.377906976744186e-07,
+      "loss": 0.0005,
+      "step": 14000
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 1.3777906976744186e-07,
+      "loss": 0.0006,
+      "step": 14001
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 1.377674418604651e-07,
+      "loss": 0.0006,
+      "step": 14002
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.3775581395348838e-07,
+      "loss": 0.0008,
+      "step": 14003
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.3774418604651162e-07,
+      "loss": 0.0008,
+      "step": 14004
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.3773255813953488e-07,
+      "loss": 0.0006,
+      "step": 14005
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.3772093023255814e-07,
+      "loss": 0.0012,
+      "step": 14006
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.377093023255814e-07,
+      "loss": 0.0006,
+      "step": 14007
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 1.3769767441860464e-07,
+      "loss": 0.0012,
+      "step": 14008
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.376860465116279e-07,
+      "loss": 0.0005,
+      "step": 14009
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.3767441860465116e-07,
+      "loss": 0.0007,
+      "step": 14010
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.3766279069767442e-07,
+      "loss": 0.0006,
+      "step": 14011
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.3765116279069766e-07,
+      "loss": 0.0013,
+      "step": 14012
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 1.3763953488372092e-07,
+      "loss": 0.0006,
+      "step": 14013
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 1.3762790697674418e-07,
+      "loss": 0.0006,
+      "step": 14014
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 1.3761627906976744e-07,
+      "loss": 0.0006,
+      "step": 14015
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 1.376046511627907e-07,
+      "loss": 0.0006,
+      "step": 14016
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 1.3759302325581394e-07,
+      "loss": 0.0006,
+      "step": 14017
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 1.375813953488372e-07,
+      "loss": 0.0007,
+      "step": 14018
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 1.3756976744186046e-07,
+      "loss": 0.0005,
+      "step": 14019
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 1.3755813953488372e-07,
+      "loss": 0.0007,
+      "step": 14020
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 1.3754651162790696e-07,
+      "loss": 0.0012,
+      "step": 14021
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 1.3753488372093022e-07,
+      "loss": 0.0006,
+      "step": 14022
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 1.3752325581395348e-07,
+      "loss": 1.1357,
+      "step": 14023
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 1.3751162790697675e-07,
+      "loss": 0.003,
+      "step": 14024
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 1.3749999999999998e-07,
+      "loss": 0.0006,
+      "step": 14025
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 1.3748837209302324e-07,
+      "loss": 0.0008,
+      "step": 14026
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 1.374767441860465e-07,
+      "loss": 0.0089,
+      "step": 14027
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 1.3746511627906977e-07,
+      "loss": 0.0007,
+      "step": 14028
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 1.37453488372093e-07,
+      "loss": 0.0006,
+      "step": 14029
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 1.374418604651163e-07,
+      "loss": 0.0008,
+      "step": 14030
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 1.3743023255813952e-07,
+      "loss": 0.0006,
+      "step": 14031
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 1.3741860465116279e-07,
+      "loss": 0.0005,
+      "step": 14032
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 1.3740697674418605e-07,
+      "loss": 0.0009,
+      "step": 14033
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.373953488372093e-07,
+      "loss": 0.0006,
+      "step": 14034
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.3738372093023254e-07,
+      "loss": 0.0008,
+      "step": 14035
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.373720930232558e-07,
+      "loss": 0.0008,
+      "step": 14036
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.3736046511627907e-07,
+      "loss": 0.0006,
+      "step": 14037
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.3734883720930233e-07,
+      "loss": 0.0009,
+      "step": 14038
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 1.3733720930232557e-07,
+      "loss": 0.0007,
+      "step": 14039
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.3732558139534883e-07,
+      "loss": 0.0047,
+      "step": 14040
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.373139534883721e-07,
+      "loss": 0.0008,
+      "step": 14041
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.3730232558139535e-07,
+      "loss": 0.0006,
+      "step": 14042
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.3729069767441859e-07,
+      "loss": 0.0009,
+      "step": 14043
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 1.3727906976744185e-07,
+      "loss": 0.0014,
+      "step": 14044
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.372674418604651e-07,
+      "loss": 0.0009,
+      "step": 14045
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.3725581395348837e-07,
+      "loss": 0.0007,
+      "step": 14046
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.3724418604651163e-07,
+      "loss": 0.0007,
+      "step": 14047
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.3723255813953487e-07,
+      "loss": 0.0008,
+      "step": 14048
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 1.3722093023255813e-07,
+      "loss": 0.0008,
+      "step": 14049
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 1.372093023255814e-07,
+      "loss": 0.0006,
+      "step": 14050
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 1.3719767441860465e-07,
+      "loss": 0.0005,
+      "step": 14051
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 1.371860465116279e-07,
+      "loss": 0.0006,
+      "step": 14052
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 1.3717441860465115e-07,
+      "loss": 0.0007,
+      "step": 14053
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 1.371627906976744e-07,
+      "loss": 0.0006,
+      "step": 14054
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 1.3715116279069767e-07,
+      "loss": 0.0095,
+      "step": 14055
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 1.371395348837209e-07,
+      "loss": 0.0005,
+      "step": 14056
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 1.371279069767442e-07,
+      "loss": 0.0008,
+      "step": 14057
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 1.3711627906976743e-07,
+      "loss": 0.0006,
+      "step": 14058
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 1.371046511627907e-07,
+      "loss": 0.0015,
+      "step": 14059
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 1.3709302325581396e-07,
+      "loss": 0.001,
+      "step": 14060
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 1.3708139534883722e-07,
+      "loss": 0.0009,
+      "step": 14061
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 1.3706976744186045e-07,
+      "loss": 0.0005,
+      "step": 14062
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 1.3705813953488371e-07,
+      "loss": 0.0007,
+      "step": 14063
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 1.3704651162790698e-07,
+      "loss": 0.0006,
+      "step": 14064
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.3703488372093024e-07,
+      "loss": 0.0009,
+      "step": 14065
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.3702325581395347e-07,
+      "loss": 0.0006,
+      "step": 14066
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.3701162790697673e-07,
+      "loss": 0.0011,
+      "step": 14067
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.37e-07,
+      "loss": 0.0013,
+      "step": 14068
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.3698837209302326e-07,
+      "loss": 0.0007,
+      "step": 14069
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 1.369767441860465e-07,
+      "loss": 0.0008,
+      "step": 14070
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 1.3696511627906975e-07,
+      "loss": 0.0006,
+      "step": 14071
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 1.3695348837209302e-07,
+      "loss": 3.3045,
+      "step": 14072
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 1.3694186046511628e-07,
+      "loss": 0.0008,
+      "step": 14073
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 1.3693023255813954e-07,
+      "loss": 0.0006,
+      "step": 14074
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 1.3691860465116277e-07,
+      "loss": 0.0006,
+      "step": 14075
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.3690697674418604e-07,
+      "loss": 0.0005,
+      "step": 14076
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.368953488372093e-07,
+      "loss": 0.0008,
+      "step": 14077
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.3688372093023256e-07,
+      "loss": 0.0007,
+      "step": 14078
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.368720930232558e-07,
+      "loss": 0.0006,
+      "step": 14079
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 1.3686046511627906e-07,
+      "loss": 0.0006,
+      "step": 14080
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.3684883720930232e-07,
+      "loss": 0.0008,
+      "step": 14081
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.3683720930232558e-07,
+      "loss": 0.0007,
+      "step": 14082
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.3682558139534882e-07,
+      "loss": 0.0007,
+      "step": 14083
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.368139534883721e-07,
+      "loss": 0.0005,
+      "step": 14084
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.3680232558139534e-07,
+      "loss": 0.0011,
+      "step": 14085
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 1.367906976744186e-07,
+      "loss": 0.001,
+      "step": 14086
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 1.3677906976744186e-07,
+      "loss": 0.0017,
+      "step": 14087
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 1.3676744186046512e-07,
+      "loss": 0.0007,
+      "step": 14088
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 1.3675581395348836e-07,
+      "loss": 0.0022,
+      "step": 14089
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 1.3674418604651162e-07,
+      "loss": 0.0134,
+      "step": 14090
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.3673255813953488e-07,
+      "loss": 0.2303,
+      "step": 14091
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.3672093023255814e-07,
+      "loss": 0.0005,
+      "step": 14092
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.3670930232558138e-07,
+      "loss": 0.0007,
+      "step": 14093
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.3669767441860464e-07,
+      "loss": 0.0005,
+      "step": 14094
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.366860465116279e-07,
+      "loss": 0.0005,
+      "step": 14095
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 1.3667441860465117e-07,
+      "loss": 0.0017,
+      "step": 14096
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 1.366627906976744e-07,
+      "loss": 0.0007,
+      "step": 14097
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 1.3665116279069766e-07,
+      "loss": 0.0011,
+      "step": 14098
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 1.3663953488372092e-07,
+      "loss": 0.0015,
+      "step": 14099
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 1.3662790697674419e-07,
+      "loss": 0.3855,
+      "step": 14100
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 1.3661627906976745e-07,
+      "loss": 0.0009,
+      "step": 14101
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 1.3660465116279068e-07,
+      "loss": 0.0007,
+      "step": 14102
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 1.3659302325581394e-07,
+      "loss": 0.0007,
+      "step": 14103
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 1.365813953488372e-07,
+      "loss": 0.001,
+      "step": 14104
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 1.3656976744186047e-07,
+      "loss": 0.0006,
+      "step": 14105
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 1.365581395348837e-07,
+      "loss": 0.0005,
+      "step": 14106
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.3654651162790696e-07,
+      "loss": 0.0112,
+      "step": 14107
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.3653488372093023e-07,
+      "loss": 0.0008,
+      "step": 14108
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.365232558139535e-07,
+      "loss": 0.0009,
+      "step": 14109
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.3651162790697672e-07,
+      "loss": 0.0007,
+      "step": 14110
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 1.365e-07,
+      "loss": 0.0005,
+      "step": 14111
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 1.3648837209302325e-07,
+      "loss": 0.0009,
+      "step": 14112
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 1.364767441860465e-07,
+      "loss": 0.0006,
+      "step": 14113
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 1.3646511627906977e-07,
+      "loss": 0.001,
+      "step": 14114
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 1.3645348837209303e-07,
+      "loss": 0.0937,
+      "step": 14115
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 1.3644186046511627e-07,
+      "loss": 0.0006,
+      "step": 14116
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 1.3643023255813953e-07,
+      "loss": 0.0011,
+      "step": 14117
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 1.364186046511628e-07,
+      "loss": 0.0007,
+      "step": 14118
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 1.3640697674418605e-07,
+      "loss": 0.0006,
+      "step": 14119
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 1.363953488372093e-07,
+      "loss": 0.0005,
+      "step": 14120
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 1.3638372093023255e-07,
+      "loss": 1.8889,
+      "step": 14121
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.363720930232558e-07,
+      "loss": 0.0007,
+      "step": 14122
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.3636046511627905e-07,
+      "loss": 0.0006,
+      "step": 14123
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.363488372093023e-07,
+      "loss": 0.0009,
+      "step": 14124
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.3633720930232557e-07,
+      "loss": 0.0023,
+      "step": 14125
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.3632558139534883e-07,
+      "loss": 0.0013,
+      "step": 14126
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 1.363139534883721e-07,
+      "loss": 0.0006,
+      "step": 14127
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 1.3630232558139535e-07,
+      "loss": 0.0006,
+      "step": 14128
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 1.362906976744186e-07,
+      "loss": 0.0005,
+      "step": 14129
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 1.3627906976744185e-07,
+      "loss": 0.0008,
+      "step": 14130
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 1.362674418604651e-07,
+      "loss": 0.0009,
+      "step": 14131
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 1.3625581395348837e-07,
+      "loss": 0.0007,
+      "step": 14132
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 1.362441860465116e-07,
+      "loss": 0.0007,
+      "step": 14133
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 1.3623255813953487e-07,
+      "loss": 0.0007,
+      "step": 14134
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 1.3622093023255813e-07,
+      "loss": 0.0007,
+      "step": 14135
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 1.362093023255814e-07,
+      "loss": 0.0007,
+      "step": 14136
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 1.3619767441860463e-07,
+      "loss": 0.0005,
+      "step": 14137
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.3618604651162792e-07,
+      "loss": 0.0006,
+      "step": 14138
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.3617441860465115e-07,
+      "loss": 0.086,
+      "step": 14139
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.3616279069767442e-07,
+      "loss": 0.0778,
+      "step": 14140
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.3615116279069768e-07,
+      "loss": 0.0006,
+      "step": 14141
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 1.3613953488372094e-07,
+      "loss": 0.0009,
+      "step": 14142
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 1.3612790697674417e-07,
+      "loss": 0.0006,
+      "step": 14143
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 1.3611627906976744e-07,
+      "loss": 0.0006,
+      "step": 14144
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 1.361046511627907e-07,
+      "loss": 0.0025,
+      "step": 14145
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 1.3609302325581396e-07,
+      "loss": 0.0006,
+      "step": 14146
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 1.360813953488372e-07,
+      "loss": 0.0008,
+      "step": 14147
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 1.3606976744186046e-07,
+      "loss": 0.0006,
+      "step": 14148
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 1.3605813953488372e-07,
+      "loss": 0.0007,
+      "step": 14149
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 1.3604651162790695e-07,
+      "loss": 0.0007,
+      "step": 14150
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 1.3603488372093022e-07,
+      "loss": 0.0015,
+      "step": 14151
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 1.3602325581395348e-07,
+      "loss": 0.0006,
+      "step": 14152
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.3601162790697674e-07,
+      "loss": 0.0005,
+      "step": 14153
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.3599999999999997e-07,
+      "loss": 0.0006,
+      "step": 14154
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.3598837209302326e-07,
+      "loss": 0.0011,
+      "step": 14155
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.359767441860465e-07,
+      "loss": 0.0006,
+      "step": 14156
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.3596511627906976e-07,
+      "loss": 0.0007,
+      "step": 14157
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 1.3595348837209302e-07,
+      "loss": 0.0052,
+      "step": 14158
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 1.3594186046511628e-07,
+      "loss": 0.0007,
+      "step": 14159
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 1.3593023255813952e-07,
+      "loss": 0.0006,
+      "step": 14160
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 1.3591860465116278e-07,
+      "loss": 0.0006,
+      "step": 14161
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 1.3590697674418604e-07,
+      "loss": 0.0021,
+      "step": 14162
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 1.358953488372093e-07,
+      "loss": 0.0007,
+      "step": 14163
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.3588372093023254e-07,
+      "loss": 0.0019,
+      "step": 14164
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.358720930232558e-07,
+      "loss": 0.0006,
+      "step": 14165
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.3586046511627906e-07,
+      "loss": 0.1733,
+      "step": 14166
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.3584883720930232e-07,
+      "loss": 0.0006,
+      "step": 14167
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 1.3583720930232558e-07,
+      "loss": 0.0016,
+      "step": 14168
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 1.3582558139534885e-07,
+      "loss": 0.0005,
+      "step": 14169
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 1.3581395348837208e-07,
+      "loss": 0.0017,
+      "step": 14170
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 1.3580232558139534e-07,
+      "loss": 0.0006,
+      "step": 14171
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 1.357906976744186e-07,
+      "loss": 0.0008,
+      "step": 14172
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 1.3577906976744187e-07,
+      "loss": 3.2857,
+      "step": 14173
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.357674418604651e-07,
+      "loss": 0.0172,
+      "step": 14174
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.3575581395348836e-07,
+      "loss": 0.0008,
+      "step": 14175
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.3574418604651163e-07,
+      "loss": 0.0006,
+      "step": 14176
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.3573255813953486e-07,
+      "loss": 0.0007,
+      "step": 14177
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 1.3572093023255812e-07,
+      "loss": 0.0006,
+      "step": 14178
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 1.3570930232558138e-07,
+      "loss": 0.0008,
+      "step": 14179
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 1.3569767441860465e-07,
+      "loss": 0.0006,
+      "step": 14180
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 1.3568604651162788e-07,
+      "loss": 0.5745,
+      "step": 14181
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 1.3567441860465117e-07,
+      "loss": 0.0008,
+      "step": 14182
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 1.356627906976744e-07,
+      "loss": 0.0013,
+      "step": 14183
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.3565116279069767e-07,
+      "loss": 2.3219,
+      "step": 14184
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.3563953488372093e-07,
+      "loss": 0.0007,
+      "step": 14185
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.356279069767442e-07,
+      "loss": 0.0006,
+      "step": 14186
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.3561627906976742e-07,
+      "loss": 0.0005,
+      "step": 14187
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.3560465116279069e-07,
+      "loss": 0.0007,
+      "step": 14188
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 1.3559302325581395e-07,
+      "loss": 0.0005,
+      "step": 14189
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 1.355813953488372e-07,
+      "loss": 0.0009,
+      "step": 14190
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 1.3556976744186045e-07,
+      "loss": 0.0009,
+      "step": 14191
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 1.355581395348837e-07,
+      "loss": 0.0007,
+      "step": 14192
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 1.3554651162790697e-07,
+      "loss": 0.0007,
+      "step": 14193
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 1.3553488372093023e-07,
+      "loss": 0.0005,
+      "step": 14194
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 1.355232558139535e-07,
+      "loss": 0.0007,
+      "step": 14195
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 1.3551162790697675e-07,
+      "loss": 0.0004,
+      "step": 14196
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 1.355e-07,
+      "loss": 0.0004,
+      "step": 14197
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 1.3548837209302325e-07,
+      "loss": 0.0006,
+      "step": 14198
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 1.354767441860465e-07,
+      "loss": 0.0007,
+      "step": 14199
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 1.3546511627906977e-07,
+      "loss": 0.0005,
+      "step": 14200
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 1.35453488372093e-07,
+      "loss": 0.0005,
+      "step": 14201
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 1.3544186046511627e-07,
+      "loss": 0.001,
+      "step": 14202
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 1.3543023255813953e-07,
+      "loss": 0.0202,
+      "step": 14203
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 1.3541860465116277e-07,
+      "loss": 0.0394,
+      "step": 14204
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.3540697674418603e-07,
+      "loss": 0.0005,
+      "step": 14205
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.353953488372093e-07,
+      "loss": 0.0007,
+      "step": 14206
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.3538372093023255e-07,
+      "loss": 0.003,
+      "step": 14207
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.353720930232558e-07,
+      "loss": 0.0009,
+      "step": 14208
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 1.3536046511627908e-07,
+      "loss": 0.0011,
+      "step": 14209
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 1.353488372093023e-07,
+      "loss": 0.0008,
+      "step": 14210
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 1.3533720930232557e-07,
+      "loss": 0.0005,
+      "step": 14211
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 1.3532558139534884e-07,
+      "loss": 0.0008,
+      "step": 14212
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 1.353139534883721e-07,
+      "loss": 0.1592,
+      "step": 14213
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 1.3530232558139533e-07,
+      "loss": 0.0005,
+      "step": 14214
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.352906976744186e-07,
+      "loss": 0.0005,
+      "step": 14215
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.3527906976744186e-07,
+      "loss": 0.0008,
+      "step": 14216
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.3526744186046512e-07,
+      "loss": 0.0008,
+      "step": 14217
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.3525581395348835e-07,
+      "loss": 0.0006,
+      "step": 14218
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.3524418604651161e-07,
+      "loss": 0.0005,
+      "step": 14219
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 1.3523255813953488e-07,
+      "loss": 0.0006,
+      "step": 14220
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 1.3522093023255814e-07,
+      "loss": 0.0006,
+      "step": 14221
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 1.352093023255814e-07,
+      "loss": 0.0008,
+      "step": 14222
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 1.3519767441860466e-07,
+      "loss": 0.0007,
+      "step": 14223
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 1.351860465116279e-07,
+      "loss": 0.0006,
+      "step": 14224
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 1.3517441860465116e-07,
+      "loss": 0.0005,
+      "step": 14225
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 1.3516279069767442e-07,
+      "loss": 0.0006,
+      "step": 14226
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 1.3515116279069768e-07,
+      "loss": 0.0008,
+      "step": 14227
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 1.3513953488372092e-07,
+      "loss": 0.0005,
+      "step": 14228
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 1.3512790697674418e-07,
+      "loss": 0.0013,
+      "step": 14229
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 1.3511627906976744e-07,
+      "loss": 0.0084,
+      "step": 14230
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 1.3510465116279068e-07,
+      "loss": 0.0007,
+      "step": 14231
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 1.3509302325581394e-07,
+      "loss": 0.0006,
+      "step": 14232
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 1.350813953488372e-07,
+      "loss": 0.0016,
+      "step": 14233
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 1.3506976744186046e-07,
+      "loss": 0.0009,
+      "step": 14234
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 1.350581395348837e-07,
+      "loss": 0.0007,
+      "step": 14235
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 1.3504651162790698e-07,
+      "loss": 0.0004,
+      "step": 14236
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 1.3503488372093022e-07,
+      "loss": 0.0008,
+      "step": 14237
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 1.3502325581395348e-07,
+      "loss": 0.0225,
+      "step": 14238
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 1.3501162790697674e-07,
+      "loss": 0.0007,
+      "step": 14239
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 1.35e-07,
+      "loss": 0.2176,
+      "step": 14240
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 1.3498837209302324e-07,
+      "loss": 0.0005,
+      "step": 14241
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 1.349767441860465e-07,
+      "loss": 0.0008,
+      "step": 14242
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 1.3496511627906976e-07,
+      "loss": 0.0125,
+      "step": 14243
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 1.3495348837209302e-07,
+      "loss": 0.0009,
+      "step": 14244
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 1.3494186046511626e-07,
+      "loss": 0.0005,
+      "step": 14245
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 1.3493023255813952e-07,
+      "loss": 0.0008,
+      "step": 14246
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 1.3491860465116278e-07,
+      "loss": 0.0005,
+      "step": 14247
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 1.3490697674418605e-07,
+      "loss": 0.0006,
+      "step": 14248
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 1.348953488372093e-07,
+      "loss": 0.0008,
+      "step": 14249
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 1.3488372093023257e-07,
+      "loss": 0.0008,
+      "step": 14250
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 1.348720930232558e-07,
+      "loss": 0.0005,
+      "step": 14251
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 1.3486046511627907e-07,
+      "loss": 0.0008,
+      "step": 14252
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 1.3484883720930233e-07,
+      "loss": 0.0007,
+      "step": 14253
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 1.348372093023256e-07,
+      "loss": 0.0013,
+      "step": 14254
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 1.3482558139534882e-07,
+      "loss": 0.0005,
+      "step": 14255
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 1.3481395348837209e-07,
+      "loss": 0.0023,
+      "step": 14256
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 1.3480232558139535e-07,
+      "loss": 0.0016,
+      "step": 14257
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 1.3479069767441858e-07,
+      "loss": 0.0008,
+      "step": 14258
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 1.3477906976744184e-07,
+      "loss": 0.0007,
+      "step": 14259
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 1.347674418604651e-07,
+      "loss": 0.0005,
+      "step": 14260
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 1.3475581395348837e-07,
+      "loss": 0.0006,
+      "step": 14261
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 1.347441860465116e-07,
+      "loss": 0.0006,
+      "step": 14262
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 1.347325581395349e-07,
+      "loss": 0.001,
+      "step": 14263
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 1.3472093023255813e-07,
+      "loss": 0.0006,
+      "step": 14264
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 1.347093023255814e-07,
+      "loss": 0.0008,
+      "step": 14265
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 1.3469767441860465e-07,
+      "loss": 0.0094,
+      "step": 14266
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 1.346860465116279e-07,
+      "loss": 0.0049,
+      "step": 14267
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 1.3467441860465115e-07,
+      "loss": 0.0007,
+      "step": 14268
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 1.346627906976744e-07,
+      "loss": 0.0008,
+      "step": 14269
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 1.3465116279069767e-07,
+      "loss": 0.0008,
+      "step": 14270
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 1.3463953488372093e-07,
+      "loss": 0.001,
+      "step": 14271
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 1.3462790697674417e-07,
+      "loss": 0.0005,
+      "step": 14272
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 1.3461627906976743e-07,
+      "loss": 0.0006,
+      "step": 14273
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 1.346046511627907e-07,
+      "loss": 0.0008,
+      "step": 14274
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 1.3459302325581395e-07,
+      "loss": 0.0007,
+      "step": 14275
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 1.3458139534883721e-07,
+      "loss": 0.0008,
+      "step": 14276
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 1.3456976744186048e-07,
+      "loss": 0.0007,
+      "step": 14277
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 1.345581395348837e-07,
+      "loss": 0.0006,
+      "step": 14278
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 1.3454651162790697e-07,
+      "loss": 0.0006,
+      "step": 14279
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 1.3453488372093023e-07,
+      "loss": 0.0006,
+      "step": 14280
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 1.345232558139535e-07,
+      "loss": 0.0006,
+      "step": 14281
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 1.3451162790697673e-07,
+      "loss": 0.0005,
+      "step": 14282
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 1.345e-07,
+      "loss": 0.0006,
+      "step": 14283
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 1.3448837209302326e-07,
+      "loss": 0.0007,
+      "step": 14284
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 1.344767441860465e-07,
+      "loss": 0.0008,
+      "step": 14285
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 1.3446511627906975e-07,
+      "loss": 0.0005,
+      "step": 14286
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 1.3445348837209301e-07,
+      "loss": 2.5605,
+      "step": 14287
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 1.3444186046511628e-07,
+      "loss": 0.0008,
+      "step": 14288
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 1.344302325581395e-07,
+      "loss": 0.0007,
+      "step": 14289
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 1.344186046511628e-07,
+      "loss": 0.0007,
+      "step": 14290
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 1.3440697674418603e-07,
+      "loss": 0.0005,
+      "step": 14291
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 1.343953488372093e-07,
+      "loss": 0.0006,
+      "step": 14292
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 1.3438372093023256e-07,
+      "loss": 0.312,
+      "step": 14293
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 1.3437209302325582e-07,
+      "loss": 0.0007,
+      "step": 14294
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 1.3436046511627905e-07,
+      "loss": 0.0007,
+      "step": 14295
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 1.3434883720930232e-07,
+      "loss": 0.0005,
+      "step": 14296
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 1.3433720930232558e-07,
+      "loss": 0.0007,
+      "step": 14297
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 1.3432558139534884e-07,
+      "loss": 0.0005,
+      "step": 14298
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 1.3431395348837207e-07,
+      "loss": 0.0029,
+      "step": 14299
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 1.3430232558139534e-07,
+      "loss": 0.0006,
+      "step": 14300
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 1.342906976744186e-07,
+      "loss": 0.0005,
+      "step": 14301
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 1.3427906976744186e-07,
+      "loss": 0.0005,
+      "step": 14302
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 1.342674418604651e-07,
+      "loss": 0.0005,
+      "step": 14303
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 1.3425581395348838e-07,
+      "loss": 0.0005,
+      "step": 14304
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 1.3424418604651162e-07,
+      "loss": 0.0008,
+      "step": 14305
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 1.3423255813953488e-07,
+      "loss": 0.0007,
+      "step": 14306
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 1.3422093023255814e-07,
+      "loss": 1.657,
+      "step": 14307
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.342093023255814e-07,
+      "loss": 0.0008,
+      "step": 14308
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.3419767441860464e-07,
+      "loss": 0.0008,
+      "step": 14309
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.341860465116279e-07,
+      "loss": 0.0012,
+      "step": 14310
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.3417441860465116e-07,
+      "loss": 0.0007,
+      "step": 14311
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.341627906976744e-07,
+      "loss": 0.0021,
+      "step": 14312
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 1.3415116279069766e-07,
+      "loss": 0.0005,
+      "step": 14313
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 1.3413953488372092e-07,
+      "loss": 0.0031,
+      "step": 14314
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 1.3412790697674418e-07,
+      "loss": 0.0007,
+      "step": 14315
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 1.3411627906976742e-07,
+      "loss": 0.0011,
+      "step": 14316
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 1.341046511627907e-07,
+      "loss": 0.0059,
+      "step": 14317
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 1.3409302325581394e-07,
+      "loss": 0.0007,
+      "step": 14318
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 1.340813953488372e-07,
+      "loss": 0.0006,
+      "step": 14319
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 1.3406976744186046e-07,
+      "loss": 0.0007,
+      "step": 14320
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 1.3405813953488373e-07,
+      "loss": 0.0007,
+      "step": 14321
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 1.3404651162790696e-07,
+      "loss": 0.0006,
+      "step": 14322
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 1.3403488372093022e-07,
+      "loss": 0.0006,
+      "step": 14323
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 1.3402325581395349e-07,
+      "loss": 0.0005,
+      "step": 14324
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 1.3401162790697675e-07,
+      "loss": 0.0006,
+      "step": 14325
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 1.3399999999999998e-07,
+      "loss": 0.0008,
+      "step": 14326
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 1.3398837209302324e-07,
+      "loss": 0.0008,
+      "step": 14327
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 1.339767441860465e-07,
+      "loss": 0.0111,
+      "step": 14328
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 1.3396511627906977e-07,
+      "loss": 0.0006,
+      "step": 14329
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 1.33953488372093e-07,
+      "loss": 0.0005,
+      "step": 14330
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 1.339418604651163e-07,
+      "loss": 0.0006,
+      "step": 14331
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 1.3393023255813953e-07,
+      "loss": 0.0005,
+      "step": 14332
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 1.339186046511628e-07,
+      "loss": 0.0007,
+      "step": 14333
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 1.3390697674418605e-07,
+      "loss": 0.0007,
+      "step": 14334
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 1.338953488372093e-07,
+      "loss": 0.0007,
+      "step": 14335
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 1.3388372093023255e-07,
+      "loss": 0.0008,
+      "step": 14336
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 1.338720930232558e-07,
+      "loss": 0.0005,
+      "step": 14337
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 1.3386046511627907e-07,
+      "loss": 0.0005,
+      "step": 14338
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.338488372093023e-07,
+      "loss": 0.0005,
+      "step": 14339
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.3383720930232557e-07,
+      "loss": 0.0008,
+      "step": 14340
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.3382558139534883e-07,
+      "loss": 0.0006,
+      "step": 14341
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.338139534883721e-07,
+      "loss": 0.0007,
+      "step": 14342
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.3380232558139533e-07,
+      "loss": 0.0008,
+      "step": 14343
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 1.3379069767441861e-07,
+      "loss": 0.0008,
+      "step": 14344
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 1.3377906976744185e-07,
+      "loss": 0.0006,
+      "step": 14345
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 1.337674418604651e-07,
+      "loss": 0.0007,
+      "step": 14346
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 1.3375581395348837e-07,
+      "loss": 0.0005,
+      "step": 14347
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 1.3374418604651163e-07,
+      "loss": 0.0006,
+      "step": 14348
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 1.3373255813953487e-07,
+      "loss": 0.0006,
+      "step": 14349
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 1.3372093023255813e-07,
+      "loss": 0.0005,
+      "step": 14350
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 1.337093023255814e-07,
+      "loss": 0.4535,
+      "step": 14351
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 1.3369767441860465e-07,
+      "loss": 0.0006,
+      "step": 14352
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 1.336860465116279e-07,
+      "loss": 0.0009,
+      "step": 14353
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 1.3367441860465115e-07,
+      "loss": 0.0006,
+      "step": 14354
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 1.336627906976744e-07,
+      "loss": 0.0006,
+      "step": 14355
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 1.3365116279069767e-07,
+      "loss": 0.0006,
+      "step": 14356
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 1.336395348837209e-07,
+      "loss": 0.0007,
+      "step": 14357
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 1.336279069767442e-07,
+      "loss": 0.0007,
+      "step": 14358
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 1.3361627906976743e-07,
+      "loss": 0.0007,
+      "step": 14359
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 1.336046511627907e-07,
+      "loss": 0.0008,
+      "step": 14360
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 1.3359302325581396e-07,
+      "loss": 0.0006,
+      "step": 14361
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 1.3358139534883722e-07,
+      "loss": 3.0501,
+      "step": 14362
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 1.3356976744186045e-07,
+      "loss": 0.0014,
+      "step": 14363
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 1.3355813953488372e-07,
+      "loss": 0.0006,
+      "step": 14364
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 1.3354651162790698e-07,
+      "loss": 0.0004,
+      "step": 14365
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 1.335348837209302e-07,
+      "loss": 0.0007,
+      "step": 14366
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 1.3352325581395347e-07,
+      "loss": 0.0008,
+      "step": 14367
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 1.3351162790697674e-07,
+      "loss": 0.0012,
+      "step": 14368
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 1.335e-07,
+      "loss": 0.0007,
+      "step": 14369
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 1.3348837209302323e-07,
+      "loss": 0.0009,
+      "step": 14370
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 1.3347674418604652e-07,
+      "loss": 0.0005,
+      "step": 14371
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 1.3346511627906976e-07,
+      "loss": 0.0005,
+      "step": 14372
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 1.3345348837209302e-07,
+      "loss": 0.0007,
+      "step": 14373
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 1.3344186046511628e-07,
+      "loss": 0.0009,
+      "step": 14374
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 1.3343023255813954e-07,
+      "loss": 0.0009,
+      "step": 14375
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 1.3341860465116278e-07,
+      "loss": 0.0006,
+      "step": 14376
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 1.3340697674418604e-07,
+      "loss": 0.0006,
+      "step": 14377
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 1.333953488372093e-07,
+      "loss": 0.0005,
+      "step": 14378
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 1.3338372093023256e-07,
+      "loss": 0.0571,
+      "step": 14379
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 1.333720930232558e-07,
+      "loss": 0.2747,
+      "step": 14380
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 1.3336046511627906e-07,
+      "loss": 0.0006,
+      "step": 14381
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 1.3334883720930232e-07,
+      "loss": 0.0005,
+      "step": 14382
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 1.3333720930232558e-07,
+      "loss": 0.0008,
+      "step": 14383
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 1.3332558139534882e-07,
+      "loss": 0.0007,
+      "step": 14384
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 1.333139534883721e-07,
+      "loss": 0.001,
+      "step": 14385
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 1.3330232558139534e-07,
+      "loss": 0.0006,
+      "step": 14386
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 1.332906976744186e-07,
+      "loss": 0.0009,
+      "step": 14387
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 1.3327906976744186e-07,
+      "loss": 0.0007,
+      "step": 14388
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 1.3326744186046513e-07,
+      "loss": 0.0007,
+      "step": 14389
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 1.3325581395348836e-07,
+      "loss": 0.0008,
+      "step": 14390
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 1.3324418604651162e-07,
+      "loss": 0.0005,
+      "step": 14391
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 1.3323255813953488e-07,
+      "loss": 0.0005,
+      "step": 14392
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 1.3322093023255812e-07,
+      "loss": 0.0006,
+      "step": 14393
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 1.3320930232558138e-07,
+      "loss": 1.77,
+      "step": 14394
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 1.3319767441860464e-07,
+      "loss": 0.0005,
+      "step": 14395
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 1.331860465116279e-07,
+      "loss": 0.0008,
+      "step": 14396
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 1.3317441860465114e-07,
+      "loss": 0.0005,
+      "step": 14397
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 1.3316279069767443e-07,
+      "loss": 0.0006,
+      "step": 14398
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 1.3315116279069766e-07,
+      "loss": 0.0006,
+      "step": 14399
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 1.3313953488372093e-07,
+      "loss": 0.0007,
+      "step": 14400
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 1.331279069767442e-07,
+      "loss": 0.0005,
+      "step": 14401
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 1.3311627906976745e-07,
+      "loss": 0.0008,
+      "step": 14402
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 1.3310465116279068e-07,
+      "loss": 0.0005,
+      "step": 14403
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 1.3309302325581395e-07,
+      "loss": 0.0007,
+      "step": 14404
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 1.330813953488372e-07,
+      "loss": 0.0011,
+      "step": 14405
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 1.3306976744186047e-07,
+      "loss": 0.0007,
+      "step": 14406
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 1.330581395348837e-07,
+      "loss": 0.0005,
+      "step": 14407
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 1.3304651162790697e-07,
+      "loss": 0.001,
+      "step": 14408
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 1.3303488372093023e-07,
+      "loss": 0.0009,
+      "step": 14409
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 1.330232558139535e-07,
+      "loss": 0.0007,
+      "step": 14410
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 1.3301162790697672e-07,
+      "loss": 0.0005,
+      "step": 14411
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 1.33e-07,
+      "loss": 0.0004,
+      "step": 14412
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 1.3298837209302325e-07,
+      "loss": 0.0019,
+      "step": 14413
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 1.329767441860465e-07,
+      "loss": 3.56,
+      "step": 14414
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 1.3296511627906977e-07,
+      "loss": 0.001,
+      "step": 14415
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 1.32953488372093e-07,
+      "loss": 0.0007,
+      "step": 14416
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 1.3294186046511627e-07,
+      "loss": 0.0006,
+      "step": 14417
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 1.3293023255813953e-07,
+      "loss": 0.0025,
+      "step": 14418
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 1.329186046511628e-07,
+      "loss": 0.0007,
+      "step": 14419
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 1.3290697674418603e-07,
+      "loss": 0.0007,
+      "step": 14420
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 1.328953488372093e-07,
+      "loss": 0.0005,
+      "step": 14421
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 1.3288372093023255e-07,
+      "loss": 0.0011,
+      "step": 14422
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 1.328720930232558e-07,
+      "loss": 0.0007,
+      "step": 14423
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 1.3286046511627905e-07,
+      "loss": 0.0006,
+      "step": 14424
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 1.328488372093023e-07,
+      "loss": 0.0239,
+      "step": 14425
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 1.3283720930232557e-07,
+      "loss": 0.1739,
+      "step": 14426
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 1.3282558139534883e-07,
+      "loss": 0.0006,
+      "step": 14427
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 1.328139534883721e-07,
+      "loss": 0.0017,
+      "step": 14428
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 1.3280232558139536e-07,
+      "loss": 0.0006,
+      "step": 14429
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 1.327906976744186e-07,
+      "loss": 0.0006,
+      "step": 14430
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 1.3277906976744185e-07,
+      "loss": 0.0007,
+      "step": 14431
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 1.3276744186046511e-07,
+      "loss": 0.0011,
+      "step": 14432
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 1.3275581395348838e-07,
+      "loss": 0.0008,
+      "step": 14433
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 1.327441860465116e-07,
+      "loss": 0.0008,
+      "step": 14434
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 1.3273255813953487e-07,
+      "loss": 0.001,
+      "step": 14435
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 1.3272093023255814e-07,
+      "loss": 0.0009,
+      "step": 14436
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 1.327093023255814e-07,
+      "loss": 0.0007,
+      "step": 14437
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 1.3269767441860463e-07,
+      "loss": 0.0008,
+      "step": 14438
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 1.3268604651162792e-07,
+      "loss": 0.0006,
+      "step": 14439
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 1.3267441860465116e-07,
+      "loss": 0.0006,
+      "step": 14440
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 1.3266279069767442e-07,
+      "loss": 0.0011,
+      "step": 14441
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 1.3265116279069768e-07,
+      "loss": 0.0006,
+      "step": 14442
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 1.3263953488372091e-07,
+      "loss": 0.0005,
+      "step": 14443
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 1.3262790697674418e-07,
+      "loss": 0.0006,
+      "step": 14444
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 1.3261627906976744e-07,
+      "loss": 0.0007,
+      "step": 14445
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 1.326046511627907e-07,
+      "loss": 0.0007,
+      "step": 14446
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 1.3259302325581393e-07,
+      "loss": 0.0005,
+      "step": 14447
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 1.325813953488372e-07,
+      "loss": 0.0007,
+      "step": 14448
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 1.3256976744186046e-07,
+      "loss": 0.0013,
+      "step": 14449
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 1.3255813953488372e-07,
+      "loss": 0.0006,
+      "step": 14450
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 1.3254651162790696e-07,
+      "loss": 2.1879,
+      "step": 14451
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 1.3253488372093022e-07,
+      "loss": 0.0005,
+      "step": 14452
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 1.3252325581395348e-07,
+      "loss": 0.0006,
+      "step": 14453
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 1.3251162790697674e-07,
+      "loss": 0.0013,
+      "step": 14454
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 1.325e-07,
+      "loss": 0.0005,
+      "step": 14455
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 1.3248837209302326e-07,
+      "loss": 0.0008,
+      "step": 14456
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 1.324767441860465e-07,
+      "loss": 0.0005,
+      "step": 14457
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 1.3246511627906976e-07,
+      "loss": 0.0009,
+      "step": 14458
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 1.3245348837209302e-07,
+      "loss": 0.0006,
+      "step": 14459
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 1.3244186046511628e-07,
+      "loss": 0.0005,
+      "step": 14460
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 1.3243023255813952e-07,
+      "loss": 0.0008,
+      "step": 14461
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 1.3241860465116278e-07,
+      "loss": 0.0008,
+      "step": 14462
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 1.3240697674418604e-07,
+      "loss": 0.0009,
+      "step": 14463
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 1.323953488372093e-07,
+      "loss": 0.0004,
+      "step": 14464
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 1.3238372093023254e-07,
+      "loss": 0.0005,
+      "step": 14465
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 1.3237209302325583e-07,
+      "loss": 0.0005,
+      "step": 14466
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 1.3236046511627906e-07,
+      "loss": 0.0005,
+      "step": 14467
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 1.3234883720930232e-07,
+      "loss": 0.0007,
+      "step": 14468
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 1.3233720930232559e-07,
+      "loss": 0.0005,
+      "step": 14469
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 1.3232558139534882e-07,
+      "loss": 0.0009,
+      "step": 14470
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 1.3231395348837208e-07,
+      "loss": 0.0005,
+      "step": 14471
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 1.3230232558139535e-07,
+      "loss": 0.0006,
+      "step": 14472
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 1.322906976744186e-07,
+      "loss": 0.0008,
+      "step": 14473
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 1.3227906976744184e-07,
+      "loss": 0.0009,
+      "step": 14474
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 1.322674418604651e-07,
+      "loss": 0.0006,
+      "step": 14475
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 1.3225581395348837e-07,
+      "loss": 0.0007,
+      "step": 14476
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.15031574666500092,
+      "eval_runtime": 1.5595,
+      "eval_samples_per_second": 663.015,
+      "eval_steps_per_second": 10.901,
+      "step": 14476
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 1.3224418604651163e-07,
+      "loss": 0.001,
+      "step": 14477
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 1.3223255813953486e-07,
+      "loss": 0.0008,
+      "step": 14478
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 1.3222093023255812e-07,
+      "loss": 0.0007,
+      "step": 14479
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 1.3220930232558139e-07,
+      "loss": 0.0006,
+      "step": 14480
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 1.3219767441860465e-07,
+      "loss": 0.2489,
+      "step": 14481
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 1.321860465116279e-07,
+      "loss": 0.0006,
+      "step": 14482
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 1.3217441860465117e-07,
+      "loss": 0.0005,
+      "step": 14483
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 1.321627906976744e-07,
+      "loss": 0.0005,
+      "step": 14484
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 1.3215116279069767e-07,
+      "loss": 0.0006,
+      "step": 14485
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 1.3213953488372093e-07,
+      "loss": 0.0028,
+      "step": 14486
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 1.321279069767442e-07,
+      "loss": 0.0005,
+      "step": 14487
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 1.3211627906976743e-07,
+      "loss": 0.0007,
+      "step": 14488
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 1.321046511627907e-07,
+      "loss": 0.0007,
+      "step": 14489
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 1.3209302325581395e-07,
+      "loss": 0.0007,
+      "step": 14490
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 1.320813953488372e-07,
+      "loss": 0.0005,
+      "step": 14491
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 1.3206976744186045e-07,
+      "loss": 0.0007,
+      "step": 14492
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 1.3205813953488374e-07,
+      "loss": 0.0006,
+      "step": 14493
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 1.3204651162790697e-07,
+      "loss": 0.0006,
+      "step": 14494
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 1.3203488372093023e-07,
+      "loss": 0.0006,
+      "step": 14495
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 1.320232558139535e-07,
+      "loss": 0.0005,
+      "step": 14496
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 1.3201162790697673e-07,
+      "loss": 0.0005,
+      "step": 14497
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 1.32e-07,
+      "loss": 0.0007,
+      "step": 14498
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 1.3198837209302325e-07,
+      "loss": 0.0008,
+      "step": 14499
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 1.3197674418604651e-07,
+      "loss": 0.0009,
+      "step": 14500
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 1.3196511627906975e-07,
+      "loss": 0.0005,
+      "step": 14501
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 1.31953488372093e-07,
+      "loss": 0.0007,
+      "step": 14502
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 1.3194186046511627e-07,
+      "loss": 0.0006,
+      "step": 14503
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 1.3193023255813953e-07,
+      "loss": 0.0005,
+      "step": 14504
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.3191860465116277e-07,
+      "loss": 0.0006,
+      "step": 14505
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.3190697674418603e-07,
+      "loss": 0.0006,
+      "step": 14506
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.318953488372093e-07,
+      "loss": 0.0243,
+      "step": 14507
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.3188372093023256e-07,
+      "loss": 0.0006,
+      "step": 14508
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.3187209302325582e-07,
+      "loss": 0.0006,
+      "step": 14509
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 1.3186046511627908e-07,
+      "loss": 0.0006,
+      "step": 14510
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 1.3184883720930231e-07,
+      "loss": 0.0005,
+      "step": 14511
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 1.3183720930232558e-07,
+      "loss": 1.2754,
+      "step": 14512
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 1.3182558139534884e-07,
+      "loss": 0.0016,
+      "step": 14513
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 1.318139534883721e-07,
+      "loss": 0.0013,
+      "step": 14514
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 1.3180232558139533e-07,
+      "loss": 0.0007,
+      "step": 14515
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 1.317906976744186e-07,
+      "loss": 0.0006,
+      "step": 14516
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 1.3177906976744186e-07,
+      "loss": 0.0006,
+      "step": 14517
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 1.3176744186046512e-07,
+      "loss": 0.0009,
+      "step": 14518
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 1.3175581395348835e-07,
+      "loss": 0.0012,
+      "step": 14519
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.3174418604651164e-07,
+      "loss": 0.0008,
+      "step": 14520
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.3173255813953488e-07,
+      "loss": 0.0006,
+      "step": 14521
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.3172093023255814e-07,
+      "loss": 0.0007,
+      "step": 14522
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.317093023255814e-07,
+      "loss": 0.0008,
+      "step": 14523
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.3169767441860464e-07,
+      "loss": 0.0004,
+      "step": 14524
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.316860465116279e-07,
+      "loss": 0.0038,
+      "step": 14525
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 1.3167441860465116e-07,
+      "loss": 0.0006,
+      "step": 14526
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 1.3166279069767442e-07,
+      "loss": 0.0014,
+      "step": 14527
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 1.3165116279069766e-07,
+      "loss": 0.0006,
+      "step": 14528
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 1.3163953488372092e-07,
+      "loss": 0.0004,
+      "step": 14529
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 1.3162790697674418e-07,
+      "loss": 0.0005,
+      "step": 14530
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 1.3161627906976744e-07,
+      "loss": 0.0007,
+      "step": 14531
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 1.3160465116279068e-07,
+      "loss": 0.0006,
+      "step": 14532
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 1.3159302325581394e-07,
+      "loss": 0.0005,
+      "step": 14533
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 1.315813953488372e-07,
+      "loss": 0.0007,
+      "step": 14534
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 1.3156976744186046e-07,
+      "loss": 0.0007,
+      "step": 14535
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 1.315581395348837e-07,
+      "loss": 0.0037,
+      "step": 14536
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 1.3154651162790699e-07,
+      "loss": 0.0007,
+      "step": 14537
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 1.3153488372093022e-07,
+      "loss": 0.1001,
+      "step": 14538
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 1.3152325581395348e-07,
+      "loss": 0.0007,
+      "step": 14539
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 1.3151162790697674e-07,
+      "loss": 0.0006,
+      "step": 14540
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.315e-07,
+      "loss": 0.0007,
+      "step": 14541
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.3148837209302324e-07,
+      "loss": 0.0177,
+      "step": 14542
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.314767441860465e-07,
+      "loss": 0.0007,
+      "step": 14543
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.3146511627906976e-07,
+      "loss": 0.0005,
+      "step": 14544
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.3145348837209303e-07,
+      "loss": 0.0007,
+      "step": 14545
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 1.3144186046511626e-07,
+      "loss": 0.0009,
+      "step": 14546
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 1.3143023255813952e-07,
+      "loss": 0.0004,
+      "step": 14547
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 1.3141860465116279e-07,
+      "loss": 0.0008,
+      "step": 14548
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 1.3140697674418605e-07,
+      "loss": 0.0006,
+      "step": 14549
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 1.313953488372093e-07,
+      "loss": 1.6374,
+      "step": 14550
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.3138372093023254e-07,
+      "loss": 0.0005,
+      "step": 14551
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.313720930232558e-07,
+      "loss": 0.0009,
+      "step": 14552
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.3136046511627907e-07,
+      "loss": 0.0006,
+      "step": 14553
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.3134883720930233e-07,
+      "loss": 0.027,
+      "step": 14554
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.3133720930232556e-07,
+      "loss": 0.0007,
+      "step": 14555
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.3132558139534883e-07,
+      "loss": 0.0005,
+      "step": 14556
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 1.313139534883721e-07,
+      "loss": 0.0007,
+      "step": 14557
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 1.3130232558139535e-07,
+      "loss": 0.0007,
+      "step": 14558
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 1.3129069767441858e-07,
+      "loss": 3.3517,
+      "step": 14559
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 1.3127906976744185e-07,
+      "loss": 0.0009,
+      "step": 14560
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 1.312674418604651e-07,
+      "loss": 0.0006,
+      "step": 14561
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 1.3125581395348837e-07,
+      "loss": 0.001,
+      "step": 14562
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 1.312441860465116e-07,
+      "loss": 0.0007,
+      "step": 14563
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 1.312325581395349e-07,
+      "loss": 0.0007,
+      "step": 14564
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 1.3122093023255813e-07,
+      "loss": 0.0006,
+      "step": 14565
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 1.312093023255814e-07,
+      "loss": 0.0007,
+      "step": 14566
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 1.3119767441860465e-07,
+      "loss": 0.0005,
+      "step": 14567
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 1.3118604651162791e-07,
+      "loss": 0.0029,
+      "step": 14568
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 1.3117441860465115e-07,
+      "loss": 0.0011,
+      "step": 14569
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 1.311627906976744e-07,
+      "loss": 0.0004,
+      "step": 14570
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 1.3115116279069767e-07,
+      "loss": 0.0005,
+      "step": 14571
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.3113953488372093e-07,
+      "loss": 0.0012,
+      "step": 14572
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.3112790697674417e-07,
+      "loss": 0.0007,
+      "step": 14573
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.3111627906976743e-07,
+      "loss": 0.0006,
+      "step": 14574
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.311046511627907e-07,
+      "loss": 0.0006,
+      "step": 14575
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.3109302325581393e-07,
+      "loss": 0.0005,
+      "step": 14576
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 1.3108139534883722e-07,
+      "loss": 0.0007,
+      "step": 14577
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 1.3106976744186045e-07,
+      "loss": 0.0007,
+      "step": 14578
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 1.310581395348837e-07,
+      "loss": 0.0006,
+      "step": 14579
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 1.3104651162790697e-07,
+      "loss": 0.0006,
+      "step": 14580
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 1.3103488372093024e-07,
+      "loss": 0.0005,
+      "step": 14581
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 1.3102325581395347e-07,
+      "loss": 0.0006,
+      "step": 14582
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 1.3101162790697673e-07,
+      "loss": 0.0006,
+      "step": 14583
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 1.31e-07,
+      "loss": 0.0007,
+      "step": 14584
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 1.3098837209302326e-07,
+      "loss": 0.0007,
+      "step": 14585
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 1.309767441860465e-07,
+      "loss": 0.0008,
+      "step": 14586
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 1.3096511627906975e-07,
+      "loss": 0.0006,
+      "step": 14587
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 1.3095348837209302e-07,
+      "loss": 0.0007,
+      "step": 14588
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 1.3094186046511628e-07,
+      "loss": 0.0004,
+      "step": 14589
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 1.309302325581395e-07,
+      "loss": 0.001,
+      "step": 14590
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 1.309186046511628e-07,
+      "loss": 0.0007,
+      "step": 14591
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 1.3090697674418604e-07,
+      "loss": 2.0985,
+      "step": 14592
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 1.308953488372093e-07,
+      "loss": 0.0006,
+      "step": 14593
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 1.3088372093023256e-07,
+      "loss": 2.6642,
+      "step": 14594
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 1.3087209302325582e-07,
+      "loss": 0.0007,
+      "step": 14595
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 1.3086046511627906e-07,
+      "loss": 0.0005,
+      "step": 14596
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 1.3084883720930232e-07,
+      "loss": 0.0006,
+      "step": 14597
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 1.3083720930232558e-07,
+      "loss": 0.0006,
+      "step": 14598
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 1.3082558139534884e-07,
+      "loss": 0.0005,
+      "step": 14599
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 1.3081395348837208e-07,
+      "loss": 0.4873,
+      "step": 14600
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 1.3080232558139534e-07,
+      "loss": 0.0009,
+      "step": 14601
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 1.307906976744186e-07,
+      "loss": 0.0006,
+      "step": 14602
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 1.3077906976744184e-07,
+      "loss": 0.0005,
+      "step": 14603
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 1.3076744186046512e-07,
+      "loss": 0.0008,
+      "step": 14604
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 1.3075581395348836e-07,
+      "loss": 0.0009,
+      "step": 14605
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 1.3074418604651162e-07,
+      "loss": 0.0005,
+      "step": 14606
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 1.3073255813953488e-07,
+      "loss": 0.0006,
+      "step": 14607
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.3072093023255814e-07,
+      "loss": 0.0012,
+      "step": 14608
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.3070930232558138e-07,
+      "loss": 0.001,
+      "step": 14609
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.3069767441860464e-07,
+      "loss": 0.0008,
+      "step": 14610
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.306860465116279e-07,
+      "loss": 0.0008,
+      "step": 14611
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.3067441860465116e-07,
+      "loss": 0.0007,
+      "step": 14612
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.306627906976744e-07,
+      "loss": 0.0006,
+      "step": 14613
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 1.3065116279069766e-07,
+      "loss": 0.0007,
+      "step": 14614
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 1.3063953488372092e-07,
+      "loss": 0.0005,
+      "step": 14615
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 1.3062790697674418e-07,
+      "loss": 0.0006,
+      "step": 14616
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 1.3061627906976742e-07,
+      "loss": 0.0008,
+      "step": 14617
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 1.306046511627907e-07,
+      "loss": 0.0006,
+      "step": 14618
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 1.3059302325581394e-07,
+      "loss": 0.0005,
+      "step": 14619
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 1.305813953488372e-07,
+      "loss": 0.0006,
+      "step": 14620
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 1.3056976744186047e-07,
+      "loss": 0.1631,
+      "step": 14621
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 1.3055813953488373e-07,
+      "loss": 0.0006,
+      "step": 14622
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 1.3054651162790696e-07,
+      "loss": 0.0007,
+      "step": 14623
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 1.3053488372093023e-07,
+      "loss": 0.0006,
+      "step": 14624
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 1.305232558139535e-07,
+      "loss": 0.0011,
+      "step": 14625
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 1.3051162790697675e-07,
+      "loss": 0.0006,
+      "step": 14626
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 1.3049999999999998e-07,
+      "loss": 0.0005,
+      "step": 14627
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 1.3048837209302325e-07,
+      "loss": 0.0005,
+      "step": 14628
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 1.304767441860465e-07,
+      "loss": 0.0005,
+      "step": 14629
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 1.3046511627906974e-07,
+      "loss": 0.0005,
+      "step": 14630
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 1.3045348837209303e-07,
+      "loss": 0.0006,
+      "step": 14631
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 1.3044186046511627e-07,
+      "loss": 0.0006,
+      "step": 14632
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 1.3043023255813953e-07,
+      "loss": 0.0005,
+      "step": 14633
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 1.304186046511628e-07,
+      "loss": 0.0008,
+      "step": 14634
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 1.3040697674418605e-07,
+      "loss": 0.0007,
+      "step": 14635
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 1.3039534883720929e-07,
+      "loss": 0.0008,
+      "step": 14636
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 1.3038372093023255e-07,
+      "loss": 0.0007,
+      "step": 14637
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 1.303720930232558e-07,
+      "loss": 0.0007,
+      "step": 14638
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.3036046511627907e-07,
+      "loss": 0.0012,
+      "step": 14639
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.303488372093023e-07,
+      "loss": 0.0007,
+      "step": 14640
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.3033720930232557e-07,
+      "loss": 0.0008,
+      "step": 14641
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.3032558139534883e-07,
+      "loss": 0.0005,
+      "step": 14642
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.303139534883721e-07,
+      "loss": 0.0006,
+      "step": 14643
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.3030232558139533e-07,
+      "loss": 1.3396,
+      "step": 14644
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 1.3029069767441862e-07,
+      "loss": 0.0008,
+      "step": 14645
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 1.3027906976744185e-07,
+      "loss": 0.0013,
+      "step": 14646
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 1.302674418604651e-07,
+      "loss": 0.0007,
+      "step": 14647
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 1.3025581395348837e-07,
+      "loss": 0.0005,
+      "step": 14648
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 1.3024418604651164e-07,
+      "loss": 0.0006,
+      "step": 14649
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 1.3023255813953487e-07,
+      "loss": 0.0004,
+      "step": 14650
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 1.3022093023255813e-07,
+      "loss": 0.0007,
+      "step": 14651
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 1.302093023255814e-07,
+      "loss": 0.0012,
+      "step": 14652
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 1.3019767441860466e-07,
+      "loss": 0.0006,
+      "step": 14653
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 1.301860465116279e-07,
+      "loss": 0.0007,
+      "step": 14654
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 1.3017441860465115e-07,
+      "loss": 0.0005,
+      "step": 14655
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 1.3016279069767441e-07,
+      "loss": 0.0007,
+      "step": 14656
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 1.3015116279069765e-07,
+      "loss": 0.0099,
+      "step": 14657
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 1.301395348837209e-07,
+      "loss": 0.0006,
+      "step": 14658
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 1.3012790697674417e-07,
+      "loss": 0.0006,
+      "step": 14659
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 1.3011627906976744e-07,
+      "loss": 0.0005,
+      "step": 14660
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 1.301046511627907e-07,
+      "loss": 0.0012,
+      "step": 14661
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 1.3009302325581396e-07,
+      "loss": 0.0006,
+      "step": 14662
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 1.300813953488372e-07,
+      "loss": 0.0007,
+      "step": 14663
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 1.3006976744186046e-07,
+      "loss": 0.0008,
+      "step": 14664
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 1.3005813953488372e-07,
+      "loss": 0.0007,
+      "step": 14665
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 1.3004651162790698e-07,
+      "loss": 0.0006,
+      "step": 14666
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 1.3003488372093021e-07,
+      "loss": 0.0005,
+      "step": 14667
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 1.3002325581395348e-07,
+      "loss": 0.0007,
+      "step": 14668
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 1.3001162790697674e-07,
+      "loss": 0.0006,
+      "step": 14669
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.3e-07,
+      "loss": 0.0944,
+      "step": 14670
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.2998837209302323e-07,
+      "loss": 0.0007,
+      "step": 14671
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.2997674418604652e-07,
+      "loss": 0.0005,
+      "step": 14672
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.2996511627906976e-07,
+      "loss": 0.0006,
+      "step": 14673
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.2995348837209302e-07,
+      "loss": 0.0007,
+      "step": 14674
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.2994186046511628e-07,
+      "loss": 0.0005,
+      "step": 14675
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.2993023255813954e-07,
+      "loss": 0.0007,
+      "step": 14676
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.2991860465116278e-07,
+      "loss": 0.0006,
+      "step": 14677
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.2990697674418604e-07,
+      "loss": 0.0008,
+      "step": 14678
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.298953488372093e-07,
+      "loss": 0.0006,
+      "step": 14679
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.2988372093023256e-07,
+      "loss": 0.0007,
+      "step": 14680
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 1.298720930232558e-07,
+      "loss": 0.0007,
+      "step": 14681
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 1.2986046511627906e-07,
+      "loss": 0.0007,
+      "step": 14682
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 1.2984883720930232e-07,
+      "loss": 0.0007,
+      "step": 14683
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 1.2983720930232556e-07,
+      "loss": 0.0007,
+      "step": 14684
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 1.2982558139534882e-07,
+      "loss": 0.0008,
+      "step": 14685
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 1.2981395348837208e-07,
+      "loss": 0.0007,
+      "step": 14686
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 1.2980232558139534e-07,
+      "loss": 0.0006,
+      "step": 14687
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 1.297906976744186e-07,
+      "loss": 0.0005,
+      "step": 14688
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 1.2977906976744187e-07,
+      "loss": 0.0007,
+      "step": 14689
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 1.297674418604651e-07,
+      "loss": 1.6942,
+      "step": 14690
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 1.2975581395348836e-07,
+      "loss": 0.0007,
+      "step": 14691
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 1.2974418604651162e-07,
+      "loss": 0.0006,
+      "step": 14692
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 1.2973255813953489e-07,
+      "loss": 0.0017,
+      "step": 14693
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 1.2972093023255812e-07,
+      "loss": 0.0033,
+      "step": 14694
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 1.2970930232558138e-07,
+      "loss": 0.0008,
+      "step": 14695
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 1.2969767441860465e-07,
+      "loss": 0.0007,
+      "step": 14696
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 1.296860465116279e-07,
+      "loss": 0.0008,
+      "step": 14697
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 1.2967441860465114e-07,
+      "loss": 0.001,
+      "step": 14698
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 1.2966279069767443e-07,
+      "loss": 0.0008,
+      "step": 14699
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 1.2965116279069767e-07,
+      "loss": 0.0007,
+      "step": 14700
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 1.2963953488372093e-07,
+      "loss": 0.0252,
+      "step": 14701
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 1.296279069767442e-07,
+      "loss": 0.0006,
+      "step": 14702
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 1.2961627906976745e-07,
+      "loss": 0.0006,
+      "step": 14703
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 1.2960465116279069e-07,
+      "loss": 0.0005,
+      "step": 14704
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 1.2959302325581395e-07,
+      "loss": 0.0008,
+      "step": 14705
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 1.295813953488372e-07,
+      "loss": 0.0006,
+      "step": 14706
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.2956976744186047e-07,
+      "loss": 0.0007,
+      "step": 14707
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.295581395348837e-07,
+      "loss": 0.0006,
+      "step": 14708
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.2954651162790697e-07,
+      "loss": 0.0379,
+      "step": 14709
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.2953488372093023e-07,
+      "loss": 0.0006,
+      "step": 14710
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.2952325581395346e-07,
+      "loss": 0.0006,
+      "step": 14711
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 1.2951162790697673e-07,
+      "loss": 0.0006,
+      "step": 14712
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 1.295e-07,
+      "loss": 0.0007,
+      "step": 14713
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 1.2948837209302325e-07,
+      "loss": 0.0007,
+      "step": 14714
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 1.294767441860465e-07,
+      "loss": 0.0007,
+      "step": 14715
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 1.2946511627906977e-07,
+      "loss": 0.0025,
+      "step": 14716
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 1.29453488372093e-07,
+      "loss": 0.0006,
+      "step": 14717
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 1.2944186046511627e-07,
+      "loss": 0.0005,
+      "step": 14718
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 1.2943023255813953e-07,
+      "loss": 0.0006,
+      "step": 14719
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 1.294186046511628e-07,
+      "loss": 0.0005,
+      "step": 14720
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 1.2940697674418603e-07,
+      "loss": 0.0005,
+      "step": 14721
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 1.293953488372093e-07,
+      "loss": 0.001,
+      "step": 14722
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 1.2938372093023255e-07,
+      "loss": 0.0007,
+      "step": 14723
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 1.2937209302325581e-07,
+      "loss": 0.0006,
+      "step": 14724
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 1.2936046511627905e-07,
+      "loss": 0.0014,
+      "step": 14725
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 1.2934883720930234e-07,
+      "loss": 0.0005,
+      "step": 14726
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 1.2933720930232557e-07,
+      "loss": 0.0007,
+      "step": 14727
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 1.2932558139534883e-07,
+      "loss": 0.0005,
+      "step": 14728
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 1.293139534883721e-07,
+      "loss": 0.0005,
+      "step": 14729
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 1.2930232558139536e-07,
+      "loss": 0.0006,
+      "step": 14730
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 1.292906976744186e-07,
+      "loss": 0.0027,
+      "step": 14731
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 1.2927906976744185e-07,
+      "loss": 0.0011,
+      "step": 14732
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 1.2926744186046512e-07,
+      "loss": 0.0006,
+      "step": 14733
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 1.2925581395348838e-07,
+      "loss": 0.0006,
+      "step": 14734
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 1.2924418604651161e-07,
+      "loss": 1.813,
+      "step": 14735
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 1.2923255813953488e-07,
+      "loss": 0.0007,
+      "step": 14736
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 1.2922093023255814e-07,
+      "loss": 0.0007,
+      "step": 14737
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.2920930232558137e-07,
+      "loss": 0.0006,
+      "step": 14738
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.2919767441860463e-07,
+      "loss": 0.0005,
+      "step": 14739
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.291860465116279e-07,
+      "loss": 0.001,
+      "step": 14740
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.2917441860465116e-07,
+      "loss": 0.0005,
+      "step": 14741
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.2916279069767442e-07,
+      "loss": 0.0019,
+      "step": 14742
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.2915116279069768e-07,
+      "loss": 0.0006,
+      "step": 14743
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.2913953488372092e-07,
+      "loss": 0.0007,
+      "step": 14744
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.2912790697674418e-07,
+      "loss": 0.4519,
+      "step": 14745
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.2911627906976744e-07,
+      "loss": 0.0007,
+      "step": 14746
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.291046511627907e-07,
+      "loss": 0.0006,
+      "step": 14747
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.2909302325581394e-07,
+      "loss": 0.0006,
+      "step": 14748
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.290813953488372e-07,
+      "loss": 0.0006,
+      "step": 14749
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.2906976744186046e-07,
+      "loss": 0.0007,
+      "step": 14750
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.2905813953488372e-07,
+      "loss": 0.0004,
+      "step": 14751
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.2904651162790696e-07,
+      "loss": 0.0006,
+      "step": 14752
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.2903488372093025e-07,
+      "loss": 0.009,
+      "step": 14753
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.2902325581395348e-07,
+      "loss": 0.0006,
+      "step": 14754
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.2901162790697674e-07,
+      "loss": 0.0008,
+      "step": 14755
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.29e-07,
+      "loss": 0.0006,
+      "step": 14756
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.2898837209302327e-07,
+      "loss": 0.0006,
+      "step": 14757
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.289767441860465e-07,
+      "loss": 0.0004,
+      "step": 14758
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.2896511627906976e-07,
+      "loss": 0.0008,
+      "step": 14759
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.2895348837209302e-07,
+      "loss": 0.0005,
+      "step": 14760
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.2894186046511629e-07,
+      "loss": 0.0008,
+      "step": 14761
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.2893023255813952e-07,
+      "loss": 0.0006,
+      "step": 14762
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.2891860465116278e-07,
+      "loss": 0.0006,
+      "step": 14763
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.2890697674418604e-07,
+      "loss": 0.0006,
+      "step": 14764
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.2889534883720928e-07,
+      "loss": 0.0007,
+      "step": 14765
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.2888372093023254e-07,
+      "loss": 0.0008,
+      "step": 14766
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.288720930232558e-07,
+      "loss": 0.0008,
+      "step": 14767
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.2886046511627906e-07,
+      "loss": 0.0007,
+      "step": 14768
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.2884883720930233e-07,
+      "loss": 0.0007,
+      "step": 14769
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.288372093023256e-07,
+      "loss": 0.0006,
+      "step": 14770
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.2882558139534882e-07,
+      "loss": 0.0012,
+      "step": 14771
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.2881395348837209e-07,
+      "loss": 2.5545,
+      "step": 14772
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.2880232558139535e-07,
+      "loss": 0.0008,
+      "step": 14773
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.287906976744186e-07,
+      "loss": 0.0008,
+      "step": 14774
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.2877906976744184e-07,
+      "loss": 0.0007,
+      "step": 14775
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.287674418604651e-07,
+      "loss": 0.0006,
+      "step": 14776
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.2875581395348837e-07,
+      "loss": 0.0009,
+      "step": 14777
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.2874418604651163e-07,
+      "loss": 0.0005,
+      "step": 14778
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.2873255813953486e-07,
+      "loss": 0.0006,
+      "step": 14779
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.2872093023255815e-07,
+      "loss": 0.001,
+      "step": 14780
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.287093023255814e-07,
+      "loss": 0.0004,
+      "step": 14781
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.2869767441860465e-07,
+      "loss": 0.0009,
+      "step": 14782
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.286860465116279e-07,
+      "loss": 0.0006,
+      "step": 14783
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.2867441860465117e-07,
+      "loss": 0.0005,
+      "step": 14784
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.286627906976744e-07,
+      "loss": 0.0006,
+      "step": 14785
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.2865116279069767e-07,
+      "loss": 0.0006,
+      "step": 14786
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.2863953488372093e-07,
+      "loss": 0.0006,
+      "step": 14787
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.286279069767442e-07,
+      "loss": 0.0006,
+      "step": 14788
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.2861627906976743e-07,
+      "loss": 0.22,
+      "step": 14789
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.286046511627907e-07,
+      "loss": 0.0005,
+      "step": 14790
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.2859302325581395e-07,
+      "loss": 0.0006,
+      "step": 14791
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.285813953488372e-07,
+      "loss": 0.0007,
+      "step": 14792
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.2856976744186045e-07,
+      "loss": 0.0007,
+      "step": 14793
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.285581395348837e-07,
+      "loss": 0.0005,
+      "step": 14794
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.2854651162790697e-07,
+      "loss": 0.0007,
+      "step": 14795
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.285348837209302e-07,
+      "loss": 0.0049,
+      "step": 14796
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.285232558139535e-07,
+      "loss": 0.0029,
+      "step": 14797
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.2851162790697673e-07,
+      "loss": 0.0006,
+      "step": 14798
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.285e-07,
+      "loss": 0.0008,
+      "step": 14799
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.2848837209302325e-07,
+      "loss": 0.0007,
+      "step": 14800
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.2847674418604652e-07,
+      "loss": 0.0006,
+      "step": 14801
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.2846511627906975e-07,
+      "loss": 0.0008,
+      "step": 14802
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.28453488372093e-07,
+      "loss": 0.0006,
+      "step": 14803
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.2844186046511627e-07,
+      "loss": 0.0009,
+      "step": 14804
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.2843023255813954e-07,
+      "loss": 0.0005,
+      "step": 14805
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.2841860465116277e-07,
+      "loss": 0.0008,
+      "step": 14806
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.2840697674418603e-07,
+      "loss": 0.0006,
+      "step": 14807
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.283953488372093e-07,
+      "loss": 0.0005,
+      "step": 14808
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.2838372093023256e-07,
+      "loss": 0.0006,
+      "step": 14809
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.2837209302325582e-07,
+      "loss": 0.0007,
+      "step": 14810
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.2836046511627908e-07,
+      "loss": 0.0006,
+      "step": 14811
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.2834883720930232e-07,
+      "loss": 0.0006,
+      "step": 14812
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.2833720930232558e-07,
+      "loss": 0.0006,
+      "step": 14813
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.2832558139534884e-07,
+      "loss": 0.0007,
+      "step": 14814
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.283139534883721e-07,
+      "loss": 0.0005,
+      "step": 14815
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.2830232558139534e-07,
+      "loss": 0.0006,
+      "step": 14816
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.282906976744186e-07,
+      "loss": 0.0006,
+      "step": 14817
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.2827906976744186e-07,
+      "loss": 0.0005,
+      "step": 14818
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.282674418604651e-07,
+      "loss": 0.0005,
+      "step": 14819
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.2825581395348836e-07,
+      "loss": 0.0007,
+      "step": 14820
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.2824418604651162e-07,
+      "loss": 0.0011,
+      "step": 14821
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.2823255813953488e-07,
+      "loss": 0.0005,
+      "step": 14822
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.2822093023255811e-07,
+      "loss": 0.0007,
+      "step": 14823
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.282093023255814e-07,
+      "loss": 0.0006,
+      "step": 14824
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.2819767441860464e-07,
+      "loss": 0.0008,
+      "step": 14825
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.281860465116279e-07,
+      "loss": 0.0005,
+      "step": 14826
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.2817441860465116e-07,
+      "loss": 0.0005,
+      "step": 14827
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.2816279069767442e-07,
+      "loss": 0.0006,
+      "step": 14828
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.2815116279069766e-07,
+      "loss": 0.001,
+      "step": 14829
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.2813953488372092e-07,
+      "loss": 0.0008,
+      "step": 14830
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.2812790697674418e-07,
+      "loss": 0.0007,
+      "step": 14831
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.2811627906976744e-07,
+      "loss": 0.0007,
+      "step": 14832
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.2810465116279068e-07,
+      "loss": 0.0009,
+      "step": 14833
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.2809302325581394e-07,
+      "loss": 0.0007,
+      "step": 14834
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.280813953488372e-07,
+      "loss": 0.0012,
+      "step": 14835
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.2806976744186046e-07,
+      "loss": 0.0005,
+      "step": 14836
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.2805813953488373e-07,
+      "loss": 0.0007,
+      "step": 14837
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.28046511627907e-07,
+      "loss": 0.0006,
+      "step": 14838
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.2803488372093022e-07,
+      "loss": 0.0008,
+      "step": 14839
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.2802325581395348e-07,
+      "loss": 0.0007,
+      "step": 14840
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.2801162790697675e-07,
+      "loss": 0.0006,
+      "step": 14841
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.28e-07,
+      "loss": 0.0007,
+      "step": 14842
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.2798837209302324e-07,
+      "loss": 0.0007,
+      "step": 14843
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.279767441860465e-07,
+      "loss": 0.0103,
+      "step": 14844
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.2796511627906977e-07,
+      "loss": 0.0006,
+      "step": 14845
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.27953488372093e-07,
+      "loss": 0.0008,
+      "step": 14846
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.2794186046511626e-07,
+      "loss": 0.0024,
+      "step": 14847
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.2793023255813953e-07,
+      "loss": 0.0005,
+      "step": 14848
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.279186046511628e-07,
+      "loss": 0.0006,
+      "step": 14849
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.2790697674418602e-07,
+      "loss": 0.0006,
+      "step": 14850
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.278953488372093e-07,
+      "loss": 0.0005,
+      "step": 14851
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.2788372093023255e-07,
+      "loss": 0.0005,
+      "step": 14852
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.278720930232558e-07,
+      "loss": 0.0004,
+      "step": 14853
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.2786046511627907e-07,
+      "loss": 0.0006,
+      "step": 14854
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.2784883720930233e-07,
+      "loss": 0.0004,
+      "step": 14855
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.2783720930232557e-07,
+      "loss": 0.0014,
+      "step": 14856
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.2782558139534883e-07,
+      "loss": 0.0006,
+      "step": 14857
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.278139534883721e-07,
+      "loss": 0.0006,
+      "step": 14858
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.2780232558139535e-07,
+      "loss": 0.001,
+      "step": 14859
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.2779069767441859e-07,
+      "loss": 0.4288,
+      "step": 14860
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.2777906976744185e-07,
+      "loss": 0.0012,
+      "step": 14861
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.277674418604651e-07,
+      "loss": 0.0005,
+      "step": 14862
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.2775581395348837e-07,
+      "loss": 0.0004,
+      "step": 14863
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.2774418604651163e-07,
+      "loss": 0.0007,
+      "step": 14864
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.277325581395349e-07,
+      "loss": 0.0006,
+      "step": 14865
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.2772093023255813e-07,
+      "loss": 0.0008,
+      "step": 14866
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.277093023255814e-07,
+      "loss": 0.0007,
+      "step": 14867
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.2769767441860465e-07,
+      "loss": 0.0007,
+      "step": 14868
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.2768604651162792e-07,
+      "loss": 0.0006,
+      "step": 14869
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.2767441860465115e-07,
+      "loss": 0.0008,
+      "step": 14870
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.276627906976744e-07,
+      "loss": 0.001,
+      "step": 14871
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.2765116279069767e-07,
+      "loss": 0.0006,
+      "step": 14872
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.276395348837209e-07,
+      "loss": 0.0005,
+      "step": 14873
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.2762790697674417e-07,
+      "loss": 0.0113,
+      "step": 14874
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.2761627906976743e-07,
+      "loss": 0.0009,
+      "step": 14875
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.276046511627907e-07,
+      "loss": 0.0009,
+      "step": 14876
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.2759302325581393e-07,
+      "loss": 0.0008,
+      "step": 14877
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.2758139534883722e-07,
+      "loss": 0.0006,
+      "step": 14878
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.2756976744186045e-07,
+      "loss": 0.0007,
+      "step": 14879
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.2755813953488371e-07,
+      "loss": 0.0005,
+      "step": 14880
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.2754651162790698e-07,
+      "loss": 0.0006,
+      "step": 14881
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.2753488372093024e-07,
+      "loss": 0.0006,
+      "step": 14882
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.2752325581395347e-07,
+      "loss": 0.0007,
+      "step": 14883
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.2751162790697674e-07,
+      "loss": 0.0005,
+      "step": 14884
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.275e-07,
+      "loss": 0.0006,
+      "step": 14885
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.2748837209302326e-07,
+      "loss": 0.0176,
+      "step": 14886
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.274767441860465e-07,
+      "loss": 0.0009,
+      "step": 14887
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.2746511627906976e-07,
+      "loss": 0.0008,
+      "step": 14888
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.2745348837209302e-07,
+      "loss": 0.0103,
+      "step": 14889
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.2744186046511628e-07,
+      "loss": 0.0005,
+      "step": 14890
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.2743023255813954e-07,
+      "loss": 0.0013,
+      "step": 14891
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.274186046511628e-07,
+      "loss": 0.0048,
+      "step": 14892
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.2740697674418604e-07,
+      "loss": 0.0005,
+      "step": 14893
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.273953488372093e-07,
+      "loss": 0.3188,
+      "step": 14894
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.2738372093023256e-07,
+      "loss": 0.0007,
+      "step": 14895
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.273720930232558e-07,
+      "loss": 0.0006,
+      "step": 14896
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.2736046511627906e-07,
+      "loss": 0.0314,
+      "step": 14897
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.2734883720930232e-07,
+      "loss": 0.0005,
+      "step": 14898
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.2733720930232558e-07,
+      "loss": 0.0005,
+      "step": 14899
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.2732558139534882e-07,
+      "loss": 0.5544,
+      "step": 14900
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.2731395348837208e-07,
+      "loss": 0.0005,
+      "step": 14901
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.2730232558139534e-07,
+      "loss": 0.0006,
+      "step": 14902
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.272906976744186e-07,
+      "loss": 0.0013,
+      "step": 14903
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.2727906976744184e-07,
+      "loss": 0.0006,
+      "step": 14904
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.2726744186046513e-07,
+      "loss": 0.0009,
+      "step": 14905
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.2725581395348836e-07,
+      "loss": 0.0007,
+      "step": 14906
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.2724418604651162e-07,
+      "loss": 0.0006,
+      "step": 14907
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.2723255813953488e-07,
+      "loss": 0.0017,
+      "step": 14908
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.2722093023255815e-07,
+      "loss": 0.0006,
+      "step": 14909
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.2720930232558138e-07,
+      "loss": 0.0011,
+      "step": 14910
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.2719767441860464e-07,
+      "loss": 0.0006,
+      "step": 14911
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.271860465116279e-07,
+      "loss": 0.0006,
+      "step": 14912
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.2717441860465117e-07,
+      "loss": 0.0009,
+      "step": 14913
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.271627906976744e-07,
+      "loss": 0.0031,
+      "step": 14914
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.2715116279069766e-07,
+      "loss": 0.0008,
+      "step": 14915
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.2713953488372092e-07,
+      "loss": 1.304,
+      "step": 14916
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.2712790697674419e-07,
+      "loss": 0.001,
+      "step": 14917
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.2711627906976742e-07,
+      "loss": 0.0005,
+      "step": 14918
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.271046511627907e-07,
+      "loss": 0.0004,
+      "step": 14919
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.2709302325581395e-07,
+      "loss": 0.0005,
+      "step": 14920
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.270813953488372e-07,
+      "loss": 0.0006,
+      "step": 14921
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.2706976744186047e-07,
+      "loss": 0.0005,
+      "step": 14922
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.270581395348837e-07,
+      "loss": 0.0006,
+      "step": 14923
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.2704651162790697e-07,
+      "loss": 0.0005,
+      "step": 14924
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.2703488372093023e-07,
+      "loss": 0.0006,
+      "step": 14925
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.270232558139535e-07,
+      "loss": 0.0008,
+      "step": 14926
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.2701162790697672e-07,
+      "loss": 0.0011,
+      "step": 14927
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.2699999999999999e-07,
+      "loss": 0.0096,
+      "step": 14928
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.2698837209302325e-07,
+      "loss": 0.0004,
+      "step": 14929
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.269767441860465e-07,
+      "loss": 0.0005,
+      "step": 14930
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.2696511627906974e-07,
+      "loss": 0.0007,
+      "step": 14931
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.2695348837209303e-07,
+      "loss": 0.0005,
+      "step": 14932
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.2694186046511627e-07,
+      "loss": 0.0005,
+      "step": 14933
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.2693023255813953e-07,
+      "loss": 0.0013,
+      "step": 14934
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.269186046511628e-07,
+      "loss": 0.0006,
+      "step": 14935
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.2690697674418605e-07,
+      "loss": 0.0012,
+      "step": 14936
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.268953488372093e-07,
+      "loss": 0.0005,
+      "step": 14937
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.2688372093023255e-07,
+      "loss": 0.0023,
+      "step": 14938
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.268720930232558e-07,
+      "loss": 0.0008,
+      "step": 14939
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.2686046511627907e-07,
+      "loss": 0.0008,
+      "step": 14940
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.268488372093023e-07,
+      "loss": 0.0008,
+      "step": 14941
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.2683720930232557e-07,
+      "loss": 0.0004,
+      "step": 14942
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.2682558139534883e-07,
+      "loss": 0.0007,
+      "step": 14943
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.268139534883721e-07,
+      "loss": 0.0009,
+      "step": 14944
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.2680232558139533e-07,
+      "loss": 0.0005,
+      "step": 14945
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.2679069767441862e-07,
+      "loss": 0.0006,
+      "step": 14946
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.2677906976744185e-07,
+      "loss": 0.0006,
+      "step": 14947
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.2676744186046511e-07,
+      "loss": 0.0008,
+      "step": 14948
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.2675581395348838e-07,
+      "loss": 0.0007,
+      "step": 14949
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.267441860465116e-07,
+      "loss": 0.0006,
+      "step": 14950
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.2673255813953487e-07,
+      "loss": 0.0005,
+      "step": 14951
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.2672093023255813e-07,
+      "loss": 0.0008,
+      "step": 14952
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.267093023255814e-07,
+      "loss": 0.0006,
+      "step": 14953
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.2669767441860463e-07,
+      "loss": 0.0005,
+      "step": 14954
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.266860465116279e-07,
+      "loss": 0.0009,
+      "step": 14955
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.2667441860465115e-07,
+      "loss": 0.0008,
+      "step": 14956
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.2666279069767442e-07,
+      "loss": 0.0006,
+      "step": 14957
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.2665116279069765e-07,
+      "loss": 0.0011,
+      "step": 14958
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.2663953488372094e-07,
+      "loss": 0.0007,
+      "step": 14959
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.2662790697674418e-07,
+      "loss": 0.0006,
+      "step": 14960
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.2661627906976744e-07,
+      "loss": 0.0008,
+      "step": 14961
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.266046511627907e-07,
+      "loss": 0.0006,
+      "step": 14962
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.2659302325581396e-07,
+      "loss": 0.0008,
+      "step": 14963
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.265813953488372e-07,
+      "loss": 0.0006,
+      "step": 14964
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.2656976744186046e-07,
+      "loss": 0.0009,
+      "step": 14965
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.2655813953488372e-07,
+      "loss": 3.3184,
+      "step": 14966
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.2654651162790698e-07,
+      "loss": 0.0926,
+      "step": 14967
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.2653488372093022e-07,
+      "loss": 0.001,
+      "step": 14968
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.2652325581395348e-07,
+      "loss": 0.0007,
+      "step": 14969
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.2651162790697674e-07,
+      "loss": 0.0006,
+      "step": 14970
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.265e-07,
+      "loss": 0.001,
+      "step": 14971
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.2648837209302324e-07,
+      "loss": 0.0006,
+      "step": 14972
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.2647674418604652e-07,
+      "loss": 0.0009,
+      "step": 14973
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.2646511627906976e-07,
+      "loss": 0.0006,
+      "step": 14974
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.2645348837209302e-07,
+      "loss": 0.1844,
+      "step": 14975
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.2644186046511628e-07,
+      "loss": 0.0005,
+      "step": 14976
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.2643023255813952e-07,
+      "loss": 0.0004,
+      "step": 14977
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.2641860465116278e-07,
+      "loss": 0.0006,
+      "step": 14978
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.2640697674418604e-07,
+      "loss": 0.0007,
+      "step": 14979
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.263953488372093e-07,
+      "loss": 0.0006,
+      "step": 14980
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.2638372093023254e-07,
+      "loss": 0.0005,
+      "step": 14981
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.263720930232558e-07,
+      "loss": 0.0007,
+      "step": 14982
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.2636046511627906e-07,
+      "loss": 0.0007,
+      "step": 14983
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.2634883720930232e-07,
+      "loss": 0.0007,
+      "step": 14984
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.2633720930232556e-07,
+      "loss": 0.0005,
+      "step": 14985
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.2632558139534885e-07,
+      "loss": 0.0007,
+      "step": 14986
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.2631395348837208e-07,
+      "loss": 0.0016,
+      "step": 14987
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.2630232558139534e-07,
+      "loss": 0.0006,
+      "step": 14988
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.262906976744186e-07,
+      "loss": 0.0007,
+      "step": 14989
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.2627906976744187e-07,
+      "loss": 0.0004,
+      "step": 14990
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.262674418604651e-07,
+      "loss": 0.001,
+      "step": 14991
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.2625581395348836e-07,
+      "loss": 0.5682,
+      "step": 14992
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.2624418604651163e-07,
+      "loss": 0.0008,
+      "step": 14993
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.1476403772830963,
+      "eval_runtime": 1.5632,
+      "eval_samples_per_second": 661.48,
+      "eval_steps_per_second": 10.875,
+      "step": 14993
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.262325581395349e-07,
+      "loss": 0.0012,
+      "step": 14994
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.2622093023255812e-07,
+      "loss": 0.0005,
+      "step": 14995
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.2620930232558139e-07,
+      "loss": 0.0009,
+      "step": 14996
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.2619767441860465e-07,
+      "loss": 0.0016,
+      "step": 14997
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.261860465116279e-07,
+      "loss": 0.0005,
+      "step": 14998
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.2617441860465114e-07,
+      "loss": 0.0006,
+      "step": 14999
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.2616279069767443e-07,
+      "loss": 0.0005,
+      "step": 15000
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.2615116279069767e-07,
+      "loss": 0.0004,
+      "step": 15001
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.2613953488372093e-07,
+      "loss": 0.001,
+      "step": 15002
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.261279069767442e-07,
+      "loss": 0.0007,
+      "step": 15003
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.2611627906976743e-07,
+      "loss": 0.0005,
+      "step": 15004
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.261046511627907e-07,
+      "loss": 0.0007,
+      "step": 15005
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.2609302325581395e-07,
+      "loss": 0.0019,
+      "step": 15006
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.260813953488372e-07,
+      "loss": 0.0006,
+      "step": 15007
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.2606976744186045e-07,
+      "loss": 0.0006,
+      "step": 15008
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.260581395348837e-07,
+      "loss": 0.0007,
+      "step": 15009
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.2604651162790697e-07,
+      "loss": 0.0009,
+      "step": 15010
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.2603488372093023e-07,
+      "loss": 0.0006,
+      "step": 15011
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.2602325581395347e-07,
+      "loss": 0.0006,
+      "step": 15012
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.2601162790697675e-07,
+      "loss": 0.0009,
+      "step": 15013
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.26e-07,
+      "loss": 0.0008,
+      "step": 15014
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.2598837209302325e-07,
+      "loss": 0.0008,
+      "step": 15015
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.2597674418604651e-07,
+      "loss": 0.0007,
+      "step": 15016
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2596511627906978e-07,
+      "loss": 0.0008,
+      "step": 15017
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.25953488372093e-07,
+      "loss": 0.0008,
+      "step": 15018
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2594186046511627e-07,
+      "loss": 0.0006,
+      "step": 15019
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2593023255813953e-07,
+      "loss": 0.0004,
+      "step": 15020
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.259186046511628e-07,
+      "loss": 0.0007,
+      "step": 15021
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2590697674418603e-07,
+      "loss": 1.406,
+      "step": 15022
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.258953488372093e-07,
+      "loss": 0.0007,
+      "step": 15023
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2588372093023255e-07,
+      "loss": 0.0007,
+      "step": 15024
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2587209302325582e-07,
+      "loss": 0.0006,
+      "step": 15025
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2586046511627905e-07,
+      "loss": 0.0012,
+      "step": 15026
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2584883720930234e-07,
+      "loss": 0.0006,
+      "step": 15027
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2583720930232557e-07,
+      "loss": 0.0015,
+      "step": 15028
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2582558139534884e-07,
+      "loss": 0.0005,
+      "step": 15029
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.258139534883721e-07,
+      "loss": 0.0007,
+      "step": 15030
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2580232558139533e-07,
+      "loss": 0.0004,
+      "step": 15031
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.257906976744186e-07,
+      "loss": 0.0006,
+      "step": 15032
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2577906976744186e-07,
+      "loss": 0.0005,
+      "step": 15033
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2576744186046512e-07,
+      "loss": 0.0007,
+      "step": 15034
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2575581395348835e-07,
+      "loss": 0.0006,
+      "step": 15035
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2574418604651162e-07,
+      "loss": 0.0008,
+      "step": 15036
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2573255813953488e-07,
+      "loss": 0.0005,
+      "step": 15037
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2572093023255814e-07,
+      "loss": 0.0008,
+      "step": 15038
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2570930232558137e-07,
+      "loss": 0.0006,
+      "step": 15039
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2569767441860464e-07,
+      "loss": 0.0019,
+      "step": 15040
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.256860465116279e-07,
+      "loss": 0.0139,
+      "step": 15041
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2567441860465116e-07,
+      "loss": 0.0005,
+      "step": 15042
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2566279069767442e-07,
+      "loss": 0.0006,
+      "step": 15043
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2565116279069768e-07,
+      "loss": 0.0005,
+      "step": 15044
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2563953488372092e-07,
+      "loss": 0.0007,
+      "step": 15045
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2562790697674418e-07,
+      "loss": 0.0005,
+      "step": 15046
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2561627906976744e-07,
+      "loss": 0.0009,
+      "step": 15047
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.256046511627907e-07,
+      "loss": 0.0008,
+      "step": 15048
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.2559302325581394e-07,
+      "loss": 0.0007,
+      "step": 15049
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.255813953488372e-07,
+      "loss": 0.0005,
+      "step": 15050
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.2556976744186046e-07,
+      "loss": 0.0006,
+      "step": 15051
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.2555813953488372e-07,
+      "loss": 0.0009,
+      "step": 15052
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.2554651162790696e-07,
+      "loss": 0.0009,
+      "step": 15053
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.2553488372093025e-07,
+      "loss": 0.0005,
+      "step": 15054
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.2552325581395348e-07,
+      "loss": 0.0006,
+      "step": 15055
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.2551162790697672e-07,
+      "loss": 0.0005,
+      "step": 15056
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.255e-07,
+      "loss": 0.0085,
+      "step": 15057
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.2548837209302324e-07,
+      "loss": 0.0005,
+      "step": 15058
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.254767441860465e-07,
+      "loss": 0.0008,
+      "step": 15059
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.2546511627906976e-07,
+      "loss": 0.0006,
+      "step": 15060
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.2545348837209303e-07,
+      "loss": 0.0006,
+      "step": 15061
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.2544186046511626e-07,
+      "loss": 0.0005,
+      "step": 15062
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.2543023255813952e-07,
+      "loss": 0.0007,
+      "step": 15063
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.2541860465116278e-07,
+      "loss": 0.0005,
+      "step": 15064
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.2540697674418605e-07,
+      "loss": 0.0005,
+      "step": 15065
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.2539534883720928e-07,
+      "loss": 0.0006,
+      "step": 15066
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.2538372093023254e-07,
+      "loss": 0.0006,
+      "step": 15067
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.253720930232558e-07,
+      "loss": 0.0005,
+      "step": 15068
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.2536046511627907e-07,
+      "loss": 0.0021,
+      "step": 15069
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.2534883720930233e-07,
+      "loss": 0.0004,
+      "step": 15070
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.253372093023256e-07,
+      "loss": 0.001,
+      "step": 15071
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.2532558139534883e-07,
+      "loss": 0.0007,
+      "step": 15072
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.253139534883721e-07,
+      "loss": 0.0006,
+      "step": 15073
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.2530232558139535e-07,
+      "loss": 0.0016,
+      "step": 15074
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.252906976744186e-07,
+      "loss": 0.0005,
+      "step": 15075
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.2527906976744185e-07,
+      "loss": 0.0004,
+      "step": 15076
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.252674418604651e-07,
+      "loss": 0.0033,
+      "step": 15077
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.2525581395348837e-07,
+      "loss": 0.002,
+      "step": 15078
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.2524418604651163e-07,
+      "loss": 0.0008,
+      "step": 15079
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.2523255813953487e-07,
+      "loss": 0.0005,
+      "step": 15080
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.2522093023255815e-07,
+      "loss": 0.0008,
+      "step": 15081
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.252093023255814e-07,
+      "loss": 0.0004,
+      "step": 15082
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.2519767441860462e-07,
+      "loss": 0.0008,
+      "step": 15083
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.251860465116279e-07,
+      "loss": 0.0016,
+      "step": 15084
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.2517441860465115e-07,
+      "loss": 0.0879,
+      "step": 15085
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.251627906976744e-07,
+      "loss": 0.0007,
+      "step": 15086
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.2515116279069767e-07,
+      "loss": 0.0006,
+      "step": 15087
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.2513953488372093e-07,
+      "loss": 0.0012,
+      "step": 15088
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.2512790697674417e-07,
+      "loss": 0.0005,
+      "step": 15089
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.2511627906976743e-07,
+      "loss": 0.0007,
+      "step": 15090
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.251046511627907e-07,
+      "loss": 0.0008,
+      "step": 15091
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.2509302325581395e-07,
+      "loss": 0.0005,
+      "step": 15092
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.250813953488372e-07,
+      "loss": 0.0009,
+      "step": 15093
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.2506976744186045e-07,
+      "loss": 0.0006,
+      "step": 15094
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.250581395348837e-07,
+      "loss": 0.0007,
+      "step": 15095
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.2504651162790697e-07,
+      "loss": 0.0005,
+      "step": 15096
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.2503488372093024e-07,
+      "loss": 0.0008,
+      "step": 15097
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.250232558139535e-07,
+      "loss": 0.0005,
+      "step": 15098
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.2501162790697673e-07,
+      "loss": 0.0006,
+      "step": 15099
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.25e-07,
+      "loss": 0.0006,
+      "step": 15100
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.2498837209302326e-07,
+      "loss": 0.0004,
+      "step": 15101
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.2497674418604652e-07,
+      "loss": 0.0005,
+      "step": 15102
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.2496511627906975e-07,
+      "loss": 0.0008,
+      "step": 15103
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.2495348837209301e-07,
+      "loss": 0.001,
+      "step": 15104
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.2494186046511628e-07,
+      "loss": 0.0004,
+      "step": 15105
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.2493023255813954e-07,
+      "loss": 1.4269,
+      "step": 15106
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.2491860465116277e-07,
+      "loss": 0.0008,
+      "step": 15107
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.2490697674418606e-07,
+      "loss": 0.0007,
+      "step": 15108
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.248953488372093e-07,
+      "loss": 0.0005,
+      "step": 15109
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.2488372093023253e-07,
+      "loss": 0.0005,
+      "step": 15110
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.2487209302325582e-07,
+      "loss": 0.0006,
+      "step": 15111
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.2486046511627906e-07,
+      "loss": 0.0006,
+      "step": 15112
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.2484883720930232e-07,
+      "loss": 0.0006,
+      "step": 15113
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.2483720930232558e-07,
+      "loss": 0.0009,
+      "step": 15114
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.2482558139534884e-07,
+      "loss": 0.0007,
+      "step": 15115
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.2481395348837208e-07,
+      "loss": 0.0005,
+      "step": 15116
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.2480232558139534e-07,
+      "loss": 0.0012,
+      "step": 15117
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.247906976744186e-07,
+      "loss": 0.0007,
+      "step": 15118
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.2477906976744186e-07,
+      "loss": 0.0007,
+      "step": 15119
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.247674418604651e-07,
+      "loss": 0.0009,
+      "step": 15120
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.2475581395348836e-07,
+      "loss": 0.0006,
+      "step": 15121
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.2474418604651162e-07,
+      "loss": 0.0006,
+      "step": 15122
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.2473255813953488e-07,
+      "loss": 0.0007,
+      "step": 15123
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.2472093023255814e-07,
+      "loss": 0.0008,
+      "step": 15124
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.247093023255814e-07,
+      "loss": 0.0006,
+      "step": 15125
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.2469767441860464e-07,
+      "loss": 0.0043,
+      "step": 15126
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.246860465116279e-07,
+      "loss": 0.0007,
+      "step": 15127
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.2467441860465116e-07,
+      "loss": 0.0005,
+      "step": 15128
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.2466279069767443e-07,
+      "loss": 0.0006,
+      "step": 15129
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.2465116279069766e-07,
+      "loss": 0.0006,
+      "step": 15130
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 1.2463953488372092e-07,
+      "loss": 0.0011,
+      "step": 15131
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 1.2462790697674418e-07,
+      "loss": 0.0009,
+      "step": 15132
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 1.2461627906976745e-07,
+      "loss": 0.0007,
+      "step": 15133
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 1.2460465116279068e-07,
+      "loss": 1.6758,
+      "step": 15134
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 1.2459302325581397e-07,
+      "loss": 0.0004,
+      "step": 15135
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 1.245813953488372e-07,
+      "loss": 0.0007,
+      "step": 15136
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 1.2456976744186044e-07,
+      "loss": 0.0006,
+      "step": 15137
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 1.2455813953488373e-07,
+      "loss": 0.0007,
+      "step": 15138
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 1.2454651162790696e-07,
+      "loss": 0.0008,
+      "step": 15139
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 1.2453488372093022e-07,
+      "loss": 0.0015,
+      "step": 15140
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.2452325581395349e-07,
+      "loss": 0.0006,
+      "step": 15141
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.2451162790697675e-07,
+      "loss": 0.0009,
+      "step": 15142
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.2449999999999998e-07,
+      "loss": 0.0116,
+      "step": 15143
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.2448837209302324e-07,
+      "loss": 0.0004,
+      "step": 15144
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.244767441860465e-07,
+      "loss": 0.0006,
+      "step": 15145
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 1.2446511627906977e-07,
+      "loss": 0.4686,
+      "step": 15146
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 1.24453488372093e-07,
+      "loss": 0.0005,
+      "step": 15147
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 1.2444186046511627e-07,
+      "loss": 0.0009,
+      "step": 15148
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 1.2443023255813953e-07,
+      "loss": 0.0009,
+      "step": 15149
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 1.244186046511628e-07,
+      "loss": 0.001,
+      "step": 15150
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 1.2440697674418605e-07,
+      "loss": 0.0005,
+      "step": 15151
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 1.243953488372093e-07,
+      "loss": 0.0009,
+      "step": 15152
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 1.2438372093023255e-07,
+      "loss": 0.0009,
+      "step": 15153
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 1.243720930232558e-07,
+      "loss": 0.0005,
+      "step": 15154
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 1.2436046511627907e-07,
+      "loss": 0.0008,
+      "step": 15155
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 1.2434883720930233e-07,
+      "loss": 0.0004,
+      "step": 15156
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 1.2433720930232557e-07,
+      "loss": 0.0005,
+      "step": 15157
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 1.2432558139534883e-07,
+      "loss": 0.0006,
+      "step": 15158
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 1.243139534883721e-07,
+      "loss": 0.0006,
+      "step": 15159
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 1.2430232558139535e-07,
+      "loss": 0.0007,
+      "step": 15160
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 1.242906976744186e-07,
+      "loss": 0.0008,
+      "step": 15161
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 1.2427906976744185e-07,
+      "loss": 0.0005,
+      "step": 15162
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 1.242674418604651e-07,
+      "loss": 0.0005,
+      "step": 15163
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 1.2425581395348835e-07,
+      "loss": 0.0009,
+      "step": 15164
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 1.2424418604651164e-07,
+      "loss": 0.0005,
+      "step": 15165
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 1.2423255813953487e-07,
+      "loss": 0.0004,
+      "step": 15166
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 1.2422093023255813e-07,
+      "loss": 0.0012,
+      "step": 15167
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 1.242093023255814e-07,
+      "loss": 0.0008,
+      "step": 15168
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 1.2419767441860466e-07,
+      "loss": 0.0554,
+      "step": 15169
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 1.241860465116279e-07,
+      "loss": 0.0006,
+      "step": 15170
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 1.2417441860465115e-07,
+      "loss": 0.0008,
+      "step": 15171
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.2416279069767441e-07,
+      "loss": 0.0007,
+      "step": 15172
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.2415116279069768e-07,
+      "loss": 0.0006,
+      "step": 15173
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.241395348837209e-07,
+      "loss": 0.0006,
+      "step": 15174
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.2412790697674417e-07,
+      "loss": 0.0008,
+      "step": 15175
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.2411627906976743e-07,
+      "loss": 0.0006,
+      "step": 15176
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 1.241046511627907e-07,
+      "loss": 0.0006,
+      "step": 15177
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 1.2409302325581393e-07,
+      "loss": 0.0005,
+      "step": 15178
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 1.2408139534883722e-07,
+      "loss": 0.0005,
+      "step": 15179
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 1.2406976744186045e-07,
+      "loss": 0.0006,
+      "step": 15180
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 1.2405813953488372e-07,
+      "loss": 0.0005,
+      "step": 15181
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 1.2404651162790698e-07,
+      "loss": 0.0006,
+      "step": 15182
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 1.2403488372093024e-07,
+      "loss": 0.0008,
+      "step": 15183
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 1.2402325581395348e-07,
+      "loss": 0.0006,
+      "step": 15184
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 1.2401162790697674e-07,
+      "loss": 0.0005,
+      "step": 15185
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 1.24e-07,
+      "loss": 0.0008,
+      "step": 15186
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.2398837209302326e-07,
+      "loss": 0.0005,
+      "step": 15187
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.239767441860465e-07,
+      "loss": 0.0006,
+      "step": 15188
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.2396511627906976e-07,
+      "loss": 0.0008,
+      "step": 15189
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.2395348837209302e-07,
+      "loss": 0.0008,
+      "step": 15190
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.2394186046511625e-07,
+      "loss": 0.0005,
+      "step": 15191
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.2393023255813954e-07,
+      "loss": 0.0009,
+      "step": 15192
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 1.2391860465116278e-07,
+      "loss": 0.0007,
+      "step": 15193
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 1.2390697674418604e-07,
+      "loss": 0.0008,
+      "step": 15194
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 1.238953488372093e-07,
+      "loss": 0.0006,
+      "step": 15195
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 1.2388372093023256e-07,
+      "loss": 0.0005,
+      "step": 15196
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 1.238720930232558e-07,
+      "loss": 0.0007,
+      "step": 15197
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 1.2386046511627906e-07,
+      "loss": 0.0007,
+      "step": 15198
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 1.2384883720930232e-07,
+      "loss": 3.1854,
+      "step": 15199
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 1.2383720930232558e-07,
+      "loss": 0.0005,
+      "step": 15200
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 1.2382558139534882e-07,
+      "loss": 0.0004,
+      "step": 15201
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 1.2381395348837208e-07,
+      "loss": 0.0006,
+      "step": 15202
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 1.2380232558139534e-07,
+      "loss": 0.0004,
+      "step": 15203
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 1.237906976744186e-07,
+      "loss": 0.0007,
+      "step": 15204
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 1.2377906976744184e-07,
+      "loss": 0.0007,
+      "step": 15205
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 1.2376744186046513e-07,
+      "loss": 0.001,
+      "step": 15206
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 1.2375581395348836e-07,
+      "loss": 0.0007,
+      "step": 15207
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.2374418604651162e-07,
+      "loss": 0.0013,
+      "step": 15208
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.2373255813953489e-07,
+      "loss": 0.0009,
+      "step": 15209
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.2372093023255815e-07,
+      "loss": 0.0005,
+      "step": 15210
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.2370930232558138e-07,
+      "loss": 0.0006,
+      "step": 15211
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.2369767441860464e-07,
+      "loss": 0.0008,
+      "step": 15212
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 1.236860465116279e-07,
+      "loss": 0.0005,
+      "step": 15213
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 1.2367441860465117e-07,
+      "loss": 0.0006,
+      "step": 15214
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 1.236627906976744e-07,
+      "loss": 0.0007,
+      "step": 15215
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 1.2365116279069766e-07,
+      "loss": 0.0006,
+      "step": 15216
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 1.2363953488372093e-07,
+      "loss": 0.0005,
+      "step": 15217
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.2362790697674416e-07,
+      "loss": 0.0005,
+      "step": 15218
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.2361627906976745e-07,
+      "loss": 0.0007,
+      "step": 15219
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.2360465116279069e-07,
+      "loss": 0.0007,
+      "step": 15220
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.2359302325581395e-07,
+      "loss": 0.0006,
+      "step": 15221
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.235813953488372e-07,
+      "loss": 0.001,
+      "step": 15222
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.2356976744186047e-07,
+      "loss": 0.0004,
+      "step": 15223
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 1.235581395348837e-07,
+      "loss": 0.191,
+      "step": 15224
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 1.2354651162790697e-07,
+      "loss": 0.0005,
+      "step": 15225
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 1.2353488372093023e-07,
+      "loss": 0.0019,
+      "step": 15226
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 1.235232558139535e-07,
+      "loss": 0.0007,
+      "step": 15227
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 1.2351162790697673e-07,
+      "loss": 0.0006,
+      "step": 15228
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 1.235e-07,
+      "loss": 0.6439,
+      "step": 15229
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 1.2348837209302325e-07,
+      "loss": 0.0005,
+      "step": 15230
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 1.234767441860465e-07,
+      "loss": 0.0006,
+      "step": 15231
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 1.2346511627906975e-07,
+      "loss": 0.0028,
+      "step": 15232
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 1.2345348837209303e-07,
+      "loss": 0.0007,
+      "step": 15233
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 1.2344186046511627e-07,
+      "loss": 0.0009,
+      "step": 15234
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 1.2343023255813953e-07,
+      "loss": 0.001,
+      "step": 15235
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 1.234186046511628e-07,
+      "loss": 0.0009,
+      "step": 15236
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 1.2340697674418605e-07,
+      "loss": 0.0007,
+      "step": 15237
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 1.233953488372093e-07,
+      "loss": 0.0008,
+      "step": 15238
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 1.2338372093023255e-07,
+      "loss": 0.0006,
+      "step": 15239
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 1.2337209302325581e-07,
+      "loss": 0.0006,
+      "step": 15240
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 1.2336046511627908e-07,
+      "loss": 0.0008,
+      "step": 15241
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 1.233488372093023e-07,
+      "loss": 0.0006,
+      "step": 15242
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 1.2333720930232557e-07,
+      "loss": 0.0004,
+      "step": 15243
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 1.2332558139534883e-07,
+      "loss": 0.0007,
+      "step": 15244
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 1.2331395348837207e-07,
+      "loss": 0.0005,
+      "step": 15245
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 1.2330232558139536e-07,
+      "loss": 1.7617,
+      "step": 15246
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 1.232906976744186e-07,
+      "loss": 0.0007,
+      "step": 15247
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 1.2327906976744185e-07,
+      "loss": 0.0005,
+      "step": 15248
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 1.2326744186046512e-07,
+      "loss": 0.0009,
+      "step": 15249
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 1.2325581395348838e-07,
+      "loss": 0.0006,
+      "step": 15250
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 1.232441860465116e-07,
+      "loss": 0.0005,
+      "step": 15251
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 1.2323255813953487e-07,
+      "loss": 0.0006,
+      "step": 15252
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 1.2322093023255814e-07,
+      "loss": 0.0009,
+      "step": 15253
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 1.232093023255814e-07,
+      "loss": 0.0009,
+      "step": 15254
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 1.2319767441860463e-07,
+      "loss": 0.3458,
+      "step": 15255
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 1.231860465116279e-07,
+      "loss": 0.0008,
+      "step": 15256
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 1.2317441860465116e-07,
+      "loss": 0.0006,
+      "step": 15257
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 1.2316279069767442e-07,
+      "loss": 0.0008,
+      "step": 15258
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 1.2315116279069765e-07,
+      "loss": 0.0004,
+      "step": 15259
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 1.2313953488372094e-07,
+      "loss": 0.0006,
+      "step": 15260
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 1.2312790697674418e-07,
+      "loss": 0.0006,
+      "step": 15261
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 1.2311627906976744e-07,
+      "loss": 0.0007,
+      "step": 15262
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 1.231046511627907e-07,
+      "loss": 0.0004,
+      "step": 15263
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 1.2309302325581396e-07,
+      "loss": 0.0006,
+      "step": 15264
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 1.230813953488372e-07,
+      "loss": 0.0004,
+      "step": 15265
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 1.2306976744186046e-07,
+      "loss": 0.0007,
+      "step": 15266
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 1.2305813953488372e-07,
+      "loss": 0.0005,
+      "step": 15267
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 1.2304651162790698e-07,
+      "loss": 0.0004,
+      "step": 15268
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 1.2303488372093022e-07,
+      "loss": 0.0006,
+      "step": 15269
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 1.2302325581395348e-07,
+      "loss": 0.0006,
+      "step": 15270
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 1.2301162790697674e-07,
+      "loss": 0.0026,
+      "step": 15271
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 1.2299999999999998e-07,
+      "loss": 0.0005,
+      "step": 15272
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 1.2298837209302326e-07,
+      "loss": 0.0005,
+      "step": 15273
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 1.229767441860465e-07,
+      "loss": 0.0006,
+      "step": 15274
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.2296511627906976e-07,
+      "loss": 0.0008,
+      "step": 15275
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.2295348837209302e-07,
+      "loss": 0.0051,
+      "step": 15276
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.2294186046511628e-07,
+      "loss": 0.0007,
+      "step": 15277
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.2293023255813952e-07,
+      "loss": 0.0006,
+      "step": 15278
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 1.2291860465116278e-07,
+      "loss": 0.0005,
+      "step": 15279
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.2290697674418604e-07,
+      "loss": 0.0008,
+      "step": 15280
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.228953488372093e-07,
+      "loss": 0.0005,
+      "step": 15281
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.2288372093023254e-07,
+      "loss": 0.0038,
+      "step": 15282
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.228720930232558e-07,
+      "loss": 0.0008,
+      "step": 15283
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.2286046511627906e-07,
+      "loss": 0.0004,
+      "step": 15284
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 1.2284883720930233e-07,
+      "loss": 0.0006,
+      "step": 15285
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 1.2283720930232556e-07,
+      "loss": 0.0007,
+      "step": 15286
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 1.2282558139534885e-07,
+      "loss": 0.0009,
+      "step": 15287
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 1.2281395348837208e-07,
+      "loss": 0.0004,
+      "step": 15288
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 1.2280232558139535e-07,
+      "loss": 0.0005,
+      "step": 15289
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 1.227906976744186e-07,
+      "loss": 0.0004,
+      "step": 15290
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 1.2277906976744187e-07,
+      "loss": 0.0006,
+      "step": 15291
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 1.227674418604651e-07,
+      "loss": 0.0007,
+      "step": 15292
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 1.2275581395348837e-07,
+      "loss": 0.0007,
+      "step": 15293
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 1.2274418604651163e-07,
+      "loss": 0.0006,
+      "step": 15294
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 1.227325581395349e-07,
+      "loss": 0.0005,
+      "step": 15295
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 1.2272093023255813e-07,
+      "loss": 0.0005,
+      "step": 15296
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 1.227093023255814e-07,
+      "loss": 0.0005,
+      "step": 15297
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 1.2269767441860465e-07,
+      "loss": 0.0006,
+      "step": 15298
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 1.2268604651162788e-07,
+      "loss": 0.3608,
+      "step": 15299
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 1.2267441860465115e-07,
+      "loss": 3.6172,
+      "step": 15300
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 1.226627906976744e-07,
+      "loss": 0.0008,
+      "step": 15301
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 1.2265116279069767e-07,
+      "loss": 0.0007,
+      "step": 15302
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 1.2263953488372093e-07,
+      "loss": 0.0247,
+      "step": 15303
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 1.226279069767442e-07,
+      "loss": 0.0005,
+      "step": 15304
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 1.2261627906976743e-07,
+      "loss": 0.0005,
+      "step": 15305
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.226046511627907e-07,
+      "loss": 0.0004,
+      "step": 15306
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.2259302325581395e-07,
+      "loss": 0.0012,
+      "step": 15307
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.225813953488372e-07,
+      "loss": 0.0009,
+      "step": 15308
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.2256976744186045e-07,
+      "loss": 0.0007,
+      "step": 15309
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 1.225581395348837e-07,
+      "loss": 0.0006,
+      "step": 15310
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.2254651162790697e-07,
+      "loss": 0.0005,
+      "step": 15311
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.2253488372093023e-07,
+      "loss": 0.0014,
+      "step": 15312
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.2252325581395347e-07,
+      "loss": 0.0007,
+      "step": 15313
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.2251162790697676e-07,
+      "loss": 0.0007,
+      "step": 15314
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.225e-07,
+      "loss": 0.0006,
+      "step": 15315
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 1.2248837209302325e-07,
+      "loss": 0.0006,
+      "step": 15316
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.2247674418604652e-07,
+      "loss": 0.0007,
+      "step": 15317
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.2246511627906978e-07,
+      "loss": 0.0006,
+      "step": 15318
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.22453488372093e-07,
+      "loss": 0.0005,
+      "step": 15319
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.2244186046511627e-07,
+      "loss": 0.0009,
+      "step": 15320
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 1.2243023255813954e-07,
+      "loss": 0.0006,
+      "step": 15321
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 1.224186046511628e-07,
+      "loss": 0.0114,
+      "step": 15322
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 1.2240697674418603e-07,
+      "loss": 0.0006,
+      "step": 15323
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 1.223953488372093e-07,
+      "loss": 0.0005,
+      "step": 15324
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 1.2238372093023256e-07,
+      "loss": 0.002,
+      "step": 15325
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 1.223720930232558e-07,
+      "loss": 0.0006,
+      "step": 15326
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 1.2236046511627905e-07,
+      "loss": 0.0005,
+      "step": 15327
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 1.2234883720930231e-07,
+      "loss": 0.0007,
+      "step": 15328
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 1.2233720930232558e-07,
+      "loss": 0.0033,
+      "step": 15329
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 1.2232558139534884e-07,
+      "loss": 0.0006,
+      "step": 15330
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 1.223139534883721e-07,
+      "loss": 0.0005,
+      "step": 15331
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 1.2230232558139534e-07,
+      "loss": 0.0006,
+      "step": 15332
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 1.222906976744186e-07,
+      "loss": 0.0006,
+      "step": 15333
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 1.2227906976744186e-07,
+      "loss": 0.0008,
+      "step": 15334
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 1.2226744186046512e-07,
+      "loss": 0.4381,
+      "step": 15335
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 1.2225581395348836e-07,
+      "loss": 0.0008,
+      "step": 15336
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 1.2224418604651162e-07,
+      "loss": 0.0007,
+      "step": 15337
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 1.2223255813953488e-07,
+      "loss": 0.0005,
+      "step": 15338
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 1.2222093023255814e-07,
+      "loss": 0.0007,
+      "step": 15339
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 1.2220930232558138e-07,
+      "loss": 0.0005,
+      "step": 15340
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 1.2219767441860466e-07,
+      "loss": 0.0019,
+      "step": 15341
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.221860465116279e-07,
+      "loss": 0.0008,
+      "step": 15342
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.2217441860465116e-07,
+      "loss": 0.0015,
+      "step": 15343
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.2216279069767442e-07,
+      "loss": 0.0005,
+      "step": 15344
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.2215116279069768e-07,
+      "loss": 0.0005,
+      "step": 15345
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.2213953488372092e-07,
+      "loss": 0.0007,
+      "step": 15346
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 1.2212790697674418e-07,
+      "loss": 0.0008,
+      "step": 15347
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 1.2211627906976744e-07,
+      "loss": 0.0013,
+      "step": 15348
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 1.2210465116279068e-07,
+      "loss": 0.0005,
+      "step": 15349
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 1.2209302325581394e-07,
+      "loss": 0.0005,
+      "step": 15350
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 1.220813953488372e-07,
+      "loss": 0.0006,
+      "step": 15351
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 1.2206976744186046e-07,
+      "loss": 0.0007,
+      "step": 15352
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 1.220581395348837e-07,
+      "loss": 0.0006,
+      "step": 15353
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 1.2204651162790696e-07,
+      "loss": 0.0007,
+      "step": 15354
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 1.2203488372093022e-07,
+      "loss": 0.0008,
+      "step": 15355
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 1.2202325581395348e-07,
+      "loss": 0.0006,
+      "step": 15356
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 1.2201162790697675e-07,
+      "loss": 0.0005,
+      "step": 15357
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 1.22e-07,
+      "loss": 0.0005,
+      "step": 15358
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 1.2198837209302324e-07,
+      "loss": 0.0006,
+      "step": 15359
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 1.219767441860465e-07,
+      "loss": 0.0004,
+      "step": 15360
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 1.2196511627906977e-07,
+      "loss": 0.0006,
+      "step": 15361
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 1.2195348837209303e-07,
+      "loss": 0.0007,
+      "step": 15362
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 1.2194186046511626e-07,
+      "loss": 0.0006,
+      "step": 15363
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 1.2193023255813952e-07,
+      "loss": 0.0005,
+      "step": 15364
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 1.2191860465116279e-07,
+      "loss": 0.0261,
+      "step": 15365
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 1.2190697674418605e-07,
+      "loss": 0.0006,
+      "step": 15366
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 1.2189534883720928e-07,
+      "loss": 0.0006,
+      "step": 15367
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 1.2188372093023257e-07,
+      "loss": 0.0004,
+      "step": 15368
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 1.218720930232558e-07,
+      "loss": 0.0005,
+      "step": 15369
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 1.2186046511627907e-07,
+      "loss": 0.0008,
+      "step": 15370
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 1.2184883720930233e-07,
+      "loss": 0.2112,
+      "step": 15371
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 1.218372093023256e-07,
+      "loss": 0.0008,
+      "step": 15372
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.2182558139534883e-07,
+      "loss": 0.0006,
+      "step": 15373
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.218139534883721e-07,
+      "loss": 0.0008,
+      "step": 15374
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.2180232558139535e-07,
+      "loss": 0.0006,
+      "step": 15375
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.2179069767441859e-07,
+      "loss": 0.0005,
+      "step": 15376
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.2177906976744185e-07,
+      "loss": 0.0007,
+      "step": 15377
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 1.217674418604651e-07,
+      "loss": 0.0007,
+      "step": 15378
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 1.2175581395348837e-07,
+      "loss": 0.2959,
+      "step": 15379
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 1.217441860465116e-07,
+      "loss": 0.006,
+      "step": 15380
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 1.2173255813953487e-07,
+      "loss": 0.0005,
+      "step": 15381
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 1.2172093023255813e-07,
+      "loss": 0.0005,
+      "step": 15382
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 1.217093023255814e-07,
+      "loss": 1.5396,
+      "step": 15383
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 1.2169767441860465e-07,
+      "loss": 0.0005,
+      "step": 15384
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 1.2168604651162791e-07,
+      "loss": 0.0009,
+      "step": 15385
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 1.2167441860465115e-07,
+      "loss": 0.0008,
+      "step": 15386
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 1.216627906976744e-07,
+      "loss": 0.0004,
+      "step": 15387
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 1.2165116279069767e-07,
+      "loss": 0.0005,
+      "step": 15388
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 1.2163953488372093e-07,
+      "loss": 0.0005,
+      "step": 15389
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 1.2162790697674417e-07,
+      "loss": 0.0006,
+      "step": 15390
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 1.2161627906976743e-07,
+      "loss": 0.0006,
+      "step": 15391
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 1.216046511627907e-07,
+      "loss": 0.0007,
+      "step": 15392
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 1.2159302325581396e-07,
+      "loss": 0.0005,
+      "step": 15393
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 1.215813953488372e-07,
+      "loss": 0.0004,
+      "step": 15394
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 1.2156976744186048e-07,
+      "loss": 0.0006,
+      "step": 15395
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 1.2155813953488371e-07,
+      "loss": 0.0005,
+      "step": 15396
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 1.2154651162790698e-07,
+      "loss": 0.0005,
+      "step": 15397
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 1.2153488372093024e-07,
+      "loss": 0.0005,
+      "step": 15398
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 1.215232558139535e-07,
+      "loss": 0.0006,
+      "step": 15399
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 1.2151162790697673e-07,
+      "loss": 0.0005,
+      "step": 15400
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 1.215e-07,
+      "loss": 0.0005,
+      "step": 15401
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 1.2148837209302326e-07,
+      "loss": 0.0014,
+      "step": 15402
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 1.214767441860465e-07,
+      "loss": 0.0008,
+      "step": 15403
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 1.2146511627906975e-07,
+      "loss": 0.0006,
+      "step": 15404
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 1.2145348837209302e-07,
+      "loss": 0.0004,
+      "step": 15405
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 1.2144186046511628e-07,
+      "loss": 0.0006,
+      "step": 15406
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 1.2143023255813951e-07,
+      "loss": 0.0007,
+      "step": 15407
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 1.2141860465116278e-07,
+      "loss": 0.0009,
+      "step": 15408
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 1.2140697674418604e-07,
+      "loss": 0.0007,
+      "step": 15409
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 1.213953488372093e-07,
+      "loss": 0.0006,
+      "step": 15410
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 1.2138372093023256e-07,
+      "loss": 0.0006,
+      "step": 15411
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 1.2137209302325582e-07,
+      "loss": 0.0004,
+      "step": 15412
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 1.2136046511627906e-07,
+      "loss": 2.2548,
+      "step": 15413
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 1.2134883720930232e-07,
+      "loss": 0.0007,
+      "step": 15414
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 1.2133720930232558e-07,
+      "loss": 0.0006,
+      "step": 15415
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 1.2132558139534884e-07,
+      "loss": 0.0006,
+      "step": 15416
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 1.2131395348837208e-07,
+      "loss": 0.0007,
+      "step": 15417
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 1.2130232558139534e-07,
+      "loss": 0.0004,
+      "step": 15418
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 1.212906976744186e-07,
+      "loss": 0.0005,
+      "step": 15419
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 1.2127906976744186e-07,
+      "loss": 0.7001,
+      "step": 15420
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 1.212674418604651e-07,
+      "loss": 0.0008,
+      "step": 15421
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 1.2125581395348836e-07,
+      "loss": 1.2564,
+      "step": 15422
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 1.2124418604651162e-07,
+      "loss": 0.0007,
+      "step": 15423
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 1.2123255813953488e-07,
+      "loss": 0.0005,
+      "step": 15424
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 1.2122093023255814e-07,
+      "loss": 0.0006,
+      "step": 15425
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 1.212093023255814e-07,
+      "loss": 0.0004,
+      "step": 15426
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 1.2119767441860464e-07,
+      "loss": 0.0009,
+      "step": 15427
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 1.211860465116279e-07,
+      "loss": 0.0009,
+      "step": 15428
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 1.2117441860465117e-07,
+      "loss": 0.0006,
+      "step": 15429
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.211627906976744e-07,
+      "loss": 0.0006,
+      "step": 15430
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.2115116279069766e-07,
+      "loss": 0.0006,
+      "step": 15431
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.2113953488372092e-07,
+      "loss": 0.0007,
+      "step": 15432
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.2112790697674419e-07,
+      "loss": 0.0102,
+      "step": 15433
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.2111627906976742e-07,
+      "loss": 0.0005,
+      "step": 15434
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.2110465116279068e-07,
+      "loss": 0.0011,
+      "step": 15435
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.2109302325581394e-07,
+      "loss": 0.0005,
+      "step": 15436
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.210813953488372e-07,
+      "loss": 1.8244,
+      "step": 15437
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.2106976744186044e-07,
+      "loss": 0.0006,
+      "step": 15438
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.2105813953488373e-07,
+      "loss": 0.0004,
+      "step": 15439
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.2104651162790696e-07,
+      "loss": 0.0006,
+      "step": 15440
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.2103488372093023e-07,
+      "loss": 0.0005,
+      "step": 15441
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.210232558139535e-07,
+      "loss": 0.0004,
+      "step": 15442
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.2101162790697675e-07,
+      "loss": 0.0006,
+      "step": 15443
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.2099999999999998e-07,
+      "loss": 0.0005,
+      "step": 15444
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.2098837209302325e-07,
+      "loss": 0.0006,
+      "step": 15445
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.209767441860465e-07,
+      "loss": 0.0006,
+      "step": 15446
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.2096511627906977e-07,
+      "loss": 0.0008,
+      "step": 15447
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.20953488372093e-07,
+      "loss": 0.0007,
+      "step": 15448
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.2094186046511627e-07,
+      "loss": 0.0007,
+      "step": 15449
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.2093023255813953e-07,
+      "loss": 0.0006,
+      "step": 15450
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.209186046511628e-07,
+      "loss": 0.0031,
+      "step": 15451
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.2090697674418605e-07,
+      "loss": 0.0008,
+      "step": 15452
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.2089534883720931e-07,
+      "loss": 0.0007,
+      "step": 15453
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.2088372093023255e-07,
+      "loss": 0.0007,
+      "step": 15454
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.208720930232558e-07,
+      "loss": 0.0027,
+      "step": 15455
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.2086046511627907e-07,
+      "loss": 0.0037,
+      "step": 15456
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.208488372093023e-07,
+      "loss": 0.0006,
+      "step": 15457
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.2083720930232557e-07,
+      "loss": 0.0005,
+      "step": 15458
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.2082558139534883e-07,
+      "loss": 0.0005,
+      "step": 15459
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.208139534883721e-07,
+      "loss": 0.0006,
+      "step": 15460
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.2080232558139533e-07,
+      "loss": 0.0014,
+      "step": 15461
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.207906976744186e-07,
+      "loss": 0.0006,
+      "step": 15462
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.2077906976744185e-07,
+      "loss": 0.0005,
+      "step": 15463
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.2076744186046511e-07,
+      "loss": 0.0011,
+      "step": 15464
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.2075581395348835e-07,
+      "loss": 0.0006,
+      "step": 15465
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.2074418604651164e-07,
+      "loss": 0.0004,
+      "step": 15466
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.2073255813953487e-07,
+      "loss": 0.0026,
+      "step": 15467
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.2072093023255813e-07,
+      "loss": 0.0005,
+      "step": 15468
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.207093023255814e-07,
+      "loss": 0.0005,
+      "step": 15469
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.2069767441860466e-07,
+      "loss": 0.001,
+      "step": 15470
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.206860465116279e-07,
+      "loss": 0.0006,
+      "step": 15471
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.2067441860465115e-07,
+      "loss": 0.0006,
+      "step": 15472
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.2066279069767442e-07,
+      "loss": 0.0006,
+      "step": 15473
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.2065116279069768e-07,
+      "loss": 0.0008,
+      "step": 15474
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.206395348837209e-07,
+      "loss": 0.0007,
+      "step": 15475
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.2062790697674417e-07,
+      "loss": 0.0009,
+      "step": 15476
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 1.2061627906976744e-07,
+      "loss": 0.0006,
+      "step": 15477
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 1.206046511627907e-07,
+      "loss": 0.0008,
+      "step": 15478
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 1.2059302325581396e-07,
+      "loss": 0.0007,
+      "step": 15479
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 1.2058139534883722e-07,
+      "loss": 0.0007,
+      "step": 15480
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 1.2056976744186046e-07,
+      "loss": 0.0006,
+      "step": 15481
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 1.2055813953488372e-07,
+      "loss": 0.0024,
+      "step": 15482
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 1.2054651162790698e-07,
+      "loss": 0.0009,
+      "step": 15483
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 1.2053488372093022e-07,
+      "loss": 0.0005,
+      "step": 15484
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 1.2052325581395348e-07,
+      "loss": 0.0005,
+      "step": 15485
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 1.2051162790697674e-07,
+      "loss": 0.0016,
+      "step": 15486
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 1.205e-07,
+      "loss": 0.0012,
+      "step": 15487
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 1.2048837209302324e-07,
+      "loss": 0.0009,
+      "step": 15488
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 1.204767441860465e-07,
+      "loss": 0.001,
+      "step": 15489
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 1.2046511627906976e-07,
+      "loss": 0.0006,
+      "step": 15490
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 1.2045348837209302e-07,
+      "loss": 0.0007,
+      "step": 15491
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 1.2044186046511626e-07,
+      "loss": 0.0005,
+      "step": 15492
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 1.2043023255813954e-07,
+      "loss": 0.0197,
+      "step": 15493
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 1.2041860465116278e-07,
+      "loss": 0.0007,
+      "step": 15494
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 1.2040697674418604e-07,
+      "loss": 0.0005,
+      "step": 15495
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 1.203953488372093e-07,
+      "loss": 0.0006,
+      "step": 15496
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 1.2038372093023256e-07,
+      "loss": 0.0005,
+      "step": 15497
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 1.203720930232558e-07,
+      "loss": 0.3507,
+      "step": 15498
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 1.2036046511627906e-07,
+      "loss": 0.0008,
+      "step": 15499
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 1.2034883720930232e-07,
+      "loss": 0.0005,
+      "step": 15500
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 1.2033720930232558e-07,
+      "loss": 0.0004,
+      "step": 15501
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 1.2032558139534882e-07,
+      "loss": 0.0005,
+      "step": 15502
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.2031395348837208e-07,
+      "loss": 0.3216,
+      "step": 15503
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.2030232558139534e-07,
+      "loss": 0.0007,
+      "step": 15504
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.202906976744186e-07,
+      "loss": 0.0005,
+      "step": 15505
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.2027906976744187e-07,
+      "loss": 0.0006,
+      "step": 15506
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.2026744186046513e-07,
+      "loss": 1.5992,
+      "step": 15507
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 1.2025581395348836e-07,
+      "loss": 0.0004,
+      "step": 15508
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 1.2024418604651163e-07,
+      "loss": 0.0005,
+      "step": 15509
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 1.202325581395349e-07,
+      "loss": 0.0041,
+      "step": 15510
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.15230439603328705,
+      "eval_runtime": 1.5553,
+      "eval_samples_per_second": 664.816,
+      "eval_steps_per_second": 10.93,
+      "step": 15510
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 1.2022093023255812e-07,
+      "loss": 0.0006,
+      "step": 15511
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 1.2020930232558138e-07,
+      "loss": 0.0006,
+      "step": 15512
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 1.2019767441860465e-07,
+      "loss": 0.0006,
+      "step": 15513
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 1.201860465116279e-07,
+      "loss": 0.0006,
+      "step": 15514
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 1.2017441860465114e-07,
+      "loss": 0.0006,
+      "step": 15515
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 1.201627906976744e-07,
+      "loss": 0.0006,
+      "step": 15516
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 1.2015116279069767e-07,
+      "loss": 0.0008,
+      "step": 15517
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 1.2013953488372093e-07,
+      "loss": 0.0006,
+      "step": 15518
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 1.2012790697674416e-07,
+      "loss": 0.0008,
+      "step": 15519
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 1.2011627906976745e-07,
+      "loss": 0.0006,
+      "step": 15520
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 1.201046511627907e-07,
+      "loss": 2.2698,
+      "step": 15521
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 1.2009302325581395e-07,
+      "loss": 0.0014,
+      "step": 15522
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 1.200813953488372e-07,
+      "loss": 0.0007,
+      "step": 15523
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 1.2006976744186047e-07,
+      "loss": 0.0009,
+      "step": 15524
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 1.200581395348837e-07,
+      "loss": 0.0006,
+      "step": 15525
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 1.2004651162790697e-07,
+      "loss": 0.0005,
+      "step": 15526
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 1.2003488372093023e-07,
+      "loss": 0.0006,
+      "step": 15527
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 1.200232558139535e-07,
+      "loss": 0.0005,
+      "step": 15528
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 1.2001162790697673e-07,
+      "loss": 0.0005,
+      "step": 15529
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 1.2e-07,
+      "loss": 0.0007,
+      "step": 15530
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 1.1998837209302325e-07,
+      "loss": 0.0006,
+      "step": 15531
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 1.199767441860465e-07,
+      "loss": 0.0005,
+      "step": 15532
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 1.1996511627906977e-07,
+      "loss": 0.0037,
+      "step": 15533
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 1.1995348837209304e-07,
+      "loss": 0.0005,
+      "step": 15534
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 1.1994186046511627e-07,
+      "loss": 0.0166,
+      "step": 15535
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 1.1993023255813953e-07,
+      "loss": 0.0006,
+      "step": 15536
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 1.199186046511628e-07,
+      "loss": 0.0004,
+      "step": 15537
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 1.1990697674418603e-07,
+      "loss": 0.0005,
+      "step": 15538
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 1.198953488372093e-07,
+      "loss": 0.0005,
+      "step": 15539
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 1.1988372093023255e-07,
+      "loss": 0.0006,
+      "step": 15540
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 1.1987209302325582e-07,
+      "loss": 0.0106,
+      "step": 15541
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 1.1986046511627905e-07,
+      "loss": 0.0006,
+      "step": 15542
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 1.198488372093023e-07,
+      "loss": 0.0007,
+      "step": 15543
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 1.1983720930232557e-07,
+      "loss": 0.0006,
+      "step": 15544
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 1.1982558139534884e-07,
+      "loss": 0.0004,
+      "step": 15545
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 1.1981395348837207e-07,
+      "loss": 0.0022,
+      "step": 15546
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 1.1980232558139536e-07,
+      "loss": 0.0006,
+      "step": 15547
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 1.197906976744186e-07,
+      "loss": 0.0007,
+      "step": 15548
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 1.1977906976744186e-07,
+      "loss": 0.0005,
+      "step": 15549
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 1.1976744186046512e-07,
+      "loss": 0.0005,
+      "step": 15550
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 1.1975581395348838e-07,
+      "loss": 0.0005,
+      "step": 15551
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 1.1974418604651161e-07,
+      "loss": 0.0007,
+      "step": 15552
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 1.1973255813953488e-07,
+      "loss": 0.0006,
+      "step": 15553
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 1.1972093023255814e-07,
+      "loss": 0.0007,
+      "step": 15554
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 1.197093023255814e-07,
+      "loss": 0.0005,
+      "step": 15555
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 1.1969767441860463e-07,
+      "loss": 0.0006,
+      "step": 15556
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 1.196860465116279e-07,
+      "loss": 0.0009,
+      "step": 15557
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 1.1967441860465116e-07,
+      "loss": 0.0006,
+      "step": 15558
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 1.1966279069767442e-07,
+      "loss": 0.0008,
+      "step": 15559
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 1.1965116279069766e-07,
+      "loss": 0.0006,
+      "step": 15560
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 1.1963953488372094e-07,
+      "loss": 0.0041,
+      "step": 15561
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 1.1962790697674418e-07,
+      "loss": 0.0004,
+      "step": 15562
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 1.1961627906976744e-07,
+      "loss": 0.0005,
+      "step": 15563
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 1.196046511627907e-07,
+      "loss": 2.3909,
+      "step": 15564
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 1.1959302325581394e-07,
+      "loss": 0.0006,
+      "step": 15565
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 1.195813953488372e-07,
+      "loss": 0.0005,
+      "step": 15566
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 1.1956976744186046e-07,
+      "loss": 0.0008,
+      "step": 15567
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 1.1955813953488372e-07,
+      "loss": 0.0006,
+      "step": 15568
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 1.1954651162790696e-07,
+      "loss": 0.0005,
+      "step": 15569
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 1.1953488372093022e-07,
+      "loss": 0.0004,
+      "step": 15570
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 1.1952325581395348e-07,
+      "loss": 0.0005,
+      "step": 15571
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 1.1951162790697674e-07,
+      "loss": 0.6078,
+      "step": 15572
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 1.1949999999999998e-07,
+      "loss": 0.001,
+      "step": 15573
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 1.1948837209302327e-07,
+      "loss": 0.0028,
+      "step": 15574
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 1.194767441860465e-07,
+      "loss": 0.0005,
+      "step": 15575
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 1.1946511627906976e-07,
+      "loss": 0.0007,
+      "step": 15576
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 1.1945348837209303e-07,
+      "loss": 0.0006,
+      "step": 15577
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 1.1944186046511629e-07,
+      "loss": 0.0011,
+      "step": 15578
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 1.1943023255813952e-07,
+      "loss": 0.0006,
+      "step": 15579
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 1.1941860465116278e-07,
+      "loss": 0.0006,
+      "step": 15580
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 1.1940697674418605e-07,
+      "loss": 0.0005,
+      "step": 15581
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 1.193953488372093e-07,
+      "loss": 0.0006,
+      "step": 15582
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 1.1938372093023254e-07,
+      "loss": 0.0004,
+      "step": 15583
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 1.193720930232558e-07,
+      "loss": 0.0008,
+      "step": 15584
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 1.1936046511627907e-07,
+      "loss": 0.0015,
+      "step": 15585
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 1.1934883720930233e-07,
+      "loss": 0.0009,
+      "step": 15586
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 1.1933720930232556e-07,
+      "loss": 0.0028,
+      "step": 15587
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 1.1932558139534885e-07,
+      "loss": 0.0004,
+      "step": 15588
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 1.1931395348837209e-07,
+      "loss": 0.0015,
+      "step": 15589
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 1.1930232558139535e-07,
+      "loss": 0.0007,
+      "step": 15590
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 1.192906976744186e-07,
+      "loss": 0.0007,
+      "step": 15591
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 1.1927906976744184e-07,
+      "loss": 0.0006,
+      "step": 15592
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 1.192674418604651e-07,
+      "loss": 0.0006,
+      "step": 15593
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 1.1925581395348837e-07,
+      "loss": 0.0011,
+      "step": 15594
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 1.1924418604651163e-07,
+      "loss": 0.0005,
+      "step": 15595
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 1.1923255813953487e-07,
+      "loss": 0.1688,
+      "step": 15596
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 1.1922093023255813e-07,
+      "loss": 0.0007,
+      "step": 15597
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 1.192093023255814e-07,
+      "loss": 0.0007,
+      "step": 15598
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 1.1919767441860465e-07,
+      "loss": 0.0009,
+      "step": 15599
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 1.191860465116279e-07,
+      "loss": 0.0005,
+      "step": 15600
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 1.1917441860465116e-07,
+      "loss": 0.0005,
+      "step": 15601
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 1.1916279069767441e-07,
+      "loss": 0.0006,
+      "step": 15602
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 1.1915116279069767e-07,
+      "loss": 0.0004,
+      "step": 15603
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 1.1913953488372092e-07,
+      "loss": 0.0006,
+      "step": 15604
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 1.1912790697674418e-07,
+      "loss": 0.0009,
+      "step": 15605
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 1.1911627906976743e-07,
+      "loss": 0.0007,
+      "step": 15606
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 1.191046511627907e-07,
+      "loss": 0.0006,
+      "step": 15607
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 1.1909302325581394e-07,
+      "loss": 0.0005,
+      "step": 15608
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 1.1908139534883721e-07,
+      "loss": 0.0006,
+      "step": 15609
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 1.1906976744186046e-07,
+      "loss": 2.3653,
+      "step": 15610
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 1.1905813953488372e-07,
+      "loss": 0.001,
+      "step": 15611
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 1.1904651162790697e-07,
+      "loss": 0.0005,
+      "step": 15612
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 1.1903488372093023e-07,
+      "loss": 0.0007,
+      "step": 15613
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 1.1902325581395348e-07,
+      "loss": 0.0007,
+      "step": 15614
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 1.1901162790697675e-07,
+      "loss": 0.0007,
+      "step": 15615
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 1.19e-07,
+      "loss": 0.0006,
+      "step": 15616
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 1.1898837209302324e-07,
+      "loss": 0.0006,
+      "step": 15617
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 1.189767441860465e-07,
+      "loss": 0.0011,
+      "step": 15618
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 1.1896511627906975e-07,
+      "loss": 1.0249,
+      "step": 15619
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 1.1895348837209301e-07,
+      "loss": 0.0006,
+      "step": 15620
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 1.1894186046511626e-07,
+      "loss": 0.0007,
+      "step": 15621
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 1.1893023255813954e-07,
+      "loss": 0.0005,
+      "step": 15622
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 1.1891860465116279e-07,
+      "loss": 0.0005,
+      "step": 15623
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 1.1890697674418605e-07,
+      "loss": 0.0005,
+      "step": 15624
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 1.188953488372093e-07,
+      "loss": 0.0006,
+      "step": 15625
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 1.1888372093023256e-07,
+      "loss": 0.0006,
+      "step": 15626
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 1.188720930232558e-07,
+      "loss": 0.0005,
+      "step": 15627
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 1.1886046511627907e-07,
+      "loss": 0.0011,
+      "step": 15628
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 1.1884883720930232e-07,
+      "loss": 0.0006,
+      "step": 15629
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 1.1883720930232558e-07,
+      "loss": 0.0005,
+      "step": 15630
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 1.1882558139534883e-07,
+      "loss": 0.0007,
+      "step": 15631
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 1.1881395348837209e-07,
+      "loss": 0.0006,
+      "step": 15632
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 1.1880232558139534e-07,
+      "loss": 0.0005,
+      "step": 15633
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 1.1879069767441861e-07,
+      "loss": 0.0005,
+      "step": 15634
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 1.1877906976744185e-07,
+      "loss": 0.0006,
+      "step": 15635
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 1.1876744186046512e-07,
+      "loss": 0.0005,
+      "step": 15636
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 1.1875581395348837e-07,
+      "loss": 0.0006,
+      "step": 15637
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 1.1874418604651163e-07,
+      "loss": 0.0009,
+      "step": 15638
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 1.1873255813953488e-07,
+      "loss": 0.0007,
+      "step": 15639
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 1.1872093023255814e-07,
+      "loss": 0.0012,
+      "step": 15640
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 1.1870930232558139e-07,
+      "loss": 0.0005,
+      "step": 15641
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.1869767441860464e-07,
+      "loss": 0.0006,
+      "step": 15642
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.186860465116279e-07,
+      "loss": 0.0016,
+      "step": 15643
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.1867441860465115e-07,
+      "loss": 0.0004,
+      "step": 15644
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.1866279069767441e-07,
+      "loss": 0.0016,
+      "step": 15645
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.1865116279069766e-07,
+      "loss": 0.0004,
+      "step": 15646
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 1.1863953488372092e-07,
+      "loss": 0.0004,
+      "step": 15647
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 1.1862790697674417e-07,
+      "loss": 0.0006,
+      "step": 15648
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 1.1861627906976744e-07,
+      "loss": 1.7861,
+      "step": 15649
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 1.1860465116279068e-07,
+      "loss": 0.0008,
+      "step": 15650
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 1.1859302325581395e-07,
+      "loss": 0.0007,
+      "step": 15651
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 1.185813953488372e-07,
+      "loss": 0.0006,
+      "step": 15652
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 1.1856976744186047e-07,
+      "loss": 0.0006,
+      "step": 15653
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 1.1855813953488371e-07,
+      "loss": 0.0006,
+      "step": 15654
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 1.1854651162790698e-07,
+      "loss": 0.0005,
+      "step": 15655
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 1.1853488372093022e-07,
+      "loss": 0.0005,
+      "step": 15656
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 1.1852325581395349e-07,
+      "loss": 0.0006,
+      "step": 15657
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 1.1851162790697673e-07,
+      "loss": 0.0006,
+      "step": 15658
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 1.185e-07,
+      "loss": 0.0006,
+      "step": 15659
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 1.1848837209302324e-07,
+      "loss": 0.0006,
+      "step": 15660
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 1.184767441860465e-07,
+      "loss": 0.0006,
+      "step": 15661
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 1.1846511627906975e-07,
+      "loss": 0.0005,
+      "step": 15662
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 1.1845348837209303e-07,
+      "loss": 0.0006,
+      "step": 15663
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 1.1844186046511628e-07,
+      "loss": 0.0004,
+      "step": 15664
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 1.1843023255813954e-07,
+      "loss": 0.0005,
+      "step": 15665
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 1.1841860465116279e-07,
+      "loss": 0.0005,
+      "step": 15666
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 1.1840697674418605e-07,
+      "loss": 0.0006,
+      "step": 15667
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 1.183953488372093e-07,
+      "loss": 3.3665,
+      "step": 15668
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 1.1838372093023255e-07,
+      "loss": 0.0013,
+      "step": 15669
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 1.1837209302325581e-07,
+      "loss": 0.0006,
+      "step": 15670
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 1.1836046511627906e-07,
+      "loss": 0.0004,
+      "step": 15671
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 1.1834883720930232e-07,
+      "loss": 0.0007,
+      "step": 15672
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.1833720930232557e-07,
+      "loss": 0.0007,
+      "step": 15673
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.1832558139534883e-07,
+      "loss": 0.0007,
+      "step": 15674
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.1831395348837208e-07,
+      "loss": 0.0006,
+      "step": 15675
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.1830232558139535e-07,
+      "loss": 0.0005,
+      "step": 15676
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.1829069767441859e-07,
+      "loss": 0.0005,
+      "step": 15677
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 1.1827906976744186e-07,
+      "loss": 0.0009,
+      "step": 15678
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 1.1826744186046511e-07,
+      "loss": 0.0004,
+      "step": 15679
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 1.1825581395348837e-07,
+      "loss": 0.0047,
+      "step": 15680
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 1.1824418604651162e-07,
+      "loss": 0.0005,
+      "step": 15681
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 1.1823255813953488e-07,
+      "loss": 0.0007,
+      "step": 15682
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 1.1822093023255813e-07,
+      "loss": 0.0006,
+      "step": 15683
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 1.1820930232558139e-07,
+      "loss": 0.0006,
+      "step": 15684
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 1.1819767441860464e-07,
+      "loss": 0.0004,
+      "step": 15685
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 1.181860465116279e-07,
+      "loss": 0.0004,
+      "step": 15686
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 1.1817441860465115e-07,
+      "loss": 0.0005,
+      "step": 15687
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 1.1816279069767441e-07,
+      "loss": 0.0007,
+      "step": 15688
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 1.1815116279069766e-07,
+      "loss": 0.0005,
+      "step": 15689
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 1.1813953488372094e-07,
+      "loss": 0.0007,
+      "step": 15690
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 1.1812790697674419e-07,
+      "loss": 0.0004,
+      "step": 15691
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 1.1811627906976745e-07,
+      "loss": 0.0007,
+      "step": 15692
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 1.181046511627907e-07,
+      "loss": 0.0005,
+      "step": 15693
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 1.1809302325581396e-07,
+      "loss": 0.0007,
+      "step": 15694
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 1.180813953488372e-07,
+      "loss": 0.0005,
+      "step": 15695
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 1.1806976744186045e-07,
+      "loss": 0.0008,
+      "step": 15696
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 1.1805813953488372e-07,
+      "loss": 0.0005,
+      "step": 15697
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 1.1804651162790696e-07,
+      "loss": 0.0006,
+      "step": 15698
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 1.1803488372093023e-07,
+      "loss": 0.0007,
+      "step": 15699
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 1.1802325581395347e-07,
+      "loss": 0.0007,
+      "step": 15700
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 1.1801162790697674e-07,
+      "loss": 0.0006,
+      "step": 15701
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 1.1799999999999998e-07,
+      "loss": 0.0005,
+      "step": 15702
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 1.1798837209302325e-07,
+      "loss": 0.0005,
+      "step": 15703
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 1.179767441860465e-07,
+      "loss": 0.0006,
+      "step": 15704
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 1.1796511627906977e-07,
+      "loss": 0.0005,
+      "step": 15705
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 1.1795348837209302e-07,
+      "loss": 0.0043,
+      "step": 15706
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 1.1794186046511628e-07,
+      "loss": 0.0007,
+      "step": 15707
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 1.1793023255813953e-07,
+      "loss": 0.0005,
+      "step": 15708
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 1.1791860465116279e-07,
+      "loss": 0.0004,
+      "step": 15709
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 1.1790697674418604e-07,
+      "loss": 0.0153,
+      "step": 15710
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 1.178953488372093e-07,
+      "loss": 0.2526,
+      "step": 15711
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 1.1788372093023255e-07,
+      "loss": 0.0005,
+      "step": 15712
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 1.1787209302325581e-07,
+      "loss": 0.0004,
+      "step": 15713
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 1.1786046511627906e-07,
+      "loss": 0.0017,
+      "step": 15714
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 1.1784883720930232e-07,
+      "loss": 0.0006,
+      "step": 15715
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 1.1783720930232557e-07,
+      "loss": 0.0007,
+      "step": 15716
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 1.1782558139534884e-07,
+      "loss": 0.0006,
+      "step": 15717
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 1.1781395348837209e-07,
+      "loss": 0.0006,
+      "step": 15718
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 1.1780232558139535e-07,
+      "loss": 0.0006,
+      "step": 15719
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 1.177906976744186e-07,
+      "loss": 0.8632,
+      "step": 15720
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 1.1777906976744186e-07,
+      "loss": 0.0005,
+      "step": 15721
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 1.1776744186046511e-07,
+      "loss": 0.0005,
+      "step": 15722
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 1.1775581395348836e-07,
+      "loss": 0.0007,
+      "step": 15723
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 1.1774418604651162e-07,
+      "loss": 0.0006,
+      "step": 15724
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 1.1773255813953487e-07,
+      "loss": 0.0008,
+      "step": 15725
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 1.1772093023255813e-07,
+      "loss": 0.0008,
+      "step": 15726
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 1.1770930232558138e-07,
+      "loss": 0.3824,
+      "step": 15727
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 1.1769767441860464e-07,
+      "loss": 0.0031,
+      "step": 15728
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 1.1768604651162789e-07,
+      "loss": 0.0005,
+      "step": 15729
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 1.1767441860465115e-07,
+      "loss": 0.0007,
+      "step": 15730
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 1.176627906976744e-07,
+      "loss": 0.0006,
+      "step": 15731
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 1.1765116279069768e-07,
+      "loss": 0.0011,
+      "step": 15732
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 1.1763953488372093e-07,
+      "loss": 0.0009,
+      "step": 15733
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 1.1762790697674419e-07,
+      "loss": 0.0013,
+      "step": 15734
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 1.1761627906976744e-07,
+      "loss": 0.0008,
+      "step": 15735
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 1.176046511627907e-07,
+      "loss": 0.0014,
+      "step": 15736
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 1.1759302325581395e-07,
+      "loss": 0.001,
+      "step": 15737
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 1.1758139534883721e-07,
+      "loss": 0.0005,
+      "step": 15738
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 1.1756976744186046e-07,
+      "loss": 0.0007,
+      "step": 15739
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 1.1755813953488372e-07,
+      "loss": 0.0005,
+      "step": 15740
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 1.1754651162790697e-07,
+      "loss": 0.0005,
+      "step": 15741
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 1.1753488372093023e-07,
+      "loss": 0.0005,
+      "step": 15742
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 1.1752325581395348e-07,
+      "loss": 0.0007,
+      "step": 15743
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 1.1751162790697675e-07,
+      "loss": 0.0005,
+      "step": 15744
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 1.175e-07,
+      "loss": 0.0006,
+      "step": 15745
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 1.1748837209302326e-07,
+      "loss": 0.0006,
+      "step": 15746
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 1.1747674418604651e-07,
+      "loss": 0.0006,
+      "step": 15747
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 1.1746511627906977e-07,
+      "loss": 0.0011,
+      "step": 15748
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 1.1745348837209302e-07,
+      "loss": 0.001,
+      "step": 15749
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 1.1744186046511627e-07,
+      "loss": 0.0005,
+      "step": 15750
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 1.1743023255813953e-07,
+      "loss": 0.0007,
+      "step": 15751
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 1.1741860465116278e-07,
+      "loss": 0.0005,
+      "step": 15752
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 1.1740697674418604e-07,
+      "loss": 0.0006,
+      "step": 15753
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 1.1739534883720929e-07,
+      "loss": 0.0007,
+      "step": 15754
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 1.1738372093023255e-07,
+      "loss": 0.0006,
+      "step": 15755
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 1.173720930232558e-07,
+      "loss": 0.0006,
+      "step": 15756
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 1.1736046511627906e-07,
+      "loss": 0.0013,
+      "step": 15757
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 1.1734883720930231e-07,
+      "loss": 0.0005,
+      "step": 15758
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 1.1733720930232558e-07,
+      "loss": 0.0006,
+      "step": 15759
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 1.1732558139534883e-07,
+      "loss": 0.0005,
+      "step": 15760
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 1.173139534883721e-07,
+      "loss": 0.0005,
+      "step": 15761
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 1.1730232558139534e-07,
+      "loss": 0.0005,
+      "step": 15762
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 1.172906976744186e-07,
+      "loss": 0.0004,
+      "step": 15763
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 1.1727906976744185e-07,
+      "loss": 0.0005,
+      "step": 15764
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 1.1726744186046512e-07,
+      "loss": 0.0004,
+      "step": 15765
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 1.1725581395348836e-07,
+      "loss": 0.0005,
+      "step": 15766
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 1.1724418604651163e-07,
+      "loss": 0.0005,
+      "step": 15767
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 1.1723255813953487e-07,
+      "loss": 0.0005,
+      "step": 15768
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 1.1722093023255814e-07,
+      "loss": 0.0005,
+      "step": 15769
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 1.1720930232558138e-07,
+      "loss": 0.3641,
+      "step": 15770
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 1.1719767441860466e-07,
+      "loss": 0.0005,
+      "step": 15771
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 1.171860465116279e-07,
+      "loss": 0.0008,
+      "step": 15772
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 1.1717441860465117e-07,
+      "loss": 0.0005,
+      "step": 15773
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 1.1716279069767442e-07,
+      "loss": 0.0009,
+      "step": 15774
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 1.1715116279069768e-07,
+      "loss": 0.0006,
+      "step": 15775
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 1.1713953488372093e-07,
+      "loss": 0.0005,
+      "step": 15776
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 1.1712790697674418e-07,
+      "loss": 0.0005,
+      "step": 15777
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 1.1711627906976744e-07,
+      "loss": 0.0004,
+      "step": 15778
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 1.1710465116279069e-07,
+      "loss": 0.0005,
+      "step": 15779
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 1.1709302325581395e-07,
+      "loss": 0.0005,
+      "step": 15780
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 1.170813953488372e-07,
+      "loss": 0.0005,
+      "step": 15781
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 1.1706976744186046e-07,
+      "loss": 0.0006,
+      "step": 15782
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 1.1705813953488371e-07,
+      "loss": 0.0006,
+      "step": 15783
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 1.1704651162790697e-07,
+      "loss": 0.0004,
+      "step": 15784
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 1.1703488372093022e-07,
+      "loss": 0.0032,
+      "step": 15785
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 1.1702325581395349e-07,
+      "loss": 0.0008,
+      "step": 15786
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 1.1701162790697674e-07,
+      "loss": 1.2015,
+      "step": 15787
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 1.17e-07,
+      "loss": 0.0009,
+      "step": 15788
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 1.1698837209302325e-07,
+      "loss": 0.0005,
+      "step": 15789
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 1.1697674418604651e-07,
+      "loss": 0.0006,
+      "step": 15790
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 1.1696511627906976e-07,
+      "loss": 0.0004,
+      "step": 15791
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 1.1695348837209302e-07,
+      "loss": 0.0004,
+      "step": 15792
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 1.1694186046511627e-07,
+      "loss": 0.0007,
+      "step": 15793
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 1.1693023255813953e-07,
+      "loss": 0.0004,
+      "step": 15794
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 1.1691860465116278e-07,
+      "loss": 0.0005,
+      "step": 15795
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 1.1690697674418604e-07,
+      "loss": 0.0007,
+      "step": 15796
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 1.1689534883720929e-07,
+      "loss": 0.0006,
+      "step": 15797
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 1.1688372093023257e-07,
+      "loss": 0.0004,
+      "step": 15798
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 1.168720930232558e-07,
+      "loss": 0.0005,
+      "step": 15799
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 1.1686046511627908e-07,
+      "loss": 0.0005,
+      "step": 15800
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 1.1684883720930232e-07,
+      "loss": 0.0007,
+      "step": 15801
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 1.1683720930232557e-07,
+      "loss": 0.0007,
+      "step": 15802
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 1.1682558139534884e-07,
+      "loss": 0.0005,
+      "step": 15803
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 1.1681395348837208e-07,
+      "loss": 0.0008,
+      "step": 15804
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 1.1680232558139535e-07,
+      "loss": 0.0284,
+      "step": 15805
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 1.167906976744186e-07,
+      "loss": 0.0005,
+      "step": 15806
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 1.1677906976744186e-07,
+      "loss": 0.0005,
+      "step": 15807
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 1.167674418604651e-07,
+      "loss": 0.0004,
+      "step": 15808
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 1.1675581395348837e-07,
+      "loss": 0.0007,
+      "step": 15809
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 1.1674418604651161e-07,
+      "loss": 2.707,
+      "step": 15810
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 1.1673255813953488e-07,
+      "loss": 0.0031,
+      "step": 15811
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 1.1672093023255812e-07,
+      "loss": 0.0008,
+      "step": 15812
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 1.167093023255814e-07,
+      "loss": 0.0004,
+      "step": 15813
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 1.1669767441860465e-07,
+      "loss": 0.0005,
+      "step": 15814
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 1.1668604651162791e-07,
+      "loss": 0.0004,
+      "step": 15815
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 1.1667441860465116e-07,
+      "loss": 0.0005,
+      "step": 15816
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 1.1666279069767442e-07,
+      "loss": 0.0066,
+      "step": 15817
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 1.1665116279069767e-07,
+      "loss": 0.0006,
+      "step": 15818
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 1.1663953488372093e-07,
+      "loss": 0.0007,
+      "step": 15819
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 1.1662790697674418e-07,
+      "loss": 0.0006,
+      "step": 15820
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 1.1661627906976744e-07,
+      "loss": 0.0005,
+      "step": 15821
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 1.1660465116279069e-07,
+      "loss": 0.0007,
+      "step": 15822
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 1.1659302325581395e-07,
+      "loss": 0.0006,
+      "step": 15823
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 1.165813953488372e-07,
+      "loss": 0.0006,
+      "step": 15824
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 1.1656976744186046e-07,
+      "loss": 0.0005,
+      "step": 15825
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 1.1655813953488371e-07,
+      "loss": 0.0005,
+      "step": 15826
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 1.1654651162790698e-07,
+      "loss": 0.0006,
+      "step": 15827
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 1.1653488372093023e-07,
+      "loss": 0.0009,
+      "step": 15828
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 1.1652325581395348e-07,
+      "loss": 0.0008,
+      "step": 15829
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 1.1651162790697674e-07,
+      "loss": 0.0006,
+      "step": 15830
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 1.1649999999999999e-07,
+      "loss": 0.0006,
+      "step": 15831
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 1.1648837209302325e-07,
+      "loss": 0.0008,
+      "step": 15832
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 1.164767441860465e-07,
+      "loss": 0.0007,
+      "step": 15833
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 1.1646511627906976e-07,
+      "loss": 0.0015,
+      "step": 15834
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 1.1645348837209301e-07,
+      "loss": 0.0006,
+      "step": 15835
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 1.1644186046511627e-07,
+      "loss": 2.936,
+      "step": 15836
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 1.1643023255813952e-07,
+      "loss": 0.0006,
+      "step": 15837
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 1.1641860465116278e-07,
+      "loss": 0.0008,
+      "step": 15838
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 1.1640697674418603e-07,
+      "loss": 0.0005,
+      "step": 15839
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 1.1639534883720931e-07,
+      "loss": 0.0003,
+      "step": 15840
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 1.1638372093023254e-07,
+      "loss": 0.0005,
+      "step": 15841
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 1.1637209302325582e-07,
+      "loss": 0.0006,
+      "step": 15842
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 1.1636046511627907e-07,
+      "loss": 0.0008,
+      "step": 15843
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 1.1634883720930233e-07,
+      "loss": 0.0008,
+      "step": 15844
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 1.1633720930232558e-07,
+      "loss": 0.0007,
+      "step": 15845
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 1.1632558139534884e-07,
+      "loss": 0.0006,
+      "step": 15846
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 1.1631395348837209e-07,
+      "loss": 0.0005,
+      "step": 15847
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 1.1630232558139535e-07,
+      "loss": 0.0009,
+      "step": 15848
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 1.162906976744186e-07,
+      "loss": 0.0004,
+      "step": 15849
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 1.1627906976744186e-07,
+      "loss": 0.0006,
+      "step": 15850
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 1.162674418604651e-07,
+      "loss": 0.0005,
+      "step": 15851
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 1.1625581395348837e-07,
+      "loss": 0.0015,
+      "step": 15852
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 1.1624418604651162e-07,
+      "loss": 0.0005,
+      "step": 15853
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 1.1623255813953489e-07,
+      "loss": 0.004,
+      "step": 15854
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 1.1622093023255814e-07,
+      "loss": 0.001,
+      "step": 15855
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 1.1620930232558139e-07,
+      "loss": 0.0004,
+      "step": 15856
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 1.1619767441860465e-07,
+      "loss": 0.0011,
+      "step": 15857
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 1.161860465116279e-07,
+      "loss": 0.0007,
+      "step": 15858
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 1.1617441860465116e-07,
+      "loss": 0.0005,
+      "step": 15859
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 1.1616279069767441e-07,
+      "loss": 0.0007,
+      "step": 15860
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 1.1615116279069767e-07,
+      "loss": 0.0006,
+      "step": 15861
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 1.1613953488372092e-07,
+      "loss": 0.011,
+      "step": 15862
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 1.1612790697674418e-07,
+      "loss": 0.0006,
+      "step": 15863
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 1.1611627906976743e-07,
+      "loss": 0.0004,
+      "step": 15864
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 1.1610465116279069e-07,
+      "loss": 0.0005,
+      "step": 15865
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 1.1609302325581394e-07,
+      "loss": 0.0004,
+      "step": 15866
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 1.1608139534883721e-07,
+      "loss": 0.0006,
+      "step": 15867
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 1.1606976744186045e-07,
+      "loss": 0.0005,
+      "step": 15868
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 1.1605813953488372e-07,
+      "loss": 0.0006,
+      "step": 15869
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 1.1604651162790697e-07,
+      "loss": 0.001,
+      "step": 15870
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 1.1603488372093023e-07,
+      "loss": 0.0005,
+      "step": 15871
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 1.1602325581395348e-07,
+      "loss": 0.0018,
+      "step": 15872
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 1.1601162790697674e-07,
+      "loss": 0.0007,
+      "step": 15873
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 1.1599999999999999e-07,
+      "loss": 0.0007,
+      "step": 15874
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 1.1598837209302325e-07,
+      "loss": 0.0013,
+      "step": 15875
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 1.159767441860465e-07,
+      "loss": 0.0008,
+      "step": 15876
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 1.1596511627906977e-07,
+      "loss": 0.0007,
+      "step": 15877
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 1.1595348837209301e-07,
+      "loss": 0.0007,
+      "step": 15878
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 1.1594186046511628e-07,
+      "loss": 2.1251,
+      "step": 15879
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 1.1593023255813952e-07,
+      "loss": 0.0006,
+      "step": 15880
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 1.159186046511628e-07,
+      "loss": 0.0008,
+      "step": 15881
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 1.1590697674418605e-07,
+      "loss": 0.0007,
+      "step": 15882
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 1.158953488372093e-07,
+      "loss": 0.0005,
+      "step": 15883
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 1.1588372093023256e-07,
+      "loss": 0.0012,
+      "step": 15884
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 1.158720930232558e-07,
+      "loss": 0.0006,
+      "step": 15885
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 1.1586046511627907e-07,
+      "loss": 0.0063,
+      "step": 15886
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 1.1584883720930232e-07,
+      "loss": 0.0688,
+      "step": 15887
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 1.1583720930232558e-07,
+      "loss": 0.0006,
+      "step": 15888
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 1.1582558139534883e-07,
+      "loss": 0.0005,
+      "step": 15889
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 1.1581395348837209e-07,
+      "loss": 0.0006,
+      "step": 15890
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 1.1580232558139534e-07,
+      "loss": 0.0006,
+      "step": 15891
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 1.157906976744186e-07,
+      "loss": 0.0006,
+      "step": 15892
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 1.1577906976744185e-07,
+      "loss": 0.0005,
+      "step": 15893
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 1.1576744186046511e-07,
+      "loss": 0.0012,
+      "step": 15894
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 1.1575581395348836e-07,
+      "loss": 0.0009,
+      "step": 15895
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 1.1574418604651163e-07,
+      "loss": 0.0008,
+      "step": 15896
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 1.1573255813953488e-07,
+      "loss": 0.0008,
+      "step": 15897
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 1.1572093023255814e-07,
+      "loss": 0.0006,
+      "step": 15898
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 1.1570930232558139e-07,
+      "loss": 0.0005,
+      "step": 15899
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 1.1569767441860465e-07,
+      "loss": 0.0005,
+      "step": 15900
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 1.156860465116279e-07,
+      "loss": 0.0005,
+      "step": 15901
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 1.1567441860465116e-07,
+      "loss": 0.0009,
+      "step": 15902
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 1.1566279069767441e-07,
+      "loss": 0.001,
+      "step": 15903
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 1.1565116279069767e-07,
+      "loss": 0.0005,
+      "step": 15904
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 1.1563953488372092e-07,
+      "loss": 0.0003,
+      "step": 15905
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 1.1562790697674418e-07,
+      "loss": 0.0029,
+      "step": 15906
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 1.1561627906976743e-07,
+      "loss": 0.0008,
+      "step": 15907
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 1.156046511627907e-07,
+      "loss": 1.3099,
+      "step": 15908
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 1.1559302325581395e-07,
+      "loss": 0.0016,
+      "step": 15909
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 1.1558139534883719e-07,
+      "loss": 0.0006,
+      "step": 15910
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 1.1556976744186046e-07,
+      "loss": 0.0004,
+      "step": 15911
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 1.1555813953488371e-07,
+      "loss": 0.0005,
+      "step": 15912
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 1.1554651162790697e-07,
+      "loss": 0.0004,
+      "step": 15913
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 1.1553488372093022e-07,
+      "loss": 0.0008,
+      "step": 15914
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 1.1552325581395349e-07,
+      "loss": 0.0011,
+      "step": 15915
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.1551162790697673e-07,
+      "loss": 0.0005,
+      "step": 15916
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.155e-07,
+      "loss": 0.0006,
+      "step": 15917
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.1548837209302324e-07,
+      "loss": 0.0007,
+      "step": 15918
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.154767441860465e-07,
+      "loss": 0.0006,
+      "step": 15919
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.1546511627906975e-07,
+      "loss": 0.0004,
+      "step": 15920
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.1545348837209302e-07,
+      "loss": 0.0012,
+      "step": 15921
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 1.1544186046511626e-07,
+      "loss": 0.0007,
+      "step": 15922
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 1.1543023255813954e-07,
+      "loss": 0.0006,
+      "step": 15923
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 1.1541860465116279e-07,
+      "loss": 0.0008,
+      "step": 15924
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 1.1540697674418605e-07,
+      "loss": 0.0006,
+      "step": 15925
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 1.153953488372093e-07,
+      "loss": 0.0005,
+      "step": 15926
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 1.1538372093023256e-07,
+      "loss": 0.0009,
+      "step": 15927
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 1.1537209302325581e-07,
+      "loss": 0.0008,
+      "step": 15928
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 1.1536046511627907e-07,
+      "loss": 0.0005,
+      "step": 15929
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 1.1534883720930232e-07,
+      "loss": 0.0005,
+      "step": 15930
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 1.1533720930232558e-07,
+      "loss": 0.0007,
+      "step": 15931
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 1.1532558139534883e-07,
+      "loss": 0.0005,
+      "step": 15932
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 1.1531395348837209e-07,
+      "loss": 0.0009,
+      "step": 15933
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 1.1530232558139534e-07,
+      "loss": 0.0006,
+      "step": 15934
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 1.1529069767441861e-07,
+      "loss": 0.0012,
+      "step": 15935
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 1.1527906976744186e-07,
+      "loss": 0.0006,
+      "step": 15936
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 1.152674418604651e-07,
+      "loss": 0.0006,
+      "step": 15937
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 1.1525581395348837e-07,
+      "loss": 0.0006,
+      "step": 15938
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 1.1524418604651162e-07,
+      "loss": 0.001,
+      "step": 15939
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 1.1523255813953488e-07,
+      "loss": 0.0006,
+      "step": 15940
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 1.1522093023255813e-07,
+      "loss": 0.0005,
+      "step": 15941
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 1.1520930232558139e-07,
+      "loss": 0.0006,
+      "step": 15942
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 1.1519767441860464e-07,
+      "loss": 0.0006,
+      "step": 15943
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 1.151860465116279e-07,
+      "loss": 0.0005,
+      "step": 15944
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 1.1517441860465115e-07,
+      "loss": 0.001,
+      "step": 15945
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 1.1516279069767441e-07,
+      "loss": 0.0008,
+      "step": 15946
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 1.1515116279069766e-07,
+      "loss": 0.0009,
+      "step": 15947
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 1.1513953488372092e-07,
+      "loss": 0.0005,
+      "step": 15948
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 1.1512790697674417e-07,
+      "loss": 0.0004,
+      "step": 15949
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 1.1511627906976745e-07,
+      "loss": 0.0005,
+      "step": 15950
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 1.151046511627907e-07,
+      "loss": 0.0006,
+      "step": 15951
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 1.1509302325581396e-07,
+      "loss": 0.0003,
+      "step": 15952
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 1.150813953488372e-07,
+      "loss": 0.0005,
+      "step": 15953
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 1.1506976744186047e-07,
+      "loss": 0.0005,
+      "step": 15954
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 1.1505813953488372e-07,
+      "loss": 0.0006,
+      "step": 15955
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 1.1504651162790698e-07,
+      "loss": 0.0004,
+      "step": 15956
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 1.1503488372093023e-07,
+      "loss": 0.0004,
+      "step": 15957
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 1.1502325581395349e-07,
+      "loss": 0.0004,
+      "step": 15958
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 1.1501162790697674e-07,
+      "loss": 0.0102,
+      "step": 15959
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 1.15e-07,
+      "loss": 0.0004,
+      "step": 15960
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 1.1498837209302325e-07,
+      "loss": 0.0004,
+      "step": 15961
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 1.149767441860465e-07,
+      "loss": 0.0005,
+      "step": 15962
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 1.1496511627906976e-07,
+      "loss": 0.0008,
+      "step": 15963
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 1.14953488372093e-07,
+      "loss": 0.0006,
+      "step": 15964
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 1.1494186046511628e-07,
+      "loss": 0.0006,
+      "step": 15965
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 1.1493023255813953e-07,
+      "loss": 0.0006,
+      "step": 15966
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 1.1491860465116279e-07,
+      "loss": 0.0008,
+      "step": 15967
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 1.1490697674418604e-07,
+      "loss": 0.0006,
+      "step": 15968
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 1.148953488372093e-07,
+      "loss": 0.0006,
+      "step": 15969
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 1.1488372093023255e-07,
+      "loss": 0.0005,
+      "step": 15970
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 1.1487209302325581e-07,
+      "loss": 0.0007,
+      "step": 15971
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 1.1486046511627906e-07,
+      "loss": 0.5281,
+      "step": 15972
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 1.1484883720930232e-07,
+      "loss": 0.0006,
+      "step": 15973
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 1.1483720930232557e-07,
+      "loss": 0.0005,
+      "step": 15974
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 1.1482558139534883e-07,
+      "loss": 0.0006,
+      "step": 15975
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 1.1481395348837208e-07,
+      "loss": 0.0014,
+      "step": 15976
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 1.1480232558139535e-07,
+      "loss": 0.0006,
+      "step": 15977
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 1.147906976744186e-07,
+      "loss": 0.0007,
+      "step": 15978
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 1.1477906976744186e-07,
+      "loss": 1.6841,
+      "step": 15979
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 1.1476744186046511e-07,
+      "loss": 0.0006,
+      "step": 15980
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 1.1475581395348837e-07,
+      "loss": 0.0382,
+      "step": 15981
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 1.1474418604651162e-07,
+      "loss": 0.0007,
+      "step": 15982
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 1.1473255813953488e-07,
+      "loss": 0.0555,
+      "step": 15983
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 1.1472093023255813e-07,
+      "loss": 0.0056,
+      "step": 15984
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 1.147093023255814e-07,
+      "loss": 0.0007,
+      "step": 15985
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 1.1469767441860464e-07,
+      "loss": 0.0007,
+      "step": 15986
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 1.146860465116279e-07,
+      "loss": 0.0006,
+      "step": 15987
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 1.1467441860465115e-07,
+      "loss": 0.0006,
+      "step": 15988
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 1.146627906976744e-07,
+      "loss": 0.0006,
+      "step": 15989
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 1.1465116279069766e-07,
+      "loss": 0.0004,
+      "step": 15990
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 1.1463953488372091e-07,
+      "loss": 0.0005,
+      "step": 15991
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 1.1462790697674419e-07,
+      "loss": 0.0004,
+      "step": 15992
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 1.1461627906976744e-07,
+      "loss": 0.0008,
+      "step": 15993
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 1.146046511627907e-07,
+      "loss": 0.0005,
+      "step": 15994
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 1.1459302325581395e-07,
+      "loss": 0.0019,
+      "step": 15995
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 1.1458139534883721e-07,
+      "loss": 0.0007,
+      "step": 15996
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 1.1456976744186046e-07,
+      "loss": 0.0005,
+      "step": 15997
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 1.1455813953488372e-07,
+      "loss": 0.0009,
+      "step": 15998
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 1.1454651162790697e-07,
+      "loss": 0.0006,
+      "step": 15999
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 1.1453488372093023e-07,
+      "loss": 0.0009,
+      "step": 16000
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 1.1452325581395348e-07,
+      "loss": 0.0005,
+      "step": 16001
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 1.1451162790697674e-07,
+      "loss": 0.0005,
+      "step": 16002
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 1.1449999999999999e-07,
+      "loss": 0.0006,
+      "step": 16003
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 1.1448837209302326e-07,
+      "loss": 0.0009,
+      "step": 16004
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 1.1447674418604651e-07,
+      "loss": 0.0006,
+      "step": 16005
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 1.1446511627906977e-07,
+      "loss": 0.0006,
+      "step": 16006
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 1.1445348837209302e-07,
+      "loss": 0.0172,
+      "step": 16007
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 1.1444186046511628e-07,
+      "loss": 0.0008,
+      "step": 16008
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 1.1443023255813953e-07,
+      "loss": 0.0005,
+      "step": 16009
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 1.1441860465116279e-07,
+      "loss": 0.0013,
+      "step": 16010
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 1.1440697674418604e-07,
+      "loss": 0.0006,
+      "step": 16011
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 1.143953488372093e-07,
+      "loss": 0.0004,
+      "step": 16012
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 1.1438372093023255e-07,
+      "loss": 0.0009,
+      "step": 16013
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 1.1437209302325581e-07,
+      "loss": 0.0005,
+      "step": 16014
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 1.1436046511627906e-07,
+      "loss": 0.0009,
+      "step": 16015
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 1.1434883720930231e-07,
+      "loss": 0.0005,
+      "step": 16016
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 1.1433720930232557e-07,
+      "loss": 0.0005,
+      "step": 16017
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 1.1432558139534882e-07,
+      "loss": 0.0007,
+      "step": 16018
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 1.143139534883721e-07,
+      "loss": 0.0005,
+      "step": 16019
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 1.1430232558139534e-07,
+      "loss": 0.0004,
+      "step": 16020
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 1.142906976744186e-07,
+      "loss": 0.0005,
+      "step": 16021
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 1.1427906976744185e-07,
+      "loss": 0.0008,
+      "step": 16022
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 1.1426744186046511e-07,
+      "loss": 0.0645,
+      "step": 16023
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 1.1425581395348836e-07,
+      "loss": 0.0007,
+      "step": 16024
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 1.1424418604651162e-07,
+      "loss": 0.0005,
+      "step": 16025
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 1.1423255813953487e-07,
+      "loss": 0.0008,
+      "step": 16026
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 1.1422093023255814e-07,
+      "loss": 0.0024,
+      "step": 16027
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.15163522958755493,
+      "eval_runtime": 1.5583,
+      "eval_samples_per_second": 663.545,
+      "eval_steps_per_second": 10.909,
+      "step": 16027
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 1.1420930232558138e-07,
+      "loss": 0.0004,
+      "step": 16028
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 1.1419767441860465e-07,
+      "loss": 0.0004,
+      "step": 16029
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 1.141860465116279e-07,
+      "loss": 0.0006,
+      "step": 16030
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 1.1417441860465117e-07,
+      "loss": 0.0005,
+      "step": 16031
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 1.141627906976744e-07,
+      "loss": 0.0009,
+      "step": 16032
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 1.1415116279069768e-07,
+      "loss": 0.0007,
+      "step": 16033
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 1.1413953488372093e-07,
+      "loss": 0.0007,
+      "step": 16034
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 1.1412790697674419e-07,
+      "loss": 0.0006,
+      "step": 16035
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 1.1411627906976744e-07,
+      "loss": 0.0004,
+      "step": 16036
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 1.141046511627907e-07,
+      "loss": 0.0005,
+      "step": 16037
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 1.1409302325581395e-07,
+      "loss": 0.0005,
+      "step": 16038
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 1.1408139534883721e-07,
+      "loss": 0.0005,
+      "step": 16039
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.1406976744186046e-07,
+      "loss": 0.2287,
+      "step": 16040
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.1405813953488372e-07,
+      "loss": 0.0006,
+      "step": 16041
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.1404651162790697e-07,
+      "loss": 0.0006,
+      "step": 16042
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.1403488372093022e-07,
+      "loss": 0.0005,
+      "step": 16043
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.1402325581395348e-07,
+      "loss": 0.0004,
+      "step": 16044
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 1.1401162790697673e-07,
+      "loss": 0.0004,
+      "step": 16045
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 1.14e-07,
+      "loss": 0.0004,
+      "step": 16046
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 1.1398837209302325e-07,
+      "loss": 0.0005,
+      "step": 16047
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 1.1397674418604651e-07,
+      "loss": 0.0005,
+      "step": 16048
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 1.1396511627906976e-07,
+      "loss": 0.0003,
+      "step": 16049
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 1.1395348837209302e-07,
+      "loss": 0.0004,
+      "step": 16050
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 1.1394186046511627e-07,
+      "loss": 0.0006,
+      "step": 16051
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 1.1393023255813953e-07,
+      "loss": 0.0005,
+      "step": 16052
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 1.1391860465116278e-07,
+      "loss": 0.0005,
+      "step": 16053
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 1.1390697674418604e-07,
+      "loss": 0.0007,
+      "step": 16054
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 1.1389534883720929e-07,
+      "loss": 0.0005,
+      "step": 16055
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 1.1388372093023255e-07,
+      "loss": 0.0005,
+      "step": 16056
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 1.138720930232558e-07,
+      "loss": 0.0005,
+      "step": 16057
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 1.1386046511627908e-07,
+      "loss": 0.0471,
+      "step": 16058
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 1.1384883720930231e-07,
+      "loss": 0.0004,
+      "step": 16059
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 1.1383720930232559e-07,
+      "loss": 0.0015,
+      "step": 16060
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 1.1382558139534883e-07,
+      "loss": 0.0007,
+      "step": 16061
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 1.138139534883721e-07,
+      "loss": 0.0005,
+      "step": 16062
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 1.1380232558139534e-07,
+      "loss": 0.001,
+      "step": 16063
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 1.1379069767441861e-07,
+      "loss": 0.0004,
+      "step": 16064
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 1.1377906976744186e-07,
+      "loss": 0.0008,
+      "step": 16065
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 1.1376744186046512e-07,
+      "loss": 0.001,
+      "step": 16066
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 1.1375581395348837e-07,
+      "loss": 0.0007,
+      "step": 16067
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 1.1374418604651163e-07,
+      "loss": 0.0005,
+      "step": 16068
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 1.1373255813953488e-07,
+      "loss": 0.0005,
+      "step": 16069
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 1.1372093023255812e-07,
+      "loss": 0.0007,
+      "step": 16070
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 1.1370930232558139e-07,
+      "loss": 0.0006,
+      "step": 16071
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 1.1369767441860463e-07,
+      "loss": 0.0006,
+      "step": 16072
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 1.1368604651162791e-07,
+      "loss": 0.0006,
+      "step": 16073
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 1.1367441860465116e-07,
+      "loss": 0.0004,
+      "step": 16074
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 1.1366279069767442e-07,
+      "loss": 0.0007,
+      "step": 16075
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 1.1365116279069767e-07,
+      "loss": 0.0009,
+      "step": 16076
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 1.1363953488372093e-07,
+      "loss": 0.1189,
+      "step": 16077
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 1.1362790697674418e-07,
+      "loss": 0.0005,
+      "step": 16078
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 1.1361627906976744e-07,
+      "loss": 0.0007,
+      "step": 16079
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 1.1360465116279069e-07,
+      "loss": 0.0004,
+      "step": 16080
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 1.1359302325581395e-07,
+      "loss": 0.0008,
+      "step": 16081
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 1.135813953488372e-07,
+      "loss": 0.0005,
+      "step": 16082
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 1.1356976744186046e-07,
+      "loss": 0.0008,
+      "step": 16083
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 1.1355813953488371e-07,
+      "loss": 0.0008,
+      "step": 16084
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 1.1354651162790697e-07,
+      "loss": 0.0006,
+      "step": 16085
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 1.1353488372093022e-07,
+      "loss": 0.0005,
+      "step": 16086
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 1.135232558139535e-07,
+      "loss": 0.0009,
+      "step": 16087
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 1.1351162790697674e-07,
+      "loss": 0.0004,
+      "step": 16088
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 1.135e-07,
+      "loss": 0.0006,
+      "step": 16089
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 1.1348837209302325e-07,
+      "loss": 1.7234,
+      "step": 16090
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 1.1347674418604651e-07,
+      "loss": 0.0005,
+      "step": 16091
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 1.1346511627906976e-07,
+      "loss": 0.0005,
+      "step": 16092
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 1.1345348837209302e-07,
+      "loss": 0.0005,
+      "step": 16093
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 1.1344186046511627e-07,
+      "loss": 0.0005,
+      "step": 16094
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 1.1343023255813953e-07,
+      "loss": 0.0006,
+      "step": 16095
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 1.1341860465116278e-07,
+      "loss": 0.0045,
+      "step": 16096
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 1.1340697674418603e-07,
+      "loss": 0.0491,
+      "step": 16097
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 1.1339534883720929e-07,
+      "loss": 0.0005,
+      "step": 16098
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 1.1338372093023254e-07,
+      "loss": 0.0007,
+      "step": 16099
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 1.1337209302325582e-07,
+      "loss": 0.0007,
+      "step": 16100
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 1.1336046511627905e-07,
+      "loss": 0.0005,
+      "step": 16101
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 1.1334883720930233e-07,
+      "loss": 1.71,
+      "step": 16102
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 1.1333720930232558e-07,
+      "loss": 0.0006,
+      "step": 16103
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 1.1332558139534884e-07,
+      "loss": 0.0005,
+      "step": 16104
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 1.1331395348837209e-07,
+      "loss": 0.0005,
+      "step": 16105
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 1.1330232558139535e-07,
+      "loss": 0.0005,
+      "step": 16106
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 1.132906976744186e-07,
+      "loss": 0.0009,
+      "step": 16107
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 1.1327906976744186e-07,
+      "loss": 0.0005,
+      "step": 16108
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 1.132674418604651e-07,
+      "loss": 0.0007,
+      "step": 16109
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 1.1325581395348837e-07,
+      "loss": 0.7909,
+      "step": 16110
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 1.1324418604651162e-07,
+      "loss": 0.0006,
+      "step": 16111
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 1.1323255813953488e-07,
+      "loss": 0.0005,
+      "step": 16112
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 1.1322093023255813e-07,
+      "loss": 0.0009,
+      "step": 16113
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 1.132093023255814e-07,
+      "loss": 0.0006,
+      "step": 16114
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 1.1319767441860465e-07,
+      "loss": 0.0009,
+      "step": 16115
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 1.1318604651162791e-07,
+      "loss": 0.001,
+      "step": 16116
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 1.1317441860465116e-07,
+      "loss": 0.0006,
+      "step": 16117
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 1.1316279069767442e-07,
+      "loss": 0.0007,
+      "step": 16118
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 1.1315116279069767e-07,
+      "loss": 0.0006,
+      "step": 16119
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 1.1313953488372093e-07,
+      "loss": 0.0008,
+      "step": 16120
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 1.1312790697674418e-07,
+      "loss": 2.0135,
+      "step": 16121
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 1.1311627906976743e-07,
+      "loss": 0.0005,
+      "step": 16122
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 1.1310465116279069e-07,
+      "loss": 0.0009,
+      "step": 16123
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 1.1309302325581394e-07,
+      "loss": 0.0006,
+      "step": 16124
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 1.130813953488372e-07,
+      "loss": 0.0005,
+      "step": 16125
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 1.1306976744186045e-07,
+      "loss": 0.0007,
+      "step": 16126
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 1.1305813953488372e-07,
+      "loss": 0.0007,
+      "step": 16127
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 1.1304651162790696e-07,
+      "loss": 0.0007,
+      "step": 16128
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 1.1303488372093023e-07,
+      "loss": 0.0005,
+      "step": 16129
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 1.1302325581395348e-07,
+      "loss": 0.0005,
+      "step": 16130
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 1.1301162790697674e-07,
+      "loss": 0.0004,
+      "step": 16131
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 1.1299999999999999e-07,
+      "loss": 0.0006,
+      "step": 16132
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 1.1298837209302325e-07,
+      "loss": 3.0566,
+      "step": 16133
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 1.129767441860465e-07,
+      "loss": 0.0006,
+      "step": 16134
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 1.1296511627906976e-07,
+      "loss": 0.0005,
+      "step": 16135
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 1.1295348837209301e-07,
+      "loss": 0.0004,
+      "step": 16136
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 1.1294186046511627e-07,
+      "loss": 0.0004,
+      "step": 16137
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 1.1293023255813952e-07,
+      "loss": 0.0008,
+      "step": 16138
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 1.1291860465116279e-07,
+      "loss": 0.0005,
+      "step": 16139
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 1.1290697674418603e-07,
+      "loss": 0.0006,
+      "step": 16140
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 1.1289534883720931e-07,
+      "loss": 0.0008,
+      "step": 16141
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 1.1288372093023256e-07,
+      "loss": 0.0007,
+      "step": 16142
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 1.1287209302325582e-07,
+      "loss": 0.0004,
+      "step": 16143
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 1.1286046511627907e-07,
+      "loss": 0.0006,
+      "step": 16144
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 1.1284883720930233e-07,
+      "loss": 0.0005,
+      "step": 16145
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 1.1283720930232558e-07,
+      "loss": 0.0009,
+      "step": 16146
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 1.1282558139534884e-07,
+      "loss": 0.0303,
+      "step": 16147
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 1.1281395348837209e-07,
+      "loss": 0.0007,
+      "step": 16148
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 1.1280232558139534e-07,
+      "loss": 0.0009,
+      "step": 16149
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 1.127906976744186e-07,
+      "loss": 0.0005,
+      "step": 16150
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 1.1277906976744185e-07,
+      "loss": 0.0007,
+      "step": 16151
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 1.1276744186046511e-07,
+      "loss": 0.0004,
+      "step": 16152
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 1.1275581395348836e-07,
+      "loss": 0.0005,
+      "step": 16153
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 1.1274418604651162e-07,
+      "loss": 0.0006,
+      "step": 16154
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 1.1273255813953487e-07,
+      "loss": 0.0005,
+      "step": 16155
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 1.1272093023255814e-07,
+      "loss": 0.3565,
+      "step": 16156
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 1.1270930232558139e-07,
+      "loss": 0.0006,
+      "step": 16157
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 1.1269767441860465e-07,
+      "loss": 0.0008,
+      "step": 16158
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 1.126860465116279e-07,
+      "loss": 0.0005,
+      "step": 16159
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 1.1267441860465116e-07,
+      "loss": 0.0007,
+      "step": 16160
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 1.1266279069767441e-07,
+      "loss": 0.0008,
+      "step": 16161
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 1.1265116279069767e-07,
+      "loss": 0.0006,
+      "step": 16162
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 1.1263953488372092e-07,
+      "loss": 0.0004,
+      "step": 16163
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 1.1262790697674418e-07,
+      "loss": 0.0008,
+      "step": 16164
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 1.1261627906976743e-07,
+      "loss": 0.0006,
+      "step": 16165
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 1.1260465116279069e-07,
+      "loss": 0.0007,
+      "step": 16166
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 1.1259302325581394e-07,
+      "loss": 0.0005,
+      "step": 16167
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 1.1258139534883722e-07,
+      "loss": 0.0005,
+      "step": 16168
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 1.1256976744186046e-07,
+      "loss": 0.0008,
+      "step": 16169
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 1.1255813953488373e-07,
+      "loss": 0.0006,
+      "step": 16170
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 1.1254651162790697e-07,
+      "loss": 0.0005,
+      "step": 16171
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 1.1253488372093024e-07,
+      "loss": 0.0005,
+      "step": 16172
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 1.1252325581395348e-07,
+      "loss": 0.0008,
+      "step": 16173
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 1.1251162790697675e-07,
+      "loss": 0.0005,
+      "step": 16174
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 1.125e-07,
+      "loss": 0.0005,
+      "step": 16175
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 1.1248837209302324e-07,
+      "loss": 0.0006,
+      "step": 16176
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 1.124767441860465e-07,
+      "loss": 0.0005,
+      "step": 16177
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 1.1246511627906975e-07,
+      "loss": 0.0006,
+      "step": 16178
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 1.1245348837209302e-07,
+      "loss": 0.0006,
+      "step": 16179
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 1.1244186046511626e-07,
+      "loss": 0.0004,
+      "step": 16180
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 1.1243023255813953e-07,
+      "loss": 0.0006,
+      "step": 16181
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 1.1241860465116277e-07,
+      "loss": 0.0006,
+      "step": 16182
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 1.1240697674418605e-07,
+      "loss": 0.0007,
+      "step": 16183
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 1.123953488372093e-07,
+      "loss": 0.0005,
+      "step": 16184
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 1.1238372093023256e-07,
+      "loss": 0.0005,
+      "step": 16185
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 1.1237209302325581e-07,
+      "loss": 0.0007,
+      "step": 16186
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 1.1236046511627907e-07,
+      "loss": 0.0019,
+      "step": 16187
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 1.1234883720930232e-07,
+      "loss": 0.0006,
+      "step": 16188
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 1.1233720930232558e-07,
+      "loss": 0.0006,
+      "step": 16189
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 1.1232558139534883e-07,
+      "loss": 0.0004,
+      "step": 16190
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 1.1231395348837209e-07,
+      "loss": 0.0004,
+      "step": 16191
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 1.1230232558139534e-07,
+      "loss": 0.0006,
+      "step": 16192
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 1.122906976744186e-07,
+      "loss": 0.0004,
+      "step": 16193
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 1.1227906976744185e-07,
+      "loss": 0.0005,
+      "step": 16194
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 1.1226744186046512e-07,
+      "loss": 0.0006,
+      "step": 16195
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 1.1225581395348837e-07,
+      "loss": 0.0006,
+      "step": 16196
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 1.1224418604651163e-07,
+      "loss": 1.8822,
+      "step": 16197
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 1.1223255813953488e-07,
+      "loss": 0.0006,
+      "step": 16198
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 1.1222093023255814e-07,
+      "loss": 0.0006,
+      "step": 16199
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 1.1220930232558139e-07,
+      "loss": 0.0005,
+      "step": 16200
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 1.1219767441860465e-07,
+      "loss": 0.0008,
+      "step": 16201
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 1.121860465116279e-07,
+      "loss": 0.0007,
+      "step": 16202
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 1.1217441860465115e-07,
+      "loss": 0.0006,
+      "step": 16203
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 1.1216279069767441e-07,
+      "loss": 0.0006,
+      "step": 16204
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 1.1215116279069766e-07,
+      "loss": 0.0005,
+      "step": 16205
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 1.1213953488372092e-07,
+      "loss": 0.0005,
+      "step": 16206
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 1.1212790697674417e-07,
+      "loss": 0.0007,
+      "step": 16207
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 1.1211627906976743e-07,
+      "loss": 0.0004,
+      "step": 16208
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 1.1210465116279068e-07,
+      "loss": 0.0006,
+      "step": 16209
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 1.1209302325581396e-07,
+      "loss": 0.0019,
+      "step": 16210
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 1.120813953488372e-07,
+      "loss": 0.0048,
+      "step": 16211
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 1.1206976744186047e-07,
+      "loss": 0.0004,
+      "step": 16212
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 1.1205813953488371e-07,
+      "loss": 0.0005,
+      "step": 16213
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 1.1204651162790698e-07,
+      "loss": 0.0005,
+      "step": 16214
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 1.1203488372093023e-07,
+      "loss": 0.0005,
+      "step": 16215
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 1.1202325581395349e-07,
+      "loss": 0.001,
+      "step": 16216
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 1.1201162790697674e-07,
+      "loss": 0.0005,
+      "step": 16217
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 1.12e-07,
+      "loss": 0.0009,
+      "step": 16218
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 1.1198837209302325e-07,
+      "loss": 0.0005,
+      "step": 16219
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 1.1197674418604651e-07,
+      "loss": 0.0007,
+      "step": 16220
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 1.1196511627906976e-07,
+      "loss": 0.0006,
+      "step": 16221
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 1.1195348837209303e-07,
+      "loss": 0.0004,
+      "step": 16222
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 1.1194186046511627e-07,
+      "loss": 0.0003,
+      "step": 16223
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 1.1193023255813954e-07,
+      "loss": 0.0006,
+      "step": 16224
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 1.1191860465116279e-07,
+      "loss": 0.0006,
+      "step": 16225
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 1.1190697674418605e-07,
+      "loss": 0.0009,
+      "step": 16226
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 1.118953488372093e-07,
+      "loss": 0.0007,
+      "step": 16227
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 1.1188372093023256e-07,
+      "loss": 0.0008,
+      "step": 16228
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 1.1187209302325581e-07,
+      "loss": 0.0006,
+      "step": 16229
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 1.1186046511627906e-07,
+      "loss": 0.0004,
+      "step": 16230
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 1.1184883720930232e-07,
+      "loss": 0.0006,
+      "step": 16231
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 1.1183720930232557e-07,
+      "loss": 0.0006,
+      "step": 16232
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 1.1182558139534883e-07,
+      "loss": 0.0005,
+      "step": 16233
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 1.1181395348837208e-07,
+      "loss": 0.0004,
+      "step": 16234
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 1.1180232558139534e-07,
+      "loss": 0.0005,
+      "step": 16235
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 1.1179069767441859e-07,
+      "loss": 0.0005,
+      "step": 16236
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 1.1177906976744186e-07,
+      "loss": 0.0006,
+      "step": 16237
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 1.1176744186046511e-07,
+      "loss": 0.0007,
+      "step": 16238
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 1.1175581395348837e-07,
+      "loss": 0.0005,
+      "step": 16239
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 1.1174418604651162e-07,
+      "loss": 0.0006,
+      "step": 16240
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 1.1173255813953488e-07,
+      "loss": 0.0008,
+      "step": 16241
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 1.1172093023255813e-07,
+      "loss": 0.0005,
+      "step": 16242
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 1.117093023255814e-07,
+      "loss": 0.0227,
+      "step": 16243
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 1.1169767441860464e-07,
+      "loss": 0.0007,
+      "step": 16244
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 1.116860465116279e-07,
+      "loss": 0.0004,
+      "step": 16245
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 1.1167441860465115e-07,
+      "loss": 0.0006,
+      "step": 16246
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 1.1166279069767441e-07,
+      "loss": 0.0006,
+      "step": 16247
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 1.1165116279069766e-07,
+      "loss": 0.0006,
+      "step": 16248
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 1.1163953488372094e-07,
+      "loss": 0.0005,
+      "step": 16249
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 1.1162790697674417e-07,
+      "loss": 0.0006,
+      "step": 16250
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 1.1161627906976745e-07,
+      "loss": 0.0006,
+      "step": 16251
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 1.116046511627907e-07,
+      "loss": 0.0007,
+      "step": 16252
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 1.1159302325581396e-07,
+      "loss": 0.0006,
+      "step": 16253
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 1.1158139534883721e-07,
+      "loss": 0.001,
+      "step": 16254
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 1.1156976744186046e-07,
+      "loss": 0.0005,
+      "step": 16255
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 1.1155813953488372e-07,
+      "loss": 0.0005,
+      "step": 16256
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 1.1154651162790697e-07,
+      "loss": 0.0008,
+      "step": 16257
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 1.1153488372093023e-07,
+      "loss": 0.0005,
+      "step": 16258
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 1.1152325581395348e-07,
+      "loss": 0.0006,
+      "step": 16259
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 1.1151162790697674e-07,
+      "loss": 0.0004,
+      "step": 16260
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 1.1149999999999999e-07,
+      "loss": 0.0006,
+      "step": 16261
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 1.1148837209302325e-07,
+      "loss": 0.0006,
+      "step": 16262
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 1.114767441860465e-07,
+      "loss": 0.0011,
+      "step": 16263
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 1.1146511627906977e-07,
+      "loss": 0.0006,
+      "step": 16264
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 1.1145348837209302e-07,
+      "loss": 2.4706,
+      "step": 16265
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 1.1144186046511628e-07,
+      "loss": 0.0004,
+      "step": 16266
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 1.1143023255813953e-07,
+      "loss": 0.0011,
+      "step": 16267
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 1.1141860465116279e-07,
+      "loss": 0.0011,
+      "step": 16268
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 1.1140697674418604e-07,
+      "loss": 0.0004,
+      "step": 16269
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 1.113953488372093e-07,
+      "loss": 1.4912,
+      "step": 16270
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 1.1138372093023255e-07,
+      "loss": 0.0006,
+      "step": 16271
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 1.1137209302325581e-07,
+      "loss": 0.0007,
+      "step": 16272
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 1.1136046511627906e-07,
+      "loss": 0.0006,
+      "step": 16273
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 1.1134883720930232e-07,
+      "loss": 0.0007,
+      "step": 16274
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 1.1133720930232557e-07,
+      "loss": 0.0007,
+      "step": 16275
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 1.1132558139534883e-07,
+      "loss": 0.0052,
+      "step": 16276
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 1.1131395348837208e-07,
+      "loss": 0.0007,
+      "step": 16277
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 1.1130232558139536e-07,
+      "loss": 0.0091,
+      "step": 16278
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 1.112906976744186e-07,
+      "loss": 0.0005,
+      "step": 16279
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 1.1127906976744187e-07,
+      "loss": 0.0008,
+      "step": 16280
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 1.1126744186046511e-07,
+      "loss": 0.0005,
+      "step": 16281
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 1.1125581395348836e-07,
+      "loss": 0.0007,
+      "step": 16282
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 1.1124418604651162e-07,
+      "loss": 0.0005,
+      "step": 16283
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 1.1123255813953487e-07,
+      "loss": 0.0006,
+      "step": 16284
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 1.1122093023255813e-07,
+      "loss": 0.003,
+      "step": 16285
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 1.1120930232558138e-07,
+      "loss": 0.0006,
+      "step": 16286
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 1.1119767441860464e-07,
+      "loss": 0.0005,
+      "step": 16287
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 1.1118604651162789e-07,
+      "loss": 0.0008,
+      "step": 16288
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 1.1117441860465116e-07,
+      "loss": 0.0006,
+      "step": 16289
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 1.111627906976744e-07,
+      "loss": 0.0005,
+      "step": 16290
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 1.1115116279069768e-07,
+      "loss": 0.0005,
+      "step": 16291
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 1.1113953488372091e-07,
+      "loss": 0.0007,
+      "step": 16292
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 1.1112790697674419e-07,
+      "loss": 0.0281,
+      "step": 16293
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 1.1111627906976744e-07,
+      "loss": 0.0008,
+      "step": 16294
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 1.111046511627907e-07,
+      "loss": 0.0007,
+      "step": 16295
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 1.1109302325581395e-07,
+      "loss": 0.0007,
+      "step": 16296
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 1.1108139534883721e-07,
+      "loss": 0.0069,
+      "step": 16297
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 1.1106976744186046e-07,
+      "loss": 0.0005,
+      "step": 16298
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 1.1105813953488372e-07,
+      "loss": 0.0026,
+      "step": 16299
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 1.1104651162790697e-07,
+      "loss": 0.6846,
+      "step": 16300
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 1.1103488372093023e-07,
+      "loss": 0.0005,
+      "step": 16301
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 1.1102325581395348e-07,
+      "loss": 0.0006,
+      "step": 16302
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 1.1101162790697674e-07,
+      "loss": 0.0008,
+      "step": 16303
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 1.1099999999999999e-07,
+      "loss": 0.0016,
+      "step": 16304
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 1.1098837209302326e-07,
+      "loss": 0.0005,
+      "step": 16305
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 1.1097674418604651e-07,
+      "loss": 0.0008,
+      "step": 16306
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 1.1096511627906977e-07,
+      "loss": 0.0015,
+      "step": 16307
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 1.1095348837209302e-07,
+      "loss": 0.0005,
+      "step": 16308
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 1.1094186046511627e-07,
+      "loss": 0.0038,
+      "step": 16309
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 1.1093023255813953e-07,
+      "loss": 0.0005,
+      "step": 16310
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 1.1091860465116278e-07,
+      "loss": 0.0008,
+      "step": 16311
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 1.1090697674418604e-07,
+      "loss": 0.0009,
+      "step": 16312
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 1.1089534883720929e-07,
+      "loss": 0.0005,
+      "step": 16313
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 1.1088372093023255e-07,
+      "loss": 0.0005,
+      "step": 16314
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 1.108720930232558e-07,
+      "loss": 0.0014,
+      "step": 16315
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 1.1086046511627906e-07,
+      "loss": 0.0005,
+      "step": 16316
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 1.1084883720930231e-07,
+      "loss": 0.0008,
+      "step": 16317
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 1.1083720930232559e-07,
+      "loss": 0.006,
+      "step": 16318
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 1.1082558139534882e-07,
+      "loss": 0.0003,
+      "step": 16319
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 1.108139534883721e-07,
+      "loss": 0.0013,
+      "step": 16320
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 1.1080232558139534e-07,
+      "loss": 0.0012,
+      "step": 16321
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 1.107906976744186e-07,
+      "loss": 0.0004,
+      "step": 16322
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 1.1077906976744185e-07,
+      "loss": 0.1209,
+      "step": 16323
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 1.1076744186046512e-07,
+      "loss": 0.0005,
+      "step": 16324
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 1.1075581395348836e-07,
+      "loss": 0.0005,
+      "step": 16325
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 1.1074418604651163e-07,
+      "loss": 0.0005,
+      "step": 16326
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 1.1073255813953488e-07,
+      "loss": 0.0005,
+      "step": 16327
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 1.1072093023255814e-07,
+      "loss": 0.0006,
+      "step": 16328
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 1.1070930232558139e-07,
+      "loss": 0.0004,
+      "step": 16329
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 1.1069767441860465e-07,
+      "loss": 0.0004,
+      "step": 16330
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 1.106860465116279e-07,
+      "loss": 0.0007,
+      "step": 16331
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 1.1067441860465117e-07,
+      "loss": 0.0005,
+      "step": 16332
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 1.1066279069767442e-07,
+      "loss": 0.0006,
+      "step": 16333
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 1.1065116279069768e-07,
+      "loss": 0.0004,
+      "step": 16334
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 1.1063953488372093e-07,
+      "loss": 0.0005,
+      "step": 16335
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 1.1062790697674418e-07,
+      "loss": 0.0005,
+      "step": 16336
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 1.1061627906976744e-07,
+      "loss": 0.0006,
+      "step": 16337
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 1.1060465116279069e-07,
+      "loss": 0.0005,
+      "step": 16338
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 1.1059302325581395e-07,
+      "loss": 0.0007,
+      "step": 16339
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 1.105813953488372e-07,
+      "loss": 0.1294,
+      "step": 16340
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 1.1056976744186046e-07,
+      "loss": 0.0005,
+      "step": 16341
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 1.1055813953488371e-07,
+      "loss": 0.0005,
+      "step": 16342
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 1.1054651162790697e-07,
+      "loss": 0.0006,
+      "step": 16343
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 1.1053488372093022e-07,
+      "loss": 0.0006,
+      "step": 16344
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 1.1052325581395348e-07,
+      "loss": 0.0006,
+      "step": 16345
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 1.1051162790697673e-07,
+      "loss": 0.5121,
+      "step": 16346
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 1.105e-07,
+      "loss": 0.0006,
+      "step": 16347
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 1.1048837209302325e-07,
+      "loss": 0.0009,
+      "step": 16348
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 1.1047674418604651e-07,
+      "loss": 0.0009,
+      "step": 16349
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 1.1046511627906976e-07,
+      "loss": 0.0006,
+      "step": 16350
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 1.1045348837209302e-07,
+      "loss": 0.0006,
+      "step": 16351
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 1.1044186046511627e-07,
+      "loss": 0.0006,
+      "step": 16352
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 1.1043023255813953e-07,
+      "loss": 0.001,
+      "step": 16353
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 1.1041860465116278e-07,
+      "loss": 0.0005,
+      "step": 16354
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 1.1040697674418604e-07,
+      "loss": 0.0009,
+      "step": 16355
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 1.1039534883720929e-07,
+      "loss": 0.0007,
+      "step": 16356
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 1.1038372093023255e-07,
+      "loss": 0.0013,
+      "step": 16357
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 1.103720930232558e-07,
+      "loss": 0.0008,
+      "step": 16358
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 1.1036046511627908e-07,
+      "loss": 0.0005,
+      "step": 16359
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 1.1034883720930233e-07,
+      "loss": 0.0005,
+      "step": 16360
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 1.1033720930232559e-07,
+      "loss": 0.0008,
+      "step": 16361
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 1.1032558139534884e-07,
+      "loss": 0.0004,
+      "step": 16362
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 1.1031395348837208e-07,
+      "loss": 0.0229,
+      "step": 16363
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 1.1030232558139535e-07,
+      "loss": 0.002,
+      "step": 16364
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 1.102906976744186e-07,
+      "loss": 0.0005,
+      "step": 16365
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 1.1027906976744186e-07,
+      "loss": 0.0005,
+      "step": 16366
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 1.102674418604651e-07,
+      "loss": 0.0006,
+      "step": 16367
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 1.1025581395348837e-07,
+      "loss": 0.0009,
+      "step": 16368
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 1.1024418604651162e-07,
+      "loss": 0.0005,
+      "step": 16369
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 1.1023255813953488e-07,
+      "loss": 0.0006,
+      "step": 16370
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 1.1022093023255813e-07,
+      "loss": 0.0006,
+      "step": 16371
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 1.1020930232558139e-07,
+      "loss": 0.0005,
+      "step": 16372
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 1.1019767441860464e-07,
+      "loss": 0.0064,
+      "step": 16373
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 1.1018604651162791e-07,
+      "loss": 0.0005,
+      "step": 16374
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 1.1017441860465116e-07,
+      "loss": 0.0006,
+      "step": 16375
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.1016279069767442e-07,
+      "loss": 0.0006,
+      "step": 16376
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.1015116279069767e-07,
+      "loss": 0.0007,
+      "step": 16377
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.1013953488372093e-07,
+      "loss": 0.0005,
+      "step": 16378
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.1012790697674418e-07,
+      "loss": 1.5699,
+      "step": 16379
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.1011627906976744e-07,
+      "loss": 0.0007,
+      "step": 16380
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 1.1010465116279069e-07,
+      "loss": 0.0012,
+      "step": 16381
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 1.1009302325581395e-07,
+      "loss": 1.2465,
+      "step": 16382
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 1.100813953488372e-07,
+      "loss": 0.0014,
+      "step": 16383
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 1.1006976744186046e-07,
+      "loss": 0.0008,
+      "step": 16384
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 1.1005813953488371e-07,
+      "loss": 0.0005,
+      "step": 16385
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 1.1004651162790699e-07,
+      "loss": 0.0005,
+      "step": 16386
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 1.1003488372093023e-07,
+      "loss": 0.0004,
+      "step": 16387
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 1.100232558139535e-07,
+      "loss": 0.0007,
+      "step": 16388
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 1.1001162790697674e-07,
+      "loss": 0.0007,
+      "step": 16389
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 1.0999999999999999e-07,
+      "loss": 0.0005,
+      "step": 16390
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 1.0998837209302325e-07,
+      "loss": 0.0289,
+      "step": 16391
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 1.099767441860465e-07,
+      "loss": 0.0005,
+      "step": 16392
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 1.0996511627906976e-07,
+      "loss": 0.0006,
+      "step": 16393
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 1.0995348837209301e-07,
+      "loss": 0.0011,
+      "step": 16394
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 1.0994186046511627e-07,
+      "loss": 0.0004,
+      "step": 16395
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 1.0993023255813952e-07,
+      "loss": 0.0005,
+      "step": 16396
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 1.0991860465116278e-07,
+      "loss": 0.0005,
+      "step": 16397
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 1.0990697674418603e-07,
+      "loss": 0.0005,
+      "step": 16398
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 1.098953488372093e-07,
+      "loss": 0.0007,
+      "step": 16399
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 1.0988372093023254e-07,
+      "loss": 0.0007,
+      "step": 16400
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 1.0987209302325582e-07,
+      "loss": 0.0005,
+      "step": 16401
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 1.0986046511627907e-07,
+      "loss": 0.0008,
+      "step": 16402
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 1.0984883720930233e-07,
+      "loss": 0.0006,
+      "step": 16403
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 1.0983720930232558e-07,
+      "loss": 0.0006,
+      "step": 16404
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 1.0982558139534884e-07,
+      "loss": 0.0011,
+      "step": 16405
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 1.0981395348837209e-07,
+      "loss": 0.0005,
+      "step": 16406
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.0980232558139535e-07,
+      "loss": 0.0006,
+      "step": 16407
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.097906976744186e-07,
+      "loss": 0.0005,
+      "step": 16408
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.0977906976744186e-07,
+      "loss": 0.0005,
+      "step": 16409
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.0976744186046511e-07,
+      "loss": 0.0359,
+      "step": 16410
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.0975581395348837e-07,
+      "loss": 0.0007,
+      "step": 16411
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 1.0974418604651162e-07,
+      "loss": 0.0005,
+      "step": 16412
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 1.0973255813953489e-07,
+      "loss": 0.0006,
+      "step": 16413
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 1.0972093023255813e-07,
+      "loss": 0.0006,
+      "step": 16414
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 1.0970930232558138e-07,
+      "loss": 0.0006,
+      "step": 16415
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 1.0969767441860465e-07,
+      "loss": 0.0004,
+      "step": 16416
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 1.096860465116279e-07,
+      "loss": 0.0052,
+      "step": 16417
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 1.0967441860465116e-07,
+      "loss": 0.0009,
+      "step": 16418
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 1.0966279069767441e-07,
+      "loss": 0.0006,
+      "step": 16419
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 1.0965116279069767e-07,
+      "loss": 0.0005,
+      "step": 16420
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 1.0963953488372092e-07,
+      "loss": 0.0004,
+      "step": 16421
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 1.0962790697674418e-07,
+      "loss": 0.0009,
+      "step": 16422
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 1.0961627906976743e-07,
+      "loss": 0.0005,
+      "step": 16423
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 1.0960465116279069e-07,
+      "loss": 0.0006,
+      "step": 16424
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 1.0959302325581394e-07,
+      "loss": 0.0009,
+      "step": 16425
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 1.095813953488372e-07,
+      "loss": 0.0006,
+      "step": 16426
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 1.0956976744186045e-07,
+      "loss": 0.0004,
+      "step": 16427
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 1.0955813953488373e-07,
+      "loss": 0.0008,
+      "step": 16428
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 1.0954651162790697e-07,
+      "loss": 0.0006,
+      "step": 16429
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 1.0953488372093024e-07,
+      "loss": 0.0005,
+      "step": 16430
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 1.0952325581395348e-07,
+      "loss": 0.0006,
+      "step": 16431
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 1.0951162790697675e-07,
+      "loss": 0.0005,
+      "step": 16432
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 1.095e-07,
+      "loss": 0.0005,
+      "step": 16433
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 1.0948837209302326e-07,
+      "loss": 0.0007,
+      "step": 16434
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 1.094767441860465e-07,
+      "loss": 2.125,
+      "step": 16435
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 1.0946511627906977e-07,
+      "loss": 0.0006,
+      "step": 16436
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 1.0945348837209301e-07,
+      "loss": 0.0006,
+      "step": 16437
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 1.0944186046511628e-07,
+      "loss": 0.0006,
+      "step": 16438
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 1.0943023255813953e-07,
+      "loss": 0.0008,
+      "step": 16439
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 1.094186046511628e-07,
+      "loss": 0.0006,
+      "step": 16440
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 1.0940697674418604e-07,
+      "loss": 0.0005,
+      "step": 16441
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 1.0939534883720928e-07,
+      "loss": 1.8955,
+      "step": 16442
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 1.0938372093023256e-07,
+      "loss": 0.4138,
+      "step": 16443
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 1.0937209302325581e-07,
+      "loss": 0.0008,
+      "step": 16444
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 1.0936046511627907e-07,
+      "loss": 0.0006,
+      "step": 16445
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 1.0934883720930232e-07,
+      "loss": 0.0005,
+      "step": 16446
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 1.0933720930232558e-07,
+      "loss": 0.0007,
+      "step": 16447
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 1.0932558139534883e-07,
+      "loss": 0.0005,
+      "step": 16448
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 1.0931395348837209e-07,
+      "loss": 0.0017,
+      "step": 16449
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 1.0930232558139534e-07,
+      "loss": 0.0008,
+      "step": 16450
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 1.092906976744186e-07,
+      "loss": 0.0004,
+      "step": 16451
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 1.0927906976744185e-07,
+      "loss": 0.0004,
+      "step": 16452
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 1.0926744186046511e-07,
+      "loss": 0.0007,
+      "step": 16453
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 1.0925581395348836e-07,
+      "loss": 0.0006,
+      "step": 16454
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 1.0924418604651163e-07,
+      "loss": 0.0005,
+      "step": 16455
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 1.0923255813953488e-07,
+      "loss": 0.0006,
+      "step": 16456
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 1.0922093023255814e-07,
+      "loss": 0.0005,
+      "step": 16457
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 1.0920930232558139e-07,
+      "loss": 0.0005,
+      "step": 16458
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 1.0919767441860465e-07,
+      "loss": 0.0006,
+      "step": 16459
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 1.091860465116279e-07,
+      "loss": 0.4487,
+      "step": 16460
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 1.0917441860465116e-07,
+      "loss": 1.3839,
+      "step": 16461
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 1.0916279069767441e-07,
+      "loss": 0.0005,
+      "step": 16462
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 1.0915116279069767e-07,
+      "loss": 0.0005,
+      "step": 16463
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 1.0913953488372092e-07,
+      "loss": 0.0007,
+      "step": 16464
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 1.0912790697674418e-07,
+      "loss": 0.0004,
+      "step": 16465
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 1.0911627906976743e-07,
+      "loss": 0.0004,
+      "step": 16466
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 1.091046511627907e-07,
+      "loss": 0.0007,
+      "step": 16467
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 1.0909302325581394e-07,
+      "loss": 0.0004,
+      "step": 16468
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 1.0908139534883719e-07,
+      "loss": 0.0007,
+      "step": 16469
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 1.0906976744186047e-07,
+      "loss": 0.0006,
+      "step": 16470
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 1.0905813953488371e-07,
+      "loss": 0.0005,
+      "step": 16471
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 1.0904651162790698e-07,
+      "loss": 0.0005,
+      "step": 16472
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 1.0903488372093022e-07,
+      "loss": 0.001,
+      "step": 16473
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 1.0902325581395349e-07,
+      "loss": 0.0004,
+      "step": 16474
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 1.0901162790697673e-07,
+      "loss": 0.0051,
+      "step": 16475
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 1.09e-07,
+      "loss": 0.0006,
+      "step": 16476
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 1.0898837209302325e-07,
+      "loss": 0.0005,
+      "step": 16477
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 1.0897674418604651e-07,
+      "loss": 0.0003,
+      "step": 16478
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 1.0896511627906976e-07,
+      "loss": 0.0008,
+      "step": 16479
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 1.0895348837209302e-07,
+      "loss": 0.0041,
+      "step": 16480
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 1.0894186046511627e-07,
+      "loss": 0.0006,
+      "step": 16481
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 1.0893023255813954e-07,
+      "loss": 0.0005,
+      "step": 16482
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 1.0891860465116278e-07,
+      "loss": 0.0004,
+      "step": 16483
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 1.0890697674418605e-07,
+      "loss": 0.0005,
+      "step": 16484
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 1.088953488372093e-07,
+      "loss": 0.0005,
+      "step": 16485
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 1.0888372093023256e-07,
+      "loss": 0.0006,
+      "step": 16486
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 1.0887209302325581e-07,
+      "loss": 0.0005,
+      "step": 16487
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 1.0886046511627907e-07,
+      "loss": 0.0008,
+      "step": 16488
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 1.0884883720930232e-07,
+      "loss": 0.0006,
+      "step": 16489
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 1.0883720930232558e-07,
+      "loss": 0.0007,
+      "step": 16490
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 1.0882558139534883e-07,
+      "loss": 0.0005,
+      "step": 16491
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 1.0881395348837209e-07,
+      "loss": 0.0004,
+      "step": 16492
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 1.0880232558139534e-07,
+      "loss": 0.0006,
+      "step": 16493
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 1.087906976744186e-07,
+      "loss": 0.0006,
+      "step": 16494
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 1.0877906976744185e-07,
+      "loss": 0.0004,
+      "step": 16495
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 1.087674418604651e-07,
+      "loss": 0.0008,
+      "step": 16496
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 1.0875581395348837e-07,
+      "loss": 0.0005,
+      "step": 16497
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 1.0874418604651162e-07,
+      "loss": 0.0047,
+      "step": 16498
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 1.0873255813953488e-07,
+      "loss": 0.0006,
+      "step": 16499
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 1.0872093023255813e-07,
+      "loss": 0.0005,
+      "step": 16500
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 1.087093023255814e-07,
+      "loss": 0.0004,
+      "step": 16501
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 1.0869767441860464e-07,
+      "loss": 0.0006,
+      "step": 16502
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 1.086860465116279e-07,
+      "loss": 0.0006,
+      "step": 16503
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 1.0867441860465115e-07,
+      "loss": 0.0005,
+      "step": 16504
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 1.0866279069767441e-07,
+      "loss": 0.0006,
+      "step": 16505
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 1.0865116279069766e-07,
+      "loss": 0.0005,
+      "step": 16506
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 1.0863953488372092e-07,
+      "loss": 0.0005,
+      "step": 16507
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 1.0862790697674417e-07,
+      "loss": 0.0006,
+      "step": 16508
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 1.0861627906976745e-07,
+      "loss": 0.0005,
+      "step": 16509
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 1.0860465116279068e-07,
+      "loss": 0.0006,
+      "step": 16510
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 1.0859302325581396e-07,
+      "loss": 0.0006,
+      "step": 16511
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 1.085813953488372e-07,
+      "loss": 0.0004,
+      "step": 16512
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 1.0856976744186047e-07,
+      "loss": 0.0006,
+      "step": 16513
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 1.0855813953488372e-07,
+      "loss": 0.0005,
+      "step": 16514
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 1.0854651162790698e-07,
+      "loss": 0.0005,
+      "step": 16515
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 1.0853488372093023e-07,
+      "loss": 0.0014,
+      "step": 16516
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 1.0852325581395349e-07,
+      "loss": 0.0007,
+      "step": 16517
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 1.0851162790697674e-07,
+      "loss": 0.0012,
+      "step": 16518
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 1.085e-07,
+      "loss": 0.0007,
+      "step": 16519
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 1.0848837209302325e-07,
+      "loss": 0.0009,
+      "step": 16520
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 1.0847674418604651e-07,
+      "loss": 0.0005,
+      "step": 16521
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 1.0846511627906976e-07,
+      "loss": 0.0008,
+      "step": 16522
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 1.08453488372093e-07,
+      "loss": 0.011,
+      "step": 16523
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 1.0844186046511628e-07,
+      "loss": 0.0005,
+      "step": 16524
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 1.0843023255813953e-07,
+      "loss": 0.0005,
+      "step": 16525
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 1.0841860465116279e-07,
+      "loss": 0.0007,
+      "step": 16526
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 1.0840697674418604e-07,
+      "loss": 0.0004,
+      "step": 16527
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 1.083953488372093e-07,
+      "loss": 0.0005,
+      "step": 16528
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 1.0838372093023255e-07,
+      "loss": 0.0004,
+      "step": 16529
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 1.0837209302325581e-07,
+      "loss": 0.0007,
+      "step": 16530
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 1.0836046511627906e-07,
+      "loss": 0.0007,
+      "step": 16531
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 1.0834883720930232e-07,
+      "loss": 0.0007,
+      "step": 16532
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 1.0833720930232557e-07,
+      "loss": 0.0006,
+      "step": 16533
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 1.0832558139534883e-07,
+      "loss": 0.0005,
+      "step": 16534
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 1.0831395348837208e-07,
+      "loss": 0.0004,
+      "step": 16535
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 1.0830232558139534e-07,
+      "loss": 0.0005,
+      "step": 16536
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 1.0829069767441859e-07,
+      "loss": 0.0007,
+      "step": 16537
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 1.0827906976744187e-07,
+      "loss": 0.0017,
+      "step": 16538
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 1.0826744186046511e-07,
+      "loss": 0.0022,
+      "step": 16539
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 1.0825581395348838e-07,
+      "loss": 0.0009,
+      "step": 16540
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 1.0824418604651162e-07,
+      "loss": 0.0004,
+      "step": 16541
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 1.0823255813953489e-07,
+      "loss": 0.0012,
+      "step": 16542
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 1.0822093023255813e-07,
+      "loss": 0.0005,
+      "step": 16543
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 1.082093023255814e-07,
+      "loss": 0.0005,
+      "step": 16544
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.15137770771980286,
+      "eval_runtime": 1.5632,
+      "eval_samples_per_second": 661.462,
+      "eval_steps_per_second": 10.875,
+      "step": 16544
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 1.0819767441860464e-07,
+      "loss": 0.0005,
+      "step": 16545
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 1.081860465116279e-07,
+      "loss": 0.0005,
+      "step": 16546
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 1.0817441860465115e-07,
+      "loss": 0.0006,
+      "step": 16547
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 1.0816279069767442e-07,
+      "loss": 0.0007,
+      "step": 16548
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 1.0815116279069766e-07,
+      "loss": 0.2408,
+      "step": 16549
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 1.0813953488372091e-07,
+      "loss": 1.5145,
+      "step": 16550
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 1.0812790697674419e-07,
+      "loss": 0.0016,
+      "step": 16551
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 1.0811627906976742e-07,
+      "loss": 0.0004,
+      "step": 16552
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 1.081046511627907e-07,
+      "loss": 0.0005,
+      "step": 16553
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 1.0809302325581395e-07,
+      "loss": 0.0003,
+      "step": 16554
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 1.0808139534883721e-07,
+      "loss": 0.1727,
+      "step": 16555
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 1.0806976744186046e-07,
+      "loss": 0.1181,
+      "step": 16556
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 1.0805813953488372e-07,
+      "loss": 0.0004,
+      "step": 16557
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 1.0804651162790697e-07,
+      "loss": 0.0004,
+      "step": 16558
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 1.0803488372093023e-07,
+      "loss": 0.0006,
+      "step": 16559
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 1.0802325581395348e-07,
+      "loss": 0.0006,
+      "step": 16560
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 1.0801162790697674e-07,
+      "loss": 0.0005,
+      "step": 16561
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 1.0799999999999999e-07,
+      "loss": 0.0004,
+      "step": 16562
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 1.0798837209302325e-07,
+      "loss": 0.0005,
+      "step": 16563
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 1.079767441860465e-07,
+      "loss": 0.0005,
+      "step": 16564
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 1.0796511627906977e-07,
+      "loss": 0.001,
+      "step": 16565
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 1.0795348837209302e-07,
+      "loss": 0.0023,
+      "step": 16566
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 1.0794186046511628e-07,
+      "loss": 0.0005,
+      "step": 16567
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 1.0793023255813953e-07,
+      "loss": 0.0007,
+      "step": 16568
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 1.0791860465116279e-07,
+      "loss": 0.0006,
+      "step": 16569
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 1.0790697674418604e-07,
+      "loss": 0.0007,
+      "step": 16570
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 1.078953488372093e-07,
+      "loss": 2.7469,
+      "step": 16571
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 1.0788372093023255e-07,
+      "loss": 0.0007,
+      "step": 16572
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 1.0787209302325581e-07,
+      "loss": 0.0035,
+      "step": 16573
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 1.0786046511627906e-07,
+      "loss": 0.0006,
+      "step": 16574
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 1.0784883720930231e-07,
+      "loss": 0.0006,
+      "step": 16575
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 1.0783720930232557e-07,
+      "loss": 1.3887,
+      "step": 16576
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 1.0782558139534882e-07,
+      "loss": 0.0004,
+      "step": 16577
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 1.078139534883721e-07,
+      "loss": 0.0006,
+      "step": 16578
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 1.0780232558139533e-07,
+      "loss": 0.0008,
+      "step": 16579
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 1.077906976744186e-07,
+      "loss": 0.0005,
+      "step": 16580
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 1.0777906976744185e-07,
+      "loss": 0.0875,
+      "step": 16581
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 1.0776744186046512e-07,
+      "loss": 0.0004,
+      "step": 16582
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 1.0775581395348836e-07,
+      "loss": 0.0007,
+      "step": 16583
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 1.0774418604651163e-07,
+      "loss": 0.0005,
+      "step": 16584
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 1.0773255813953487e-07,
+      "loss": 0.0006,
+      "step": 16585
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 1.0772093023255814e-07,
+      "loss": 0.0005,
+      "step": 16586
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 1.0770930232558138e-07,
+      "loss": 0.0005,
+      "step": 16587
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 1.0769767441860465e-07,
+      "loss": 0.0004,
+      "step": 16588
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 1.076860465116279e-07,
+      "loss": 0.0004,
+      "step": 16589
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 1.0767441860465116e-07,
+      "loss": 0.0005,
+      "step": 16590
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 1.076627906976744e-07,
+      "loss": 0.0004,
+      "step": 16591
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 1.0765116279069768e-07,
+      "loss": 0.0006,
+      "step": 16592
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 1.0763953488372093e-07,
+      "loss": 0.0004,
+      "step": 16593
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 1.0762790697674419e-07,
+      "loss": 0.0007,
+      "step": 16594
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 1.0761627906976744e-07,
+      "loss": 0.0006,
+      "step": 16595
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 1.076046511627907e-07,
+      "loss": 0.0006,
+      "step": 16596
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 1.0759302325581395e-07,
+      "loss": 0.0005,
+      "step": 16597
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 1.0758139534883721e-07,
+      "loss": 0.0006,
+      "step": 16598
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 1.0756976744186046e-07,
+      "loss": 0.0005,
+      "step": 16599
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 1.0755813953488372e-07,
+      "loss": 0.0008,
+      "step": 16600
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 1.0754651162790697e-07,
+      "loss": 0.0006,
+      "step": 16601
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 1.0753488372093022e-07,
+      "loss": 0.0093,
+      "step": 16602
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 1.0752325581395348e-07,
+      "loss": 0.0008,
+      "step": 16603
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 1.0751162790697673e-07,
+      "loss": 0.0004,
+      "step": 16604
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 1.0749999999999999e-07,
+      "loss": 0.0008,
+      "step": 16605
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 1.0748837209302324e-07,
+      "loss": 0.0006,
+      "step": 16606
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 1.0747674418604651e-07,
+      "loss": 0.0005,
+      "step": 16607
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 1.0746511627906976e-07,
+      "loss": 0.0007,
+      "step": 16608
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 1.0745348837209302e-07,
+      "loss": 2.4845,
+      "step": 16609
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 1.0744186046511627e-07,
+      "loss": 0.0007,
+      "step": 16610
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 1.0743023255813953e-07,
+      "loss": 0.0004,
+      "step": 16611
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 1.0741860465116278e-07,
+      "loss": 0.0006,
+      "step": 16612
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 1.0740697674418604e-07,
+      "loss": 0.0007,
+      "step": 16613
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 1.0739534883720929e-07,
+      "loss": 0.0005,
+      "step": 16614
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 1.0738372093023255e-07,
+      "loss": 0.0005,
+      "step": 16615
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 1.073720930232558e-07,
+      "loss": 0.0004,
+      "step": 16616
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 1.0736046511627906e-07,
+      "loss": 0.0005,
+      "step": 16617
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 1.0734883720930231e-07,
+      "loss": 0.0006,
+      "step": 16618
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 1.0733720930232559e-07,
+      "loss": 0.0006,
+      "step": 16619
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 1.0732558139534884e-07,
+      "loss": 0.0006,
+      "step": 16620
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 1.073139534883721e-07,
+      "loss": 0.0006,
+      "step": 16621
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 1.0730232558139535e-07,
+      "loss": 0.0004,
+      "step": 16622
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 1.0729069767441861e-07,
+      "loss": 0.0007,
+      "step": 16623
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 1.0727906976744186e-07,
+      "loss": 0.0005,
+      "step": 16624
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 1.0726744186046512e-07,
+      "loss": 0.0005,
+      "step": 16625
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 1.0725581395348837e-07,
+      "loss": 0.0008,
+      "step": 16626
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 1.0724418604651163e-07,
+      "loss": 0.0007,
+      "step": 16627
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 1.0723255813953488e-07,
+      "loss": 0.0005,
+      "step": 16628
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 1.0722093023255813e-07,
+      "loss": 0.0005,
+      "step": 16629
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 1.0720930232558139e-07,
+      "loss": 0.0007,
+      "step": 16630
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 1.0719767441860464e-07,
+      "loss": 0.0005,
+      "step": 16631
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 1.071860465116279e-07,
+      "loss": 0.0006,
+      "step": 16632
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 1.0717441860465115e-07,
+      "loss": 0.0004,
+      "step": 16633
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 1.0716279069767442e-07,
+      "loss": 0.0005,
+      "step": 16634
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 1.0715116279069767e-07,
+      "loss": 0.0004,
+      "step": 16635
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 1.0713953488372093e-07,
+      "loss": 0.0105,
+      "step": 16636
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 1.0712790697674418e-07,
+      "loss": 0.0006,
+      "step": 16637
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 1.0711627906976744e-07,
+      "loss": 0.0011,
+      "step": 16638
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 1.0710465116279069e-07,
+      "loss": 0.0048,
+      "step": 16639
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 1.0709302325581395e-07,
+      "loss": 0.0007,
+      "step": 16640
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 1.070813953488372e-07,
+      "loss": 0.0004,
+      "step": 16641
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 1.0706976744186046e-07,
+      "loss": 0.0005,
+      "step": 16642
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 1.0705813953488371e-07,
+      "loss": 0.0004,
+      "step": 16643
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 1.0704651162790697e-07,
+      "loss": 0.0006,
+      "step": 16644
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 1.0703488372093022e-07,
+      "loss": 0.0005,
+      "step": 16645
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 1.070232558139535e-07,
+      "loss": 0.0005,
+      "step": 16646
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 1.0701162790697674e-07,
+      "loss": 0.0006,
+      "step": 16647
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 1.07e-07,
+      "loss": 0.0004,
+      "step": 16648
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 1.0698837209302325e-07,
+      "loss": 0.0005,
+      "step": 16649
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 1.0697674418604652e-07,
+      "loss": 0.0005,
+      "step": 16650
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 1.0696511627906976e-07,
+      "loss": 0.0003,
+      "step": 16651
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 1.0695348837209303e-07,
+      "loss": 0.0005,
+      "step": 16652
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 1.0694186046511627e-07,
+      "loss": 0.0006,
+      "step": 16653
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 1.0693023255813954e-07,
+      "loss": 0.0008,
+      "step": 16654
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 1.0691860465116278e-07,
+      "loss": 0.0004,
+      "step": 16655
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 1.0690697674418603e-07,
+      "loss": 0.0005,
+      "step": 16656
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 1.068953488372093e-07,
+      "loss": 0.0004,
+      "step": 16657
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 1.0688372093023254e-07,
+      "loss": 0.0005,
+      "step": 16658
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 1.068720930232558e-07,
+      "loss": 0.0006,
+      "step": 16659
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 1.0686046511627905e-07,
+      "loss": 0.0006,
+      "step": 16660
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 1.0684883720930233e-07,
+      "loss": 0.0006,
+      "step": 16661
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 1.0683720930232558e-07,
+      "loss": 0.0005,
+      "step": 16662
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 1.0682558139534884e-07,
+      "loss": 0.0005,
+      "step": 16663
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 1.0681395348837209e-07,
+      "loss": 0.0006,
+      "step": 16664
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 1.0680232558139535e-07,
+      "loss": 0.0007,
+      "step": 16665
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 1.067906976744186e-07,
+      "loss": 0.0005,
+      "step": 16666
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 1.0677906976744186e-07,
+      "loss": 0.0005,
+      "step": 16667
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 1.0676744186046511e-07,
+      "loss": 0.0005,
+      "step": 16668
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 1.0675581395348837e-07,
+      "loss": 0.0006,
+      "step": 16669
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 1.0674418604651162e-07,
+      "loss": 0.0005,
+      "step": 16670
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 1.0673255813953488e-07,
+      "loss": 0.0006,
+      "step": 16671
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 1.0672093023255813e-07,
+      "loss": 0.0005,
+      "step": 16672
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 1.067093023255814e-07,
+      "loss": 0.0005,
+      "step": 16673
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 1.0669767441860464e-07,
+      "loss": 0.0007,
+      "step": 16674
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 1.0668604651162791e-07,
+      "loss": 0.0004,
+      "step": 16675
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.0667441860465116e-07,
+      "loss": 0.0007,
+      "step": 16676
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.0666279069767442e-07,
+      "loss": 0.0004,
+      "step": 16677
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.0665116279069767e-07,
+      "loss": 0.0009,
+      "step": 16678
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.0663953488372093e-07,
+      "loss": 0.0005,
+      "step": 16679
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.0662790697674418e-07,
+      "loss": 0.0004,
+      "step": 16680
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 1.0661627906976744e-07,
+      "loss": 0.0007,
+      "step": 16681
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 1.0660465116279069e-07,
+      "loss": 0.0005,
+      "step": 16682
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 1.0659302325581394e-07,
+      "loss": 0.0008,
+      "step": 16683
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 1.065813953488372e-07,
+      "loss": 0.0006,
+      "step": 16684
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 1.0656976744186045e-07,
+      "loss": 0.0005,
+      "step": 16685
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 1.0655813953488371e-07,
+      "loss": 0.0005,
+      "step": 16686
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 1.0654651162790696e-07,
+      "loss": 0.001,
+      "step": 16687
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 1.0653488372093024e-07,
+      "loss": 0.0077,
+      "step": 16688
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 1.0652325581395348e-07,
+      "loss": 0.0006,
+      "step": 16689
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 1.0651162790697675e-07,
+      "loss": 0.0008,
+      "step": 16690
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 1.065e-07,
+      "loss": 0.0006,
+      "step": 16691
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 1.0648837209302326e-07,
+      "loss": 0.0008,
+      "step": 16692
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 1.064767441860465e-07,
+      "loss": 0.001,
+      "step": 16693
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 1.0646511627906977e-07,
+      "loss": 0.0009,
+      "step": 16694
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 1.0645348837209301e-07,
+      "loss": 0.0005,
+      "step": 16695
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 1.0644186046511628e-07,
+      "loss": 0.0005,
+      "step": 16696
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 1.0643023255813952e-07,
+      "loss": 0.0004,
+      "step": 16697
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 1.0641860465116279e-07,
+      "loss": 0.0006,
+      "step": 16698
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 1.0640697674418603e-07,
+      "loss": 0.0005,
+      "step": 16699
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 1.0639534883720931e-07,
+      "loss": 0.0005,
+      "step": 16700
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 1.0638372093023255e-07,
+      "loss": 0.0005,
+      "step": 16701
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 1.0637209302325582e-07,
+      "loss": 0.0005,
+      "step": 16702
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 1.0636046511627907e-07,
+      "loss": 0.0006,
+      "step": 16703
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 1.0634883720930233e-07,
+      "loss": 0.0005,
+      "step": 16704
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 1.0633720930232558e-07,
+      "loss": 0.0004,
+      "step": 16705
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 1.0632558139534884e-07,
+      "loss": 0.0006,
+      "step": 16706
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.0631395348837209e-07,
+      "loss": 0.0005,
+      "step": 16707
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.0630232558139535e-07,
+      "loss": 0.0008,
+      "step": 16708
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.062906976744186e-07,
+      "loss": 0.0006,
+      "step": 16709
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.0627906976744185e-07,
+      "loss": 0.0007,
+      "step": 16710
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.0626744186046511e-07,
+      "loss": 0.0007,
+      "step": 16711
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 1.0625581395348836e-07,
+      "loss": 0.0004,
+      "step": 16712
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 1.0624418604651162e-07,
+      "loss": 0.0005,
+      "step": 16713
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 1.0623255813953487e-07,
+      "loss": 0.0006,
+      "step": 16714
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 1.0622093023255814e-07,
+      "loss": 0.0004,
+      "step": 16715
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 1.0620930232558138e-07,
+      "loss": 0.0006,
+      "step": 16716
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 1.0619767441860465e-07,
+      "loss": 0.0065,
+      "step": 16717
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 1.061860465116279e-07,
+      "loss": 0.0005,
+      "step": 16718
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 1.0617441860465116e-07,
+      "loss": 0.0007,
+      "step": 16719
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 1.0616279069767441e-07,
+      "loss": 0.0005,
+      "step": 16720
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 1.0615116279069767e-07,
+      "loss": 0.0005,
+      "step": 16721
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 1.0613953488372092e-07,
+      "loss": 0.0145,
+      "step": 16722
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 1.0612790697674418e-07,
+      "loss": 0.0004,
+      "step": 16723
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 1.0611627906976743e-07,
+      "loss": 0.0062,
+      "step": 16724
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 1.061046511627907e-07,
+      "loss": 0.0052,
+      "step": 16725
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 1.0609302325581394e-07,
+      "loss": 0.0007,
+      "step": 16726
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 1.060813953488372e-07,
+      "loss": 0.0006,
+      "step": 16727
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 1.0606976744186045e-07,
+      "loss": 2.3088,
+      "step": 16728
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 1.0605813953488373e-07,
+      "loss": 0.0006,
+      "step": 16729
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 1.0604651162790698e-07,
+      "loss": 1.245,
+      "step": 16730
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 1.0603488372093024e-07,
+      "loss": 0.0039,
+      "step": 16731
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 1.0602325581395349e-07,
+      "loss": 0.0005,
+      "step": 16732
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 1.0601162790697675e-07,
+      "loss": 0.0005,
+      "step": 16733
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 1.06e-07,
+      "loss": 0.0006,
+      "step": 16734
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 1.0598837209302324e-07,
+      "loss": 0.0006,
+      "step": 16735
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 1.059767441860465e-07,
+      "loss": 0.0007,
+      "step": 16736
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 1.0596511627906975e-07,
+      "loss": 0.0005,
+      "step": 16737
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 1.0595348837209302e-07,
+      "loss": 0.0006,
+      "step": 16738
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 1.0594186046511627e-07,
+      "loss": 0.0007,
+      "step": 16739
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 1.0593023255813953e-07,
+      "loss": 0.0009,
+      "step": 16740
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 1.0591860465116278e-07,
+      "loss": 0.0009,
+      "step": 16741
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 1.0590697674418605e-07,
+      "loss": 0.0005,
+      "step": 16742
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 1.0589534883720929e-07,
+      "loss": 0.0006,
+      "step": 16743
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 1.0588372093023256e-07,
+      "loss": 0.0013,
+      "step": 16744
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 1.0587209302325581e-07,
+      "loss": 0.0006,
+      "step": 16745
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 1.0586046511627907e-07,
+      "loss": 0.0006,
+      "step": 16746
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 1.0584883720930232e-07,
+      "loss": 0.0004,
+      "step": 16747
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 1.0583720930232558e-07,
+      "loss": 0.0004,
+      "step": 16748
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 1.0582558139534883e-07,
+      "loss": 0.0007,
+      "step": 16749
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 1.0581395348837209e-07,
+      "loss": 0.0005,
+      "step": 16750
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 1.0580232558139534e-07,
+      "loss": 0.0007,
+      "step": 16751
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 1.057906976744186e-07,
+      "loss": 2.4797,
+      "step": 16752
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 1.0577906976744185e-07,
+      "loss": 0.0005,
+      "step": 16753
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 1.0576744186046511e-07,
+      "loss": 0.0009,
+      "step": 16754
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 1.0575581395348836e-07,
+      "loss": 0.0007,
+      "step": 16755
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 1.0574418604651163e-07,
+      "loss": 0.0008,
+      "step": 16756
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 1.0573255813953488e-07,
+      "loss": 0.0005,
+      "step": 16757
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 1.0572093023255814e-07,
+      "loss": 0.0005,
+      "step": 16758
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 1.057093023255814e-07,
+      "loss": 0.0009,
+      "step": 16759
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 1.0569767441860466e-07,
+      "loss": 0.0004,
+      "step": 16760
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 1.056860465116279e-07,
+      "loss": 0.0006,
+      "step": 16761
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 1.0567441860465115e-07,
+      "loss": 0.0003,
+      "step": 16762
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 1.0566279069767441e-07,
+      "loss": 0.0004,
+      "step": 16763
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 1.0565116279069766e-07,
+      "loss": 0.0004,
+      "step": 16764
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 1.0563953488372092e-07,
+      "loss": 0.0005,
+      "step": 16765
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 1.0562790697674417e-07,
+      "loss": 0.0005,
+      "step": 16766
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 1.0561627906976743e-07,
+      "loss": 0.0005,
+      "step": 16767
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 1.0560465116279068e-07,
+      "loss": 0.0036,
+      "step": 16768
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 1.0559302325581396e-07,
+      "loss": 0.0008,
+      "step": 16769
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 1.0558139534883719e-07,
+      "loss": 0.0004,
+      "step": 16770
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 1.0556976744186047e-07,
+      "loss": 0.0006,
+      "step": 16771
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 1.0555813953488372e-07,
+      "loss": 0.0005,
+      "step": 16772
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 1.0554651162790698e-07,
+      "loss": 0.0004,
+      "step": 16773
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 1.0553488372093023e-07,
+      "loss": 0.2403,
+      "step": 16774
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 1.0552325581395349e-07,
+      "loss": 0.0009,
+      "step": 16775
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 1.0551162790697674e-07,
+      "loss": 0.0005,
+      "step": 16776
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 1.055e-07,
+      "loss": 0.0006,
+      "step": 16777
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 1.0548837209302325e-07,
+      "loss": 0.0008,
+      "step": 16778
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 1.0547674418604651e-07,
+      "loss": 0.0006,
+      "step": 16779
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 1.0546511627906976e-07,
+      "loss": 0.0006,
+      "step": 16780
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 1.0545348837209302e-07,
+      "loss": 0.0005,
+      "step": 16781
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 1.0544186046511627e-07,
+      "loss": 0.0011,
+      "step": 16782
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 1.0543023255813954e-07,
+      "loss": 0.0006,
+      "step": 16783
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 1.0541860465116279e-07,
+      "loss": 2.503,
+      "step": 16784
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 1.0540697674418605e-07,
+      "loss": 0.0006,
+      "step": 16785
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 1.053953488372093e-07,
+      "loss": 0.0007,
+      "step": 16786
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 1.0538372093023256e-07,
+      "loss": 0.0006,
+      "step": 16787
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 1.0537209302325581e-07,
+      "loss": 0.0003,
+      "step": 16788
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 1.0536046511627906e-07,
+      "loss": 0.0004,
+      "step": 16789
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 1.0534883720930232e-07,
+      "loss": 0.0011,
+      "step": 16790
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 1.0533720930232557e-07,
+      "loss": 0.0007,
+      "step": 16791
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 1.0532558139534883e-07,
+      "loss": 0.0005,
+      "step": 16792
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 1.0531395348837208e-07,
+      "loss": 0.0006,
+      "step": 16793
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 1.0530232558139534e-07,
+      "loss": 0.0005,
+      "step": 16794
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 1.0529069767441859e-07,
+      "loss": 0.0004,
+      "step": 16795
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 1.0527906976744185e-07,
+      "loss": 0.0004,
+      "step": 16796
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 1.052674418604651e-07,
+      "loss": 0.0005,
+      "step": 16797
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 1.0525581395348838e-07,
+      "loss": 0.0006,
+      "step": 16798
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 1.0524418604651162e-07,
+      "loss": 0.0005,
+      "step": 16799
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 1.0523255813953489e-07,
+      "loss": 0.0006,
+      "step": 16800
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 1.0522093023255813e-07,
+      "loss": 0.0005,
+      "step": 16801
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 1.052093023255814e-07,
+      "loss": 0.0026,
+      "step": 16802
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 1.0519767441860464e-07,
+      "loss": 0.0007,
+      "step": 16803
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 1.051860465116279e-07,
+      "loss": 0.0006,
+      "step": 16804
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 1.0517441860465115e-07,
+      "loss": 0.0005,
+      "step": 16805
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 1.0516279069767442e-07,
+      "loss": 0.0005,
+      "step": 16806
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 1.0515116279069766e-07,
+      "loss": 0.0006,
+      "step": 16807
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 1.0513953488372093e-07,
+      "loss": 0.0005,
+      "step": 16808
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 1.0512790697674417e-07,
+      "loss": 0.0006,
+      "step": 16809
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 1.0511627906976745e-07,
+      "loss": 0.4485,
+      "step": 16810
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 1.051046511627907e-07,
+      "loss": 0.0007,
+      "step": 16811
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 1.0509302325581396e-07,
+      "loss": 0.0006,
+      "step": 16812
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 1.0508139534883721e-07,
+      "loss": 0.0008,
+      "step": 16813
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 1.0506976744186047e-07,
+      "loss": 0.0006,
+      "step": 16814
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 1.0505813953488372e-07,
+      "loss": 0.0004,
+      "step": 16815
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 1.0504651162790697e-07,
+      "loss": 0.0008,
+      "step": 16816
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 1.0503488372093023e-07,
+      "loss": 0.0004,
+      "step": 16817
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 1.0502325581395348e-07,
+      "loss": 0.0008,
+      "step": 16818
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 1.0501162790697674e-07,
+      "loss": 0.0006,
+      "step": 16819
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 1.0499999999999999e-07,
+      "loss": 0.0004,
+      "step": 16820
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 1.0498837209302325e-07,
+      "loss": 0.0006,
+      "step": 16821
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 1.049767441860465e-07,
+      "loss": 0.0005,
+      "step": 16822
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 1.0496511627906976e-07,
+      "loss": 0.0004,
+      "step": 16823
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 1.0495348837209301e-07,
+      "loss": 0.0007,
+      "step": 16824
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 1.0494186046511628e-07,
+      "loss": 0.0006,
+      "step": 16825
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 1.0493023255813953e-07,
+      "loss": 0.0005,
+      "step": 16826
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 1.0491860465116279e-07,
+      "loss": 0.0004,
+      "step": 16827
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 1.0490697674418604e-07,
+      "loss": 0.0005,
+      "step": 16828
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 1.048953488372093e-07,
+      "loss": 0.0005,
+      "step": 16829
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 1.0488372093023255e-07,
+      "loss": 0.0005,
+      "step": 16830
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 1.0487209302325581e-07,
+      "loss": 0.0006,
+      "step": 16831
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 1.0486046511627906e-07,
+      "loss": 0.1666,
+      "step": 16832
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 1.0484883720930232e-07,
+      "loss": 0.0013,
+      "step": 16833
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 1.0483720930232557e-07,
+      "loss": 0.0005,
+      "step": 16834
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 1.0482558139534883e-07,
+      "loss": 0.0005,
+      "step": 16835
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 1.0481395348837208e-07,
+      "loss": 0.9371,
+      "step": 16836
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 1.0480232558139536e-07,
+      "loss": 0.0005,
+      "step": 16837
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 1.047906976744186e-07,
+      "loss": 0.002,
+      "step": 16838
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 1.0477906976744187e-07,
+      "loss": 0.0005,
+      "step": 16839
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 1.0476744186046512e-07,
+      "loss": 0.0008,
+      "step": 16840
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 1.0475581395348838e-07,
+      "loss": 0.0004,
+      "step": 16841
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 1.0474418604651163e-07,
+      "loss": 0.0005,
+      "step": 16842
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 1.0473255813953487e-07,
+      "loss": 0.0006,
+      "step": 16843
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 1.0472093023255814e-07,
+      "loss": 0.0005,
+      "step": 16844
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 1.0470930232558138e-07,
+      "loss": 0.0004,
+      "step": 16845
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 1.0469767441860465e-07,
+      "loss": 0.0006,
+      "step": 16846
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 1.046860465116279e-07,
+      "loss": 0.0011,
+      "step": 16847
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 1.0467441860465116e-07,
+      "loss": 0.0004,
+      "step": 16848
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 1.046627906976744e-07,
+      "loss": 0.0008,
+      "step": 16849
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 1.0465116279069767e-07,
+      "loss": 0.001,
+      "step": 16850
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 1.0463953488372092e-07,
+      "loss": 0.0005,
+      "step": 16851
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 1.0462790697674419e-07,
+      "loss": 0.0005,
+      "step": 16852
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 1.0461627906976744e-07,
+      "loss": 0.0009,
+      "step": 16853
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 1.046046511627907e-07,
+      "loss": 0.0005,
+      "step": 16854
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 1.0459302325581395e-07,
+      "loss": 0.0023,
+      "step": 16855
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 1.0458139534883721e-07,
+      "loss": 0.0236,
+      "step": 16856
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 1.0456976744186046e-07,
+      "loss": 0.0007,
+      "step": 16857
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 1.0455813953488372e-07,
+      "loss": 0.0006,
+      "step": 16858
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 1.0454651162790697e-07,
+      "loss": 0.0006,
+      "step": 16859
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 1.0453488372093023e-07,
+      "loss": 0.0004,
+      "step": 16860
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 1.0452325581395348e-07,
+      "loss": 0.0004,
+      "step": 16861
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 1.0451162790697674e-07,
+      "loss": 0.0005,
+      "step": 16862
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 1.0449999999999999e-07,
+      "loss": 0.0007,
+      "step": 16863
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 1.0448837209302326e-07,
+      "loss": 0.0005,
+      "step": 16864
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 1.044767441860465e-07,
+      "loss": 0.001,
+      "step": 16865
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 1.0446511627906977e-07,
+      "loss": 0.0006,
+      "step": 16866
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 1.0445348837209302e-07,
+      "loss": 0.0007,
+      "step": 16867
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 1.0444186046511627e-07,
+      "loss": 0.0008,
+      "step": 16868
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 1.0443023255813953e-07,
+      "loss": 0.0003,
+      "step": 16869
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 1.0441860465116278e-07,
+      "loss": 0.0014,
+      "step": 16870
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 1.0440697674418604e-07,
+      "loss": 0.0042,
+      "step": 16871
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 1.0439534883720929e-07,
+      "loss": 0.0004,
+      "step": 16872
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 1.0438372093023255e-07,
+      "loss": 0.0004,
+      "step": 16873
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 1.043720930232558e-07,
+      "loss": 0.0005,
+      "step": 16874
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 1.0436046511627906e-07,
+      "loss": 0.0006,
+      "step": 16875
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 1.0434883720930231e-07,
+      "loss": 0.0009,
+      "step": 16876
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 1.0433720930232557e-07,
+      "loss": 0.0004,
+      "step": 16877
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 1.0432558139534882e-07,
+      "loss": 0.001,
+      "step": 16878
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 1.043139534883721e-07,
+      "loss": 0.0005,
+      "step": 16879
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 1.0430232558139535e-07,
+      "loss": 0.0004,
+      "step": 16880
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 1.0429069767441861e-07,
+      "loss": 0.0007,
+      "step": 16881
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 1.0427906976744186e-07,
+      "loss": 0.001,
+      "step": 16882
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 1.0426744186046512e-07,
+      "loss": 0.0005,
+      "step": 16883
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 1.0425581395348837e-07,
+      "loss": 0.0006,
+      "step": 16884
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 1.0424418604651163e-07,
+      "loss": 0.0006,
+      "step": 16885
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 1.0423255813953488e-07,
+      "loss": 0.0005,
+      "step": 16886
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 1.0422093023255814e-07,
+      "loss": 1.9563,
+      "step": 16887
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 1.0420930232558139e-07,
+      "loss": 0.0004,
+      "step": 16888
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 1.0419767441860465e-07,
+      "loss": 0.0006,
+      "step": 16889
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 1.041860465116279e-07,
+      "loss": 0.0005,
+      "step": 16890
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 1.0417441860465117e-07,
+      "loss": 0.0005,
+      "step": 16891
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 1.0416279069767441e-07,
+      "loss": 0.0433,
+      "step": 16892
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 1.0415116279069768e-07,
+      "loss": 0.0005,
+      "step": 16893
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 1.0413953488372093e-07,
+      "loss": 0.0004,
+      "step": 16894
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 1.0412790697674418e-07,
+      "loss": 0.0006,
+      "step": 16895
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 1.0411627906976744e-07,
+      "loss": 0.0005,
+      "step": 16896
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 1.0410465116279069e-07,
+      "loss": 0.0005,
+      "step": 16897
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 1.0409302325581395e-07,
+      "loss": 0.0007,
+      "step": 16898
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 1.040813953488372e-07,
+      "loss": 0.0011,
+      "step": 16899
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 1.0406976744186046e-07,
+      "loss": 0.0007,
+      "step": 16900
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 1.0405813953488371e-07,
+      "loss": 0.0006,
+      "step": 16901
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 1.0404651162790697e-07,
+      "loss": 0.0012,
+      "step": 16902
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 1.0403488372093022e-07,
+      "loss": 0.0009,
+      "step": 16903
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 1.0402325581395348e-07,
+      "loss": 0.0004,
+      "step": 16904
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 1.0401162790697673e-07,
+      "loss": 0.0005,
+      "step": 16905
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 1.04e-07,
+      "loss": 0.1057,
+      "step": 16906
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 1.0398837209302324e-07,
+      "loss": 0.0006,
+      "step": 16907
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 1.0397674418604652e-07,
+      "loss": 0.0006,
+      "step": 16908
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 1.0396511627906976e-07,
+      "loss": 0.0006,
+      "step": 16909
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 1.0395348837209303e-07,
+      "loss": 0.0004,
+      "step": 16910
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 1.0394186046511627e-07,
+      "loss": 0.0021,
+      "step": 16911
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 1.0393023255813954e-07,
+      "loss": 0.0005,
+      "step": 16912
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 1.0391860465116278e-07,
+      "loss": 0.0008,
+      "step": 16913
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 1.0390697674418605e-07,
+      "loss": 0.0005,
+      "step": 16914
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 1.038953488372093e-07,
+      "loss": 0.0006,
+      "step": 16915
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 1.0388372093023256e-07,
+      "loss": 0.512,
+      "step": 16916
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 1.038720930232558e-07,
+      "loss": 0.001,
+      "step": 16917
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 1.0386046511627907e-07,
+      "loss": 0.0255,
+      "step": 16918
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 1.0384883720930231e-07,
+      "loss": 0.0009,
+      "step": 16919
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 1.0383720930232559e-07,
+      "loss": 0.0004,
+      "step": 16920
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 1.0382558139534884e-07,
+      "loss": 0.0009,
+      "step": 16921
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 1.0381395348837209e-07,
+      "loss": 0.0008,
+      "step": 16922
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 1.0380232558139535e-07,
+      "loss": 0.0004,
+      "step": 16923
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 1.037906976744186e-07,
+      "loss": 0.0006,
+      "step": 16924
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 1.0377906976744186e-07,
+      "loss": 0.0006,
+      "step": 16925
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 1.0376744186046511e-07,
+      "loss": 0.0007,
+      "step": 16926
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 1.0375581395348837e-07,
+      "loss": 0.0011,
+      "step": 16927
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 1.0374418604651162e-07,
+      "loss": 0.0004,
+      "step": 16928
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 1.0373255813953488e-07,
+      "loss": 0.0005,
+      "step": 16929
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 1.0372093023255813e-07,
+      "loss": 0.0005,
+      "step": 16930
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 1.0370930232558139e-07,
+      "loss": 0.0158,
+      "step": 16931
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 1.0369767441860464e-07,
+      "loss": 0.0005,
+      "step": 16932
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 1.0368604651162791e-07,
+      "loss": 0.0062,
+      "step": 16933
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 1.0367441860465115e-07,
+      "loss": 0.0007,
+      "step": 16934
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 1.0366279069767442e-07,
+      "loss": 0.001,
+      "step": 16935
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 1.0365116279069767e-07,
+      "loss": 0.0005,
+      "step": 16936
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 1.0363953488372093e-07,
+      "loss": 0.0006,
+      "step": 16937
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 1.0362790697674418e-07,
+      "loss": 0.0005,
+      "step": 16938
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 1.0361627906976744e-07,
+      "loss": 0.0007,
+      "step": 16939
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 1.0360465116279069e-07,
+      "loss": 0.0006,
+      "step": 16940
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 1.0359302325581395e-07,
+      "loss": 0.0006,
+      "step": 16941
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 1.035813953488372e-07,
+      "loss": 0.0007,
+      "step": 16942
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 1.0356976744186046e-07,
+      "loss": 0.0007,
+      "step": 16943
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 1.0355813953488371e-07,
+      "loss": 0.003,
+      "step": 16944
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 1.0354651162790697e-07,
+      "loss": 0.0007,
+      "step": 16945
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 1.0353488372093022e-07,
+      "loss": 0.0008,
+      "step": 16946
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 1.035232558139535e-07,
+      "loss": 0.0004,
+      "step": 16947
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 1.0351162790697675e-07,
+      "loss": 0.0007,
+      "step": 16948
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 1.035e-07,
+      "loss": 0.0005,
+      "step": 16949
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 1.0348837209302326e-07,
+      "loss": 0.0005,
+      "step": 16950
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 1.034767441860465e-07,
+      "loss": 0.0005,
+      "step": 16951
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 1.0346511627906977e-07,
+      "loss": 0.0008,
+      "step": 16952
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 1.0345348837209301e-07,
+      "loss": 0.0005,
+      "step": 16953
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 1.0344186046511628e-07,
+      "loss": 0.0014,
+      "step": 16954
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 1.0343023255813952e-07,
+      "loss": 0.0004,
+      "step": 16955
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 1.0341860465116279e-07,
+      "loss": 0.0005,
+      "step": 16956
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 1.0340697674418603e-07,
+      "loss": 0.0006,
+      "step": 16957
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 1.033953488372093e-07,
+      "loss": 0.0006,
+      "step": 16958
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 1.0338372093023254e-07,
+      "loss": 0.0004,
+      "step": 16959
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 1.0337209302325582e-07,
+      "loss": 0.3577,
+      "step": 16960
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 1.0336046511627905e-07,
+      "loss": 0.0009,
+      "step": 16961
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 1.0334883720930233e-07,
+      "loss": 0.0006,
+      "step": 16962
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 1.0333720930232558e-07,
+      "loss": 0.0007,
+      "step": 16963
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 1.0332558139534884e-07,
+      "loss": 0.0006,
+      "step": 16964
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 1.0331395348837209e-07,
+      "loss": 0.0004,
+      "step": 16965
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 1.0330232558139535e-07,
+      "loss": 0.0009,
+      "step": 16966
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 1.032906976744186e-07,
+      "loss": 0.0005,
+      "step": 16967
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 1.0327906976744186e-07,
+      "loss": 0.0007,
+      "step": 16968
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 1.0326744186046511e-07,
+      "loss": 0.0007,
+      "step": 16969
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 1.0325581395348837e-07,
+      "loss": 0.0022,
+      "step": 16970
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 1.0324418604651162e-07,
+      "loss": 0.0006,
+      "step": 16971
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 1.0323255813953488e-07,
+      "loss": 0.0005,
+      "step": 16972
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 1.0322093023255813e-07,
+      "loss": 0.0008,
+      "step": 16973
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 1.032093023255814e-07,
+      "loss": 0.0008,
+      "step": 16974
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 1.0319767441860465e-07,
+      "loss": 0.0005,
+      "step": 16975
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 1.0318604651162789e-07,
+      "loss": 0.0006,
+      "step": 16976
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 1.0317441860465116e-07,
+      "loss": 0.0004,
+      "step": 16977
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 1.0316279069767441e-07,
+      "loss": 0.0005,
+      "step": 16978
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 1.0315116279069767e-07,
+      "loss": 0.0005,
+      "step": 16979
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 1.0313953488372092e-07,
+      "loss": 0.0006,
+      "step": 16980
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.0312790697674418e-07,
+      "loss": 0.0006,
+      "step": 16981
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.0311627906976743e-07,
+      "loss": 0.0003,
+      "step": 16982
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.031046511627907e-07,
+      "loss": 0.0005,
+      "step": 16983
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.0309302325581394e-07,
+      "loss": 0.0006,
+      "step": 16984
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.030813953488372e-07,
+      "loss": 0.0009,
+      "step": 16985
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 1.0306976744186045e-07,
+      "loss": 0.0005,
+      "step": 16986
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 1.0305813953488371e-07,
+      "loss": 0.0005,
+      "step": 16987
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 1.0304651162790696e-07,
+      "loss": 0.0006,
+      "step": 16988
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 1.0303488372093024e-07,
+      "loss": 0.0004,
+      "step": 16989
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 1.0302325581395349e-07,
+      "loss": 0.0006,
+      "step": 16990
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 1.0301162790697675e-07,
+      "loss": 0.0005,
+      "step": 16991
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 1.03e-07,
+      "loss": 0.0007,
+      "step": 16992
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 1.0298837209302326e-07,
+      "loss": 0.0005,
+      "step": 16993
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 1.029767441860465e-07,
+      "loss": 0.0006,
+      "step": 16994
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 1.0296511627906977e-07,
+      "loss": 0.0007,
+      "step": 16995
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 1.0295348837209302e-07,
+      "loss": 0.0005,
+      "step": 16996
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 1.0294186046511628e-07,
+      "loss": 0.0008,
+      "step": 16997
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 1.0293023255813953e-07,
+      "loss": 1.897,
+      "step": 16998
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 1.0291860465116279e-07,
+      "loss": 0.0006,
+      "step": 16999
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 1.0290697674418604e-07,
+      "loss": 0.0005,
+      "step": 17000
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 1.0289534883720931e-07,
+      "loss": 0.0014,
+      "step": 17001
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 1.0288372093023256e-07,
+      "loss": 0.0005,
+      "step": 17002
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 1.028720930232558e-07,
+      "loss": 0.0005,
+      "step": 17003
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 1.0286046511627907e-07,
+      "loss": 0.0007,
+      "step": 17004
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 1.0284883720930232e-07,
+      "loss": 0.0073,
+      "step": 17005
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 1.0283720930232558e-07,
+      "loss": 0.0007,
+      "step": 17006
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 1.0282558139534883e-07,
+      "loss": 0.0013,
+      "step": 17007
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 1.0281395348837209e-07,
+      "loss": 0.0006,
+      "step": 17008
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 1.0280232558139534e-07,
+      "loss": 0.0006,
+      "step": 17009
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 1.027906976744186e-07,
+      "loss": 3.7551,
+      "step": 17010
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 1.0277906976744185e-07,
+      "loss": 0.0007,
+      "step": 17011
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.0276744186046511e-07,
+      "loss": 0.0005,
+      "step": 17012
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.0275581395348836e-07,
+      "loss": 0.0008,
+      "step": 17013
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.0274418604651162e-07,
+      "loss": 0.0006,
+      "step": 17014
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.0273255813953487e-07,
+      "loss": 0.0008,
+      "step": 17015
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.0272093023255814e-07,
+      "loss": 0.0005,
+      "step": 17016
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 1.0270930232558139e-07,
+      "loss": 0.0004,
+      "step": 17017
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 1.0269767441860465e-07,
+      "loss": 0.0005,
+      "step": 17018
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 1.026860465116279e-07,
+      "loss": 0.0005,
+      "step": 17019
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 1.0267441860465116e-07,
+      "loss": 0.0004,
+      "step": 17020
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 1.0266279069767441e-07,
+      "loss": 0.0005,
+      "step": 17021
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 1.0265116279069768e-07,
+      "loss": 0.0006,
+      "step": 17022
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 1.0263953488372092e-07,
+      "loss": 0.0005,
+      "step": 17023
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 1.0262790697674419e-07,
+      "loss": 0.0009,
+      "step": 17024
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 1.0261627906976743e-07,
+      "loss": 0.0007,
+      "step": 17025
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 1.026046511627907e-07,
+      "loss": 0.0004,
+      "step": 17026
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 1.0259302325581394e-07,
+      "loss": 0.0009,
+      "step": 17027
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 1.0258139534883719e-07,
+      "loss": 0.0004,
+      "step": 17028
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 1.0256976744186045e-07,
+      "loss": 0.0007,
+      "step": 17029
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 1.025581395348837e-07,
+      "loss": 0.0006,
+      "step": 17030
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 1.0254651162790698e-07,
+      "loss": 0.0005,
+      "step": 17031
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 1.0253488372093023e-07,
+      "loss": 0.0005,
+      "step": 17032
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 1.0252325581395349e-07,
+      "loss": 0.0006,
+      "step": 17033
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 1.0251162790697674e-07,
+      "loss": 0.0006,
+      "step": 17034
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 1.025e-07,
+      "loss": 0.0004,
+      "step": 17035
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 1.0248837209302325e-07,
+      "loss": 0.0005,
+      "step": 17036
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 1.0247674418604651e-07,
+      "loss": 0.0006,
+      "step": 17037
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.0246511627906976e-07,
+      "loss": 0.0005,
+      "step": 17038
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.0245348837209302e-07,
+      "loss": 0.0005,
+      "step": 17039
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.0244186046511627e-07,
+      "loss": 0.0004,
+      "step": 17040
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.0243023255813953e-07,
+      "loss": 0.0021,
+      "step": 17041
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 1.0241860465116278e-07,
+      "loss": 0.0005,
+      "step": 17042
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 1.0240697674418605e-07,
+      "loss": 0.0006,
+      "step": 17043
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 1.023953488372093e-07,
+      "loss": 0.0005,
+      "step": 17044
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 1.0238372093023256e-07,
+      "loss": 0.0004,
+      "step": 17045
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 1.0237209302325581e-07,
+      "loss": 0.3214,
+      "step": 17046
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 1.0236046511627907e-07,
+      "loss": 0.0004,
+      "step": 17047
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 1.0234883720930232e-07,
+      "loss": 0.0004,
+      "step": 17048
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 1.0233720930232558e-07,
+      "loss": 0.0004,
+      "step": 17049
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 1.0232558139534883e-07,
+      "loss": 0.0004,
+      "step": 17050
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 1.0231395348837209e-07,
+      "loss": 0.0013,
+      "step": 17051
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 1.0230232558139534e-07,
+      "loss": 0.0009,
+      "step": 17052
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 1.022906976744186e-07,
+      "loss": 0.0006,
+      "step": 17053
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 1.0227906976744185e-07,
+      "loss": 0.0005,
+      "step": 17054
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 1.022674418604651e-07,
+      "loss": 0.0005,
+      "step": 17055
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 1.0225581395348836e-07,
+      "loss": 0.0006,
+      "step": 17056
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 1.0224418604651161e-07,
+      "loss": 0.0004,
+      "step": 17057
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 1.0223255813953489e-07,
+      "loss": 0.0004,
+      "step": 17058
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 1.0222093023255813e-07,
+      "loss": 0.0007,
+      "step": 17059
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 1.022093023255814e-07,
+      "loss": 0.0005,
+      "step": 17060
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 1.0219767441860464e-07,
+      "loss": 0.0006,
+      "step": 17061
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.9903288201160542,
+      "eval_loss": 0.1485539674758911,
+      "eval_runtime": 1.5562,
+      "eval_samples_per_second": 664.43,
+      "eval_steps_per_second": 10.924,
+      "step": 17061
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 1.021860465116279e-07,
+      "loss": 0.0004,
+      "step": 17062
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 1.0217441860465115e-07,
+      "loss": 0.0006,
+      "step": 17063
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 1.0216279069767442e-07,
+      "loss": 0.0006,
+      "step": 17064
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 1.0215116279069766e-07,
+      "loss": 0.0007,
+      "step": 17065
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 1.0213953488372093e-07,
+      "loss": 0.0005,
+      "step": 17066
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 1.0212790697674417e-07,
+      "loss": 0.434,
+      "step": 17067
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 1.0211627906976744e-07,
+      "loss": 0.0006,
+      "step": 17068
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 1.0210465116279068e-07,
+      "loss": 0.0005,
+      "step": 17069
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 1.0209302325581396e-07,
+      "loss": 0.001,
+      "step": 17070
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 1.0208139534883721e-07,
+      "loss": 0.0004,
+      "step": 17071
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 1.0206976744186047e-07,
+      "loss": 0.0005,
+      "step": 17072
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 1.0205813953488372e-07,
+      "loss": 0.0008,
+      "step": 17073
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.0204651162790698e-07,
+      "loss": 0.0006,
+      "step": 17074
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.0203488372093023e-07,
+      "loss": 0.0004,
+      "step": 17075
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.0202325581395349e-07,
+      "loss": 0.0006,
+      "step": 17076
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.0201162790697674e-07,
+      "loss": 0.0004,
+      "step": 17077
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.02e-07,
+      "loss": 0.0005,
+      "step": 17078
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 1.0198837209302325e-07,
+      "loss": 0.0005,
+      "step": 17079
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 1.0197674418604651e-07,
+      "loss": 0.0005,
+      "step": 17080
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 1.0196511627906976e-07,
+      "loss": 0.0008,
+      "step": 17081
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 1.0195348837209301e-07,
+      "loss": 0.0011,
+      "step": 17082
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 1.0194186046511627e-07,
+      "loss": 0.0004,
+      "step": 17083
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 1.0193023255813952e-07,
+      "loss": 0.0004,
+      "step": 17084
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 1.0191860465116279e-07,
+      "loss": 0.0005,
+      "step": 17085
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 1.0190697674418604e-07,
+      "loss": 0.0017,
+      "step": 17086
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 1.018953488372093e-07,
+      "loss": 0.0004,
+      "step": 17087
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 1.0188372093023255e-07,
+      "loss": 0.0005,
+      "step": 17088
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 1.0187209302325581e-07,
+      "loss": 0.0006,
+      "step": 17089
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 1.0186046511627906e-07,
+      "loss": 0.0014,
+      "step": 17090
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 1.0184883720930232e-07,
+      "loss": 0.0004,
+      "step": 17091
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 1.0183720930232557e-07,
+      "loss": 0.0004,
+      "step": 17092
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 1.0182558139534883e-07,
+      "loss": 0.0004,
+      "step": 17093
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 1.0181395348837208e-07,
+      "loss": 0.0007,
+      "step": 17094
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 1.0180232558139534e-07,
+      "loss": 0.0006,
+      "step": 17095
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 1.0179069767441859e-07,
+      "loss": 0.0005,
+      "step": 17096
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 1.0177906976744187e-07,
+      "loss": 0.0005,
+      "step": 17097
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 1.017674418604651e-07,
+      "loss": 0.0004,
+      "step": 17098
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 1.0175581395348838e-07,
+      "loss": 0.0006,
+      "step": 17099
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 1.0174418604651163e-07,
+      "loss": 0.1903,
+      "step": 17100
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 1.0173255813953489e-07,
+      "loss": 0.0005,
+      "step": 17101
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 1.0172093023255814e-07,
+      "loss": 0.0012,
+      "step": 17102
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 1.017093023255814e-07,
+      "loss": 0.0006,
+      "step": 17103
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 1.0169767441860465e-07,
+      "loss": 0.0005,
+      "step": 17104
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.0168604651162791e-07,
+      "loss": 0.0005,
+      "step": 17105
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.0167441860465116e-07,
+      "loss": 0.0006,
+      "step": 17106
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.0166279069767442e-07,
+      "loss": 0.0008,
+      "step": 17107
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.0165116279069767e-07,
+      "loss": 0.9208,
+      "step": 17108
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.0163953488372091e-07,
+      "loss": 0.0004,
+      "step": 17109
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 1.0162790697674418e-07,
+      "loss": 0.0005,
+      "step": 17110
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 1.0161627906976742e-07,
+      "loss": 0.0005,
+      "step": 17111
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 1.016046511627907e-07,
+      "loss": 0.0006,
+      "step": 17112
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 1.0159302325581395e-07,
+      "loss": 0.0069,
+      "step": 17113
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 1.0158139534883721e-07,
+      "loss": 0.0008,
+      "step": 17114
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 1.0156976744186046e-07,
+      "loss": 0.0005,
+      "step": 17115
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 1.0155813953488372e-07,
+      "loss": 0.0004,
+      "step": 17116
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 1.0154651162790697e-07,
+      "loss": 0.0005,
+      "step": 17117
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 1.0153488372093023e-07,
+      "loss": 0.0006,
+      "step": 17118
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 1.0152325581395348e-07,
+      "loss": 0.0011,
+      "step": 17119
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 1.0151162790697674e-07,
+      "loss": 0.0005,
+      "step": 17120
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 1.0149999999999999e-07,
+      "loss": 0.0005,
+      "step": 17121
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 1.0148837209302325e-07,
+      "loss": 0.0005,
+      "step": 17122
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 1.014767441860465e-07,
+      "loss": 0.0028,
+      "step": 17123
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 1.0146511627906977e-07,
+      "loss": 0.0005,
+      "step": 17124
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 1.0145348837209301e-07,
+      "loss": 0.0006,
+      "step": 17125
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 1.0144186046511628e-07,
+      "loss": 0.0005,
+      "step": 17126
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 1.0143023255813953e-07,
+      "loss": 0.0009,
+      "step": 17127
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 1.014186046511628e-07,
+      "loss": 0.0005,
+      "step": 17128
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 1.0140697674418604e-07,
+      "loss": 0.0004,
+      "step": 17129
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 1.013953488372093e-07,
+      "loss": 0.0004,
+      "step": 17130
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 1.0138372093023255e-07,
+      "loss": 0.0005,
+      "step": 17131
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 1.0137209302325581e-07,
+      "loss": 0.0006,
+      "step": 17132
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 1.0136046511627906e-07,
+      "loss": 0.0005,
+      "step": 17133
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 1.0134883720930233e-07,
+      "loss": 0.0004,
+      "step": 17134
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 1.0133720930232557e-07,
+      "loss": 0.0355,
+      "step": 17135
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.0132558139534882e-07,
+      "loss": 0.0008,
+      "step": 17136
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.0131395348837208e-07,
+      "loss": 0.0007,
+      "step": 17137
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.0130232558139533e-07,
+      "loss": 0.0005,
+      "step": 17138
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.0129069767441861e-07,
+      "loss": 0.0009,
+      "step": 17139
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.0127906976744186e-07,
+      "loss": 0.0007,
+      "step": 17140
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 1.0126744186046512e-07,
+      "loss": 0.0004,
+      "step": 17141
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 1.0125581395348837e-07,
+      "loss": 0.0005,
+      "step": 17142
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 1.0124418604651163e-07,
+      "loss": 0.0005,
+      "step": 17143
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 1.0123255813953488e-07,
+      "loss": 0.0004,
+      "step": 17144
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 1.0122093023255814e-07,
+      "loss": 0.0006,
+      "step": 17145
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 1.0120930232558139e-07,
+      "loss": 0.0005,
+      "step": 17146
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 1.0119767441860465e-07,
+      "loss": 0.0008,
+      "step": 17147
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 1.011860465116279e-07,
+      "loss": 0.0004,
+      "step": 17148
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 1.0117441860465116e-07,
+      "loss": 0.0004,
+      "step": 17149
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 1.0116279069767441e-07,
+      "loss": 0.0005,
+      "step": 17150
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 1.0115116279069768e-07,
+      "loss": 0.0602,
+      "step": 17151
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 1.0113953488372092e-07,
+      "loss": 0.0007,
+      "step": 17152
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 1.0112790697674419e-07,
+      "loss": 0.0005,
+      "step": 17153
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 1.0111627906976744e-07,
+      "loss": 0.0005,
+      "step": 17154
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 1.011046511627907e-07,
+      "loss": 0.0006,
+      "step": 17155
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 1.0109302325581395e-07,
+      "loss": 0.5146,
+      "step": 17156
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 1.0108139534883721e-07,
+      "loss": 0.1209,
+      "step": 17157
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 1.0106976744186046e-07,
+      "loss": 0.0005,
+      "step": 17158
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 1.0105813953488372e-07,
+      "loss": 0.0005,
+      "step": 17159
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 1.0104651162790697e-07,
+      "loss": 0.0005,
+      "step": 17160
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 1.0103488372093023e-07,
+      "loss": 0.0006,
+      "step": 17161
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 1.0102325581395348e-07,
+      "loss": 0.0005,
+      "step": 17162
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 1.0101162790697673e-07,
+      "loss": 0.0005,
+      "step": 17163
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 1.0099999999999999e-07,
+      "loss": 0.0008,
+      "step": 17164
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 1.0098837209302324e-07,
+      "loss": 0.0005,
+      "step": 17165
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 1.0097674418604651e-07,
+      "loss": 0.0004,
+      "step": 17166
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.0096511627906975e-07,
+      "loss": 0.0005,
+      "step": 17167
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.0095348837209302e-07,
+      "loss": 0.0004,
+      "step": 17168
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.0094186046511627e-07,
+      "loss": 0.001,
+      "step": 17169
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.0093023255813953e-07,
+      "loss": 0.0005,
+      "step": 17170
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.0091860465116278e-07,
+      "loss": 0.0005,
+      "step": 17171
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 1.0090697674418605e-07,
+      "loss": 0.0015,
+      "step": 17172
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 1.008953488372093e-07,
+      "loss": 0.0005,
+      "step": 17173
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 1.0088372093023256e-07,
+      "loss": 0.0004,
+      "step": 17174
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 1.008720930232558e-07,
+      "loss": 0.0005,
+      "step": 17175
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 1.0086046511627907e-07,
+      "loss": 0.001,
+      "step": 17176
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 1.0084883720930231e-07,
+      "loss": 0.0006,
+      "step": 17177
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.0083720930232558e-07,
+      "loss": 0.0006,
+      "step": 17178
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.0082558139534882e-07,
+      "loss": 0.0005,
+      "step": 17179
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.008139534883721e-07,
+      "loss": 0.0004,
+      "step": 17180
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.0080232558139535e-07,
+      "loss": 0.0017,
+      "step": 17181
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 1.0079069767441861e-07,
+      "loss": 0.0017,
+      "step": 17182
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 1.0077906976744186e-07,
+      "loss": 0.0004,
+      "step": 17183
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 1.0076744186046512e-07,
+      "loss": 0.0006,
+      "step": 17184
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 1.0075581395348837e-07,
+      "loss": 0.0005,
+      "step": 17185
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 1.0074418604651163e-07,
+      "loss": 0.0004,
+      "step": 17186
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 1.0073255813953488e-07,
+      "loss": 0.0007,
+      "step": 17187
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 1.0072093023255813e-07,
+      "loss": 0.0005,
+      "step": 17188
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 1.0070930232558139e-07,
+      "loss": 0.0007,
+      "step": 17189
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 1.0069767441860464e-07,
+      "loss": 0.0005,
+      "step": 17190
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 1.006860465116279e-07,
+      "loss": 0.0004,
+      "step": 17191
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 1.0067441860465115e-07,
+      "loss": 0.0008,
+      "step": 17192
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.0066279069767442e-07,
+      "loss": 0.0005,
+      "step": 17193
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.0065116279069766e-07,
+      "loss": 0.0004,
+      "step": 17194
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.0063953488372093e-07,
+      "loss": 0.0009,
+      "step": 17195
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.0062790697674418e-07,
+      "loss": 0.0007,
+      "step": 17196
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.0061627906976744e-07,
+      "loss": 0.0006,
+      "step": 17197
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 1.0060465116279069e-07,
+      "loss": 0.0004,
+      "step": 17198
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 1.0059302325581395e-07,
+      "loss": 0.0006,
+      "step": 17199
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 1.005813953488372e-07,
+      "loss": 0.0007,
+      "step": 17200
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 1.0056976744186046e-07,
+      "loss": 0.0008,
+      "step": 17201
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 1.0055813953488371e-07,
+      "loss": 0.0016,
+      "step": 17202
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 1.0054651162790697e-07,
+      "loss": 0.0006,
+      "step": 17203
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 1.0053488372093022e-07,
+      "loss": 0.0004,
+      "step": 17204
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 1.0052325581395348e-07,
+      "loss": 0.0005,
+      "step": 17205
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 1.0051162790697673e-07,
+      "loss": 0.0004,
+      "step": 17206
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 1.005e-07,
+      "loss": 0.0005,
+      "step": 17207
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 1.0048837209302326e-07,
+      "loss": 0.0006,
+      "step": 17208
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 1.0047674418604652e-07,
+      "loss": 0.0005,
+      "step": 17209
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 1.0046511627906977e-07,
+      "loss": 0.0006,
+      "step": 17210
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 1.0045348837209303e-07,
+      "loss": 0.0049,
+      "step": 17211
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 1.0044186046511628e-07,
+      "loss": 0.0005,
+      "step": 17212
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 1.0043023255813954e-07,
+      "loss": 0.0005,
+      "step": 17213
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 1.0041860465116279e-07,
+      "loss": 0.0005,
+      "step": 17214
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 1.0040697674418603e-07,
+      "loss": 0.0011,
+      "step": 17215
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 1.003953488372093e-07,
+      "loss": 0.0005,
+      "step": 17216
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 1.0038372093023254e-07,
+      "loss": 0.0004,
+      "step": 17217
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 1.003720930232558e-07,
+      "loss": 0.0026,
+      "step": 17218
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 1.0036046511627905e-07,
+      "loss": 0.0007,
+      "step": 17219
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 1.0034883720930232e-07,
+      "loss": 0.0004,
+      "step": 17220
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 1.0033720930232556e-07,
+      "loss": 0.0016,
+      "step": 17221
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 1.0032558139534884e-07,
+      "loss": 0.0005,
+      "step": 17222
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 1.0031395348837209e-07,
+      "loss": 0.0005,
+      "step": 17223
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.0030232558139535e-07,
+      "loss": 0.0006,
+      "step": 17224
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.002906976744186e-07,
+      "loss": 0.0004,
+      "step": 17225
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.0027906976744186e-07,
+      "loss": 0.0005,
+      "step": 17226
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.0026744186046511e-07,
+      "loss": 0.0003,
+      "step": 17227
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.0025581395348837e-07,
+      "loss": 0.0005,
+      "step": 17228
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 1.0024418604651162e-07,
+      "loss": 0.0006,
+      "step": 17229
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 1.0023255813953488e-07,
+      "loss": 0.0005,
+      "step": 17230
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 1.0022093023255813e-07,
+      "loss": 0.0004,
+      "step": 17231
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 1.0020930232558139e-07,
+      "loss": 0.0005,
+      "step": 17232
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 1.0019767441860464e-07,
+      "loss": 0.0006,
+      "step": 17233
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 1.0018604651162791e-07,
+      "loss": 0.0006,
+      "step": 17234
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 1.0017441860465116e-07,
+      "loss": 0.0006,
+      "step": 17235
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 1.0016279069767442e-07,
+      "loss": 0.0007,
+      "step": 17236
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 1.0015116279069767e-07,
+      "loss": 0.0005,
+      "step": 17237
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 1.0013953488372093e-07,
+      "loss": 0.0005,
+      "step": 17238
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 1.0012790697674418e-07,
+      "loss": 0.0004,
+      "step": 17239
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 1.0011627906976744e-07,
+      "loss": 0.0006,
+      "step": 17240
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 1.0010465116279069e-07,
+      "loss": 0.5438,
+      "step": 17241
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 1.0009302325581394e-07,
+      "loss": 0.0003,
+      "step": 17242
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 1.000813953488372e-07,
+      "loss": 0.0006,
+      "step": 17243
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 1.0006976744186045e-07,
+      "loss": 0.0005,
+      "step": 17244
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 1.0005813953488371e-07,
+      "loss": 0.0004,
+      "step": 17245
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 1.0004651162790696e-07,
+      "loss": 0.0003,
+      "step": 17246
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 1.0003488372093022e-07,
+      "loss": 0.0006,
+      "step": 17247
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 1.0002325581395347e-07,
+      "loss": 0.0004,
+      "step": 17248
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 1.0001162790697675e-07,
+      "loss": 0.0008,
+      "step": 17249
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 1e-07,
+      "loss": 0.326,
+      "step": 17250
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 9.998837209302326e-08,
+      "loss": 0.0004,
+      "step": 17251
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 9.99767441860465e-08,
+      "loss": 0.0007,
+      "step": 17252
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 9.996511627906977e-08,
+      "loss": 0.0005,
+      "step": 17253
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 9.995348837209302e-08,
+      "loss": 0.0005,
+      "step": 17254
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 9.994186046511628e-08,
+      "loss": 0.0007,
+      "step": 17255
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 9.993023255813953e-08,
+      "loss": 0.0007,
+      "step": 17256
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 9.991860465116279e-08,
+      "loss": 0.0006,
+      "step": 17257
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 9.990697674418604e-08,
+      "loss": 0.0004,
+      "step": 17258
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 9.98953488372093e-08,
+      "loss": 0.0004,
+      "step": 17259
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 9.988372093023255e-08,
+      "loss": 0.0003,
+      "step": 17260
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 9.987209302325582e-08,
+      "loss": 0.0005,
+      "step": 17261
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 9.986046511627907e-08,
+      "loss": 0.0004,
+      "step": 17262
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 9.984883720930233e-08,
+      "loss": 0.0005,
+      "step": 17263
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 9.983720930232558e-08,
+      "loss": 0.0003,
+      "step": 17264
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 9.982558139534884e-08,
+      "loss": 0.0005,
+      "step": 17265
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 9.981395348837209e-08,
+      "loss": 0.0006,
+      "step": 17266
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 9.980232558139535e-08,
+      "loss": 0.0006,
+      "step": 17267
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 9.97906976744186e-08,
+      "loss": 0.0005,
+      "step": 17268
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 9.977906976744185e-08,
+      "loss": 0.0008,
+      "step": 17269
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 9.976744186046511e-08,
+      "loss": 0.0004,
+      "step": 17270
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 9.975581395348836e-08,
+      "loss": 0.0005,
+      "step": 17271
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 9.974418604651162e-08,
+      "loss": 0.0024,
+      "step": 17272
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 9.973255813953487e-08,
+      "loss": 0.0006,
+      "step": 17273
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 9.972093023255813e-08,
+      "loss": 0.0004,
+      "step": 17274
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 9.970930232558138e-08,
+      "loss": 0.0005,
+      "step": 17275
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 9.969767441860465e-08,
+      "loss": 0.0006,
+      "step": 17276
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 9.96860465116279e-08,
+      "loss": 0.0006,
+      "step": 17277
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 9.967441860465116e-08,
+      "loss": 0.0006,
+      "step": 17278
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 9.966279069767441e-08,
+      "loss": 0.0006,
+      "step": 17279
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 9.965116279069767e-08,
+      "loss": 0.0004,
+      "step": 17280
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 9.963953488372092e-08,
+      "loss": 0.0005,
+      "step": 17281
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 9.962790697674418e-08,
+      "loss": 0.0003,
+      "step": 17282
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 9.961627906976743e-08,
+      "loss": 0.0005,
+      "step": 17283
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 9.96046511627907e-08,
+      "loss": 0.0005,
+      "step": 17284
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 9.959302325581394e-08,
+      "loss": 0.0004,
+      "step": 17285
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 9.95813953488372e-08,
+      "loss": 0.0005,
+      "step": 17286
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 9.956976744186045e-08,
+      "loss": 0.0004,
+      "step": 17287
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 9.955813953488373e-08,
+      "loss": 0.0006,
+      "step": 17288
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 9.954651162790696e-08,
+      "loss": 0.0005,
+      "step": 17289
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 9.953488372093024e-08,
+      "loss": 0.0005,
+      "step": 17290
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 9.952325581395349e-08,
+      "loss": 1.4249,
+      "step": 17291
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 9.951162790697675e-08,
+      "loss": 0.0006,
+      "step": 17292
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 9.95e-08,
+      "loss": 0.0005,
+      "step": 17293
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 9.948837209302326e-08,
+      "loss": 0.0005,
+      "step": 17294
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 9.947674418604651e-08,
+      "loss": 0.0004,
+      "step": 17295
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 9.946511627906976e-08,
+      "loss": 0.0005,
+      "step": 17296
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 9.945348837209302e-08,
+      "loss": 0.1574,
+      "step": 17297
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 9.944186046511627e-08,
+      "loss": 0.0005,
+      "step": 17298
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 9.943023255813953e-08,
+      "loss": 0.0008,
+      "step": 17299
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 9.941860465116278e-08,
+      "loss": 0.0012,
+      "step": 17300
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 9.940697674418604e-08,
+      "loss": 0.0005,
+      "step": 17301
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 9.939534883720929e-08,
+      "loss": 0.0006,
+      "step": 17302
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 9.938372093023256e-08,
+      "loss": 0.0006,
+      "step": 17303
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 9.937209302325581e-08,
+      "loss": 0.0006,
+      "step": 17304
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 9.936046511627907e-08,
+      "loss": 0.0006,
+      "step": 17305
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 9.934883720930232e-08,
+      "loss": 0.0006,
+      "step": 17306
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 9.933720930232558e-08,
+      "loss": 0.0006,
+      "step": 17307
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 9.932558139534883e-08,
+      "loss": 0.0005,
+      "step": 17308
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 9.931395348837209e-08,
+      "loss": 0.0005,
+      "step": 17309
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 9.930232558139534e-08,
+      "loss": 1.9183,
+      "step": 17310
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 9.92906976744186e-08,
+      "loss": 0.0005,
+      "step": 17311
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 9.927906976744185e-08,
+      "loss": 0.0005,
+      "step": 17312
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 9.926744186046511e-08,
+      "loss": 0.0005,
+      "step": 17313
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 9.925581395348836e-08,
+      "loss": 0.0009,
+      "step": 17314
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 9.924418604651164e-08,
+      "loss": 0.0008,
+      "step": 17315
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 9.923255813953487e-08,
+      "loss": 0.0006,
+      "step": 17316
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 9.922093023255815e-08,
+      "loss": 0.0005,
+      "step": 17317
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 9.92093023255814e-08,
+      "loss": 0.0005,
+      "step": 17318
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 9.919767441860466e-08,
+      "loss": 0.0005,
+      "step": 17319
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 9.91860465116279e-08,
+      "loss": 0.0006,
+      "step": 17320
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 9.917441860465117e-08,
+      "loss": 0.0004,
+      "step": 17321
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 9.916279069767442e-08,
+      "loss": 0.0005,
+      "step": 17322
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 9.915116279069766e-08,
+      "loss": 0.0007,
+      "step": 17323
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 9.913953488372093e-08,
+      "loss": 0.0014,
+      "step": 17324
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 9.912790697674417e-08,
+      "loss": 0.0007,
+      "step": 17325
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 9.911627906976744e-08,
+      "loss": 0.0005,
+      "step": 17326
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 9.910465116279068e-08,
+      "loss": 2.6271,
+      "step": 17327
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 9.909302325581395e-08,
+      "loss": 0.0005,
+      "step": 17328
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 9.90813953488372e-08,
+      "loss": 0.0009,
+      "step": 17329
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 9.906976744186047e-08,
+      "loss": 0.0004,
+      "step": 17330
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 9.905813953488372e-08,
+      "loss": 0.0021,
+      "step": 17331
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 9.904651162790698e-08,
+      "loss": 0.0005,
+      "step": 17332
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 9.903488372093023e-08,
+      "loss": 0.0006,
+      "step": 17333
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 9.902325581395349e-08,
+      "loss": 0.0004,
+      "step": 17334
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 9.901162790697674e-08,
+      "loss": 0.0005,
+      "step": 17335
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 9.9e-08,
+      "loss": 0.0171,
+      "step": 17336
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 9.898837209302325e-08,
+      "loss": 0.0012,
+      "step": 17337
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 9.897674418604651e-08,
+      "loss": 0.0482,
+      "step": 17338
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 9.896511627906976e-08,
+      "loss": 0.0006,
+      "step": 17339
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 9.895348837209302e-08,
+      "loss": 0.0005,
+      "step": 17340
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 9.894186046511627e-08,
+      "loss": 0.0006,
+      "step": 17341
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 9.893023255813954e-08,
+      "loss": 0.0006,
+      "step": 17342
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 9.891860465116278e-08,
+      "loss": 0.0005,
+      "step": 17343
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 9.890697674418605e-08,
+      "loss": 0.0009,
+      "step": 17344
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 9.88953488372093e-08,
+      "loss": 0.0003,
+      "step": 17345
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 9.888372093023256e-08,
+      "loss": 0.0005,
+      "step": 17346
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 9.887209302325581e-08,
+      "loss": 0.0006,
+      "step": 17347
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 9.886046511627906e-08,
+      "loss": 0.0006,
+      "step": 17348
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 9.884883720930232e-08,
+      "loss": 0.0007,
+      "step": 17349
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 9.883720930232557e-08,
+      "loss": 0.0005,
+      "step": 17350
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 9.882558139534883e-08,
+      "loss": 0.0004,
+      "step": 17351
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 9.881395348837208e-08,
+      "loss": 0.0006,
+      "step": 17352
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 9.880232558139534e-08,
+      "loss": 0.0006,
+      "step": 17353
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 9.879069767441859e-08,
+      "loss": 0.0004,
+      "step": 17354
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 9.877906976744185e-08,
+      "loss": 0.0006,
+      "step": 17355
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 9.87674418604651e-08,
+      "loss": 0.0005,
+      "step": 17356
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 9.875581395348838e-08,
+      "loss": 0.0007,
+      "step": 17357
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 9.874418604651161e-08,
+      "loss": 0.0006,
+      "step": 17358
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 9.873255813953489e-08,
+      "loss": 0.0014,
+      "step": 17359
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 9.872093023255814e-08,
+      "loss": 1.7213,
+      "step": 17360
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 9.87093023255814e-08,
+      "loss": 0.0004,
+      "step": 17361
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 9.869767441860465e-08,
+      "loss": 0.0026,
+      "step": 17362
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 9.868604651162791e-08,
+      "loss": 0.0005,
+      "step": 17363
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 9.867441860465116e-08,
+      "loss": 0.0008,
+      "step": 17364
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 9.866279069767442e-08,
+      "loss": 0.0108,
+      "step": 17365
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 9.865116279069767e-08,
+      "loss": 3.6239,
+      "step": 17366
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 9.863953488372093e-08,
+      "loss": 0.0006,
+      "step": 17367
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 9.862790697674418e-08,
+      "loss": 0.0642,
+      "step": 17368
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 9.861627906976744e-08,
+      "loss": 0.0005,
+      "step": 17369
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 9.860465116279069e-08,
+      "loss": 0.0004,
+      "step": 17370
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 9.859302325581396e-08,
+      "loss": 0.0006,
+      "step": 17371
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 9.858139534883721e-08,
+      "loss": 0.0005,
+      "step": 17372
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 9.856976744186047e-08,
+      "loss": 0.0007,
+      "step": 17373
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 9.855813953488372e-08,
+      "loss": 0.3179,
+      "step": 17374
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 9.854651162790697e-08,
+      "loss": 0.0007,
+      "step": 17375
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 9.853488372093023e-08,
+      "loss": 0.0005,
+      "step": 17376
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 9.852325581395348e-08,
+      "loss": 0.0005,
+      "step": 17377
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 9.851162790697674e-08,
+      "loss": 0.0013,
+      "step": 17378
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.849999999999999e-08,
+      "loss": 0.0007,
+      "step": 17379
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.848837209302325e-08,
+      "loss": 0.0005,
+      "step": 17380
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.84767441860465e-08,
+      "loss": 0.0005,
+      "step": 17381
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.846511627906976e-08,
+      "loss": 0.0005,
+      "step": 17382
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.845348837209301e-08,
+      "loss": 0.0038,
+      "step": 17383
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.844186046511628e-08,
+      "loss": 0.001,
+      "step": 17384
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 9.843023255813952e-08,
+      "loss": 0.0004,
+      "step": 17385
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 9.84186046511628e-08,
+      "loss": 0.0004,
+      "step": 17386
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 9.840697674418604e-08,
+      "loss": 0.0005,
+      "step": 17387
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 9.83953488372093e-08,
+      "loss": 0.0006,
+      "step": 17388
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 9.838372093023255e-08,
+      "loss": 0.0005,
+      "step": 17389
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 9.837209302325581e-08,
+      "loss": 0.0006,
+      "step": 17390
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 9.836046511627906e-08,
+      "loss": 0.0008,
+      "step": 17391
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 9.834883720930232e-08,
+      "loss": 0.0007,
+      "step": 17392
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 9.833720930232557e-08,
+      "loss": 0.0013,
+      "step": 17393
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 9.832558139534883e-08,
+      "loss": 0.0006,
+      "step": 17394
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 9.831395348837208e-08,
+      "loss": 0.0006,
+      "step": 17395
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 9.830232558139535e-08,
+      "loss": 0.0006,
+      "step": 17396
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 9.82906976744186e-08,
+      "loss": 0.0005,
+      "step": 17397
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 9.827906976744187e-08,
+      "loss": 0.0005,
+      "step": 17398
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 9.826744186046512e-08,
+      "loss": 0.3637,
+      "step": 17399
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 9.825581395348838e-08,
+      "loss": 0.0127,
+      "step": 17400
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 9.824418604651163e-08,
+      "loss": 0.0006,
+      "step": 17401
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 9.823255813953488e-08,
+      "loss": 0.0004,
+      "step": 17402
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 9.822093023255814e-08,
+      "loss": 0.0005,
+      "step": 17403
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 9.820930232558139e-08,
+      "loss": 0.0007,
+      "step": 17404
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 9.819767441860465e-08,
+      "loss": 0.0005,
+      "step": 17405
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 9.81860465116279e-08,
+      "loss": 0.0004,
+      "step": 17406
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 9.817441860465116e-08,
+      "loss": 0.0007,
+      "step": 17407
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 9.81627906976744e-08,
+      "loss": 0.0005,
+      "step": 17408
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 9.815116279069767e-08,
+      "loss": 0.0006,
+      "step": 17409
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 9.813953488372092e-08,
+      "loss": 0.0004,
+      "step": 17410
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 9.812790697674418e-08,
+      "loss": 0.0007,
+      "step": 17411
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 9.811627906976743e-08,
+      "loss": 0.0005,
+      "step": 17412
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 9.81046511627907e-08,
+      "loss": 0.0005,
+      "step": 17413
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 9.809302325581395e-08,
+      "loss": 0.0005,
+      "step": 17414
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 9.808139534883721e-08,
+      "loss": 0.0004,
+      "step": 17415
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 9.806976744186046e-08,
+      "loss": 0.0008,
+      "step": 17416
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 9.805813953488372e-08,
+      "loss": 0.0009,
+      "step": 17417
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 9.804651162790697e-08,
+      "loss": 0.0004,
+      "step": 17418
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 9.803488372093023e-08,
+      "loss": 0.0006,
+      "step": 17419
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 9.802325581395348e-08,
+      "loss": 0.0006,
+      "step": 17420
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 9.801162790697674e-08,
+      "loss": 0.0008,
+      "step": 17421
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 9.799999999999999e-08,
+      "loss": 0.0547,
+      "step": 17422
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 9.798837209302325e-08,
+      "loss": 0.0007,
+      "step": 17423
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 9.79767441860465e-08,
+      "loss": 0.0006,
+      "step": 17424
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 9.796511627906978e-08,
+      "loss": 0.0005,
+      "step": 17425
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 9.795348837209302e-08,
+      "loss": 0.0005,
+      "step": 17426
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 9.794186046511629e-08,
+      "loss": 0.0004,
+      "step": 17427
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 9.793023255813953e-08,
+      "loss": 0.0006,
+      "step": 17428
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 9.791860465116278e-08,
+      "loss": 0.0005,
+      "step": 17429
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 9.790697674418604e-08,
+      "loss": 0.0005,
+      "step": 17430
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 9.789534883720929e-08,
+      "loss": 0.0006,
+      "step": 17431
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 9.788372093023255e-08,
+      "loss": 0.0004,
+      "step": 17432
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 9.78720930232558e-08,
+      "loss": 0.0008,
+      "step": 17433
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 9.786046511627907e-08,
+      "loss": 0.0006,
+      "step": 17434
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 9.784883720930231e-08,
+      "loss": 0.0009,
+      "step": 17435
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 9.783720930232558e-08,
+      "loss": 0.0004,
+      "step": 17436
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 9.782558139534882e-08,
+      "loss": 0.0005,
+      "step": 17437
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 9.781395348837209e-08,
+      "loss": 0.0006,
+      "step": 17438
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 9.780232558139533e-08,
+      "loss": 0.0005,
+      "step": 17439
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 9.779069767441861e-08,
+      "loss": 0.0005,
+      "step": 17440
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 9.777906976744186e-08,
+      "loss": 0.0004,
+      "step": 17441
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 9.776744186046512e-08,
+      "loss": 0.0004,
+      "step": 17442
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 9.775581395348837e-08,
+      "loss": 0.0004,
+      "step": 17443
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 9.774418604651163e-08,
+      "loss": 0.0004,
+      "step": 17444
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 9.773255813953488e-08,
+      "loss": 0.6551,
+      "step": 17445
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 9.772093023255814e-08,
+      "loss": 0.0005,
+      "step": 17446
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 9.770930232558139e-08,
+      "loss": 0.0003,
+      "step": 17447
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 9.769767441860465e-08,
+      "loss": 0.0006,
+      "step": 17448
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 9.76860465116279e-08,
+      "loss": 0.0005,
+      "step": 17449
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 9.767441860465116e-08,
+      "loss": 0.0004,
+      "step": 17450
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 9.766279069767441e-08,
+      "loss": 2.4528,
+      "step": 17451
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 9.765116279069768e-08,
+      "loss": 0.0003,
+      "step": 17452
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 9.763953488372093e-08,
+      "loss": 0.0005,
+      "step": 17453
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 9.76279069767442e-08,
+      "loss": 0.0009,
+      "step": 17454
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 9.761627906976744e-08,
+      "loss": 0.0005,
+      "step": 17455
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 9.760465116279069e-08,
+      "loss": 0.0005,
+      "step": 17456
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 9.759302325581395e-08,
+      "loss": 0.001,
+      "step": 17457
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 9.75813953488372e-08,
+      "loss": 0.0004,
+      "step": 17458
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 9.756976744186046e-08,
+      "loss": 0.0006,
+      "step": 17459
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 9.755813953488371e-08,
+      "loss": 0.0005,
+      "step": 17460
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 9.754651162790697e-08,
+      "loss": 0.0004,
+      "step": 17461
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 9.753488372093022e-08,
+      "loss": 0.0007,
+      "step": 17462
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 9.752325581395348e-08,
+      "loss": 0.0006,
+      "step": 17463
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 9.751162790697673e-08,
+      "loss": 0.0005,
+      "step": 17464
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 9.749999999999999e-08,
+      "loss": 0.0004,
+      "step": 17465
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 9.748837209302324e-08,
+      "loss": 0.0007,
+      "step": 17466
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 9.747674418604652e-08,
+      "loss": 0.0005,
+      "step": 17467
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 9.746511627906976e-08,
+      "loss": 0.0006,
+      "step": 17468
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 9.745348837209303e-08,
+      "loss": 0.0005,
+      "step": 17469
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 9.744186046511628e-08,
+      "loss": 0.0007,
+      "step": 17470
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 9.743023255813954e-08,
+      "loss": 0.0009,
+      "step": 17471
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 9.741860465116279e-08,
+      "loss": 1.6878,
+      "step": 17472
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 9.740697674418605e-08,
+      "loss": 0.0004,
+      "step": 17473
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 9.73953488372093e-08,
+      "loss": 0.0005,
+      "step": 17474
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 9.738372093023256e-08,
+      "loss": 0.0005,
+      "step": 17475
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 9.73720930232558e-08,
+      "loss": 0.0004,
+      "step": 17476
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 9.736046511627907e-08,
+      "loss": 0.0006,
+      "step": 17477
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 9.734883720930232e-08,
+      "loss": 0.0007,
+      "step": 17478
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 9.733720930232559e-08,
+      "loss": 0.0005,
+      "step": 17479
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 9.732558139534883e-08,
+      "loss": 0.0006,
+      "step": 17480
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 9.731395348837207e-08,
+      "loss": 0.0012,
+      "step": 17481
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 9.730232558139535e-08,
+      "loss": 0.0006,
+      "step": 17482
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 9.72906976744186e-08,
+      "loss": 0.0004,
+      "step": 17483
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 9.727906976744186e-08,
+      "loss": 0.0006,
+      "step": 17484
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 9.726744186046511e-08,
+      "loss": 0.0006,
+      "step": 17485
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 9.725581395348837e-08,
+      "loss": 0.0006,
+      "step": 17486
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 9.724418604651162e-08,
+      "loss": 0.0007,
+      "step": 17487
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 9.723255813953488e-08,
+      "loss": 0.0006,
+      "step": 17488
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 9.722093023255813e-08,
+      "loss": 0.0004,
+      "step": 17489
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 9.720930232558139e-08,
+      "loss": 0.0005,
+      "step": 17490
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 9.719767441860464e-08,
+      "loss": 0.0005,
+      "step": 17491
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 9.71860465116279e-08,
+      "loss": 0.0005,
+      "step": 17492
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 9.717441860465115e-08,
+      "loss": 0.0004,
+      "step": 17493
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 9.716279069767442e-08,
+      "loss": 0.0005,
+      "step": 17494
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 9.715116279069767e-08,
+      "loss": 0.0006,
+      "step": 17495
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 9.713953488372093e-08,
+      "loss": 0.0004,
+      "step": 17496
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 9.712790697674418e-08,
+      "loss": 1.643,
+      "step": 17497
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 9.711627906976744e-08,
+      "loss": 0.0004,
+      "step": 17498
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 9.710465116279069e-08,
+      "loss": 0.0005,
+      "step": 17499
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 9.709302325581395e-08,
+      "loss": 0.0004,
+      "step": 17500
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 9.70813953488372e-08,
+      "loss": 0.0004,
+      "step": 17501
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 9.706976744186046e-08,
+      "loss": 2.5557,
+      "step": 17502
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 9.705813953488371e-08,
+      "loss": 0.0006,
+      "step": 17503
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 9.704651162790697e-08,
+      "loss": 0.0005,
+      "step": 17504
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 9.703488372093022e-08,
+      "loss": 0.0008,
+      "step": 17505
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 9.70232558139535e-08,
+      "loss": 0.0004,
+      "step": 17506
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 9.701162790697673e-08,
+      "loss": 0.0006,
+      "step": 17507
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 9.699999999999998e-08,
+      "loss": 0.0004,
+      "step": 17508
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 9.698837209302326e-08,
+      "loss": 0.0007,
+      "step": 17509
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 9.69767441860465e-08,
+      "loss": 0.0005,
+      "step": 17510
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 9.696511627906977e-08,
+      "loss": 0.0007,
+      "step": 17511
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 9.695348837209302e-08,
+      "loss": 0.0004,
+      "step": 17512
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 9.694186046511628e-08,
+      "loss": 0.0004,
+      "step": 17513
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 9.693023255813953e-08,
+      "loss": 0.0006,
+      "step": 17514
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 9.691860465116279e-08,
+      "loss": 0.0003,
+      "step": 17515
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 9.690697674418604e-08,
+      "loss": 0.0004,
+      "step": 17516
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 9.68953488372093e-08,
+      "loss": 0.0005,
+      "step": 17517
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 9.688372093023255e-08,
+      "loss": 0.0006,
+      "step": 17518
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 9.687209302325581e-08,
+      "loss": 0.0007,
+      "step": 17519
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 9.686046511627906e-08,
+      "loss": 0.0005,
+      "step": 17520
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 9.684883720930233e-08,
+      "loss": 0.0205,
+      "step": 17521
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 9.683720930232558e-08,
+      "loss": 0.0005,
+      "step": 17522
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 9.682558139534884e-08,
+      "loss": 0.0009,
+      "step": 17523
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 9.681395348837209e-08,
+      "loss": 0.0009,
+      "step": 17524
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 9.680232558139535e-08,
+      "loss": 0.0005,
+      "step": 17525
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 9.67906976744186e-08,
+      "loss": 0.0004,
+      "step": 17526
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 9.677906976744186e-08,
+      "loss": 0.0005,
+      "step": 17527
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 9.676744186046511e-08,
+      "loss": 0.0004,
+      "step": 17528
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 9.675581395348837e-08,
+      "loss": 0.0011,
+      "step": 17529
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 9.674418604651162e-08,
+      "loss": 0.0003,
+      "step": 17530
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 9.673255813953488e-08,
+      "loss": 0.0004,
+      "step": 17531
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 9.672093023255813e-08,
+      "loss": 0.0004,
+      "step": 17532
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 9.670930232558139e-08,
+      "loss": 0.0006,
+      "step": 17533
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 9.669767441860464e-08,
+      "loss": 0.0004,
+      "step": 17534
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 9.668604651162789e-08,
+      "loss": 0.0004,
+      "step": 17535
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 9.667441860465116e-08,
+      "loss": 0.0005,
+      "step": 17536
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 9.666279069767441e-08,
+      "loss": 0.013,
+      "step": 17537
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 9.665116279069767e-08,
+      "loss": 0.0005,
+      "step": 17538
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 9.663953488372092e-08,
+      "loss": 0.0005,
+      "step": 17539
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 9.662790697674418e-08,
+      "loss": 0.0016,
+      "step": 17540
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 9.661627906976743e-08,
+      "loss": 0.0006,
+      "step": 17541
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 9.66046511627907e-08,
+      "loss": 0.0008,
+      "step": 17542
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 9.659302325581394e-08,
+      "loss": 0.0009,
+      "step": 17543
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 9.65813953488372e-08,
+      "loss": 0.0004,
+      "step": 17544
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 9.656976744186045e-08,
+      "loss": 0.0009,
+      "step": 17545
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 9.655813953488372e-08,
+      "loss": 0.0005,
+      "step": 17546
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 9.654651162790696e-08,
+      "loss": 0.0005,
+      "step": 17547
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 9.653488372093024e-08,
+      "loss": 0.0006,
+      "step": 17548
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 9.652325581395347e-08,
+      "loss": 0.0009,
+      "step": 17549
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 9.651162790697675e-08,
+      "loss": 0.0007,
+      "step": 17550
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 9.65e-08,
+      "loss": 0.0006,
+      "step": 17551
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 9.648837209302326e-08,
+      "loss": 0.0009,
+      "step": 17552
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 9.647674418604651e-08,
+      "loss": 0.0005,
+      "step": 17553
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 9.646511627906977e-08,
+      "loss": 0.0004,
+      "step": 17554
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 9.645348837209302e-08,
+      "loss": 0.0004,
+      "step": 17555
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 9.644186046511628e-08,
+      "loss": 0.0004,
+      "step": 17556
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 9.643023255813953e-08,
+      "loss": 0.0007,
+      "step": 17557
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 9.641860465116279e-08,
+      "loss": 0.0005,
+      "step": 17558
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 9.640697674418604e-08,
+      "loss": 0.0008,
+      "step": 17559
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 9.63953488372093e-08,
+      "loss": 0.0006,
+      "step": 17560
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 9.638372093023255e-08,
+      "loss": 0.0005,
+      "step": 17561
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 9.63720930232558e-08,
+      "loss": 0.0007,
+      "step": 17562
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 9.636046511627907e-08,
+      "loss": 0.5865,
+      "step": 17563
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 9.634883720930232e-08,
+      "loss": 0.0004,
+      "step": 17564
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 9.633720930232558e-08,
+      "loss": 0.0025,
+      "step": 17565
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 9.632558139534883e-08,
+      "loss": 0.0005,
+      "step": 17566
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 9.631395348837209e-08,
+      "loss": 0.0006,
+      "step": 17567
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 9.630232558139534e-08,
+      "loss": 0.0005,
+      "step": 17568
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 9.62906976744186e-08,
+      "loss": 0.0004,
+      "step": 17569
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 9.627906976744185e-08,
+      "loss": 0.0006,
+      "step": 17570
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 9.626744186046511e-08,
+      "loss": 0.0005,
+      "step": 17571
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 9.625581395348836e-08,
+      "loss": 0.0009,
+      "step": 17572
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 9.624418604651162e-08,
+      "loss": 0.0057,
+      "step": 17573
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 9.623255813953487e-08,
+      "loss": 0.0005,
+      "step": 17574
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 9.622093023255815e-08,
+      "loss": 0.0006,
+      "step": 17575
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 9.620930232558138e-08,
+      "loss": 0.0006,
+      "step": 17576
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 9.619767441860466e-08,
+      "loss": 0.0005,
+      "step": 17577
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 9.61860465116279e-08,
+      "loss": 0.0005,
+      "step": 17578
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.9893617021276596,
+      "eval_loss": 0.14592407643795013,
+      "eval_runtime": 1.5585,
+      "eval_samples_per_second": 663.452,
+      "eval_steps_per_second": 10.908,
+      "step": 17578
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 25850,
+  "num_train_epochs": 50,
+  "save_steps": 500,
+  "total_flos": 2153690154002016.0,
+  "trial_name": null,
+  "trial_params": null
+}