{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.9989186941152006,
  "global_step": 3004,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.197802197802198e-07,
      "loss": 8.1449,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.395604395604396e-07,
      "loss": 30.168,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.593406593406594e-07,
      "loss": 7.6897,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.791208791208792e-07,
      "loss": 7.4282,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.098901098901099e-06,
      "loss": 7.4877,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3186813186813187e-06,
      "loss": 8.2692,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.5384615384615387e-06,
      "loss": 7.5981,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7582417582417585e-06,
      "loss": 7.3395,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9780219780219782e-06,
      "loss": 7.3969,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.197802197802198e-06,
      "loss": 7.2137,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4175824175824177e-06,
      "loss": 7.6093,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6373626373626375e-06,
      "loss": 7.2331,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8571428571428573e-06,
      "loss": 6.8879,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.0769230769230774e-06,
      "loss": 7.9843,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.2967032967032968e-06,
      "loss": 7.1424,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.516483516483517e-06,
      "loss": 8.0438,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.7362637362637367e-06,
      "loss": 4.2202,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.9560439560439565e-06,
      "loss": 4.8004,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.175824175824177e-06,
      "loss": 7.3806,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.395604395604396e-06,
      "loss": 7.3383,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.615384615384616e-06,
      "loss": 7.4969,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.8351648351648355e-06,
      "loss": 8.0594,
      "step": 22
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.054945054945055e-06,
      "loss": 7.2354,
      "step": 23
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.274725274725275e-06,
      "loss": 7.1653,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.494505494505495e-06,
      "loss": 7.2533,
      "step": 25
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.7142857142857145e-06,
      "loss": 6.8458,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.934065934065935e-06,
      "loss": 7.4467,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.153846153846155e-06,
      "loss": 7.3355,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.373626373626373e-06,
      "loss": 6.8368,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.5934065934065935e-06,
      "loss": 6.583,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 6.813186813186814e-06,
      "loss": 6.5583,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.032967032967034e-06,
      "loss": 6.8466,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.252747252747253e-06,
      "loss": 7.2066,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.472527472527473e-06,
      "loss": 6.9133,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.692307692307694e-06,
      "loss": 6.6295,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.912087912087913e-06,
      "loss": 7.1095,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.131868131868132e-06,
      "loss": 7.0928,
      "step": 37
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.351648351648353e-06,
      "loss": 6.475,
      "step": 38
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.571428571428571e-06,
      "loss": 7.1967,
      "step": 39
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.791208791208792e-06,
      "loss": 6.7154,
      "step": 40
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.010989010989011e-06,
      "loss": 6.621,
      "step": 41
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.230769230769232e-06,
      "loss": 6.9752,
      "step": 42
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.450549450549452e-06,
      "loss": 7.0558,
      "step": 43
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.670329670329671e-06,
      "loss": 6.7828,
      "step": 44
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.890109890109892e-06,
      "loss": 6.6859,
      "step": 45
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.010989010989011e-05,
      "loss": 6.8216,
      "step": 46
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.0329670329670332e-05,
      "loss": 6.2387,
      "step": 47
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.054945054945055e-05,
      "loss": 7.0478,
      "step": 48
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.076923076923077e-05,
      "loss": 6.0849,
      "step": 49
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.098901098901099e-05,
      "loss": 6.7734,
      "step": 50
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.120879120879121e-05,
      "loss": 6.3915,
      "step": 51
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.1428571428571429e-05,
      "loss": 6.4094,
      "step": 52
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.164835164835165e-05,
      "loss": 5.8803,
      "step": 53
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.186813186813187e-05,
      "loss": 6.525,
      "step": 54
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.2087912087912089e-05,
      "loss": 6.5583,
      "step": 55
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.230769230769231e-05,
      "loss": 6.5734,
      "step": 56
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.2527472527472529e-05,
      "loss": 6.5075,
      "step": 57
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.2747252747252747e-05,
      "loss": 5.939,
      "step": 58
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.296703296703297e-05,
      "loss": 6.2285,
      "step": 59
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.3186813186813187e-05,
      "loss": 6.5095,
      "step": 60
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.3406593406593406e-05,
      "loss": 6.5188,
      "step": 61
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.3626373626373627e-05,
      "loss": 6.1523,
      "step": 62
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.3846153846153847e-05,
      "loss": 6.1693,
      "step": 63
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.4065934065934068e-05,
      "loss": 6.6994,
      "step": 64
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.4285714285714287e-05,
      "loss": 6.3958,
      "step": 65
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.4505494505494506e-05,
      "loss": 6.2207,
      "step": 66
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.4725274725274727e-05,
      "loss": 6.3189,
      "step": 67
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.4945054945054947e-05,
      "loss": 5.9857,
      "step": 68
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.5164835164835166e-05,
      "loss": 6.1337,
      "step": 69
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.5384615384615387e-05,
      "loss": 6.4188,
      "step": 70
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.5604395604395605e-05,
      "loss": 5.8999,
      "step": 71
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.5824175824175826e-05,
      "loss": 6.4881,
      "step": 72
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.6043956043956047e-05,
      "loss": 6.4384,
      "step": 73
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.6263736263736265e-05,
      "loss": 5.6516,
      "step": 74
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.6483516483516486e-05,
      "loss": 5.7171,
      "step": 75
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.6703296703296707e-05,
      "loss": 6.4169,
      "step": 76
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.6923076923076924e-05,
      "loss": 6.2269,
      "step": 77
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.7142857142857142e-05,
      "loss": 6.2741,
      "step": 78
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.7362637362637363e-05,
      "loss": 5.7108,
      "step": 79
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.7582417582417584e-05,
      "loss": 5.5062,
      "step": 80
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.78021978021978e-05,
      "loss": 5.8336,
      "step": 81
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.8021978021978023e-05,
      "loss": 6.1578,
      "step": 82
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.8241758241758244e-05,
      "loss": 6.2734,
      "step": 83
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.8461538461538465e-05,
      "loss": 5.841,
      "step": 84
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.8681318681318682e-05,
      "loss": 5.1361,
      "step": 85
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.8901098901098903e-05,
      "loss": 5.7608,
      "step": 86
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9120879120879124e-05,
      "loss": 5.6757,
      "step": 87
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9340659340659342e-05,
      "loss": 5.1882,
      "step": 88
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9560439560439563e-05,
      "loss": 5.6622,
      "step": 89
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9780219780219784e-05,
      "loss": 5.5154,
      "step": 90
    },
    {
      "epoch": 0.06,
      "learning_rate": 2e-05,
      "loss": 5.3391,
      "step": 91
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999999418447753e-05,
      "loss": 5.7577,
      "step": 92
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999997673791687e-05,
      "loss": 5.3209,
      "step": 93
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999994766033832e-05,
      "loss": 5.1318,
      "step": 94
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9999906951775704e-05,
      "loss": 5.3308,
      "step": 95
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999985461227636e-05,
      "loss": 5.4434,
      "step": 96
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999979064190117e-05,
      "loss": 5.0025,
      "step": 97
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999971504072454e-05,
      "loss": 5.1902,
      "step": 98
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.99996278088344e-05,
      "loss": 5.6053,
      "step": 99
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9999528946332206e-05,
      "loss": 4.9806,
      "step": 100
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999941845333295e-05,
      "loss": 4.8008,
      "step": 101
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9999296329965146e-05,
      "loss": 5.0867,
      "step": 102
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999916257637084e-05,
      "loss": 5.0519,
      "step": 103
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9999017192705595e-05,
      "loss": 4.9513,
      "step": 104
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999886017913851e-05,
      "loss": 4.8193,
      "step": 105
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999869153585221e-05,
      "loss": 4.7996,
      "step": 106
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999851126304284e-05,
      "loss": 4.8903,
      "step": 107
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999831936092008e-05,
      "loss": 5.1152,
      "step": 108
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9998115829707134e-05,
      "loss": 5.1364,
      "step": 109
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9997900669640725e-05,
      "loss": 4.8692,
      "step": 110
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999767388097111e-05,
      "loss": 5.3726,
      "step": 111
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.999743546396206e-05,
      "loss": 4.9651,
      "step": 112
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9997185418890896e-05,
      "loss": 4.7553,
      "step": 113
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9996923746048434e-05,
      "loss": 4.7369,
      "step": 114
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9996650445739024e-05,
      "loss": 4.9826,
      "step": 115
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9996365518280553e-05,
      "loss": 4.648,
      "step": 116
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9996068964004415e-05,
      "loss": 4.5032,
      "step": 117
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9995760783255537e-05,
      "loss": 4.8752,
      "step": 118
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.999544097639236e-05,
      "loss": 4.8318,
      "step": 119
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9995109543786857e-05,
      "loss": 4.3524,
      "step": 120
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.999476648582452e-05,
      "loss": 4.1178,
      "step": 121
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9994411802904357e-05,
      "loss": 4.5678,
      "step": 122
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9994045495438904e-05,
      "loss": 4.6813,
      "step": 123
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9993667563854215e-05,
      "loss": 4.5295,
      "step": 124
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9993278008589866e-05,
      "loss": 4.4025,
      "step": 125
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9992876830098943e-05,
      "loss": 4.5847,
      "step": 126
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9992464028848065e-05,
      "loss": 4.2469,
      "step": 127
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9992039605317362e-05,
      "loss": 4.358,
      "step": 128
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.999160356000048e-05,
      "loss": 3.5819,
      "step": 129
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9991155893404592e-05,
      "loss": 4.381,
      "step": 130
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9990696606050374e-05,
      "loss": 4.271,
      "step": 131
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.999022569847203e-05,
      "loss": 4.1418,
      "step": 132
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.998974317121727e-05,
      "loss": 4.0589,
      "step": 133
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9989249024847324e-05,
      "loss": 3.7793,
      "step": 134
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9988743259936944e-05,
      "loss": 3.6107,
      "step": 135
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9988225877074378e-05,
      "loss": 3.958,
      "step": 136
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9987696876861396e-05,
      "loss": 3.9307,
      "step": 137
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.998715625991329e-05,
      "loss": 3.8979,
      "step": 138
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9986604026858846e-05,
      "loss": 3.9742,
      "step": 139
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9986040178340372e-05,
      "loss": 3.6317,
      "step": 140
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.998546471501368e-05,
      "loss": 3.6351,
      "step": 141
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9984877637548095e-05,
      "loss": 3.6098,
      "step": 142
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9984278946626455e-05,
      "loss": 3.6928,
      "step": 143
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9983668642945087e-05,
      "loss": 3.672,
      "step": 144
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9983046727213856e-05,
      "loss": 3.5898,
      "step": 145
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.99824132001561e-05,
      "loss": 3.5223,
      "step": 146
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.998176806250868e-05,
      "loss": 3.2073,
      "step": 147
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9981111315021966e-05,
      "loss": 3.359,
      "step": 148
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9980442958459817e-05,
      "loss": 3.3423,
      "step": 149
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9979762993599602e-05,
      "loss": 3.7208,
      "step": 150
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9979071421232192e-05,
      "loss": 3.2851,
      "step": 151
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9978368242161964e-05,
      "loss": 3.3797,
      "step": 152
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9977653457206777e-05,
      "loss": 3.414,
      "step": 153
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.997692706719801e-05,
      "loss": 3.2115,
      "step": 154
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9976189072980525e-05,
      "loss": 3.115,
      "step": 155
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.997543947541269e-05,
      "loss": 3.122,
      "step": 156
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9974678275366363e-05,
      "loss": 3.021,
      "step": 157
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9973905473726902e-05,
      "loss": 3.0613,
      "step": 158
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9973121071393152e-05,
      "loss": 2.9357,
      "step": 159
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.997232506927746e-05,
      "loss": 2.6777,
      "step": 160
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.997151746830565e-05,
      "loss": 2.74,
      "step": 161
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.997069826941706e-05,
      "loss": 2.7392,
      "step": 162
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9969867473564496e-05,
      "loss": 2.7951,
      "step": 163
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.996902508171426e-05,
      "loss": 2.6914,
      "step": 164
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9968171094846145e-05,
      "loss": 2.9255,
      "step": 165
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9967305513953424e-05,
      "loss": 2.3729,
      "step": 166
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.996642834004286e-05,
      "loss": 2.6643,
      "step": 167
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9965539574134694e-05,
      "loss": 2.6033,
      "step": 168
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.996463921726266e-05,
      "loss": 2.4345,
      "step": 169
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9963727270473958e-05,
      "loss": 2.3691,
      "step": 170
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9962803734829287e-05,
      "loss": 2.3675,
      "step": 171
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9961868611402813e-05,
      "loss": 2.2931,
      "step": 172
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9960921901282183e-05,
      "loss": 2.2009,
      "step": 173
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.995996360556851e-05,
      "loss": 2.1677,
      "step": 174
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9958993725376404e-05,
      "loss": 2.1917,
      "step": 175
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9958012261833937e-05,
      "loss": 2.1496,
      "step": 176
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9957019216082646e-05,
      "loss": 2.0209,
      "step": 177
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9956014589277547e-05,
      "loss": 1.8994,
      "step": 178
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9954998382587133e-05,
      "loss": 1.8054,
      "step": 179
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9953970597193356e-05,
      "loss": 1.8897,
      "step": 180
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9952931234291632e-05,
      "loss": 1.7484,
      "step": 181
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9951880295090856e-05,
      "loss": 1.5704,
      "step": 182
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9950817780813377e-05,
      "loss": 1.5952,
      "step": 183
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9949743692695014e-05,
      "loss": 1.5658,
      "step": 184
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9948658031985038e-05,
      "loss": 1.491,
      "step": 185
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9947560799946185e-05,
      "loss": 1.4275,
      "step": 186
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.994645199785466e-05,
      "loss": 1.3664,
      "step": 187
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9945331627000105e-05,
      "loss": 1.3207,
      "step": 188
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9944199688685632e-05,
      "loss": 1.3096,
      "step": 189
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9943056184227808e-05,
      "loss": 1.2967,
      "step": 190
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.994190111495664e-05,
      "loss": 1.2631,
      "step": 191
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9940734482215603e-05,
      "loss": 1.1747,
      "step": 192
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9939556287361607e-05,
      "loss": 1.1084,
      "step": 193
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9938366531765017e-05,
      "loss": 1.055,
      "step": 194
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9937165216809644e-05,
      "loss": 1.0318,
      "step": 195
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.993595234389274e-05,
      "loss": 0.9725,
      "step": 196
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9934727914425007e-05,
      "loss": 0.9484,
      "step": 197
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9933491929830584e-05,
      "loss": 0.8603,
      "step": 198
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9932244391547044e-05,
      "loss": 0.8481,
      "step": 199
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9930985301025415e-05,
      "loss": 0.7886,
      "step": 200
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.992971465973014e-05,
      "loss": 0.7722,
      "step": 201
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9928432469139116e-05,
      "loss": 0.7543,
      "step": 202
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.992713873074366e-05,
      "loss": 0.719,
      "step": 203
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9925833446048524e-05,
      "loss": 0.6721,
      "step": 204
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9924516616571897e-05,
      "loss": 0.6013,
      "step": 205
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9923188243845383e-05,
      "loss": 0.5811,
      "step": 206
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.992184832941402e-05,
      "loss": 0.5928,
      "step": 207
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9920496874836268e-05,
      "loss": 0.5274,
      "step": 208
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9919133881684016e-05,
      "loss": 0.4976,
      "step": 209
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9917759351542555e-05,
      "loss": 0.4778,
      "step": 210
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9916373286010616e-05,
      "loss": 0.495,
      "step": 211
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9914975686700337e-05,
      "loss": 0.4829,
      "step": 212
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9913566555237275e-05,
      "loss": 0.4754,
      "step": 213
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.991214589326039e-05,
      "loss": 0.4658,
      "step": 214
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9910713702422063e-05,
      "loss": 0.4087,
      "step": 215
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9909269984388082e-05,
      "loss": 0.3923,
      "step": 216
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9907814740837646e-05,
      "loss": 0.3899,
      "step": 217
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.990634797346335e-05,
      "loss": 0.3819,
      "step": 218
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.99048696839712e-05,
      "loss": 0.3989,
      "step": 219
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.99033798740806e-05,
      "loss": 0.4316,
      "step": 220
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9901878545524353e-05,
      "loss": 0.3251,
      "step": 221
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9900365700048666e-05,
      "loss": 0.3874,
      "step": 222
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9898841339413134e-05,
      "loss": 0.3823,
      "step": 223
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9897305465390744e-05,
      "loss": 0.3922,
      "step": 224
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9895758079767883e-05,
      "loss": 0.3777,
      "step": 225
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.989419918434432e-05,
      "loss": 0.3212,
      "step": 226
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9892628780933213e-05,
      "loss": 0.3593,
      "step": 227
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9891046871361108e-05,
      "loss": 0.3155,
      "step": 228
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9889453457467927e-05,
      "loss": 0.3956,
      "step": 229
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9887848541106982e-05,
      "loss": 0.3689,
      "step": 230
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9886232124144954e-05,
      "loss": 0.4261,
      "step": 231
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9884604208461903e-05,
      "loss": 0.3183,
      "step": 232
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.988296479595127e-05,
      "loss": 0.4037,
      "step": 233
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9881313888519863e-05,
      "loss": 0.3426,
      "step": 234
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.987965148808786e-05,
      "loss": 0.3762,
      "step": 235
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.98779775965888e-05,
      "loss": 0.3125,
      "step": 236
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9876292215969603e-05,
      "loss": 10.0582,
      "step": 237
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9874595348190532e-05,
      "loss": 0.3434,
      "step": 238
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.987288699522523e-05,
      "loss": 0.362,
      "step": 239
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9871167159060692e-05,
      "loss": 0.3967,
      "step": 240
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.986943584169726e-05,
      "loss": 0.4098,
      "step": 241
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9867693045148635e-05,
      "loss": 0.399,
      "step": 242
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.986593877144188e-05,
      "loss": 0.3312,
      "step": 243
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9864173022617397e-05,
      "loss": 0.3661,
      "step": 244
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9862395800728933e-05,
      "loss": 0.369,
      "step": 245
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9860607107843584e-05,
      "loss": 0.3637,
      "step": 246
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9858806946041785e-05,
      "loss": 0.4275,
      "step": 247
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9856995317417316e-05,
      "loss": 0.3659,
      "step": 248
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9855172224077287e-05,
      "loss": 0.3308,
      "step": 249
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.985333766814215e-05,
      "loss": 0.3449,
      "step": 250
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9851491651745683e-05,
      "loss": 0.3778,
      "step": 251
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9849634177034992e-05,
      "loss": 0.3303,
      "step": 252
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9847765246170523e-05,
      "loss": 0.3923,
      "step": 253
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9845884861326027e-05,
      "loss": 0.366,
      "step": 254
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9843993024688597e-05,
      "loss": 0.3405,
      "step": 255
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9842089738458635e-05,
      "loss": 0.3545,
      "step": 256
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.984017500484986e-05,
      "loss": 13.2311,
      "step": 257
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9838248826089306e-05,
      "loss": 0.4057,
      "step": 258
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9836311204417323e-05,
      "loss": 0.3276,
      "step": 259
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9834362142087565e-05,
      "loss": 0.3549,
      "step": 260
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9832401641366998e-05,
      "loss": 0.3208,
      "step": 261
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9830429704535888e-05,
      "loss": 0.3734,
      "step": 262
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.98284463338878e-05,
      "loss": 0.2986,
      "step": 263
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9826451531729608e-05,
      "loss": 0.3702,
      "step": 264
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.982444530038147e-05,
      "loss": 0.3559,
      "step": 265
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9822427642176845e-05,
      "loss": 0.3239,
      "step": 266
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9820398559462482e-05,
      "loss": 0.3484,
      "step": 267
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9818358054598414e-05,
      "loss": 0.3544,
      "step": 268
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.981630612995796e-05,
      "loss": 0.3278,
      "step": 269
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9814242787927722e-05,
      "loss": 0.3426,
      "step": 270
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9812168030907587e-05,
      "loss": 0.3605,
      "step": 271
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9810081861310714e-05,
      "loss": 0.3708,
      "step": 272
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9807984281563534e-05,
      "loss": 0.3474,
      "step": 273
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9805875294105748e-05,
      "loss": 0.3152,
      "step": 274
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9803754901390334e-05,
      "loss": 0.3394,
      "step": 275
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.980162310588353e-05,
      "loss": 0.3593,
      "step": 276
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9799479910064835e-05,
      "loss": 0.3594,
      "step": 277
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.979732531642701e-05,
      "loss": 0.3879,
      "step": 278
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.979515932747607e-05,
      "loss": 0.3425,
      "step": 279
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9792981945731293e-05,
      "loss": 0.3583,
      "step": 280
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9790793173725195e-05,
      "loss": 0.3163,
      "step": 281
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9788593014003555e-05,
      "loss": 0.3162,
      "step": 282
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9786381469125378e-05,
      "loss": 0.3566,
      "step": 283
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9784158541662927e-05,
      "loss": 0.3747,
      "step": 284
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9781924234201703e-05,
      "loss": 0.3123,
      "step": 285
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.977967854934043e-05,
      "loss": 0.3988,
      "step": 286
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.977742148969108e-05,
      "loss": 0.2966,
      "step": 287
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.977515305787885e-05,
      "loss": 0.3602,
      "step": 288
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.977287325654216e-05,
      "loss": 0.3552,
      "step": 289
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.977058208833266e-05,
      "loss": 0.3234,
      "step": 290
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9768279555915215e-05,
      "loss": 0.3168,
      "step": 291
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9765965661967915e-05,
      "loss": 0.3372,
      "step": 292
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9763640409182055e-05,
      "loss": 0.3471,
      "step": 293
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.976130380026215e-05,
      "loss": 0.3581,
      "step": 294
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9758955837925922e-05,
      "loss": 0.4141,
      "step": 295
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9756596524904294e-05,
      "loss": 0.3484,
      "step": 296
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9754225863941395e-05,
      "loss": 0.3368,
      "step": 297
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.975184385779455e-05,
      "loss": 0.332,
      "step": 298
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.974945050923428e-05,
      "loss": 0.3152,
      "step": 299
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9747045821044305e-05,
      "loss": 0.3683,
      "step": 300
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9744629796021523e-05,
      "loss": 0.4401,
      "step": 301
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9742202436976025e-05,
      "loss": 0.3302,
      "step": 302
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9739763746731086e-05,
      "loss": 0.4017,
      "step": 303
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9737313728123152e-05,
      "loss": 0.3861,
      "step": 304
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9734852384001856e-05,
      "loss": 0.3118,
      "step": 305
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9732379717229997e-05,
      "loss": 0.355,
      "step": 306
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9729895730683545e-05,
      "loss": 0.3132,
      "step": 307
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.972740042725163e-05,
      "loss": 0.3231,
      "step": 308
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.972489380983656e-05,
      "loss": 0.3184,
      "step": 309
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.972237588135379e-05,
      "loss": 0.3729,
      "step": 310
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.971984664473193e-05,
      "loss": 0.3296,
      "step": 311
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.971730610291275e-05,
      "loss": 0.3443,
      "step": 312
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9714754258851162e-05,
      "loss": 0.3131,
      "step": 313
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9712191115515232e-05,
      "loss": 0.3416,
      "step": 314
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9709616675886165e-05,
      "loss": 0.3214,
      "step": 315
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9707030942958295e-05,
      "loss": 0.3432,
      "step": 316
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.970443391973911e-05,
      "loss": 0.3531,
      "step": 317
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9701825609249206e-05,
      "loss": 0.3153,
      "step": 318
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9699206014522335e-05,
      "loss": 0.308,
      "step": 319
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9696575138605353e-05,
      "loss": 0.3552,
      "step": 320
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.969393298455824e-05,
      "loss": 0.3246,
      "step": 321
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.96912795554541e-05,
      "loss": 0.3392,
      "step": 322
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9688614854379154e-05,
      "loss": 0.3554,
      "step": 323
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.968593888443272e-05,
      "loss": 0.3189,
      "step": 324
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9683251648727235e-05,
      "loss": 0.3887,
      "step": 325
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.968055315038823e-05,
      "loss": 0.3382,
      "step": 326
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9677843392554344e-05,
      "loss": 0.3592,
      "step": 327
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9675122378377313e-05,
      "loss": 0.3445,
      "step": 328
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9672390111021954e-05,
      "loss": 0.31,
      "step": 329
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9669646593666183e-05,
      "loss": 0.3043,
      "step": 330
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9666891829500993e-05,
      "loss": 0.4068,
      "step": 331
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.966412582173047e-05,
      "loss": 0.407,
      "step": 332
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9661348573571764e-05,
      "loss": 0.3056,
      "step": 333
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9658560088255106e-05,
      "loss": 0.2975,
      "step": 334
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9655760369023798e-05,
      "loss": 0.4063,
      "step": 335
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9652949419134203e-05,
      "loss": 0.3348,
      "step": 336
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.965012724185575e-05,
      "loss": 0.2832,
      "step": 337
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.964729384047093e-05,
      "loss": 0.2897,
      "step": 338
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.964444921827528e-05,
      "loss": 0.3593,
      "step": 339
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9641593378577397e-05,
      "loss": 0.4144,
      "step": 340
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9638726324698916e-05,
      "loss": 0.3854,
      "step": 341
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9635848059974524e-05,
      "loss": 0.348,
      "step": 342
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9632958587751942e-05,
      "loss": 0.3231,
      "step": 343
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.963005791139193e-05,
      "loss": 0.33,
      "step": 344
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9627146034268273e-05,
      "loss": 0.3512,
      "step": 345
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.96242229597678e-05,
      "loss": 0.3784,
      "step": 346
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9621288691290336e-05,
      "loss": 0.3767,
      "step": 347
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9618343232248752e-05,
      "loss": 0.3555,
      "step": 348
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9615386586068924e-05,
      "loss": 0.3391,
      "step": 349
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9612418756189736e-05,
      "loss": 0.3882,
      "step": 350
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9609439746063092e-05,
      "loss": 0.3501,
      "step": 351
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9606449559153882e-05,
      "loss": 0.2545,
      "step": 352
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9603448198940013e-05,
      "loss": 0.3894,
      "step": 353
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.960043566891238e-05,
      "loss": 0.3782,
      "step": 354
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9597411972574868e-05,
      "loss": 0.362,
      "step": 355
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9594377113444353e-05,
      "loss": 0.3266,
      "step": 356
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.959133109505069e-05,
      "loss": 0.3302,
      "step": 357
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9588273920936724e-05,
      "loss": 0.3555,
      "step": 358
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.958520559465826e-05,
      "loss": 0.3012,
      "step": 359
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9582126119784087e-05,
      "loss": 0.3213,
      "step": 360
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9579035499895958e-05,
      "loss": 0.3354,
      "step": 361
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.957593373858858e-05,
      "loss": 0.4041,
      "step": 362
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.957282083946963e-05,
      "loss": 0.3382,
      "step": 363
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9569696806159734e-05,
      "loss": 0.3116,
      "step": 364
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9566561642292466e-05,
      "loss": 0.3352,
      "step": 365
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9563415351514357e-05,
      "loss": 0.3828,
      "step": 366
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9560257937484867e-05,
      "loss": 0.325,
      "step": 367
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9557089403876398e-05,
      "loss": 0.371,
      "step": 368
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9553909754374286e-05,
      "loss": 0.3655,
      "step": 369
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9550718992676798e-05,
      "loss": 0.3447,
      "step": 370
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.954751712249512e-05,
      "loss": 0.3613,
      "step": 371
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9544304147553363e-05,
      "loss": 0.3766,
      "step": 372
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9541080071588552e-05,
      "loss": 0.3719,
      "step": 373
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9537844898350626e-05,
      "loss": 0.3367,
      "step": 374
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.953459863160243e-05,
      "loss": 0.3054,
      "step": 375
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9531341275119706e-05,
      "loss": 0.2703,
      "step": 376
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9528072832691104e-05,
      "loss": 0.3287,
      "step": 377
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.952479330811817e-05,
      "loss": 0.3751,
      "step": 378
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.952150270521532e-05,
      "loss": 0.2963,
      "step": 379
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9518201027809882e-05,
      "loss": 0.2946,
      "step": 380
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9514888279742042e-05,
      "loss": 0.3192,
      "step": 381
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.951156446486488e-05,
      "loss": 0.4059,
      "step": 382
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.950822958704434e-05,
      "loss": 0.3192,
      "step": 383
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9504883650159224e-05,
      "loss": 0.3055,
      "step": 384
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9501526658101216e-05,
      "loss": 0.3014,
      "step": 385
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9498158614774844e-05,
      "loss": 0.3382,
      "step": 386
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9494779524097493e-05,
      "loss": 0.2894,
      "step": 387
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9491389389999405e-05,
      "loss": 0.3337,
      "step": 388
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9487988216423655e-05,
      "loss": 0.3503,
      "step": 389
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.948457600732616e-05,
      "loss": 0.3223,
      "step": 390
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9481152766675685e-05,
      "loss": 0.3291,
      "step": 391
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9477718498453808e-05,
      "loss": 0.2821,
      "step": 392
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.947427320665495e-05,
      "loss": 0.3324,
      "step": 393
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9470816895286334e-05,
      "loss": 0.3615,
      "step": 394
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9467349568368025e-05,
      "loss": 0.3194,
      "step": 395
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9463871229932873e-05,
      "loss": 0.3571,
      "step": 396
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.946038188402656e-05,
      "loss": 0.2888,
      "step": 397
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9456881534707554e-05,
      "loss": 0.3043,
      "step": 398
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9453370186047128e-05,
      "loss": 0.345,
      "step": 399
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.944984784212935e-05,
      "loss": 0.3424,
      "step": 400
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9446314507051075e-05,
      "loss": 0.3591,
      "step": 401
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9442770184921933e-05,
      "loss": 0.2526,
      "step": 402
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9439214879864344e-05,
      "loss": 0.3481,
      "step": 403
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9435648596013506e-05,
      "loss": 0.3006,
      "step": 404
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9432071337517375e-05,
      "loss": 0.3756,
      "step": 405
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9428483108536676e-05,
      "loss": 0.3885,
      "step": 406
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.942488391324489e-05,
      "loss": 0.3087,
      "step": 407
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.942127375582827e-05,
      "loss": 0.3826,
      "step": 408
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.941765264048579e-05,
      "loss": 0.333,
      "step": 409
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.94140205714292e-05,
      "loss": 0.3426,
      "step": 410
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9410377552882965e-05,
      "loss": 0.339,
      "step": 411
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9406723589084304e-05,
      "loss": 0.3666,
      "step": 412
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9403058684283155e-05,
      "loss": 0.3302,
      "step": 413
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9399382842742187e-05,
      "loss": 0.3388,
      "step": 414
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9395696068736785e-05,
      "loss": 0.3862,
      "step": 415
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9391998366555054e-05,
      "loss": 0.3218,
      "step": 416
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9388289740497807e-05,
      "loss": 0.3129,
      "step": 417
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9384570194878566e-05,
      "loss": 0.2887,
      "step": 418
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9380839734023552e-05,
      "loss": 0.2938,
      "step": 419
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9377098362271674e-05,
      "loss": 0.3416,
      "step": 420
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9373346083974547e-05,
      "loss": 0.3664,
      "step": 421
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.936958290349646e-05,
      "loss": 0.3135,
      "step": 422
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.936580882521438e-05,
      "loss": 0.358,
      "step": 423
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9362023853517962e-05,
      "loss": 0.3493,
      "step": 424
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9358227992809517e-05,
      "loss": 0.3028,
      "step": 425
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9354421247504033e-05,
      "loss": 0.3155,
      "step": 426
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9350603622029147e-05,
      "loss": 0.3368,
      "step": 427
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.934677512082516e-05,
      "loss": 0.3367,
      "step": 428
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9342935748345023e-05,
      "loss": 0.3167,
      "step": 429
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9339085509054318e-05,
      "loss": 0.2846,
      "step": 430
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9335224407431284e-05,
      "loss": 0.3416,
      "step": 431
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.933135244796678e-05,
      "loss": 0.3462,
      "step": 432
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.93274696351643e-05,
      "loss": 0.3357,
      "step": 433
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.932357597353996e-05,
      "loss": 0.3205,
      "step": 434
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9319671467622507e-05,
      "loss": 0.2722,
      "step": 435
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9315756121953275e-05,
      "loss": 0.3256,
      "step": 436
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.931182994108623e-05,
      "loss": 0.3139,
      "step": 437
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.930789292958792e-05,
      "loss": 0.3705,
      "step": 438
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.930394509203751e-05,
      "loss": 0.3322,
      "step": 439
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9299986433026743e-05,
      "loss": 0.3171,
      "step": 440
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9296016957159956e-05,
      "loss": 0.2665,
      "step": 441
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9292036669054062e-05,
      "loss": 0.3157,
      "step": 442
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9288045573338552e-05,
      "loss": 0.317,
      "step": 443
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.928404367465549e-05,
      "loss": 0.28,
      "step": 444
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.92800309776595e-05,
      "loss": 0.3307,
      "step": 445
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9276007487017767e-05,
      "loss": 0.32,
      "step": 446
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9271973207410034e-05,
      "loss": 0.295,
      "step": 447
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9267928143528587e-05,
      "loss": 0.311,
      "step": 448
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9263872300078258e-05,
      "loss": 0.3046,
      "step": 449
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.925980568177642e-05,
      "loss": 0.3454,
      "step": 450
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9255728293352974e-05,
      "loss": 0.2821,
      "step": 451
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9251640139550348e-05,
      "loss": 0.3658,
      "step": 452
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.924754122512349e-05,
      "loss": 0.3379,
      "step": 453
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9243431554839866e-05,
      "loss": 0.3416,
      "step": 454
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9239311133479454e-05,
      "loss": 0.3368,
      "step": 455
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9235179965834737e-05,
      "loss": 0.3103,
      "step": 456
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.923103805671069e-05,
      "loss": 0.2922,
      "step": 457
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.922688541092479e-05,
      "loss": 0.3689,
      "step": 458
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9222722033306994e-05,
      "loss": 0.4037,
      "step": 459
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9218547928699745e-05,
      "loss": 0.2964,
      "step": 460
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9214363101957968e-05,
      "loss": 0.3213,
      "step": 461
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.921016755794905e-05,
      "loss": 0.374,
      "step": 462
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.920596130155285e-05,
      "loss": 0.3388,
      "step": 463
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9201744337661678e-05,
      "loss": 0.3459,
      "step": 464
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.919751667118031e-05,
      "loss": 0.336,
      "step": 465
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.919327830702596e-05,
      "loss": 0.4125,
      "step": 466
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.918902925012829e-05,
      "loss": 0.3119,
      "step": 467
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9184769505429396e-05,
      "loss": 0.32,
      "step": 468
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9180499077883806e-05,
      "loss": 0.3141,
      "step": 469
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9176217972458477e-05,
      "loss": 0.2595,
      "step": 470
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.917192619413278e-05,
      "loss": 0.2831,
      "step": 471
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9167623747898497e-05,
      "loss": 0.3448,
      "step": 472
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9163310638759828e-05,
      "loss": 0.3312,
      "step": 473
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.915898687173337e-05,
      "loss": 0.3086,
      "step": 474
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9154652451848113e-05,
      "loss": 0.3225,
      "step": 475
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.915030738414544e-05,
      "loss": 0.3456,
      "step": 476
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9145951673679122e-05,
      "loss": 0.2985,
      "step": 477
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9141585325515303e-05,
      "loss": 0.3055,
      "step": 478
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.91372083447325e-05,
      "loss": 0.3144,
      "step": 479
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9132820736421607e-05,
      "loss": 0.3528,
      "step": 480
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9128422505685864e-05,
      "loss": 0.2978,
      "step": 481
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.912401365764087e-05,
      "loss": 0.3423,
      "step": 482
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9119594197414587e-05,
      "loss": 0.2969,
      "step": 483
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9115164130147298e-05,
      "loss": 0.35,
      "step": 484
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.911072346099164e-05,
      "loss": 0.351,
      "step": 485
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9106272195112576e-05,
      "loss": 0.3393,
      "step": 486
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.910181033768739e-05,
      "loss": 0.335,
      "step": 487
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.909733789390569e-05,
      "loss": 0.294,
      "step": 488
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9092854868969395e-05,
      "loss": 0.3151,
      "step": 489
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.908836126809273e-05,
      "loss": 0.3111,
      "step": 490
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9083857096502228e-05,
      "loss": 0.3216,
      "step": 491
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9079342359436704e-05,
      "loss": 0.3271,
      "step": 492
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9074817062147274e-05,
      "loss": 0.2843,
      "step": 493
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9070281209897326e-05,
      "loss": 0.3259,
      "step": 494
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.906573480796254e-05,
      "loss": 0.322,
      "step": 495
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9061177861630845e-05,
      "loss": 0.3318,
      "step": 496
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9056610376202457e-05,
      "loss": 0.3058,
      "step": 497
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.905203235698983e-05,
      "loss": 0.3175,
      "step": 498
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.904744380931768e-05,
      "loss": 0.3422,
      "step": 499
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9042844738522975e-05,
      "loss": 0.2996,
      "step": 500
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9038235149954904e-05,
      "loss": 0.3065,
      "step": 501
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.903361504897491e-05,
      "loss": 0.3085,
      "step": 502
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9028984440956645e-05,
      "loss": 0.3171,
      "step": 503
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9024343331285995e-05,
      "loss": 0.277,
      "step": 504
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9019691725361054e-05,
      "loss": 0.3295,
      "step": 505
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9015029628592125e-05,
      "loss": 0.3398,
      "step": 506
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9010357046401714e-05,
      "loss": 0.3461,
      "step": 507
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9005673984224525e-05,
      "loss": 0.2963,
      "step": 508
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9000980447507444e-05,
      "loss": 0.3467,
      "step": 509
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.8996276441709553e-05,
      "loss": 0.375,
      "step": 510
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.899156197230209e-05,
      "loss": 0.3602,
      "step": 511
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.8986837044768484e-05,
      "loss": 0.2535,
      "step": 512
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.898210166460432e-05,
      "loss": 0.3303,
      "step": 513
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.8977355837317336e-05,
      "loss": 0.371,
      "step": 514
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.8972599568427425e-05,
      "loss": 0.2853,
      "step": 515
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.896783286346663e-05,
      "loss": 0.2743,
      "step": 516
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.8963055727979117e-05,
      "loss": 0.2848,
      "step": 517
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.8958268167521206e-05,
      "loss": 0.3713,
      "step": 518
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8953470187661324e-05,
      "loss": 0.3548,
      "step": 519
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8948661793980018e-05,
      "loss": 0.2979,
      "step": 520
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8943842992069963e-05,
      "loss": 0.3205,
      "step": 521
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.893901378753592e-05,
      "loss": 0.3155,
      "step": 522
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8934174185994765e-05,
      "loss": 0.3243,
      "step": 523
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8929324193075452e-05,
      "loss": 0.3276,
      "step": 524
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.892446381441904e-05,
      "loss": 0.3323,
      "step": 525
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.891959305567865e-05,
      "loss": 0.2897,
      "step": 526
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8914711922519487e-05,
      "loss": 0.3141,
      "step": 527
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8909820420618816e-05,
      "loss": 0.2889,
      "step": 528
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.890491855566597e-05,
      "loss": 0.3549,
      "step": 529
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.890000633336232e-05,
      "loss": 0.3147,
      "step": 530
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8895083759421306e-05,
      "loss": 0.3017,
      "step": 531
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8890150839568385e-05,
      "loss": 0.3238,
      "step": 532
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.8885207579541067e-05,
      "loss": 0.3071,
      "step": 533
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8880253985088876e-05,
      "loss": 0.3387,
      "step": 534
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.887529006197336e-05,
      "loss": 0.3164,
      "step": 535
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8870315815968077e-05,
      "loss": 0.2528,
      "step": 536
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8865331252858605e-05,
      "loss": 0.3219,
      "step": 537
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8860336378442505e-05,
      "loss": 0.2907,
      "step": 538
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.885533119852933e-05,
      "loss": 0.331,
      "step": 539
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8850315718940643e-05,
      "loss": 0.3432,
      "step": 540
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.884528994550996e-05,
      "loss": 0.329,
      "step": 541
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8840253884082786e-05,
      "loss": 0.3189,
      "step": 542
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8835207540516584e-05,
      "loss": 0.357,
      "step": 543
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.883015092068078e-05,
      "loss": 0.2972,
      "step": 544
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8825084030456747e-05,
      "loss": 0.2815,
      "step": 545
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.882000687573782e-05,
      "loss": 0.2821,
      "step": 546
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8814919462429246e-05,
      "loss": 0.3201,
      "step": 547
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.8809821796448222e-05,
      "loss": 0.4246,
      "step": 548
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8804713883723872e-05,
      "loss": 0.3096,
      "step": 549
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.879959573019723e-05,
      "loss": 0.3336,
      "step": 550
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8794467341821238e-05,
      "loss": 0.3035,
      "step": 551
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8789328724560755e-05,
      "loss": 0.3138,
      "step": 552
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8784179884392526e-05,
      "loss": 0.2944,
      "step": 553
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.877902082730519e-05,
      "loss": 0.3034,
      "step": 554
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8773851559299273e-05,
      "loss": 0.3277,
      "step": 555
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8768672086387165e-05,
      "loss": 0.2646,
      "step": 556
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8763482414593146e-05,
      "loss": 0.333,
      "step": 557
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8758282549953336e-05,
      "loss": 0.2987,
      "step": 558
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.875307249851573e-05,
      "loss": 0.3517,
      "step": 559
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8747852266340158e-05,
      "loss": 0.3273,
      "step": 560
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8742621859498294e-05,
      "loss": 0.3238,
      "step": 561
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.8737381284073647e-05,
      "loss": 0.3125,
      "step": 562
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.873213054616156e-05,
      "loss": 0.3491,
      "step": 563
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8726869651869184e-05,
      "loss": 0.3186,
      "step": 564
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8721598607315493e-05,
      "loss": 0.2776,
      "step": 565
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8716317418631257e-05,
      "loss": 0.2857,
      "step": 566
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8711026091959056e-05,
      "loss": 0.2861,
      "step": 567
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8705724633453253e-05,
      "loss": 0.2853,
      "step": 568
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.870041304928e-05,
      "loss": 0.2738,
      "step": 569
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.869509134561722e-05,
      "loss": 0.3278,
      "step": 570
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8689759528654617e-05,
      "loss": 0.2686,
      "step": 571
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8684417604593642e-05,
      "loss": 0.3104,
      "step": 572
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8679065579647526e-05,
      "loss": 0.3316,
      "step": 573
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8673703460041217e-05,
      "loss": 0.2723,
      "step": 574
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8668331252011428e-05,
      "loss": 0.3306,
      "step": 575
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.86629489618066e-05,
      "loss": 0.2615,
      "step": 576
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.8657556595686896e-05,
      "loss": 0.3086,
      "step": 577
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.86521541599242e-05,
      "loss": 0.2763,
      "step": 578
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8646741660802116e-05,
      "loss": 0.3144,
      "step": 579
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8641319104615937e-05,
      "loss": 0.3493,
      "step": 580
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8635886497672667e-05,
      "loss": 0.334,
      "step": 581
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8630443846290995e-05,
      "loss": 0.3605,
      "step": 582
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8624991156801297e-05,
      "loss": 0.2655,
      "step": 583
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8619528435545616e-05,
      "loss": 0.3149,
      "step": 584
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8614055688877665e-05,
      "loss": 0.337,
      "step": 585
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8608572923162825e-05,
      "loss": 0.3276,
      "step": 586
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.860308014477813e-05,
      "loss": 0.2949,
      "step": 587
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8597577360112246e-05,
      "loss": 0.3469,
      "step": 588
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.859206457556549e-05,
      "loss": 0.3823,
      "step": 589
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.858654179754981e-05,
      "loss": 0.3188,
      "step": 590
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.858100903248877e-05,
      "loss": 0.2701,
      "step": 591
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8575466286817556e-05,
      "loss": 0.3506,
      "step": 592
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.8569913566982952e-05,
      "loss": 0.3466,
      "step": 593
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8564350879443367e-05,
      "loss": 0.3127,
      "step": 594
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8558778230668775e-05,
      "loss": 0.3739,
      "step": 595
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.855319562714075e-05,
      "loss": 0.3194,
      "step": 596
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8547603075352454e-05,
      "loss": 0.2527,
      "step": 597
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8542000581808598e-05,
      "loss": 0.2498,
      "step": 598
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8536388153025472e-05,
      "loss": 0.2673,
      "step": 599
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8530765795530915e-05,
      "loss": 0.2792,
      "step": 600
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.852513351586432e-05,
      "loss": 0.3251,
      "step": 601
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.851949132057661e-05,
      "loss": 0.2656,
      "step": 602
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8513839216230255e-05,
      "loss": 0.345,
      "step": 603
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.850817720939924e-05,
      "loss": 0.3206,
      "step": 604
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.850250530666907e-05,
      "loss": 0.3287,
      "step": 605
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.8496823514636763e-05,
      "loss": 0.3273,
      "step": 606
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.849113183991083e-05,
      "loss": 0.2601,
      "step": 607
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.848543028911129e-05,
      "loss": 0.2917,
      "step": 608
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8479718868869644e-05,
      "loss": 0.3394,
      "step": 609
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8473997585828865e-05,
      "loss": 0.3091,
      "step": 610
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8468266446643408e-05,
      "loss": 0.2682,
      "step": 611
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8462525457979178e-05,
      "loss": 0.3312,
      "step": 612
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.845677462651356e-05,
      "loss": 0.3079,
      "step": 613
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8451013958935357e-05,
      "loss": 0.2432,
      "step": 614
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8445243461944832e-05,
      "loss": 0.3401,
      "step": 615
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.843946314225368e-05,
      "loss": 0.3376,
      "step": 616
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8433673006585013e-05,
      "loss": 0.3224,
      "step": 617
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8427873061673363e-05,
      "loss": 0.2943,
      "step": 618
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8422063314264677e-05,
      "loss": 0.3195,
      "step": 619
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.841624377111629e-05,
      "loss": 0.3746,
      "step": 620
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.841041443899695e-05,
      "loss": 0.2806,
      "step": 621
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.8404575324686768e-05,
      "loss": 0.2644,
      "step": 622
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.839872643497725e-05,
      "loss": 0.3491,
      "step": 623
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8392867776671267e-05,
      "loss": 0.3087,
      "step": 624
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8386999356583052e-05,
      "loss": 0.3533,
      "step": 625
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8381121181538185e-05,
      "loss": 0.323,
      "step": 626
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.83752332583736e-05,
      "loss": 0.2996,
      "step": 627
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8369335593937564e-05,
      "loss": 0.2854,
      "step": 628
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8363428195089687e-05,
      "loss": 0.3117,
      "step": 629
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8357511068700882e-05,
      "loss": 0.325,
      "step": 630
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.835158422165339e-05,
      "loss": 0.2478,
      "step": 631
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8345647660840748e-05,
      "loss": 0.3592,
      "step": 632
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.83397013931678e-05,
      "loss": 0.2816,
      "step": 633
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.833374542555068e-05,
      "loss": 0.2579,
      "step": 634
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8327779764916796e-05,
      "loss": 0.3239,
      "step": 635
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8321804418204837e-05,
      "loss": 0.3273,
      "step": 636
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8315819392364754e-05,
      "loss": 0.3452,
      "step": 637
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.8309824694357757e-05,
      "loss": 0.3469,
      "step": 638
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8303820331156307e-05,
      "loss": 0.3443,
      "step": 639
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8297806309744105e-05,
      "loss": 0.2853,
      "step": 640
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8291782637116093e-05,
      "loss": 0.3517,
      "step": 641
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8285749320278422e-05,
      "loss": 0.3487,
      "step": 642
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8279706366248475e-05,
      "loss": 0.3578,
      "step": 643
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.827365378205484e-05,
      "loss": 0.2876,
      "step": 644
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.82675915747373e-05,
      "loss": 0.2835,
      "step": 645
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.826151975134684e-05,
      "loss": 0.3025,
      "step": 646
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8255438318945623e-05,
      "loss": 0.2888,
      "step": 647
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.824934728460699e-05,
      "loss": 0.2602,
      "step": 648
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8243246655415454e-05,
      "loss": 0.3083,
      "step": 649
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.823713643846668e-05,
      "loss": 0.3178,
      "step": 650
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8231016640867492e-05,
      "loss": 0.3261,
      "step": 651
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.822488726973585e-05,
      "loss": 0.3231,
      "step": 652
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.8218748332200857e-05,
      "loss": 0.3269,
      "step": 653
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.821259983540274e-05,
      "loss": 0.301,
      "step": 654
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8206441786492837e-05,
      "loss": 0.2441,
      "step": 655
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.820027419263361e-05,
      "loss": 0.309,
      "step": 656
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8194097060998608e-05,
      "loss": 0.3365,
      "step": 657
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8187910398772486e-05,
      "loss": 0.3061,
      "step": 658
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8181714213150977e-05,
      "loss": 0.3116,
      "step": 659
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.817550851134089e-05,
      "loss": 0.2835,
      "step": 660
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8169293300560107e-05,
      "loss": 0.3055,
      "step": 661
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8163068588037568e-05,
      "loss": 0.2717,
      "step": 662
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.815683438101326e-05,
      "loss": 0.2736,
      "step": 663
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8150590686738225e-05,
      "loss": 0.338,
      "step": 664
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8144337512474523e-05,
      "loss": 0.3429,
      "step": 665
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8138074865495257e-05,
      "loss": 0.2603,
      "step": 666
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8131802753084535e-05,
      "loss": 0.2847,
      "step": 667
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.812552118253748e-05,
      "loss": 0.305,
      "step": 668
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8119230161160215e-05,
      "loss": 0.325,
      "step": 669
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8112929696269858e-05,
      "loss": 0.3081,
      "step": 670
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8106619795194503e-05,
      "loss": 0.3138,
      "step": 671
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.810030046527323e-05,
      "loss": 0.3193,
      "step": 672
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8093971713856072e-05,
      "loss": 0.3904,
      "step": 673
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8087633548304034e-05,
      "loss": 0.2939,
      "step": 674
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8081285975989066e-05,
      "loss": 0.2928,
      "step": 675
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.807492900429406e-05,
      "loss": 0.3648,
      "step": 676
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8068562640612825e-05,
      "loss": 0.2868,
      "step": 677
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.806218689235012e-05,
      "loss": 0.2871,
      "step": 678
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.805580176692161e-05,
      "loss": 0.3134,
      "step": 679
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8049407271753848e-05,
      "loss": 0.2569,
      "step": 680
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.804300341428431e-05,
      "loss": 0.2987,
      "step": 681
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8036590201961354e-05,
      "loss": 0.272,
      "step": 682
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.8030167642244205e-05,
      "loss": 0.3068,
      "step": 683
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8023735742602978e-05,
      "loss": 0.3377,
      "step": 684
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8017294510518648e-05,
      "loss": 0.319,
      "step": 685
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8010843953483035e-05,
      "loss": 0.2847,
      "step": 686
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.8004384078998812e-05,
      "loss": 0.2968,
      "step": 687
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7997914894579493e-05,
      "loss": 0.294,
      "step": 688
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7991436407749406e-05,
      "loss": 0.3351,
      "step": 689
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7984948626043717e-05,
      "loss": 0.2964,
      "step": 690
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7978451557008392e-05,
      "loss": 0.2621,
      "step": 691
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7971945208200197e-05,
      "loss": 0.3329,
      "step": 692
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.79654295871867e-05,
      "loss": 0.2917,
      "step": 693
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7958904701546246e-05,
      "loss": 0.2968,
      "step": 694
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.795237055886796e-05,
      "loss": 0.3136,
      "step": 695
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.794582716675173e-05,
      "loss": 0.2478,
      "step": 696
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7939274532808215e-05,
      "loss": 0.3426,
      "step": 697
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.7932712664658803e-05,
      "loss": 0.274,
      "step": 698
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7926141569935636e-05,
      "loss": 0.3381,
      "step": 699
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7919561256281582e-05,
      "loss": 0.3672,
      "step": 700
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.791297173135023e-05,
      "loss": 0.3028,
      "step": 701
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7906373002805896e-05,
      "loss": 0.3335,
      "step": 702
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7899765078323583e-05,
      "loss": 0.286,
      "step": 703
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7893147965589e-05,
      "loss": 0.2975,
      "step": 704
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.788652167229854e-05,
      "loss": 0.3033,
      "step": 705
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7879886206159278e-05,
      "loss": 0.265,
      "step": 706
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7873241574888948e-05,
      "loss": 0.352,
      "step": 707
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7866587786215955e-05,
      "loss": 0.299,
      "step": 708
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.785992484787935e-05,
      "loss": 0.347,
      "step": 709
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7853252767628822e-05,
      "loss": 0.2974,
      "step": 710
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7846571553224706e-05,
      "loss": 0.2546,
      "step": 711
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.7839881212437947e-05,
      "loss": 0.3156,
      "step": 712
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.783318175305011e-05,
      "loss": 0.3353,
      "step": 713
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7826473182853372e-05,
      "loss": 0.3022,
      "step": 714
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7819755509650494e-05,
      "loss": 0.3164,
      "step": 715
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7813028741254833e-05,
      "loss": 0.3165,
      "step": 716
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7806292885490327e-05,
      "loss": 0.2991,
      "step": 717
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.779954795019148e-05,
      "loss": 0.3297,
      "step": 718
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7792793943203355e-05,
      "loss": 0.3326,
      "step": 719
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7786030872381564e-05,
      "loss": 0.3428,
      "step": 720
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.777925874559227e-05,
      "loss": 0.2492,
      "step": 721
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.777247757071217e-05,
      "loss": 0.3467,
      "step": 722
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7765687355628466e-05,
      "loss": 0.2659,
      "step": 723
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7758888108238895e-05,
      "loss": 0.3221,
      "step": 724
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.775207983645169e-05,
      "loss": 0.2908,
      "step": 725
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7745262548185585e-05,
      "loss": 0.3263,
      "step": 726
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.7738436251369795e-05,
      "loss": 0.2952,
      "step": 727
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.773160095394402e-05,
      "loss": 0.2988,
      "step": 728
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7724756663858423e-05,
      "loss": 0.2926,
      "step": 729
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.771790338907363e-05,
      "loss": 0.3275,
      "step": 730
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7711041137560715e-05,
      "loss": 0.3438,
      "step": 731
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7704169917301195e-05,
      "loss": 0.3022,
      "step": 732
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7697289736287015e-05,
      "loss": 0.287,
      "step": 733
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7690400602520545e-05,
      "loss": 0.2617,
      "step": 734
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7683502524014568e-05,
      "loss": 0.2144,
      "step": 735
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.767659550879227e-05,
      "loss": 0.2395,
      "step": 736
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.766967956488723e-05,
      "loss": 0.2043,
      "step": 737
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7662754700343416e-05,
      "loss": 0.3008,
      "step": 738
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7655820923215168e-05,
      "loss": 0.3261,
      "step": 739
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7648878241567193e-05,
      "loss": 0.3466,
      "step": 740
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7641926663474557e-05,
      "loss": 0.3328,
      "step": 741
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7634966197022666e-05,
      "loss": 0.2976,
      "step": 742
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.7627996850307275e-05,
      "loss": 0.3085,
      "step": 743
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7621018631434462e-05,
      "loss": 0.3119,
      "step": 744
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7614031548520626e-05,
      "loss": 0.3273,
      "step": 745
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.760703560969247e-05,
      "loss": 0.274,
      "step": 746
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7600030823087007e-05,
      "loss": 0.2919,
      "step": 747
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7593017196851533e-05,
      "loss": 0.2815,
      "step": 748
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7585994739143627e-05,
      "loss": 0.3424,
      "step": 749
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7578963458131144e-05,
      "loss": 0.2947,
      "step": 750
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7571923361992195e-05,
      "loss": 0.2837,
      "step": 751
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.756487445891515e-05,
      "loss": 0.2732,
      "step": 752
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.755781675709862e-05,
      "loss": 0.3187,
      "step": 753
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7550750264751452e-05,
      "loss": 0.2837,
      "step": 754
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.754367499009271e-05,
      "loss": 0.2861,
      "step": 755
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7536590941351682e-05,
      "loss": 0.3434,
      "step": 756
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.7529498126767852e-05,
      "loss": 0.265,
      "step": 757
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.752239655459091e-05,
      "loss": 0.2663,
      "step": 758
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7515286233080722e-05,
      "loss": 0.2922,
      "step": 759
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.750816717050734e-05,
      "loss": 0.3689,
      "step": 760
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.750103937515097e-05,
      "loss": 0.2793,
      "step": 761
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7493902855301992e-05,
      "loss": 0.3242,
      "step": 762
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7486757619260918e-05,
      "loss": 0.2822,
      "step": 763
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7479603675338403e-05,
      "loss": 0.3077,
      "step": 764
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7472441031855237e-05,
      "loss": 0.3116,
      "step": 765
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.746526969714232e-05,
      "loss": 0.3106,
      "step": 766
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.745808967954066e-05,
      "loss": 0.2342,
      "step": 767
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.745090098740138e-05,
      "loss": 0.3591,
      "step": 768
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7443703629085663e-05,
      "loss": 0.2464,
      "step": 769
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.74364976129648e-05,
      "loss": 0.2902,
      "step": 770
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.742928294742014e-05,
      "loss": 0.317,
      "step": 771
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7422059640843093e-05,
      "loss": 0.2991,
      "step": 772
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.7414827701635112e-05,
      "loss": 0.338,
      "step": 773
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7407587138207707e-05,
      "loss": 0.3259,
      "step": 774
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7400337958982403e-05,
      "loss": 0.2505,
      "step": 775
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.739308017239076e-05,
      "loss": 0.3027,
      "step": 776
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7385813786874334e-05,
      "loss": 0.2582,
      "step": 777
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.73785388108847e-05,
      "loss": 0.3336,
      "step": 778
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.737125525288341e-05,
      "loss": 0.3467,
      "step": 779
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7363963121341997e-05,
      "loss": 0.3059,
      "step": 780
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7356662424741983e-05,
      "loss": 0.2944,
      "step": 781
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7349353171574836e-05,
      "loss": 0.2446,
      "step": 782
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.734203537034198e-05,
      "loss": 0.2968,
      "step": 783
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.733470902955479e-05,
      "loss": 0.3094,
      "step": 784
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7327374157734552e-05,
      "loss": 0.2687,
      "step": 785
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7320030763412502e-05,
      "loss": 0.3081,
      "step": 786
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7312678855129767e-05,
      "loss": 0.3088,
      "step": 787
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.7305318441437388e-05,
      "loss": 0.3024,
      "step": 788
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.729794953089629e-05,
      "loss": 0.2763,
      "step": 789
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7290572132077295e-05,
      "loss": 0.3181,
      "step": 790
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.728318625356108e-05,
      "loss": 0.2424,
      "step": 791
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7275791903938203e-05,
      "loss": 0.3441,
      "step": 792
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7268389091809056e-05,
      "loss": 0.3118,
      "step": 793
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7260977825783886e-05,
      "loss": 0.2963,
      "step": 794
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7253558114482773e-05,
      "loss": 0.3782,
      "step": 795
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7246129966535608e-05,
      "loss": 0.2744,
      "step": 796
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7238693390582114e-05,
      "loss": 0.2976,
      "step": 797
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.72312483952718e-05,
      "loss": 0.3382,
      "step": 798
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.722379498926398e-05,
      "loss": 0.2708,
      "step": 799
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7216333181227734e-05,
      "loss": 0.2759,
      "step": 800
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.720886297984193e-05,
      "loss": 0.2923,
      "step": 801
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.720138439379519e-05,
      "loss": 0.3483,
      "step": 802
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.7193897431785897e-05,
      "loss": 0.3263,
      "step": 803
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.718640210252216e-05,
      "loss": 0.3316,
      "step": 804
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7178898414721843e-05,
      "loss": 0.2813,
      "step": 805
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.717138637711251e-05,
      "loss": 0.2991,
      "step": 806
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7163865998431448e-05,
      "loss": 0.2671,
      "step": 807
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7156337287425643e-05,
      "loss": 0.3023,
      "step": 808
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7148800252851773e-05,
      "loss": 0.2862,
      "step": 809
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7141254903476196e-05,
      "loss": 0.2866,
      "step": 810
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7133701248074945e-05,
      "loss": 0.3034,
      "step": 811
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7126139295433706e-05,
      "loss": 0.282,
      "step": 812
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7118569054347824e-05,
      "loss": 0.2475,
      "step": 813
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.711099053362228e-05,
      "loss": 0.2964,
      "step": 814
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.710340374207168e-05,
      "loss": 0.2636,
      "step": 815
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7095808688520264e-05,
      "loss": 0.2269,
      "step": 816
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7088205381801867e-05,
      "loss": 0.3187,
      "step": 817
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7080593830759928e-05,
      "loss": 0.3169,
      "step": 818
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.7072974044247483e-05,
      "loss": 0.3133,
      "step": 819
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.706534603112713e-05,
      "loss": 0.3502,
      "step": 820
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.7057709800271056e-05,
      "loss": 0.3303,
      "step": 821
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.705006536056099e-05,
      "loss": 0.2488,
      "step": 822
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.704241272088821e-05,
      "loss": 0.3143,
      "step": 823
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.7034751890153546e-05,
      "loss": 0.3388,
      "step": 824
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.7027082877267337e-05,
      "loss": 0.3375,
      "step": 825
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.701940569114945e-05,
      "loss": 0.3068,
      "step": 826
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.7011720340729252e-05,
      "loss": 0.2623,
      "step": 827
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.7004026834945607e-05,
      "loss": 0.2828,
      "step": 828
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.699632518274687e-05,
      "loss": 0.2882,
      "step": 829
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.698861539309087e-05,
      "loss": 0.3078,
      "step": 830
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.698089747494489e-05,
      "loss": 0.3066,
      "step": 831
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.697317143728568e-05,
      "loss": 0.339,
      "step": 832
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.6965437289099433e-05,
      "loss": 0.285,
      "step": 833
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.6957695039381765e-05,
      "loss": 0.2992,
      "step": 834
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.694994469713772e-05,
      "loss": 0.326,
      "step": 835
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6942186271381764e-05,
      "loss": 0.2758,
      "step": 836
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.693441977113775e-05,
      "loss": 0.2975,
      "step": 837
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6926645205438932e-05,
      "loss": 0.2856,
      "step": 838
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6918862583327943e-05,
      "loss": 0.2613,
      "step": 839
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.691107191385678e-05,
      "loss": 0.2832,
      "step": 840
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6903273206086813e-05,
      "loss": 0.2724,
      "step": 841
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6895466469088755e-05,
      "loss": 0.2896,
      "step": 842
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.688765171194265e-05,
      "loss": 0.2824,
      "step": 843
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6879828943737883e-05,
      "loss": 0.307,
      "step": 844
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6871998173573146e-05,
      "loss": 0.3141,
      "step": 845
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6864159410556444e-05,
      "loss": 0.3025,
      "step": 846
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.6856312663805083e-05,
      "loss": 0.3129,
      "step": 847
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.684845794244564e-05,
      "loss": 0.3175,
      "step": 848
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.684059525561399e-05,
      "loss": 0.2872,
      "step": 849
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6832724612455248e-05,
      "loss": 0.279,
      "step": 850
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6824846022123794e-05,
      "loss": 0.2978,
      "step": 851
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6816959493783257e-05,
      "loss": 0.2803,
      "step": 852
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6809065036606493e-05,
      "loss": 0.2883,
      "step": 853
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.680116265977558e-05,
      "loss": 0.2824,
      "step": 854
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6793252372481806e-05,
      "loss": 0.3189,
      "step": 855
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.678533418392567e-05,
      "loss": 0.2597,
      "step": 856
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6777408103316842e-05,
      "loss": 0.2782,
      "step": 857
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6769474139874184e-05,
      "loss": 0.2791,
      "step": 858
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6761532302825727e-05,
      "loss": 0.2889,
      "step": 859
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.675358260140866e-05,
      "loss": 0.3004,
      "step": 860
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.674562504486931e-05,
      "loss": 0.26,
      "step": 861
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6737659642463152e-05,
      "loss": 0.2377,
      "step": 862
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6729686403454785e-05,
      "loss": 0.2957,
      "step": 863
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.6721705337117907e-05,
      "loss": 0.3493,
      "step": 864
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.671371645273534e-05,
      "loss": 0.2808,
      "step": 865
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.670571975959899e-05,
      "loss": 0.2939,
      "step": 866
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.669771526700985e-05,
      "loss": 0.2684,
      "step": 867
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6689702984277976e-05,
      "loss": 0.2994,
      "step": 868
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6681682920722493e-05,
      "loss": 0.2308,
      "step": 869
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6673655085671575e-05,
      "loss": 0.3013,
      "step": 870
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6665619488462427e-05,
      "loss": 0.2925,
      "step": 871
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6657576138441293e-05,
      "loss": 0.2709,
      "step": 872
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6649525044963428e-05,
      "loss": 0.3151,
      "step": 873
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6641466217393097e-05,
      "loss": 0.2993,
      "step": 874
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.663339966510355e-05,
      "loss": 0.2978,
      "step": 875
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6625325397477044e-05,
      "loss": 0.2947,
      "step": 876
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.661724342390479e-05,
      "loss": 0.2533,
      "step": 877
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.660915375378696e-05,
      "loss": 0.2894,
      "step": 878
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.6601056396532696e-05,
      "loss": 0.2752,
      "step": 879
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6592951361560067e-05,
      "loss": 0.2789,
      "step": 880
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6584838658296075e-05,
      "loss": 0.3355,
      "step": 881
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6576718296176647e-05,
      "loss": 0.261,
      "step": 882
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6568590284646603e-05,
      "loss": 0.2399,
      "step": 883
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6560454633159675e-05,
      "loss": 0.33,
      "step": 884
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6552311351178475e-05,
      "loss": 0.2734,
      "step": 885
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6544160448174494e-05,
      "loss": 0.3048,
      "step": 886
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.653600193362808e-05,
      "loss": 0.2704,
      "step": 887
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.652783581702844e-05,
      "loss": 0.2901,
      "step": 888
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6519662107873617e-05,
      "loss": 0.2328,
      "step": 889
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6511480815670493e-05,
      "loss": 0.3314,
      "step": 890
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.650329194993477e-05,
      "loss": 0.3028,
      "step": 891
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6495095520190943e-05,
      "loss": 0.2427,
      "step": 892
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6486891535972323e-05,
      "loss": 0.3087,
      "step": 893
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.6478680006821004e-05,
      "loss": 0.2107,
      "step": 894
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6470460942287848e-05,
      "loss": 0.2653,
      "step": 895
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6462234351932485e-05,
      "loss": 0.2723,
      "step": 896
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6454000245323302e-05,
      "loss": 0.2644,
      "step": 897
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6445758632037424e-05,
      "loss": 0.2702,
      "step": 898
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6437509521660712e-05,
      "loss": 0.3551,
      "step": 899
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6429252923787736e-05,
      "loss": 0.2615,
      "step": 900
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6420988848021792e-05,
      "loss": 0.2239,
      "step": 901
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6412717303974855e-05,
      "loss": 0.2255,
      "step": 902
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6404438301267595e-05,
      "loss": 0.3198,
      "step": 903
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6396151849529363e-05,
      "loss": 0.2717,
      "step": 904
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6387857958398168e-05,
      "loss": 0.3224,
      "step": 905
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6379556637520667e-05,
      "loss": 0.318,
      "step": 906
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6371247896552166e-05,
      "loss": 0.2747,
      "step": 907
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6362931745156602e-05,
      "loss": 0.2879,
      "step": 908
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.6354608193006526e-05,
      "loss": 0.2704,
      "step": 909
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6346277249783094e-05,
      "loss": 0.2995,
      "step": 910
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.633793892517607e-05,
      "loss": 0.304,
      "step": 911
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6329593228883794e-05,
      "loss": 0.2709,
      "step": 912
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6321240170613183e-05,
      "loss": 0.3338,
      "step": 913
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.631287976007972e-05,
      "loss": 0.2627,
      "step": 914
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6304512007007428e-05,
      "loss": 0.2443,
      "step": 915
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6296136921128886e-05,
      "loss": 0.3059,
      "step": 916
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.628775451218519e-05,
      "loss": 0.261,
      "step": 917
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6279364789925966e-05,
      "loss": 0.3449,
      "step": 918
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6270967764109323e-05,
      "loss": 0.2884,
      "step": 919
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6262563444501892e-05,
      "loss": 0.3177,
      "step": 920
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6254151840878765e-05,
      "loss": 0.3773,
      "step": 921
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6245732963023525e-05,
      "loss": 0.3379,
      "step": 922
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.62373068207282e-05,
      "loss": 0.2734,
      "step": 923
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.6228873423793276e-05,
      "loss": 0.2757,
      "step": 924
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.622043278202768e-05,
      "loss": 0.2961,
      "step": 925
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6211984905248752e-05,
      "loss": 0.3354,
      "step": 926
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.620352980328226e-05,
      "loss": 0.3023,
      "step": 927
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.619506748596237e-05,
      "loss": 0.3643,
      "step": 928
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6186597963131644e-05,
      "loss": 0.2498,
      "step": 929
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.617812124464102e-05,
      "loss": 0.2652,
      "step": 930
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6169637340349805e-05,
      "loss": 0.3031,
      "step": 931
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6161146260125667e-05,
      "loss": 0.2963,
      "step": 932
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6152648013844625e-05,
      "loss": 0.2445,
      "step": 933
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.614414261139102e-05,
      "loss": 0.3067,
      "step": 934
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.613563006265753e-05,
      "loss": 0.337,
      "step": 935
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6127110377545133e-05,
      "loss": 0.2938,
      "step": 936
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.611858356596312e-05,
      "loss": 0.3216,
      "step": 937
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.6110049637829058e-05,
      "loss": 0.2455,
      "step": 938
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.61015086030688e-05,
      "loss": 0.2746,
      "step": 939
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.609296047161646e-05,
      "loss": 0.2975,
      "step": 940
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.608440525341441e-05,
      "loss": 0.3262,
      "step": 941
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6075842958413258e-05,
      "loss": 0.3099,
      "step": 942
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6067273596571855e-05,
      "loss": 0.3195,
      "step": 943
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6058697177857262e-05,
      "loss": 0.2634,
      "step": 944
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6050113712244746e-05,
      "loss": 0.3153,
      "step": 945
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6041523209717777e-05,
      "loss": 0.3144,
      "step": 946
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.603292568026801e-05,
      "loss": 0.2675,
      "step": 947
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6024321133895268e-05,
      "loss": 0.3441,
      "step": 948
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.6015709580607537e-05,
      "loss": 0.2611,
      "step": 949
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.600709103042095e-05,
      "loss": 0.3035,
      "step": 950
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.599846549335979e-05,
      "loss": 0.2792,
      "step": 951
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.5989832979456453e-05,
      "loss": 0.3471,
      "step": 952
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.598119349875145e-05,
      "loss": 0.2343,
      "step": 953
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.597254706129341e-05,
      "loss": 0.2627,
      "step": 954
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.596389367713903e-05,
      "loss": 0.2485,
      "step": 955
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5955233356353117e-05,
      "loss": 0.281,
      "step": 956
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5946566109008517e-05,
      "loss": 0.3009,
      "step": 957
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5937891945186147e-05,
      "loss": 0.3498,
      "step": 958
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5929210874974966e-05,
      "loss": 0.288,
      "step": 959
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5920522908471965e-05,
      "loss": 0.2741,
      "step": 960
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5911828055782157e-05,
      "loss": 0.2752,
      "step": 961
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.590312632701857e-05,
      "loss": 0.3328,
      "step": 962
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.589441773230221e-05,
      "loss": 0.2855,
      "step": 963
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.58857022817621e-05,
      "loss": 0.2885,
      "step": 964
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5876979985535207e-05,
      "loss": 0.3041,
      "step": 965
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5868250853766478e-05,
      "loss": 0.2832,
      "step": 966
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.5859514896608804e-05,
      "loss": 0.3223,
      "step": 967
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.585077212422302e-05,
      "loss": 0.3283,
      "step": 968
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.584202254677788e-05,
      "loss": 0.3419,
      "step": 969
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5833266174450054e-05,
      "loss": 0.2856,
      "step": 970
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.582450301742412e-05,
      "loss": 0.2118,
      "step": 971
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.581573308589255e-05,
      "loss": 0.3005,
      "step": 972
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.580695639005569e-05,
      "loss": 0.2345,
      "step": 973
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5798172940121747e-05,
      "loss": 0.2805,
      "step": 974
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.57893827463068e-05,
      "loss": 0.2827,
      "step": 975
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5780585818834758e-05,
      "loss": 0.2586,
      "step": 976
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5771782167937364e-05,
      "loss": 0.2934,
      "step": 977
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.576297180385419e-05,
      "loss": 0.2992,
      "step": 978
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.575415473683261e-05,
      "loss": 0.3443,
      "step": 979
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.574533097712779e-05,
      "loss": 0.3595,
      "step": 980
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5736500535002687e-05,
      "loss": 0.3019,
      "step": 981
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5727663420728028e-05,
      "loss": 0.3301,
      "step": 982
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5718819644582296e-05,
      "loss": 0.2893,
      "step": 983
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.5709969216851736e-05,
      "loss": 0.2923,
      "step": 984
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5701112147830313e-05,
      "loss": 0.2744,
      "step": 985
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5692248447819726e-05,
      "loss": 0.2621,
      "step": 986
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5683378127129382e-05,
      "loss": 0.2976,
      "step": 987
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5674501196076394e-05,
      "loss": 0.2992,
      "step": 988
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.566561766498556e-05,
      "loss": 0.2662,
      "step": 989
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5656727544189357e-05,
      "loss": 0.2793,
      "step": 990
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5647830844027917e-05,
      "loss": 0.2898,
      "step": 991
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5638927574849038e-05,
      "loss": 0.2714,
      "step": 992
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5630017747008152e-05,
      "loss": 0.3917,
      "step": 993
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.562110137086832e-05,
      "loss": 0.2892,
      "step": 994
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.561217845680022e-05,
      "loss": 0.3122,
      "step": 995
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5603249015182124e-05,
      "loss": 0.3066,
      "step": 996
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.559431305639992e-05,
      "loss": 0.3129,
      "step": 997
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5585370590847053e-05,
      "loss": 0.261,
      "step": 998
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.5576421628924545e-05,
      "loss": 0.2572,
      "step": 999
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5567466181040977e-05,
      "loss": 0.2878,
      "step": 1000
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.555850425761247e-05,
      "loss": 0.2743,
      "step": 1001
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5549535869062675e-05,
      "loss": 0.2499,
      "step": 1002
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.554056102582277e-05,
      "loss": 0.3233,
      "step": 1003
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.553157973833143e-05,
      "loss": 0.3077,
      "step": 1004
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5522592017034833e-05,
      "loss": 0.3582,
      "step": 1005
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5513597872386637e-05,
      "loss": 0.2553,
      "step": 1006
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5504597314847977e-05,
      "loss": 0.256,
      "step": 1007
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.549559035488744e-05,
      "loss": 0.2705,
      "step": 1008
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5486577002981053e-05,
      "loss": 0.3368,
      "step": 1009
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.54775572696123e-05,
      "loss": 0.2942,
      "step": 1010
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5468531165272063e-05,
      "loss": 0.2708,
      "step": 1011
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5459498700458647e-05,
      "loss": 0.2596,
      "step": 1012
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5450459885677753e-05,
      "loss": 0.2628,
      "step": 1013
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.5441414731442467e-05,
      "loss": 0.2673,
      "step": 1014
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5432363248273255e-05,
      "loss": 0.3237,
      "step": 1015
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5423305446697928e-05,
      "loss": 0.3352,
      "step": 1016
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5414241337251658e-05,
      "loss": 0.254,
      "step": 1017
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5405170930476952e-05,
      "loss": 0.3206,
      "step": 1018
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5396094236923643e-05,
      "loss": 0.2564,
      "step": 1019
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5387011267148876e-05,
      "loss": 0.2464,
      "step": 1020
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.537792203171709e-05,
      "loss": 0.3386,
      "step": 1021
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.536882654120001e-05,
      "loss": 0.2661,
      "step": 1022
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.535972480617666e-05,
      "loss": 0.2512,
      "step": 1023
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.535061683723329e-05,
      "loss": 0.272,
      "step": 1024
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5341502644963423e-05,
      "loss": 0.2512,
      "step": 1025
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5332382239967825e-05,
      "loss": 0.2983,
      "step": 1026
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.532325563285448e-05,
      "loss": 0.2567,
      "step": 1027
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.531412283423858e-05,
      "loss": 0.2879,
      "step": 1028
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.5304983854742523e-05,
      "loss": 0.2878,
      "step": 1029
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.52958387049959e-05,
      "loss": 0.2868,
      "step": 1030
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5286687395635478e-05,
      "loss": 0.2931,
      "step": 1031
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5277529937305182e-05,
      "loss": 0.2896,
      "step": 1032
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5268366340656094e-05,
      "loss": 0.2456,
      "step": 1033
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5259196616346438e-05,
      "loss": 0.2649,
      "step": 1034
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5250020775041556e-05,
      "loss": 0.2533,
      "step": 1035
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5240838827413913e-05,
      "loss": 0.303,
      "step": 1036
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5231650784143077e-05,
      "loss": 0.2844,
      "step": 1037
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5222456655915694e-05,
      "loss": 0.2922,
      "step": 1038
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5213256453425502e-05,
      "loss": 0.2592,
      "step": 1039
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5204050187373299e-05,
      "loss": 0.2741,
      "step": 1040
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.519483786846693e-05,
      "loss": 0.3028,
      "step": 1041
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5185619507421287e-05,
      "loss": 0.302,
      "step": 1042
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5176395114958289e-05,
      "loss": 0.2416,
      "step": 1043
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.5167164701806864e-05,
      "loss": 0.2622,
      "step": 1044
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5157928278702951e-05,
      "loss": 0.3176,
      "step": 1045
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5148685856389472e-05,
      "loss": 0.2919,
      "step": 1046
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5139437445616331e-05,
      "loss": 0.299,
      "step": 1047
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5130183057140397e-05,
      "loss": 0.2857,
      "step": 1048
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5120922701725492e-05,
      "loss": 0.2608,
      "step": 1049
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5111656390142374e-05,
      "loss": 0.3035,
      "step": 1050
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5102384133168732e-05,
      "loss": 0.3114,
      "step": 1051
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.509310594158917e-05,
      "loss": 0.3253,
      "step": 1052
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5083821826195197e-05,
      "loss": 0.2467,
      "step": 1053
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5074531797785207e-05,
      "loss": 0.3113,
      "step": 1054
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5065235867164476e-05,
      "loss": 0.2714,
      "step": 1055
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5055934045145137e-05,
      "loss": 0.2355,
      "step": 1056
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5046626342546187e-05,
      "loss": 0.2874,
      "step": 1057
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.5037312770193456e-05,
      "loss": 0.2625,
      "step": 1058
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.50279933389196e-05,
      "loss": 0.2991,
      "step": 1059
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.501866805956409e-05,
      "loss": 0.2586,
      "step": 1060
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.5009336942973204e-05,
      "loss": 0.2486,
      "step": 1061
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 0.3195,
      "step": 1062
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4990657241504327e-05,
      "loss": 0.2674,
      "step": 1063
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4981308678352784e-05,
      "loss": 0.2645,
      "step": 1064
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4971954321418726e-05,
      "loss": 0.3172,
      "step": 1065
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.496259418158225e-05,
      "loss": 0.2979,
      "step": 1066
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4953228269730174e-05,
      "loss": 0.2794,
      "step": 1067
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4943856596756035e-05,
      "loss": 0.2349,
      "step": 1068
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4934479173560066e-05,
      "loss": 0.2771,
      "step": 1069
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4925096011049192e-05,
      "loss": 0.2488,
      "step": 1070
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4915707120137008e-05,
      "loss": 0.2529,
      "step": 1071
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.490631251174378e-05,
      "loss": 0.2687,
      "step": 1072
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4896912196796419e-05,
      "loss": 0.299,
      "step": 1073
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.4887506186228468e-05,
      "loss": 0.228,
      "step": 1074
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4878094490980104e-05,
      "loss": 0.3083,
      "step": 1075
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4868677121998114e-05,
      "loss": 0.2708,
      "step": 1076
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.485925409023588e-05,
      "loss": 0.3281,
      "step": 1077
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4849825406653369e-05,
      "loss": 0.2362,
      "step": 1078
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.484039108221713e-05,
      "loss": 0.2934,
      "step": 1079
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4830951127900268e-05,
      "loss": 0.3012,
      "step": 1080
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4821505554682437e-05,
      "loss": 0.2858,
      "step": 1081
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.481205437354982e-05,
      "loss": 0.2834,
      "step": 1082
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4802597595495134e-05,
      "loss": 0.3005,
      "step": 1083
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4793135231517598e-05,
      "loss": 0.3149,
      "step": 1084
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.478366729262293e-05,
      "loss": 0.2273,
      "step": 1085
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4774193789823333e-05,
      "loss": 0.2915,
      "step": 1086
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.476471473413748e-05,
      "loss": 0.237,
      "step": 1087
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4755230136590503e-05,
      "loss": 0.2425,
      "step": 1088
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.4745740008213983e-05,
      "loss": 0.2768,
      "step": 1089
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4736244360045923e-05,
      "loss": 0.2796,
      "step": 1090
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4726743203130764e-05,
      "loss": 0.2554,
      "step": 1091
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.471723654851934e-05,
      "loss": 0.2637,
      "step": 1092
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4707724407268881e-05,
      "loss": 0.3018,
      "step": 1093
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4698206790443005e-05,
      "loss": 0.2722,
      "step": 1094
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4688683709111695e-05,
      "loss": 0.2661,
      "step": 1095
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4679155174351287e-05,
      "loss": 0.2426,
      "step": 1096
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4669621197244466e-05,
      "loss": 0.2678,
      "step": 1097
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4660081788880241e-05,
      "loss": 0.2715,
      "step": 1098
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4650536960353938e-05,
      "loss": 0.2628,
      "step": 1099
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4640986722767196e-05,
      "loss": 0.2955,
      "step": 1100
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4631431087227936e-05,
      "loss": 0.311,
      "step": 1101
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4621870064850361e-05,
      "loss": 0.243,
      "step": 1102
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4612303666754941e-05,
      "loss": 0.2458,
      "step": 1103
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.4602731904068395e-05,
      "loss": 0.2382,
      "step": 1104
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4593154787923682e-05,
      "loss": 0.267,
      "step": 1105
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.458357232945999e-05,
      "loss": 0.2913,
      "step": 1106
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.457398453982272e-05,
      "loss": 0.2433,
      "step": 1107
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4564391430163474e-05,
      "loss": 0.2932,
      "step": 1108
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4554793011640038e-05,
      "loss": 0.2297,
      "step": 1109
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4545189295416377e-05,
      "loss": 0.2315,
      "step": 1110
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4535580292662618e-05,
      "loss": 0.2823,
      "step": 1111
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4525966014555035e-05,
      "loss": 0.2744,
      "step": 1112
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4516346472276036e-05,
      "loss": 0.3053,
      "step": 1113
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4506721677014157e-05,
      "loss": 0.2696,
      "step": 1114
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4497091639964036e-05,
      "loss": 0.2934,
      "step": 1115
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4487456372326418e-05,
      "loss": 0.2951,
      "step": 1116
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4477815885308121e-05,
      "loss": 0.2454,
      "step": 1117
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4468170190122043e-05,
      "loss": 0.2344,
      "step": 1118
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.4458519297987135e-05,
      "loss": 0.3155,
      "step": 1119
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4448863220128388e-05,
      "loss": 0.2495,
      "step": 1120
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4439201967776835e-05,
      "loss": 0.2303,
      "step": 1121
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.442953555216952e-05,
      "loss": 0.3035,
      "step": 1122
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4419863984549494e-05,
      "loss": 0.2304,
      "step": 1123
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4410187276165801e-05,
      "loss": 0.3023,
      "step": 1124
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4400505438273467e-05,
      "loss": 0.2625,
      "step": 1125
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4390818482133475e-05,
      "loss": 0.2955,
      "step": 1126
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4381126419012772e-05,
      "loss": 0.327,
      "step": 1127
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4371429260184239e-05,
      "loss": 0.2952,
      "step": 1128
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4361727016926685e-05,
      "loss": 0.2952,
      "step": 1129
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4352019700524836e-05,
      "loss": 0.2844,
      "step": 1130
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4342307322269306e-05,
      "loss": 0.2762,
      "step": 1131
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4332589893456615e-05,
      "loss": 0.2704,
      "step": 1132
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4322867425389144e-05,
      "loss": 0.2767,
      "step": 1133
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.4313139929375141e-05,
      "loss": 0.3022,
      "step": 1134
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.43034074167287e-05,
      "loss": 0.2696,
      "step": 1135
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4293669898769749e-05,
      "loss": 0.2877,
      "step": 1136
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.428392738682404e-05,
      "loss": 0.2516,
      "step": 1137
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4274179892223133e-05,
      "loss": 0.2241,
      "step": 1138
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4264427426304382e-05,
      "loss": 0.2644,
      "step": 1139
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4254670000410923e-05,
      "loss": 0.2231,
      "step": 1140
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4244907625891661e-05,
      "loss": 0.3312,
      "step": 1141
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4235140314101263e-05,
      "loss": 0.3317,
      "step": 1142
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4225368076400127e-05,
      "loss": 0.3073,
      "step": 1143
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4215590924154392e-05,
      "loss": 0.2315,
      "step": 1144
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4205808868735904e-05,
      "loss": 0.2564,
      "step": 1145
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4196021921522216e-05,
      "loss": 0.3015,
      "step": 1146
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4186230093896574e-05,
      "loss": 0.2813,
      "step": 1147
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.4176433397247889e-05,
      "loss": 0.2558,
      "step": 1148
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.416663184297075e-05,
      "loss": 0.2168,
      "step": 1149
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4156825442465386e-05,
      "loss": 0.3046,
      "step": 1150
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4147014207137666e-05,
      "loss": 0.2647,
      "step": 1151
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4137198148399083e-05,
      "loss": 0.3004,
      "step": 1152
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4127377277666736e-05,
      "loss": 0.2775,
      "step": 1153
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4117551606363327e-05,
      "loss": 0.2878,
      "step": 1154
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4107721145917137e-05,
      "loss": 0.2803,
      "step": 1155
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.409788590776202e-05,
      "loss": 0.3172,
      "step": 1156
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4088045903337383e-05,
      "loss": 0.2822,
      "step": 1157
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4078201144088181e-05,
      "loss": 0.2604,
      "step": 1158
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4068351641464898e-05,
      "loss": 0.3042,
      "step": 1159
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4058497406923536e-05,
      "loss": 0.2126,
      "step": 1160
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4048638451925598e-05,
      "loss": 0.2767,
      "step": 1161
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4038774787938077e-05,
      "loss": 0.3063,
      "step": 1162
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4028906426433447e-05,
      "loss": 0.3388,
      "step": 1163
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.4019033378889644e-05,
      "loss": 0.2962,
      "step": 1164
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.4009155656790052e-05,
      "loss": 0.3245,
      "step": 1165
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3999273271623498e-05,
      "loss": 0.2452,
      "step": 1166
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3989386234884225e-05,
      "loss": 0.3075,
      "step": 1167
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3979494558071892e-05,
      "loss": 0.2619,
      "step": 1168
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.396959825269155e-05,
      "loss": 0.2766,
      "step": 1169
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3959697330253636e-05,
      "loss": 0.2544,
      "step": 1170
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3949791802273965e-05,
      "loss": 0.2732,
      "step": 1171
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.393988168027369e-05,
      "loss": 0.2891,
      "step": 1172
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3929966975779325e-05,
      "loss": 0.2632,
      "step": 1173
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.392004770032271e-05,
      "loss": 0.3102,
      "step": 1174
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3910123865440991e-05,
      "loss": 0.2421,
      "step": 1175
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.390019548267663e-05,
      "loss": 0.2233,
      "step": 1176
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3890262563577375e-05,
      "loss": 0.2626,
      "step": 1177
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3880325119696243e-05,
      "loss": 0.2829,
      "step": 1178
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.3870383162591525e-05,
      "loss": 0.3007,
      "step": 1179
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3860436703826754e-05,
      "loss": 0.2729,
      "step": 1180
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3850485754970701e-05,
      "loss": 0.269,
      "step": 1181
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3840530327597361e-05,
      "loss": 0.2765,
      "step": 1182
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3830570433285932e-05,
      "loss": 0.2536,
      "step": 1183
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3820606083620816e-05,
      "loss": 0.2438,
      "step": 1184
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3810637290191591e-05,
      "loss": 0.2245,
      "step": 1185
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3800664064593006e-05,
      "loss": 0.2893,
      "step": 1186
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3790686418424964e-05,
      "loss": 0.3116,
      "step": 1187
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3780704363292512e-05,
      "loss": 0.2831,
      "step": 1188
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3770717910805819e-05,
      "loss": 0.3081,
      "step": 1189
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3760727072580177e-05,
      "loss": 0.2637,
      "step": 1190
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3750731860235975e-05,
      "loss": 0.3157,
      "step": 1191
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3740732285398686e-05,
      "loss": 0.2596,
      "step": 1192
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3730728359698862e-05,
      "loss": 0.2839,
      "step": 1193
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.3720720094772113e-05,
      "loss": 0.2683,
      "step": 1194
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3710707502259099e-05,
      "loss": 0.3168,
      "step": 1195
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.370069059380551e-05,
      "loss": 0.2896,
      "step": 1196
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3690669381062059e-05,
      "loss": 0.2921,
      "step": 1197
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3680643875684464e-05,
      "loss": 0.3424,
      "step": 1198
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.367061408933343e-05,
      "loss": 0.2857,
      "step": 1199
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3660580033674652e-05,
      "loss": 0.3153,
      "step": 1200
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3650541720378782e-05,
      "loss": 0.2671,
      "step": 1201
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.364049916112143e-05,
      "loss": 0.2865,
      "step": 1202
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.363045236758314e-05,
      "loss": 0.2965,
      "step": 1203
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3620401351449384e-05,
      "loss": 0.2921,
      "step": 1204
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3610346124410542e-05,
      "loss": 0.2943,
      "step": 1205
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3600286698161898e-05,
      "loss": 0.237,
      "step": 1206
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3590223084403608e-05,
      "loss": 0.257,
      "step": 1207
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3580155294840715e-05,
      "loss": 0.2643,
      "step": 1208
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.3570083341183105e-05,
      "loss": 0.3035,
      "step": 1209
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3560007235145516e-05,
      "loss": 0.3022,
      "step": 1210
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3549926988447508e-05,
      "loss": 0.2871,
      "step": 1211
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3539842612813465e-05,
      "loss": 0.2556,
      "step": 1212
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3529754119972569e-05,
      "loss": 0.2784,
      "step": 1213
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3519661521658792e-05,
      "loss": 0.2895,
      "step": 1214
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3509564829610877e-05,
      "loss": 0.2773,
      "step": 1215
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3499464055572334e-05,
      "loss": 0.2427,
      "step": 1216
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3489359211291418e-05,
      "loss": 0.2711,
      "step": 1217
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.347925030852112e-05,
      "loss": 0.2745,
      "step": 1218
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.346913735901915e-05,
      "loss": 0.284,
      "step": 1219
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3459020374547926e-05,
      "loss": 0.2732,
      "step": 1220
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3448899366874555e-05,
      "loss": 0.2882,
      "step": 1221
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3438774347770832e-05,
      "loss": 0.2703,
      "step": 1222
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3428645329013203e-05,
      "loss": 0.2733,
      "step": 1223
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.3418512322382786e-05,
      "loss": 0.3069,
      "step": 1224
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3408375339665316e-05,
      "loss": 0.211,
      "step": 1225
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3398234392651172e-05,
      "loss": 0.3113,
      "step": 1226
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3388089493135333e-05,
      "loss": 0.2619,
      "step": 1227
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3377940652917374e-05,
      "loss": 0.2814,
      "step": 1228
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.336778788380146e-05,
      "loss": 0.2419,
      "step": 1229
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3357631197596319e-05,
      "loss": 0.308,
      "step": 1230
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.334747060611524e-05,
      "loss": 0.2875,
      "step": 1231
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3337306121176057e-05,
      "loss": 0.2769,
      "step": 1232
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.332713775460112e-05,
      "loss": 0.2742,
      "step": 1233
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3316965518217309e-05,
      "loss": 0.2687,
      "step": 1234
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3306789423855993e-05,
      "loss": 0.2515,
      "step": 1235
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3296609483353036e-05,
      "loss": 0.2483,
      "step": 1236
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3286425708548772e-05,
      "loss": 0.2702,
      "step": 1237
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3276238111287989e-05,
      "loss": 0.3126,
      "step": 1238
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.3266046703419936e-05,
      "loss": 0.2613,
      "step": 1239
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3255851496798283e-05,
      "loss": 0.2445,
      "step": 1240
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3245652503281124e-05,
      "loss": 0.26,
      "step": 1241
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3235449734730945e-05,
      "loss": 0.2515,
      "step": 1242
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3225243203014637e-05,
      "loss": 0.3055,
      "step": 1243
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3215032920003466e-05,
      "loss": 0.2872,
      "step": 1244
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3204818897573052e-05,
      "loss": 0.2499,
      "step": 1245
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3194601147603374e-05,
      "loss": 0.2898,
      "step": 1246
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3184379681978743e-05,
      "loss": 0.3018,
      "step": 1247
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.317415451258779e-05,
      "loss": 0.324,
      "step": 1248
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3163925651323457e-05,
      "loss": 0.3327,
      "step": 1249
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3153693110082977e-05,
      "loss": 0.2733,
      "step": 1250
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3143456900767865e-05,
      "loss": 0.2732,
      "step": 1251
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3133217035283904e-05,
      "loss": 0.3165,
      "step": 1252
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3122973525541127e-05,
      "loss": 0.2854,
      "step": 1253
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.3112726383453804e-05,
      "loss": 0.3092,
      "step": 1254
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3102475620940437e-05,
      "loss": 0.2749,
      "step": 1255
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3092221249923723e-05,
      "loss": 0.3038,
      "step": 1256
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3081963282330579e-05,
      "loss": 0.2517,
      "step": 1257
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3071701730092088e-05,
      "loss": 0.2437,
      "step": 1258
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3061436605143509e-05,
      "loss": 0.2613,
      "step": 1259
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3051167919424255e-05,
      "loss": 0.2832,
      "step": 1260
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3040895684877877e-05,
      "loss": 0.2834,
      "step": 1261
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.303061991345206e-05,
      "loss": 0.3031,
      "step": 1262
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3020340617098604e-05,
      "loss": 0.2668,
      "step": 1263
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.3010057807773397e-05,
      "loss": 0.3228,
      "step": 1264
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2999771497436423e-05,
      "loss": 0.2836,
      "step": 1265
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2989481698051742e-05,
      "loss": 0.2652,
      "step": 1266
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.297918842158746e-05,
      "loss": 0.2655,
      "step": 1267
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.296889168001573e-05,
      "loss": 0.2682,
      "step": 1268
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.2958591485312742e-05,
      "loss": 0.2428,
      "step": 1269
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2948287849458703e-05,
      "loss": 0.2807,
      "step": 1270
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2937980784437813e-05,
      "loss": 0.2896,
      "step": 1271
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2927670302238267e-05,
      "loss": 0.3066,
      "step": 1272
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2917356414852234e-05,
      "loss": 0.296,
      "step": 1273
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2907039134275842e-05,
      "loss": 0.3257,
      "step": 1274
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2896718472509167e-05,
      "loss": 0.2273,
      "step": 1275
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2886394441556217e-05,
      "loss": 0.2931,
      "step": 1276
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2876067053424914e-05,
      "loss": 0.2961,
      "step": 1277
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2865736320127098e-05,
      "loss": 0.288,
      "step": 1278
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2855402253678486e-05,
      "loss": 0.3087,
      "step": 1279
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.284506486609868e-05,
      "loss": 0.2995,
      "step": 1280
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.283472416941114e-05,
      "loss": 0.2355,
      "step": 1281
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2824380175643177e-05,
      "loss": 0.3035,
      "step": 1282
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2814032896825934e-05,
      "loss": 0.2634,
      "step": 1283
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.2803682344994383e-05,
      "loss": 0.2949,
      "step": 1284
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2793328532187294e-05,
      "loss": 0.265,
      "step": 1285
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2782971470447234e-05,
      "loss": 0.1999,
      "step": 1286
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2772611171820546e-05,
      "loss": 0.2601,
      "step": 1287
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2762247648357344e-05,
      "loss": 0.2943,
      "step": 1288
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2751880912111487e-05,
      "loss": 0.2747,
      "step": 1289
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2741510975140572e-05,
      "loss": 0.2473,
      "step": 1290
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.273113784950592e-05,
      "loss": 0.2463,
      "step": 1291
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2720761547272558e-05,
      "loss": 0.2624,
      "step": 1292
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.271038208050921e-05,
      "loss": 0.2647,
      "step": 1293
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2699999461288284e-05,
      "loss": 0.2308,
      "step": 1294
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.268961370168585e-05,
      "loss": 0.26,
      "step": 1295
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2679224813781628e-05,
      "loss": 0.2751,
      "step": 1296
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.2668832809658981e-05,
      "loss": 0.2336,
      "step": 1297
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.26584377014049e-05,
      "loss": 0.2945,
      "step": 1298
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.264803950110998e-05,
      "loss": 0.2907,
      "step": 1299
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2637638220868411e-05,
      "loss": 0.2839,
      "step": 1300
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2627233872777974e-05,
      "loss": 0.26,
      "step": 1301
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2616826468940008e-05,
      "loss": 0.2562,
      "step": 1302
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2606416021459418e-05,
      "loss": 0.2482,
      "step": 1303
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2596002542444636e-05,
      "loss": 0.2976,
      "step": 1304
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.258558604400763e-05,
      "loss": 0.2702,
      "step": 1305
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2575166538263873e-05,
      "loss": 0.2887,
      "step": 1306
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2564744037332339e-05,
      "loss": 0.2592,
      "step": 1307
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2554318553335491e-05,
      "loss": 0.309,
      "step": 1308
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2543890098399251e-05,
      "loss": 0.2807,
      "step": 1309
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2533458684653004e-05,
      "loss": 0.2591,
      "step": 1310
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2523024324229573e-05,
      "loss": 0.2756,
      "step": 1311
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.251258702926521e-05,
      "loss": 0.2994,
      "step": 1312
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.250214681189958e-05,
      "loss": 0.3182,
      "step": 1313
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.2491703684275747e-05,
      "loss": 0.3461,
      "step": 1314
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2481257658540159e-05,
      "loss": 0.296,
      "step": 1315
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2470808746842634e-05,
      "loss": 0.2836,
      "step": 1316
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.246035696133635e-05,
      "loss": 0.2977,
      "step": 1317
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2449902314177827e-05,
      "loss": 0.2866,
      "step": 1318
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2439444817526909e-05,
      "loss": 0.2192,
      "step": 1319
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.242898448354676e-05,
      "loss": 0.2798,
      "step": 1320
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2418521324403843e-05,
      "loss": 0.253,
      "step": 1321
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.24080553522679e-05,
      "loss": 0.2791,
      "step": 1322
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2397586579311953e-05,
      "loss": 0.3228,
      "step": 1323
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2387115017712282e-05,
      "loss": 0.2291,
      "step": 1324
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2376640679648404e-05,
      "loss": 0.2467,
      "step": 1325
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2366163577303068e-05,
      "loss": 0.2413,
      "step": 1326
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.235568372286224e-05,
      "loss": 0.3041,
      "step": 1327
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2345201128515082e-05,
      "loss": 0.3076,
      "step": 1328
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.2334715806453954e-05,
      "loss": 0.2464,
      "step": 1329
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2324227768874376e-05,
      "loss": 0.2848,
      "step": 1330
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2313737027975035e-05,
      "loss": 0.2435,
      "step": 1331
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2303243595957754e-05,
      "loss": 0.257,
      "step": 1332
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2292747485027498e-05,
      "loss": 0.2608,
      "step": 1333
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2282248707392334e-05,
      "loss": 0.2753,
      "step": 1334
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2271747275263441e-05,
      "loss": 0.2407,
      "step": 1335
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2261243200855081e-05,
      "loss": 0.2689,
      "step": 1336
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2250736496384592e-05,
      "loss": 0.2919,
      "step": 1337
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2240227174072368e-05,
      "loss": 0.2609,
      "step": 1338
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2229715246141847e-05,
      "loss": 0.2308,
      "step": 1339
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.22192007248195e-05,
      "loss": 0.2696,
      "step": 1340
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2208683622334816e-05,
      "loss": 0.2754,
      "step": 1341
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2198163950920285e-05,
      "loss": 0.2758,
      "step": 1342
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.218764172281138e-05,
      "loss": 0.2541,
      "step": 1343
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2177116950246557e-05,
      "loss": 0.2597,
      "step": 1344
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.2166589645467222e-05,
      "loss": 0.2358,
      "step": 1345
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2156059820717732e-05,
      "loss": 0.2577,
      "step": 1346
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2145527488245373e-05,
      "loss": 0.2232,
      "step": 1347
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.213499266030035e-05,
      "loss": 0.2834,
      "step": 1348
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2124455349135764e-05,
      "loss": 0.2946,
      "step": 1349
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2113915567007614e-05,
      "loss": 0.2653,
      "step": 1350
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.210337332617477e-05,
      "loss": 0.2719,
      "step": 1351
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2092828638898955e-05,
      "loss": 0.3045,
      "step": 1352
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.208228151744474e-05,
      "loss": 0.2608,
      "step": 1353
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2071731974079537e-05,
      "loss": 0.2443,
      "step": 1354
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2061180021073564e-05,
      "loss": 0.2638,
      "step": 1355
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.205062567069984e-05,
      "loss": 0.2983,
      "step": 1356
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2040068935234184e-05,
      "loss": 0.2571,
      "step": 1357
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2029509826955177e-05,
      "loss": 0.3156,
      "step": 1358
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2018948358144169e-05,
      "loss": 0.2604,
      "step": 1359
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.2008384541085251e-05,
      "loss": 0.311,
      "step": 1360
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1997818388065248e-05,
      "loss": 0.2406,
      "step": 1361
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1987249911373695e-05,
      "loss": 0.2755,
      "step": 1362
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.197667912330284e-05,
      "loss": 0.2952,
      "step": 1363
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.196610603614761e-05,
      "loss": 0.276,
      "step": 1364
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1955530662205614e-05,
      "loss": 0.2694,
      "step": 1365
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1944953013777114e-05,
      "loss": 0.257,
      "step": 1366
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1934373103165021e-05,
      "loss": 0.2918,
      "step": 1367
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.192379094267488e-05,
      "loss": 0.2218,
      "step": 1368
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1913206544614846e-05,
      "loss": 0.2814,
      "step": 1369
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1902619921295679e-05,
      "loss": 0.2409,
      "step": 1370
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1892031085030731e-05,
      "loss": 0.2198,
      "step": 1371
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1881440048135924e-05,
      "loss": 0.2788,
      "step": 1372
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.187084682292974e-05,
      "loss": 0.2497,
      "step": 1373
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1860251421733207e-05,
      "loss": 0.2921,
      "step": 1374
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.1849653856869882e-05,
      "loss": 0.3042,
      "step": 1375
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1839054140665844e-05,
      "loss": 0.2883,
      "step": 1376
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.182845228544967e-05,
      "loss": 0.2598,
      "step": 1377
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1817848303552424e-05,
      "loss": 0.2729,
      "step": 1378
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1807242207307645e-05,
      "loss": 0.2642,
      "step": 1379
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.179663400905133e-05,
      "loss": 0.2647,
      "step": 1380
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1786023721121925e-05,
      "loss": 0.2709,
      "step": 1381
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1775411355860302e-05,
      "loss": 0.299,
      "step": 1382
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1764796925609751e-05,
      "loss": 0.272,
      "step": 1383
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1754180442715963e-05,
      "loss": 0.2429,
      "step": 1384
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.174356191952702e-05,
      "loss": 0.2008,
      "step": 1385
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1732941368393371e-05,
      "loss": 0.2921,
      "step": 1386
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1722318801667825e-05,
      "loss": 0.3148,
      "step": 1387
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1711694231705541e-05,
      "loss": 0.2258,
      "step": 1388
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.1701067670864001e-05,
      "loss": 0.3173,
      "step": 1389
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.169043913150301e-05,
      "loss": 0.3016,
      "step": 1390
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1679808625984663e-05,
      "loss": 0.2481,
      "step": 1391
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1669176166673356e-05,
      "loss": 0.2854,
      "step": 1392
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1658541765935742e-05,
      "loss": 0.3107,
      "step": 1393
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.164790543614075e-05,
      "loss": 0.2569,
      "step": 1394
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1637267189659538e-05,
      "loss": 0.2761,
      "step": 1395
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1626627038865497e-05,
      "loss": 0.3021,
      "step": 1396
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1615984996134237e-05,
      "loss": 0.2317,
      "step": 1397
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1605341073843564e-05,
      "loss": 0.2201,
      "step": 1398
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1594695284373472e-05,
      "loss": 0.3078,
      "step": 1399
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1584047640106129e-05,
      "loss": 0.2404,
      "step": 1400
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1573398153425854e-05,
      "loss": 0.2999,
      "step": 1401
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1562746836719114e-05,
      "loss": 0.239,
      "step": 1402
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1552093702374507e-05,
      "loss": 0.2236,
      "step": 1403
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1541438762782735e-05,
      "loss": 0.2149,
      "step": 1404
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.1530782030336611e-05,
      "loss": 0.2918,
      "step": 1405
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1520123517431028e-05,
      "loss": 0.2792,
      "step": 1406
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.150946323646295e-05,
      "loss": 0.2331,
      "step": 1407
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1498801199831395e-05,
      "loss": 0.2342,
      "step": 1408
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1488137419937428e-05,
      "loss": 0.2569,
      "step": 1409
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.147747190918414e-05,
      "loss": 0.2201,
      "step": 1410
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1466804679976631e-05,
      "loss": 0.2206,
      "step": 1411
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1456135744722007e-05,
      "loss": 0.2727,
      "step": 1412
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1445465115829355e-05,
      "loss": 0.2174,
      "step": 1413
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.143479280570973e-05,
      "loss": 0.2415,
      "step": 1414
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1424118826776143e-05,
      "loss": 0.2508,
      "step": 1415
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1413443191443547e-05,
      "loss": 0.3245,
      "step": 1416
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1402765912128821e-05,
      "loss": 0.2868,
      "step": 1417
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.139208700125076e-05,
      "loss": 0.279,
      "step": 1418
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1381406471230047e-05,
      "loss": 0.3163,
      "step": 1419
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1370724334489262e-05,
      "loss": 0.2539,
      "step": 1420
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1360040603452842e-05,
      "loss": 0.291,
      "step": 1421
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1349355290547082e-05,
      "loss": 0.2117,
      "step": 1422
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1338668408200117e-05,
      "loss": 0.2254,
      "step": 1423
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.132797996884191e-05,
      "loss": 0.2628,
      "step": 1424
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1317289984904232e-05,
      "loss": 0.2749,
      "step": 1425
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1306598468820653e-05,
      "loss": 0.2776,
      "step": 1426
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1295905433026521e-05,
      "loss": 0.2635,
      "step": 1427
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1285210889958958e-05,
      "loss": 0.2413,
      "step": 1428
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.127451485205683e-05,
      "loss": 0.2577,
      "step": 1429
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1263817331760749e-05,
      "loss": 0.2399,
      "step": 1430
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1253118341513049e-05,
      "loss": 0.2517,
      "step": 1431
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1242417893757774e-05,
      "loss": 0.2743,
      "step": 1432
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1231716000940662e-05,
      "loss": 0.3134,
      "step": 1433
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1221012675509134e-05,
      "loss": 0.2449,
      "step": 1434
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1210307929912273e-05,
      "loss": 0.2251,
      "step": 1435
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1199601776600819e-05,
      "loss": 0.2545,
      "step": 1436
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1188894228027147e-05,
      "loss": 0.245,
      "step": 1437
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1178185296645256e-05,
      "loss": 0.2337,
      "step": 1438
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.116747499491075e-05,
      "loss": 0.2553,
      "step": 1439
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.115676333528083e-05,
      "loss": 0.2672,
      "step": 1440
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1146050330214271e-05,
      "loss": 0.3138,
      "step": 1441
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1135335992171429e-05,
      "loss": 0.2398,
      "step": 1442
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1124620333614187e-05,
      "loss": 0.2546,
      "step": 1443
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1113903367005983e-05,
      "loss": 0.3185,
      "step": 1444
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1103185104811765e-05,
      "loss": 0.2546,
      "step": 1445
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1092465559497991e-05,
      "loss": 0.2614,
      "step": 1446
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1081744743532617e-05,
      "loss": 0.2379,
      "step": 1447
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1071022669385067e-05,
      "loss": 0.2664,
      "step": 1448
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.1060299349526239e-05,
      "loss": 0.244,
      "step": 1449
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.104957479642847e-05,
      "loss": 0.2625,
      "step": 1450
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.1038849022565538e-05,
      "loss": 0.2526,
      "step": 1451
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.1028122040412639e-05,
      "loss": 0.2246,
      "step": 1452
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.1017393862446373e-05,
      "loss": 0.2729,
      "step": 1453
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.1006664501144733e-05,
      "loss": 0.2965,
      "step": 1454
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0995933968987085e-05,
      "loss": 0.2428,
      "step": 1455
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0985202278454163e-05,
      "loss": 0.2614,
      "step": 1456
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0974469442028041e-05,
      "loss": 0.2996,
      "step": 1457
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0963735472192134e-05,
      "loss": 0.243,
      "step": 1458
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0953000381431167e-05,
      "loss": 0.2542,
      "step": 1459
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0942264182231173e-05,
      "loss": 0.3189,
      "step": 1460
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.093152688707947e-05,
      "loss": 0.25,
      "step": 1461
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0920788508464662e-05,
      "loss": 0.2718,
      "step": 1462
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0910049058876599e-05,
      "loss": 0.2498,
      "step": 1463
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0899308550806384e-05,
      "loss": 0.2813,
      "step": 1464
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.0888566996746354e-05,
      "loss": 0.2267,
      "step": 1465
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0877824409190055e-05,
      "loss": 0.259,
      "step": 1466
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0867080800632238e-05,
      "loss": 0.2419,
      "step": 1467
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.085633618356885e-05,
      "loss": 0.2796,
      "step": 1468
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0845590570496993e-05,
      "loss": 0.2696,
      "step": 1469
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0834843973914943e-05,
      "loss": 0.2521,
      "step": 1470
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0824096406322117e-05,
      "loss": 0.3053,
      "step": 1471
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0813347880219053e-05,
      "loss": 0.2758,
      "step": 1472
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0802598408107416e-05,
      "loss": 0.2551,
      "step": 1473
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0791848002489967e-05,
      "loss": 0.2767,
      "step": 1474
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0781096675870541e-05,
      "loss": 0.2944,
      "step": 1475
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0770344440754063e-05,
      "loss": 0.2412,
      "step": 1476
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0759591309646504e-05,
      "loss": 0.2661,
      "step": 1477
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0748837295054876e-05,
      "loss": 0.2936,
      "step": 1478
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0738082409487227e-05,
      "loss": 0.2535,
      "step": 1479
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.0727326665452608e-05,
      "loss": 0.2655,
      "step": 1480
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0716570075461079e-05,
      "loss": 0.3016,
      "step": 1481
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.070581265202367e-05,
      "loss": 0.2564,
      "step": 1482
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0695054407652395e-05,
      "loss": 0.2302,
      "step": 1483
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0684295354860215e-05,
      "loss": 0.2663,
      "step": 1484
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.067353550616103e-05,
      "loss": 0.2929,
      "step": 1485
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0662774874069671e-05,
      "loss": 0.262,
      "step": 1486
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0652013471101876e-05,
      "loss": 0.2614,
      "step": 1487
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0641251309774283e-05,
      "loss": 0.2626,
      "step": 1488
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.063048840260441e-05,
      "loss": 0.2347,
      "step": 1489
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0619724762110644e-05,
      "loss": 0.2216,
      "step": 1490
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0608960400812218e-05,
      "loss": 0.2683,
      "step": 1491
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0598195331229214e-05,
      "loss": 0.338,
      "step": 1492
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0587429565882531e-05,
      "loss": 0.2466,
      "step": 1493
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0576663117293882e-05,
      "loss": 0.2787,
      "step": 1494
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0565895997985766e-05,
      "loss": 0.288,
      "step": 1495
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0555128220481475e-05,
      "loss": 0.2134,
      "step": 1496
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0544359797305052e-05,
      "loss": 0.2629,
      "step": 1497
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0533590740981305e-05,
      "loss": 0.2493,
      "step": 1498
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0522821064035766e-05,
      "loss": 0.2667,
      "step": 1499
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0512050778994701e-05,
      "loss": 0.2978,
      "step": 1500
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.050127989838507e-05,
      "loss": 0.2584,
      "step": 1501
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0490508434734536e-05,
      "loss": 0.2529,
      "step": 1502
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0479736400571438e-05,
      "loss": 0.24,
      "step": 1503
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0468963808424777e-05,
      "loss": 0.1874,
      "step": 1504
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0458190670824202e-05,
      "loss": 0.1784,
      "step": 1505
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0447417000299998e-05,
      "loss": 0.1812,
      "step": 1506
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.043664280938307e-05,
      "loss": 0.2459,
      "step": 1507
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0425868110604926e-05,
      "loss": 0.1983,
      "step": 1508
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0415092916497671e-05,
      "loss": 0.1692,
      "step": 1509
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.040431723959398e-05,
      "loss": 0.2615,
      "step": 1510
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0393541092427088e-05,
      "loss": 0.2177,
      "step": 1511
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0382764487530783e-05,
      "loss": 0.2013,
      "step": 1512
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.037198743743938e-05,
      "loss": 0.221,
      "step": 1513
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.036120995468772e-05,
      "loss": 0.168,
      "step": 1514
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0350432051811136e-05,
      "loss": 0.1808,
      "step": 1515
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0339653741345458e-05,
      "loss": 0.2384,
      "step": 1516
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0328875035826987e-05,
      "loss": 0.1851,
      "step": 1517
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0318095947792483e-05,
      "loss": 0.2125,
      "step": 1518
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0307316489779154e-05,
      "loss": 0.2373,
      "step": 1519
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0296536674324632e-05,
      "loss": 0.2093,
      "step": 1520
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0285756513966974e-05,
      "loss": 0.1637,
      "step": 1521
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.027497602124463e-05,
      "loss": 0.2011,
      "step": 1522
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.0264195208696437e-05,
      "loss": 0.2106,
      "step": 1523
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.025341408886161e-05,
      "loss": 0.1942,
      "step": 1524
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.024263267427972e-05,
      "loss": 0.1897,
      "step": 1525
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0231850977490674e-05,
      "loss": 0.2101,
      "step": 1526
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0221069011034715e-05,
      "loss": 0.2027,
      "step": 1527
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0210286787452392e-05,
      "loss": 0.1687,
      "step": 1528
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0199504319284565e-05,
      "loss": 0.1704,
      "step": 1529
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0188721619072368e-05,
      "loss": 0.2119,
      "step": 1530
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0177938699357207e-05,
      "loss": 0.1753,
      "step": 1531
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0167155572680743e-05,
      "loss": 0.2468,
      "step": 1532
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0156372251584882e-05,
      "loss": 0.1771,
      "step": 1533
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0145588748611753e-05,
      "loss": 0.165,
      "step": 1534
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0134805076303696e-05,
      "loss": 0.1508,
      "step": 1535
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0124021247203248e-05,
      "loss": 0.1811,
      "step": 1536
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0113237273853129e-05,
      "loss": 0.1762,
      "step": 1537
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.010245316879623e-05,
      "loss": 0.2313,
      "step": 1538
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.009166894457559e-05,
      "loss": 0.1967,
      "step": 1539
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.0080884613734388e-05,
      "loss": 0.1998,
      "step": 1540
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.0070100188815928e-05,
      "loss": 0.2039,
      "step": 1541
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.0059315682363621e-05,
      "loss": 0.1908,
      "step": 1542
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.004853110692098e-05,
      "loss": 0.1617,
      "step": 1543
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.003774647503159e-05,
      "loss": 0.2079,
      "step": 1544
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.0026961799239105e-05,
      "loss": 0.2095,
      "step": 1545
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.001617709208723e-05,
      "loss": 0.2446,
      "step": 1546
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.0005392366119706e-05,
      "loss": 0.1931,
      "step": 1547
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.994607633880298e-06,
      "loss": 0.2332,
      "step": 1548
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.983822907912774e-06,
      "loss": 0.1759,
      "step": 1549
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.9730382007609e-06,
      "loss": 0.1822,
      "step": 1550
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.962253524968414e-06,
      "loss": 0.2111,
      "step": 1551
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.951468893079026e-06,
      "loss": 0.1712,
      "step": 1552
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.940684317636384e-06,
      "loss": 0.1923,
      "step": 1553
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.929899811184079e-06,
      "loss": 0.1664,
      "step": 1554
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.919115386265619e-06,
      "loss": 0.1993,
      "step": 1555
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.908331055424416e-06,
      "loss": 0.1682,
      "step": 1556
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.897546831203776e-06,
      "loss": 0.2119,
      "step": 1557
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.886762726146876e-06,
      "loss": 0.1892,
      "step": 1558
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.875978752796757e-06,
      "loss": 0.2298,
      "step": 1559
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.86519492369631e-06,
      "loss": 0.2276,
      "step": 1560
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.854411251388252e-06,
      "loss": 0.185,
      "step": 1561
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.843627748415121e-06,
      "loss": 0.1727,
      "step": 1562
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.83284442731926e-06,
      "loss": 0.2141,
      "step": 1563
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.822061300642798e-06,
      "loss": 0.2228,
      "step": 1564
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.811278380927635e-06,
      "loss": 0.2112,
      "step": 1565
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.800495680715438e-06,
      "loss": 0.182,
      "step": 1566
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.78971321254761e-06,
      "loss": 0.2266,
      "step": 1567
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.778930988965289e-06,
      "loss": 0.204,
      "step": 1568
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.76814902250933e-06,
      "loss": 0.1913,
      "step": 1569
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.757367325720283e-06,
      "loss": 0.2287,
      "step": 1570
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.746585911138391e-06,
      "loss": 0.1715,
      "step": 1571
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.735804791303567e-06,
      "loss": 0.1846,
      "step": 1572
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.725023978755373e-06,
      "loss": 0.1954,
      "step": 1573
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.71424348603303e-06,
      "loss": 0.1985,
      "step": 1574
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.70346332567537e-06,
      "loss": 0.1467,
      "step": 1575
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.69268351022085e-06,
      "loss": 0.2006,
      "step": 1576
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.68190405220752e-06,
      "loss": 0.1987,
      "step": 1577
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.671124964173017e-06,
      "loss": 0.1673,
      "step": 1578
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.660346258654543e-06,
      "loss": 0.1874,
      "step": 1579
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.649567948188866e-06,
      "loss": 0.2059,
      "step": 1580
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.638790045312283e-06,
      "loss": 0.2495,
      "step": 1581
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.628012562560621e-06,
      "loss": 0.181,
      "step": 1582
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.61723551246922e-06,
      "loss": 0.1937,
      "step": 1583
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.606458907572915e-06,
      "loss": 0.1668,
      "step": 1584
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.595682760406025e-06,
      "loss": 0.1675,
      "step": 1585
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.58490708350233e-06,
      "loss": 0.1931,
      "step": 1586
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.574131889395076e-06,
      "loss": 0.21,
      "step": 1587
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.563357190616933e-06,
      "loss": 0.1542,
      "step": 1588
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.552582999700005e-06,
      "loss": 0.2133,
      "step": 1589
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.541809329175801e-06,
      "loss": 0.1919,
      "step": 1590
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.531036191575226e-06,
      "loss": 0.224,
      "step": 1591
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.520263599428564e-06,
      "loss": 0.1941,
      "step": 1592
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.509491565265467e-06,
      "loss": 0.1899,
      "step": 1593
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.498720101614933e-06,
      "loss": 0.161,
      "step": 1594
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.487949221005304e-06,
      "loss": 0.2195,
      "step": 1595
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.477178935964234e-06,
      "loss": 0.2426,
      "step": 1596
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.466409259018697e-06,
      "loss": 0.2029,
      "step": 1597
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.455640202694948e-06,
      "loss": 0.184,
      "step": 1598
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.444871779518527e-06,
      "loss": 0.1813,
      "step": 1599
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.434104002014233e-06,
      "loss": 0.2003,
      "step": 1600
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.42333688270612e-06,
      "loss": 0.1873,
      "step": 1601
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.412570434117469e-06,
      "loss": 0.2338,
      "step": 1602
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.401804668770787e-06,
      "loss": 0.2019,
      "step": 1603
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.391039599187782e-06,
      "loss": 0.1902,
      "step": 1604
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.380275237889358e-06,
      "loss": 0.185,
      "step": 1605
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.36951159739559e-06,
      "loss": 0.2289,
      "step": 1606
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.358748690225719e-06,
      "loss": 0.1954,
      "step": 1607
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.347986528898127e-06,
      "loss": 0.19,
      "step": 1608
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.337225125930332e-06,
      "loss": 0.1722,
      "step": 1609
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.326464493838975e-06,
      "loss": 0.1973,
      "step": 1610
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.31570464513979e-06,
      "loss": 0.208,
      "step": 1611
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.30494559234761e-06,
      "loss": 0.1772,
      "step": 1612
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.294187347976335e-06,
      "loss": 0.195,
      "step": 1613
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.283429924538928e-06,
      "loss": 0.1786,
      "step": 1614
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.272673334547395e-06,
      "loss": 0.2139,
      "step": 1615
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.261917590512778e-06,
      "loss": 0.191,
      "step": 1616
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.251162704945127e-06,
      "loss": 0.2003,
      "step": 1617
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.240408690353501e-06,
      "loss": 0.1862,
      "step": 1618
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.229655559245942e-06,
      "loss": 0.1932,
      "step": 1619
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.218903324129464e-06,
      "loss": 0.2124,
      "step": 1620
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.20815199751004e-06,
      "loss": 0.1838,
      "step": 1621
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.197401591892587e-06,
      "loss": 0.2296,
      "step": 1622
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.18665211978095e-06,
      "loss": 0.1834,
      "step": 1623
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.175903593677886e-06,
      "loss": 0.2058,
      "step": 1624
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.165156026085058e-06,
      "loss": 0.2091,
      "step": 1625
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.154409429503009e-06,
      "loss": 0.217,
      "step": 1626
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.143663816431154e-06,
      "loss": 0.1547,
      "step": 1627
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.132919199367763e-06,
      "loss": 0.1542,
      "step": 1628
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.122175590809947e-06,
      "loss": 0.1891,
      "step": 1629
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.11143300325365e-06,
      "loss": 0.1935,
      "step": 1630
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.100691449193618e-06,
      "loss": 0.1548,
      "step": 1631
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.089950941123404e-06,
      "loss": 0.1965,
      "step": 1632
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.079211491535341e-06,
      "loss": 0.1609,
      "step": 1633
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.068473112920531e-06,
      "loss": 0.1895,
      "step": 1634
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.057735817768832e-06,
      "loss": 0.2219,
      "step": 1635
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.046999618568838e-06,
      "loss": 0.2064,
      "step": 1636
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.03626452780787e-06,
      "loss": 0.1858,
      "step": 1637
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.02553055797196e-06,
      "loss": 0.1409,
      "step": 1638
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.01479772154584e-06,
      "loss": 0.1923,
      "step": 1639
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.00406603101292e-06,
      "loss": 0.1862,
      "step": 1640
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.99333549885527e-06,
      "loss": 0.1921,
      "step": 1641
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.98260613755363e-06,
      "loss": 0.1984,
      "step": 1642
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.971877959587365e-06,
      "loss": 0.1885,
      "step": 1643
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.961150977434465e-06,
      "loss": 0.1771,
      "step": 1644
    },
    {
      "epoch": 1.09,
      "learning_rate": 8.950425203571533e-06,
      "loss": 0.2207,
      "step": 1645
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.939700650473765e-06,
      "loss": 0.1798,
      "step": 1646
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.928977330614936e-06,
      "loss": 0.1738,
      "step": 1647
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.918255256467386e-06,
      "loss": 0.1768,
      "step": 1648
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.907534440502012e-06,
      "loss": 0.2035,
      "step": 1649
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.89681489518824e-06,
      "loss": 0.1868,
      "step": 1650
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.88609663299402e-06,
      "loss": 0.2174,
      "step": 1651
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.875379666385815e-06,
      "loss": 0.2113,
      "step": 1652
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.864664007828573e-06,
      "loss": 0.1844,
      "step": 1653
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.853949669785729e-06,
      "loss": 0.1897,
      "step": 1654
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.843236664719174e-06,
      "loss": 0.1946,
      "step": 1655
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.832525005089252e-06,
      "loss": 0.1909,
      "step": 1656
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.821814703354747e-06,
      "loss": 0.1933,
      "step": 1657
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.811105771972853e-06,
      "loss": 0.2281,
      "step": 1658
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.800398223399181e-06,
      "loss": 0.2154,
      "step": 1659
    },
    {
      "epoch": 1.1,
      "learning_rate": 8.789692070087727e-06,
      "loss": 0.184,
      "step": 1660
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.778987324490868e-06,
      "loss": 0.2123,
      "step": 1661
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.76828399905934e-06,
      "loss": 0.2119,
      "step": 1662
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.757582106242228e-06,
      "loss": 0.1832,
      "step": 1663
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.746881658486953e-06,
      "loss": 0.176,
      "step": 1664
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.736182668239251e-06,
      "loss": 0.223,
      "step": 1665
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.725485147943171e-06,
      "loss": 0.1703,
      "step": 1666
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.714789110041047e-06,
      "loss": 0.2125,
      "step": 1667
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.70409456697348e-06,
      "loss": 0.1898,
      "step": 1668
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.69340153117935e-06,
      "loss": 0.2206,
      "step": 1669
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.68271001509577e-06,
      "loss": 0.1884,
      "step": 1670
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.672020031158094e-06,
      "loss": 0.1878,
      "step": 1671
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.661331591799888e-06,
      "loss": 0.1816,
      "step": 1672
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.650644709452925e-06,
      "loss": 0.176,
      "step": 1673
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.639959396547165e-06,
      "loss": 0.1785,
      "step": 1674
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.629275665510743e-06,
      "loss": 0.1658,
      "step": 1675
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.618593528769957e-06,
      "loss": 0.248,
      "step": 1676
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.607912998749246e-06,
      "loss": 0.2224,
      "step": 1677
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.597234087871184e-06,
      "loss": 0.1865,
      "step": 1678
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.586556808556458e-06,
      "loss": 0.1816,
      "step": 1679
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.575881173223862e-06,
      "loss": 0.2252,
      "step": 1680
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.565207194290274e-06,
      "loss": 0.1893,
      "step": 1681
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.554534884170648e-06,
      "loss": 0.1779,
      "step": 1682
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.543864255277996e-06,
      "loss": 0.2257,
      "step": 1683
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.533195320023374e-06,
      "loss": 0.1724,
      "step": 1684
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.522528090815864e-06,
      "loss": 0.1691,
      "step": 1685
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.511862580062574e-06,
      "loss": 0.1823,
      "step": 1686
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.501198800168609e-06,
      "loss": 0.2091,
      "step": 1687
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.490536763537054e-06,
      "loss": 0.1925,
      "step": 1688
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.479876482568974e-06,
      "loss": 0.2685,
      "step": 1689
    },
    {
      "epoch": 1.12,
      "learning_rate": 8.46921796966339e-06,
      "loss": 0.2019,
      "step": 1690
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.458561237217266e-06,
      "loss": 0.2337,
      "step": 1691
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.447906297625496e-06,
      "loss": 0.1581,
      "step": 1692
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.437253163280887e-06,
      "loss": 0.1951,
      "step": 1693
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.42660184657415e-06,
      "loss": 0.2262,
      "step": 1694
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.415952359893875e-06,
      "loss": 0.16,
      "step": 1695
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.40530471562653e-06,
      "loss": 0.161,
      "step": 1696
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.394658926156439e-06,
      "loss": 0.1697,
      "step": 1697
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.384015003865766e-06,
      "loss": 0.2263,
      "step": 1698
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.373372961134506e-06,
      "loss": 0.1951,
      "step": 1699
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.362732810340466e-06,
      "loss": 0.1762,
      "step": 1700
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.352094563859253e-06,
      "loss": 0.163,
      "step": 1701
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.34145823406426e-06,
      "loss": 0.1805,
      "step": 1702
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.330823833326648e-06,
      "loss": 0.1878,
      "step": 1703
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.320191374015339e-06,
      "loss": 0.1647,
      "step": 1704
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.309560868496994e-06,
      "loss": 0.1636,
      "step": 1705
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.298932329136e-06,
      "loss": 0.1924,
      "step": 1706
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.288305768294462e-06,
      "loss": 0.1814,
      "step": 1707
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.277681198332177e-06,
      "loss": 0.1648,
      "step": 1708
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.267058631606632e-06,
      "loss": 0.2144,
      "step": 1709
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.256438080472982e-06,
      "loss": 0.2129,
      "step": 1710
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.245819557284037e-06,
      "loss": 0.2077,
      "step": 1711
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.23520307439025e-06,
      "loss": 0.1617,
      "step": 1712
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.2245886441397e-06,
      "loss": 0.1798,
      "step": 1713
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.213976278878077e-06,
      "loss": 0.1968,
      "step": 1714
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.203365990948672e-06,
      "loss": 0.1976,
      "step": 1715
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.192757792692358e-06,
      "loss": 0.1813,
      "step": 1716
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.18215169644758e-06,
      "loss": 0.2177,
      "step": 1717
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.171547714550331e-06,
      "loss": 0.2097,
      "step": 1718
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.160945859334158e-06,
      "loss": 0.1486,
      "step": 1719
    },
    {
      "epoch": 1.14,
      "learning_rate": 8.150346143130116e-06,
      "loss": 0.1956,
      "step": 1720
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.139748578266793e-06,
      "loss": 0.1738,
      "step": 1721
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.129153177070262e-06,
      "loss": 0.2269,
      "step": 1722
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.118559951864078e-06,
      "loss": 0.1852,
      "step": 1723
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.107968914969269e-06,
      "loss": 0.1853,
      "step": 1724
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.097380078704321e-06,
      "loss": 0.2411,
      "step": 1725
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.08679345538516e-06,
      "loss": 0.2078,
      "step": 1726
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.076209057325123e-06,
      "loss": 0.1868,
      "step": 1727
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.065626896834982e-06,
      "loss": 0.2252,
      "step": 1728
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.055046986222889e-06,
      "loss": 0.1961,
      "step": 1729
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.044469337794389e-06,
      "loss": 0.1594,
      "step": 1730
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.033893963852393e-06,
      "loss": 0.2034,
      "step": 1731
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.023320876697164e-06,
      "loss": 0.1746,
      "step": 1732
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.012750088626308e-06,
      "loss": 0.1657,
      "step": 1733
    },
    {
      "epoch": 1.15,
      "learning_rate": 8.002181611934757e-06,
      "loss": 0.2079,
      "step": 1734
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.991615458914752e-06,
      "loss": 0.1689,
      "step": 1735
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.981051641855834e-06,
      "loss": 0.2068,
      "step": 1736
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.970490173044827e-06,
      "loss": 0.2092,
      "step": 1737
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.959931064765823e-06,
      "loss": 0.1757,
      "step": 1738
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.949374329300165e-06,
      "loss": 0.2557,
      "step": 1739
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.938819978926443e-06,
      "loss": 0.1816,
      "step": 1740
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.928268025920466e-06,
      "loss": 0.1471,
      "step": 1741
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.917718482555261e-06,
      "loss": 0.187,
      "step": 1742
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.907171361101049e-06,
      "loss": 0.2315,
      "step": 1743
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.896626673825232e-06,
      "loss": 0.173,
      "step": 1744
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.886084432992387e-06,
      "loss": 0.2078,
      "step": 1745
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.875544650864238e-06,
      "loss": 0.1652,
      "step": 1746
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.865007339699652e-06,
      "loss": 0.1782,
      "step": 1747
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.854472511754629e-06,
      "loss": 0.215,
      "step": 1748
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.843940179282271e-06,
      "loss": 0.25,
      "step": 1749
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.833410354532782e-06,
      "loss": 0.1637,
      "step": 1750
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.822883049753448e-06,
      "loss": 0.2353,
      "step": 1751
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.812358277188622e-06,
      "loss": 0.2094,
      "step": 1752
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.801836049079719e-06,
      "loss": 0.219,
      "step": 1753
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.791316377665186e-06,
      "loss": 0.2083,
      "step": 1754
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.780799275180502e-06,
      "loss": 0.1862,
      "step": 1755
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.770284753858158e-06,
      "loss": 0.1797,
      "step": 1756
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.759772825927636e-06,
      "loss": 0.1817,
      "step": 1757
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.74926350361541e-06,
      "loss": 0.1878,
      "step": 1758
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.73875679914492e-06,
      "loss": 0.2106,
      "step": 1759
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.72825272473656e-06,
      "loss": 0.1797,
      "step": 1760
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.717751292607667e-06,
      "loss": 0.1725,
      "step": 1761
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.707252514972505e-06,
      "loss": 0.2058,
      "step": 1762
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.696756404042248e-06,
      "loss": 0.1643,
      "step": 1763
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.68626297202497e-06,
      "loss": 0.1993,
      "step": 1764
    },
    {
      "epoch": 1.17,
      "learning_rate": 7.675772231125625e-06,
      "loss": 0.1773,
      "step": 1765
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.665284193546048e-06,
      "loss": 0.1969,
      "step": 1766
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.65479887148492e-06,
      "loss": 0.2125,
      "step": 1767
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.644316277137765e-06,
      "loss": 0.2135,
      "step": 1768
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.633836422696937e-06,
      "loss": 0.2163,
      "step": 1769
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.6233593203515995e-06,
      "loss": 0.2002,
      "step": 1770
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.6128849822877205e-06,
      "loss": 0.1758,
      "step": 1771
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.602413420688048e-06,
      "loss": 0.1991,
      "step": 1772
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.5919446477321015e-06,
      "loss": 0.1796,
      "step": 1773
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.581478675596159e-06,
      "loss": 0.1648,
      "step": 1774
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.57101551645324e-06,
      "loss": 0.16,
      "step": 1775
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.560555182473092e-06,
      "loss": 0.2124,
      "step": 1776
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.550097685822175e-06,
      "loss": 0.1484,
      "step": 1777
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.539643038663651e-06,
      "loss": 0.1709,
      "step": 1778
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.529191253157369e-06,
      "loss": 0.1712,
      "step": 1779
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.518742341459845e-06,
      "loss": 0.205,
      "step": 1780
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.508296315724257e-06,
      "loss": 0.1542,
      "step": 1781
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.4978531881004215e-06,
      "loss": 0.1963,
      "step": 1782
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.487412970734791e-06,
      "loss": 0.1852,
      "step": 1783
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.476975675770428e-06,
      "loss": 0.2345,
      "step": 1784
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.466541315346998e-06,
      "loss": 0.1723,
      "step": 1785
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.456109901600755e-06,
      "loss": 0.1891,
      "step": 1786
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.445681446664515e-06,
      "loss": 0.2299,
      "step": 1787
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.435255962667664e-06,
      "loss": 0.2028,
      "step": 1788
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.424833461736133e-06,
      "loss": 0.1877,
      "step": 1789
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.414413955992377e-06,
      "loss": 0.2079,
      "step": 1790
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.40399745755537e-06,
      "loss": 0.1447,
      "step": 1791
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.393583978540586e-06,
      "loss": 0.1964,
      "step": 1792
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.383173531059994e-06,
      "loss": 0.178,
      "step": 1793
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.372766127222031e-06,
      "loss": 0.1981,
      "step": 1794
    },
    {
      "epoch": 1.19,
      "learning_rate": 7.362361779131593e-06,
      "loss": 0.2029,
      "step": 1795
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.351960498890025e-06,
      "loss": 0.1996,
      "step": 1796
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.341562298595104e-06,
      "loss": 0.1856,
      "step": 1797
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.331167190341022e-06,
      "loss": 0.1816,
      "step": 1798
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.320775186218378e-06,
      "loss": 0.2012,
      "step": 1799
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.310386298314156e-06,
      "loss": 0.1696,
      "step": 1800
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.30000053871172e-06,
      "loss": 0.2724,
      "step": 1801
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.289617919490792e-06,
      "loss": 0.1917,
      "step": 1802
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.279238452727446e-06,
      "loss": 0.1523,
      "step": 1803
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.268862150494085e-06,
      "loss": 0.1983,
      "step": 1804
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.258489024859432e-06,
      "loss": 0.1876,
      "step": 1805
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.248119087888517e-06,
      "loss": 0.1636,
      "step": 1806
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.237752351642658e-06,
      "loss": 0.1687,
      "step": 1807
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.227388828179456e-06,
      "loss": 0.1981,
      "step": 1808
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.217028529552768e-06,
      "loss": 0.2071,
      "step": 1809
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.206671467812708e-06,
      "loss": 0.1901,
      "step": 1810
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.1963176550056194e-06,
      "loss": 0.2059,
      "step": 1811
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.185967103174067e-06,
      "loss": 0.1841,
      "step": 1812
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.175619824356827e-06,
      "loss": 0.2169,
      "step": 1813
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.165275830588864e-06,
      "loss": 0.1918,
      "step": 1814
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.154935133901323e-06,
      "loss": 0.2064,
      "step": 1815
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.144597746321515e-06,
      "loss": 0.2,
      "step": 1816
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.1342636798729035e-06,
      "loss": 0.1635,
      "step": 1817
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.1239329465750875e-06,
      "loss": 0.1806,
      "step": 1818
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.113605558443788e-06,
      "loss": 0.1644,
      "step": 1819
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.103281527490838e-06,
      "loss": 0.2127,
      "step": 1820
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.092960865724161e-06,
      "loss": 0.2005,
      "step": 1821
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.082643585147768e-06,
      "loss": 0.2263,
      "step": 1822
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.072329697761735e-06,
      "loss": 0.2151,
      "step": 1823
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.06201921556219e-06,
      "loss": 0.2343,
      "step": 1824
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.0517121505413e-06,
      "loss": 0.2197,
      "step": 1825
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.04140851468726e-06,
      "loss": 0.1681,
      "step": 1826
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.031108319984272e-06,
      "loss": 0.1944,
      "step": 1827
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.020811578412545e-06,
      "loss": 0.1593,
      "step": 1828
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.010518301948261e-06,
      "loss": 0.1899,
      "step": 1829
    },
    {
      "epoch": 1.22,
      "learning_rate": 7.000228502563577e-06,
      "loss": 0.1571,
      "step": 1830
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.9899421922266045e-06,
      "loss": 0.1799,
      "step": 1831
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.979659382901398e-06,
      "loss": 0.1905,
      "step": 1832
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.9693800865479385e-06,
      "loss": 0.178,
      "step": 1833
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.959104315122122e-06,
      "loss": 0.2151,
      "step": 1834
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.948832080575747e-06,
      "loss": 0.181,
      "step": 1835
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.9385633948564925e-06,
      "loss": 0.1834,
      "step": 1836
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.9282982699079115e-06,
      "loss": 0.1653,
      "step": 1837
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.918036717669422e-06,
      "loss": 0.1927,
      "step": 1838
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.907778750076277e-06,
      "loss": 0.1732,
      "step": 1839
    },
    {
      "epoch": 1.22,
      "learning_rate": 6.897524379059568e-06,
      "loss": 0.1733,
      "step": 1840
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.8872736165461975e-06,
      "loss": 0.1675,
      "step": 1841
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.8770264744588755e-06,
      "loss": 0.2123,
      "step": 1842
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.8667829647160966e-06,
      "loss": 0.1855,
      "step": 1843
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.856543099232134e-06,
      "loss": 0.1631,
      "step": 1844
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.846306889917027e-06,
      "loss": 0.1792,
      "step": 1845
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.836074348676549e-06,
      "loss": 0.1846,
      "step": 1846
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.825845487412215e-06,
      "loss": 0.2016,
      "step": 1847
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.815620318021263e-06,
      "loss": 0.201,
      "step": 1848
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.8053988523966305e-06,
      "loss": 0.1904,
      "step": 1849
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.795181102426953e-06,
      "loss": 0.1631,
      "step": 1850
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.784967079996541e-06,
      "loss": 0.2057,
      "step": 1851
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.7747567969853665e-06,
      "loss": 0.161,
      "step": 1852
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.764550265269061e-06,
      "loss": 0.2203,
      "step": 1853
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.7543474967188805e-06,
      "loss": 0.1815,
      "step": 1854
    },
    {
      "epoch": 1.23,
      "learning_rate": 6.744148503201717e-06,
      "loss": 0.2066,
      "step": 1855
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.7339532965800645e-06,
      "loss": 0.1922,
      "step": 1856
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.723761888712012e-06,
      "loss": 0.1705,
      "step": 1857
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.713574291451234e-06,
      "loss": 0.1843,
      "step": 1858
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.703390516646968e-06,
      "loss": 0.1612,
      "step": 1859
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.693210576144012e-06,
      "loss": 0.1906,
      "step": 1860
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.683034481782695e-06,
      "loss": 0.1896,
      "step": 1861
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.6728622453988825e-06,
      "loss": 0.2249,
      "step": 1862
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.662693878823947e-06,
      "loss": 0.2095,
      "step": 1863
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.652529393884761e-06,
      "loss": 0.1606,
      "step": 1864
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.6423688024036835e-06,
      "loss": 0.1938,
      "step": 1865
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.632212116198544e-06,
      "loss": 0.2192,
      "step": 1866
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.622059347082629e-06,
      "loss": 0.1835,
      "step": 1867
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.611910506864671e-06,
      "loss": 0.1633,
      "step": 1868
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.60176560734883e-06,
      "loss": 0.1669,
      "step": 1869
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.591624660334685e-06,
      "loss": 0.1795,
      "step": 1870
    },
    {
      "epoch": 1.24,
      "learning_rate": 6.581487677617217e-06,
      "loss": 0.1831,
      "step": 1871
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.571354670986798e-06,
      "loss": 0.2155,
      "step": 1872
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.561225652229173e-06,
      "loss": 0.1898,
      "step": 1873
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.551100633125446e-06,
      "loss": 0.1981,
      "step": 1874
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.540979625452077e-06,
      "loss": 0.2052,
      "step": 1875
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.530862640980852e-06,
      "loss": 0.2226,
      "step": 1876
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.520749691478883e-06,
      "loss": 0.1662,
      "step": 1877
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.510640788708585e-06,
      "loss": 0.2262,
      "step": 1878
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.500535944427669e-06,
      "loss": 0.2035,
      "step": 1879
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.490435170389127e-06,
      "loss": 0.1818,
      "step": 1880
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.4803384783412125e-06,
      "loss": 0.2086,
      "step": 1881
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.470245880027433e-06,
      "loss": 0.1721,
      "step": 1882
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.460157387186536e-06,
      "loss": 0.1745,
      "step": 1883
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.450073011552493e-06,
      "loss": 0.1684,
      "step": 1884
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.439992764854487e-06,
      "loss": 0.1473,
      "step": 1885
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.429916658816897e-06,
      "loss": 0.1903,
      "step": 1886
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.4198447051592886e-06,
      "loss": 0.1981,
      "step": 1887
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.4097769155963944e-06,
      "loss": 0.1817,
      "step": 1888
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.399713301838104e-06,
      "loss": 0.1611,
      "step": 1889
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.389653875589459e-06,
      "loss": 0.1794,
      "step": 1890
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.379598648550618e-06,
      "loss": 0.1679,
      "step": 1891
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.369547632416861e-06,
      "loss": 0.224,
      "step": 1892
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.359500838878572e-06,
      "loss": 0.1884,
      "step": 1893
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.349458279621219e-06,
      "loss": 0.1695,
      "step": 1894
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.33941996632535e-06,
      "loss": 0.2514,
      "step": 1895
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.3293859106665716e-06,
      "loss": 0.1768,
      "step": 1896
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.319356124315539e-06,
      "loss": 0.2061,
      "step": 1897
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.309330618937941e-06,
      "loss": 0.1484,
      "step": 1898
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.299309406194489e-06,
      "loss": 0.2367,
      "step": 1899
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.289292497740902e-06,
      "loss": 0.1678,
      "step": 1900
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.279279905227889e-06,
      "loss": 0.1847,
      "step": 1901
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.26927164030114e-06,
      "loss": 0.1908,
      "step": 1902
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.259267714601317e-06,
      "loss": 0.1583,
      "step": 1903
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.249268139764029e-06,
      "loss": 0.2292,
      "step": 1904
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.239272927419826e-06,
      "loss": 0.1782,
      "step": 1905
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.229282089194184e-06,
      "loss": 0.2242,
      "step": 1906
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.219295636707494e-06,
      "loss": 0.1718,
      "step": 1907
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.2093135815750405e-06,
      "loss": 0.2004,
      "step": 1908
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.199335935406998e-06,
      "loss": 0.1851,
      "step": 1909
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.189362709808414e-06,
      "loss": 0.2108,
      "step": 1910
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.179393916379188e-06,
      "loss": 0.1747,
      "step": 1911
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.169429566714074e-06,
      "loss": 0.2121,
      "step": 1912
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.159469672402645e-06,
      "loss": 0.1983,
      "step": 1913
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.149514245029303e-06,
      "loss": 0.1771,
      "step": 1914
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.13956329617325e-06,
      "loss": 0.221,
      "step": 1915
    },
    {
      "epoch": 1.27,
      "learning_rate": 6.129616837408477e-06,
      "loss": 0.1722,
      "step": 1916
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.119674880303758e-06,
      "loss": 0.1613,
      "step": 1917
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.109737436422629e-06,
      "loss": 0.1953,
      "step": 1918
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.099804517323371e-06,
      "loss": 0.1816,
      "step": 1919
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.089876134559012e-06,
      "loss": 0.1839,
      "step": 1920
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.079952299677294e-06,
      "loss": 0.1742,
      "step": 1921
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.0700330242206765e-06,
      "loss": 0.2163,
      "step": 1922
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.060118319726311e-06,
      "loss": 0.2218,
      "step": 1923
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.0502081977260395e-06,
      "loss": 0.2161,
      "step": 1924
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.0403026697463655e-06,
      "loss": 0.1849,
      "step": 1925
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.030401747308454e-06,
      "loss": 0.1917,
      "step": 1926
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.0205054419281126e-06,
      "loss": 0.1864,
      "step": 1927
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.010613765115778e-06,
      "loss": 0.2226,
      "step": 1928
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.000726728376505e-06,
      "loss": 0.2277,
      "step": 1929
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.9908443432099506e-06,
      "loss": 0.1497,
      "step": 1930
    },
    {
      "epoch": 1.28,
      "learning_rate": 5.98096662111036e-06,
      "loss": 0.1848,
      "step": 1931
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.971093573566557e-06,
      "loss": 0.2056,
      "step": 1932
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.961225212061926e-06,
      "loss": 0.1378,
      "step": 1933
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.951361548074406e-06,
      "loss": 0.1753,
      "step": 1934
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.941502593076467e-06,
      "loss": 0.1812,
      "step": 1935
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.931648358535103e-06,
      "loss": 0.19,
      "step": 1936
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.921798855911821e-06,
      "loss": 0.2034,
      "step": 1937
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.91195409666262e-06,
      "loss": 0.1982,
      "step": 1938
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.902114092237984e-06,
      "loss": 0.2124,
      "step": 1939
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.892278854082865e-06,
      "loss": 0.2176,
      "step": 1940
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.882448393636675e-06,
      "loss": 0.1851,
      "step": 1941
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.8726227223332655e-06,
      "loss": 0.1868,
      "step": 1942
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.86280185160092e-06,
      "loss": 0.2041,
      "step": 1943
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.852985792862336e-06,
      "loss": 0.1884,
      "step": 1944
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.843174557534617e-06,
      "loss": 0.1823,
      "step": 1945
    },
    {
      "epoch": 1.29,
      "learning_rate": 5.833368157029251e-06,
      "loss": 0.1553,
      "step": 1946
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.823566602752113e-06,
      "loss": 0.1683,
      "step": 1947
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.813769906103428e-06,
      "loss": 0.1673,
      "step": 1948
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.803978078477785e-06,
      "loss": 0.1754,
      "step": 1949
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.794191131264095e-06,
      "loss": 0.1912,
      "step": 1950
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.784409075845609e-06,
      "loss": 0.146,
      "step": 1951
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.774631923599872e-06,
      "loss": 0.2033,
      "step": 1952
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.764859685898738e-06,
      "loss": 0.2016,
      "step": 1953
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.755092374108338e-06,
      "loss": 0.3189,
      "step": 1954
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.74532999958908e-06,
      "loss": 0.2173,
      "step": 1955
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.735572573695619e-06,
      "loss": 0.1666,
      "step": 1956
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.7258201077768695e-06,
      "loss": 0.1929,
      "step": 1957
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.71607261317596e-06,
      "loss": 0.1534,
      "step": 1958
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.706330101230249e-06,
      "loss": 0.1661,
      "step": 1959
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.696592583271301e-06,
      "loss": 0.2528,
      "step": 1960
    },
    {
      "epoch": 1.3,
      "learning_rate": 5.686860070624858e-06,
      "loss": 0.2545,
      "step": 1961
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.677132574610857e-06,
      "loss": 0.1711,
      "step": 1962
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.667410106543385e-06,
      "loss": 0.1781,
      "step": 1963
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.657692677730699e-06,
      "loss": 0.2767,
      "step": 1964
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.647980299475171e-06,
      "loss": 0.1944,
      "step": 1965
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.638272983073321e-06,
      "loss": 0.1808,
      "step": 1966
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.628570739815764e-06,
      "loss": 0.2323,
      "step": 1967
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.618873580987234e-06,
      "loss": 0.2211,
      "step": 1968
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.6091815178665285e-06,
      "loss": 0.1903,
      "step": 1969
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.599494561726537e-06,
      "loss": 0.1995,
      "step": 1970
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.589812723834203e-06,
      "loss": 0.1809,
      "step": 1971
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.580136015450508e-06,
      "loss": 0.1679,
      "step": 1972
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.570464447830485e-06,
      "loss": 0.1838,
      "step": 1973
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.560798032223167e-06,
      "loss": 0.1816,
      "step": 1974
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.551136779871616e-06,
      "loss": 0.1883,
      "step": 1975
    },
    {
      "epoch": 1.31,
      "learning_rate": 5.541480702012869e-06,
      "loss": 0.219,
      "step": 1976
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.53182980987796e-06,
      "loss": 0.1836,
      "step": 1977
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.522184114691879e-06,
      "loss": 0.1661,
      "step": 1978
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.512543627673587e-06,
      "loss": 0.1885,
      "step": 1979
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.502908360035966e-06,
      "loss": 0.1869,
      "step": 1980
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.4932783229858484e-06,
      "loss": 0.1961,
      "step": 1981
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.483653527723967e-06,
      "loss": 0.144,
      "step": 1982
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.47403398544497e-06,
      "loss": 0.1803,
      "step": 1983
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.4644197073373836e-06,
      "loss": 0.1849,
      "step": 1984
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.4548107045836274e-06,
      "loss": 0.1596,
      "step": 1985
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.445206988359966e-06,
      "loss": 0.1714,
      "step": 1986
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.435608569836528e-06,
      "loss": 0.1412,
      "step": 1987
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.426015460177283e-06,
      "loss": 0.1573,
      "step": 1988
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.416427670540011e-06,
      "loss": 0.1333,
      "step": 1989
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.406845212076322e-06,
      "loss": 0.1849,
      "step": 1990
    },
    {
      "epoch": 1.32,
      "learning_rate": 5.3972680959316075e-06,
      "loss": 0.165,
      "step": 1991
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.387696333245063e-06,
      "loss": 0.1827,
      "step": 1992
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.3781299351496395e-06,
      "loss": 0.1898,
      "step": 1993
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.368568912772068e-06,
      "loss": 0.2114,
      "step": 1994
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.359013277232805e-06,
      "loss": 0.2394,
      "step": 1995
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.3494630396460655e-06,
      "loss": 0.2,
      "step": 1996
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.339918211119763e-06,
      "loss": 0.1757,
      "step": 1997
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.3303788027555405e-06,
      "loss": 0.1891,
      "step": 1998
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.320844825648715e-06,
      "loss": 0.1636,
      "step": 1999
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.31131629088831e-06,
      "loss": 0.1711,
      "step": 2000
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.301793209556997e-06,
      "loss": 0.2564,
      "step": 2001
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.292275592731123e-06,
      "loss": 0.1691,
      "step": 2002
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.282763451480663e-06,
      "loss": 0.1888,
      "step": 2003
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.273256796869236e-06,
      "loss": 0.1989,
      "step": 2004
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.263755639954078e-06,
      "loss": 0.166,
      "step": 2005
    },
    {
      "epoch": 1.33,
      "learning_rate": 5.2542599917860195e-06,
      "loss": 0.2278,
      "step": 2006
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.2447698634095e-06,
      "loss": 0.1685,
      "step": 2007
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.235285265862521e-06,
      "loss": 0.2057,
      "step": 2008
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.225806210176669e-06,
      "loss": 0.1886,
      "step": 2009
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.216332707377071e-06,
      "loss": 0.1607,
      "step": 2010
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.206864768482405e-06,
      "loss": 0.1478,
      "step": 2011
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.197402404504867e-06,
      "loss": 0.1879,
      "step": 2012
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.187945626450183e-06,
      "loss": 0.1841,
      "step": 2013
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.178494445317566e-06,
      "loss": 0.2034,
      "step": 2014
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.169048872099733e-06,
      "loss": 0.1802,
      "step": 2015
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.15960891778287e-06,
      "loss": 0.1816,
      "step": 2016
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.150174593346633e-06,
      "loss": 0.1798,
      "step": 2017
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.140745909764122e-06,
      "loss": 0.188,
      "step": 2018
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.1313228780018885e-06,
      "loss": 0.1577,
      "step": 2019
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.1219055090198955e-06,
      "loss": 0.2011,
      "step": 2020
    },
    {
      "epoch": 1.34,
      "learning_rate": 5.112493813771531e-06,
      "loss": 0.1539,
      "step": 2021
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.103087803203584e-06,
      "loss": 0.1812,
      "step": 2022
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.0936874882562185e-06,
      "loss": 0.1983,
      "step": 2023
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.084292879862995e-06,
      "loss": 0.1699,
      "step": 2024
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.074903988950812e-06,
      "loss": 0.184,
      "step": 2025
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.0655208264399404e-06,
      "loss": 0.2167,
      "step": 2026
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.056143403243969e-06,
      "loss": 0.1884,
      "step": 2027
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.046771730269833e-06,
      "loss": 0.1729,
      "step": 2028
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.037405818417756e-06,
      "loss": 0.1851,
      "step": 2029
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.028045678581281e-06,
      "loss": 0.2109,
      "step": 2030
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.018691321647221e-06,
      "loss": 0.189,
      "step": 2031
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.009342758495674e-06,
      "loss": 0.1857,
      "step": 2032
    },
    {
      "epoch": 1.35,
      "learning_rate": 5.000000000000003e-06,
      "loss": 0.1781,
      "step": 2033
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.9906630570268e-06,
      "loss": 0.2158,
      "step": 2034
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.981331940435915e-06,
      "loss": 0.1968,
      "step": 2035
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.972006661080404e-06,
      "loss": 0.1948,
      "step": 2036
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.962687229806549e-06,
      "loss": 0.1825,
      "step": 2037
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.9533736574538135e-06,
      "loss": 0.1918,
      "step": 2038
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.944065954854866e-06,
      "loss": 0.1903,
      "step": 2039
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.934764132835528e-06,
      "loss": 0.2451,
      "step": 2040
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.925468202214796e-06,
      "loss": 0.159,
      "step": 2041
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.916178173804804e-06,
      "loss": 0.1591,
      "step": 2042
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.906894058410833e-06,
      "loss": 0.2046,
      "step": 2043
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.89761586683127e-06,
      "loss": 0.1573,
      "step": 2044
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.8883436098576306e-06,
      "loss": 0.1926,
      "step": 2045
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.8790772982745116e-06,
      "loss": 0.1465,
      "step": 2046
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.8698169428596065e-06,
      "loss": 0.1602,
      "step": 2047
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.860562554383671e-06,
      "loss": 0.1869,
      "step": 2048
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.851314143610528e-06,
      "loss": 0.1725,
      "step": 2049
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.842071721297051e-06,
      "loss": 0.1703,
      "step": 2050
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.832835298193136e-06,
      "loss": 0.1947,
      "step": 2051
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.8236048850417135e-06,
      "loss": 0.199,
      "step": 2052
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.8143804925787115e-06,
      "loss": 0.1795,
      "step": 2053
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.805162131533072e-06,
      "loss": 0.2069,
      "step": 2054
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.795949812626703e-06,
      "loss": 0.1736,
      "step": 2055
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.7867435465745006e-06,
      "loss": 0.2106,
      "step": 2056
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.777543344084308e-06,
      "loss": 0.1691,
      "step": 2057
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.768349215856929e-06,
      "loss": 0.1623,
      "step": 2058
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.759161172586089e-06,
      "loss": 0.2212,
      "step": 2059
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.7499792249584486e-06,
      "loss": 0.2087,
      "step": 2060
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.740803383653565e-06,
      "loss": 0.2117,
      "step": 2061
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.73163365934391e-06,
      "loss": 0.1875,
      "step": 2062
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.72247006269482e-06,
      "loss": 0.2025,
      "step": 2063
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.713312604364527e-06,
      "loss": 0.173,
      "step": 2064
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.704161295004102e-06,
      "loss": 0.1649,
      "step": 2065
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.695016145257478e-06,
      "loss": 0.201,
      "step": 2066
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.6858771657614246e-06,
      "loss": 0.21,
      "step": 2067
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.67674436714552e-06,
      "loss": 0.1797,
      "step": 2068
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.667617760032174e-06,
      "loss": 0.203,
      "step": 2069
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.658497355036574e-06,
      "loss": 0.2058,
      "step": 2070
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.649383162766714e-06,
      "loss": 0.1825,
      "step": 2071
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.6402751938233435e-06,
      "loss": 0.1801,
      "step": 2072
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.631173458799989e-06,
      "loss": 0.2048,
      "step": 2073
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.622077968282912e-06,
      "loss": 0.1848,
      "step": 2074
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.6129887328511266e-06,
      "loss": 0.1783,
      "step": 2075
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.6039057630763566e-06,
      "loss": 0.1777,
      "step": 2076
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.594829069523049e-06,
      "loss": 0.2048,
      "step": 2077
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.585758662748344e-06,
      "loss": 0.1416,
      "step": 2078
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.576694553302077e-06,
      "loss": 0.2015,
      "step": 2079
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.567636751726748e-06,
      "loss": 0.1851,
      "step": 2080
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.558585268557533e-06,
      "loss": 0.184,
      "step": 2081
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.549540114322248e-06,
      "loss": 0.1803,
      "step": 2082
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.540501299541356e-06,
      "loss": 0.1991,
      "step": 2083
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.5314688347279435e-06,
      "loss": 0.1927,
      "step": 2084
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.522442730387705e-06,
      "loss": 0.2344,
      "step": 2085
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.51342299701895e-06,
      "loss": 0.2274,
      "step": 2086
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.504409645112565e-06,
      "loss": 0.195,
      "step": 2087
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.495402685152027e-06,
      "loss": 0.1755,
      "step": 2088
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.486402127613364e-06,
      "loss": 0.2329,
      "step": 2089
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.477407982965172e-06,
      "loss": 0.1758,
      "step": 2090
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.468420261668575e-06,
      "loss": 0.2026,
      "step": 2091
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.459438974177237e-06,
      "loss": 0.1806,
      "step": 2092
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.450464130937328e-06,
      "loss": 0.199,
      "step": 2093
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.4414957423875326e-06,
      "loss": 0.1805,
      "step": 2094
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.432533818959028e-06,
      "loss": 0.197,
      "step": 2095
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.423578371075458e-06,
      "loss": 0.1837,
      "step": 2096
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.414629409152953e-06,
      "loss": 0.2016,
      "step": 2097
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.405686943600084e-06,
      "loss": 0.1846,
      "step": 2098
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.396750984817879e-06,
      "loss": 0.1411,
      "step": 2099
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.387821543199785e-06,
      "loss": 0.1855,
      "step": 2100
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.378898629131684e-06,
      "loss": 0.1548,
      "step": 2101
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3699822529918486e-06,
      "loss": 0.1885,
      "step": 2102
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.361072425150964e-06,
      "loss": 0.1557,
      "step": 2103
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.352169155972084e-06,
      "loss": 0.1511,
      "step": 2104
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.343272455810649e-06,
      "loss": 0.2172,
      "step": 2105
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.3343823350144406e-06,
      "loss": 0.2027,
      "step": 2106
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.325498803923609e-06,
      "loss": 0.1716,
      "step": 2107
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.31662187287062e-06,
      "loss": 0.1971,
      "step": 2108
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.307751552180279e-06,
      "loss": 0.2213,
      "step": 2109
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.2988878521696895e-06,
      "loss": 0.1887,
      "step": 2110
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.2900307831482655e-06,
      "loss": 0.2634,
      "step": 2111
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.281180355417706e-06,
      "loss": 0.1835,
      "step": 2112
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.272336579271975e-06,
      "loss": 0.1906,
      "step": 2113
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.2634994649973174e-06,
      "loss": 0.1734,
      "step": 2114
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.2546690228722125e-06,
      "loss": 0.1849,
      "step": 2115
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.2458452631673945e-06,
      "loss": 0.2323,
      "step": 2116
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.23702819614581e-06,
      "loss": 0.1784,
      "step": 2117
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.228217832062639e-06,
      "loss": 0.1764,
      "step": 2118
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.219414181165245e-06,
      "loss": 0.1897,
      "step": 2119
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.210617253693205e-06,
      "loss": 0.1387,
      "step": 2120
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.201827059878254e-06,
      "loss": 0.2083,
      "step": 2121
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.193043609944315e-06,
      "loss": 0.1743,
      "step": 2122
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.18426691410745e-06,
      "loss": 0.1647,
      "step": 2123
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.175496982575882e-06,
      "loss": 0.1442,
      "step": 2124
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.166733825549949e-06,
      "loss": 0.1609,
      "step": 2125
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.157977453222126e-06,
      "loss": 0.1874,
      "step": 2126
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.149227875776982e-06,
      "loss": 0.1928,
      "step": 2127
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.140485103391194e-06,
      "loss": 0.1612,
      "step": 2128
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1317491462335235e-06,
      "loss": 0.1653,
      "step": 2129
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.123020014464793e-06,
      "loss": 0.1876,
      "step": 2130
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.114297718237903e-06,
      "loss": 0.2397,
      "step": 2131
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.1055822676977886e-06,
      "loss": 0.2054,
      "step": 2132
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.096873672981435e-06,
      "loss": 0.1888,
      "step": 2133
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.088171944217843e-06,
      "loss": 0.1987,
      "step": 2134
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.079477091528038e-06,
      "loss": 0.1893,
      "step": 2135
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.070789125025036e-06,
      "loss": 0.2118,
      "step": 2136
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.062108054813857e-06,
      "loss": 0.1877,
      "step": 2137
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.053433890991484e-06,
      "loss": 0.2432,
      "step": 2138
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.044766643646885e-06,
      "loss": 0.203,
      "step": 2139
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.036106322860967e-06,
      "loss": 0.1636,
      "step": 2140
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.027452938706594e-06,
      "loss": 0.2242,
      "step": 2141
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.0188065012485535e-06,
      "loss": 0.1769,
      "step": 2142
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.010167020543552e-06,
      "loss": 0.1789,
      "step": 2143
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.001534506640215e-06,
      "loss": 0.1663,
      "step": 2144
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.992908969579051e-06,
      "loss": 0.1902,
      "step": 2145
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.9842904193924695e-06,
      "loss": 0.1887,
      "step": 2146
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.975678866104735e-06,
      "loss": 0.1805,
      "step": 2147
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.967074319731994e-06,
      "loss": 0.1934,
      "step": 2148
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.958476790282226e-06,
      "loss": 0.2062,
      "step": 2149
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.9498862877552595e-06,
      "loss": 0.1637,
      "step": 2150
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.941302822142743e-06,
      "loss": 0.1916,
      "step": 2151
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.9327264034281495e-06,
      "loss": 0.1613,
      "step": 2152
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.924157041586745e-06,
      "loss": 0.2191,
      "step": 2153
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.915594746585597e-06,
      "loss": 0.2335,
      "step": 2154
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.9070395283835436e-06,
      "loss": 0.1817,
      "step": 2155
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.898491396931202e-06,
      "loss": 0.2123,
      "step": 2156
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.889950362170944e-06,
      "loss": 0.1729,
      "step": 2157
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.8814164340368806e-06,
      "loss": 0.205,
      "step": 2158
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.872889622454867e-06,
      "loss": 0.1658,
      "step": 2159
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.86436993734247e-06,
      "loss": 0.183,
      "step": 2160
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.855857388608981e-06,
      "loss": 0.1933,
      "step": 2161
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.847351986155376e-06,
      "loss": 0.1985,
      "step": 2162
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.838853739874334e-06,
      "loss": 0.1296,
      "step": 2163
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.830362659650198e-06,
      "loss": 0.1904,
      "step": 2164
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.821878755358984e-06,
      "loss": 0.1901,
      "step": 2165
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.813402036868358e-06,
      "loss": 0.1961,
      "step": 2166
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.8049325140376326e-06,
      "loss": 0.2277,
      "step": 2167
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.796470196717741e-06,
      "loss": 0.1729,
      "step": 2168
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.7880150947512516e-06,
      "loss": 0.157,
      "step": 2169
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.779567217972323e-06,
      "loss": 0.1872,
      "step": 2170
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.771126576206726e-06,
      "loss": 0.2082,
      "step": 2171
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.7626931792718026e-06,
      "loss": 0.2156,
      "step": 2172
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.7542670369764766e-06,
      "loss": 0.1832,
      "step": 2173
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.745848159121237e-06,
      "loss": 0.2182,
      "step": 2174
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.737436555498112e-06,
      "loss": 0.2022,
      "step": 2175
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.7290322358906792e-06,
      "loss": 0.1795,
      "step": 2176
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.720635210074037e-06,
      "loss": 0.1653,
      "step": 2177
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.7122454878148107e-06,
      "loss": 0.1869,
      "step": 2178
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.703863078871115e-06,
      "loss": 0.2256,
      "step": 2179
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.6954879929925745e-06,
      "loss": 0.1672,
      "step": 2180
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.6871202399202844e-06,
      "loss": 0.2067,
      "step": 2181
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.678759829386821e-06,
      "loss": 0.2144,
      "step": 2182
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.6704067711162093e-06,
      "loss": 0.1638,
      "step": 2183
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.662061074823935e-06,
      "loss": 0.1693,
      "step": 2184
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.653722750216909e-06,
      "loss": 0.1525,
      "step": 2185
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.645391806993481e-06,
      "loss": 0.1793,
      "step": 2186
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.637068254843401e-06,
      "loss": 0.1806,
      "step": 2187
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.628752103447837e-06,
      "loss": 0.1884,
      "step": 2188
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.6204433624793343e-06,
      "loss": 0.1653,
      "step": 2189
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.6121420416018326e-06,
      "loss": 0.1827,
      "step": 2190
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.603848150470638e-06,
      "loss": 0.1605,
      "step": 2191
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.5955616987324037e-06,
      "loss": 0.2115,
      "step": 2192
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.5872826960251483e-06,
      "loss": 0.1511,
      "step": 2193
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.57901115197821e-06,
      "loss": 0.1735,
      "step": 2194
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.5707470762122644e-06,
      "loss": 0.2021,
      "step": 2195
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.562490478339289e-06,
      "loss": 0.1731,
      "step": 2196
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.5542413679625763e-06,
      "loss": 0.1971,
      "step": 2197
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.5459997546766978e-06,
      "loss": 0.2021,
      "step": 2198
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.537765648067516e-06,
      "loss": 0.1709,
      "step": 2199
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.5295390577121535e-06,
      "loss": 0.1641,
      "step": 2200
    },
    {
      "epoch": 1.46,
      "learning_rate": 3.521319993178999e-06,
      "loss": 0.1994,
      "step": 2201
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.513108464027679e-06,
      "loss": 0.212,
      "step": 2202
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.504904479809059e-06,
      "loss": 0.1982,
      "step": 2203
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4967080500652373e-06,
      "loss": 0.1488,
      "step": 2204
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4885191843295084e-06,
      "loss": 0.2078,
      "step": 2205
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4803378921263877e-06,
      "loss": 0.2152,
      "step": 2206
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.472164182971565e-06,
      "loss": 0.1841,
      "step": 2207
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.463998066371925e-06,
      "loss": 0.1791,
      "step": 2208
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4558395518255094e-06,
      "loss": 0.1592,
      "step": 2209
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4476886488215287e-06,
      "loss": 0.1595,
      "step": 2210
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.439545366840329e-06,
      "loss": 0.1588,
      "step": 2211
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.431409715353403e-06,
      "loss": 0.1578,
      "step": 2212
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.423281703823358e-06,
      "loss": 0.1763,
      "step": 2213
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4151613417039273e-06,
      "loss": 0.1798,
      "step": 2214
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.4070486384399347e-06,
      "loss": 0.1712,
      "step": 2215
    },
    {
      "epoch": 1.47,
      "learning_rate": 3.3989436034673073e-06,
      "loss": 0.2252,
      "step": 2216
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.390846246213042e-06,
      "loss": 0.1947,
      "step": 2217
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.382756576095214e-06,
      "loss": 0.1464,
      "step": 2218
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.374674602522958e-06,
      "loss": 0.1853,
      "step": 2219
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.3666003348964482e-06,
      "loss": 0.147,
      "step": 2220
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.3585337826069076e-06,
      "loss": 0.155,
      "step": 2221
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.3504749550365734e-06,
      "loss": 0.1996,
      "step": 2222
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.34242386155871e-06,
      "loss": 0.1833,
      "step": 2223
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.3343805115375736e-06,
      "loss": 0.2143,
      "step": 2224
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.326344914328429e-06,
      "loss": 0.2113,
      "step": 2225
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.3183170792775065e-06,
      "loss": 0.1782,
      "step": 2226
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.310297015722026e-06,
      "loss": 0.1873,
      "step": 2227
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.302284732990151e-06,
      "loss": 0.1908,
      "step": 2228
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.294280240401011e-06,
      "loss": 0.1689,
      "step": 2229
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.2862835472646613e-06,
      "loss": 0.2301,
      "step": 2230
    },
    {
      "epoch": 1.48,
      "learning_rate": 3.2782946628820966e-06,
      "loss": 0.1833,
      "step": 2231
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.2703135965452195e-06,
      "loss": 0.1786,
      "step": 2232
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.2623403575368493e-06,
      "loss": 0.2037,
      "step": 2233
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.2543749551306913e-06,
      "loss": 0.1817,
      "step": 2234
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.246417398591342e-06,
      "loss": 0.1844,
      "step": 2235
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.2384676971742745e-06,
      "loss": 0.1732,
      "step": 2236
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.2305258601258183e-06,
      "loss": 0.184,
      "step": 2237
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.222591896683164e-06,
      "loss": 0.1926,
      "step": 2238
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.214665816074334e-06,
      "loss": 0.2341,
      "step": 2239
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.2067476275181942e-06,
      "loss": 0.1871,
      "step": 2240
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.198837340224421e-06,
      "loss": 0.2133,
      "step": 2241
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.1909349633935096e-06,
      "loss": 0.1778,
      "step": 2242
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.1830405062167448e-06,
      "loss": 0.16,
      "step": 2243
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.1751539778762098e-06,
      "loss": 0.155,
      "step": 2244
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.167275387544757e-06,
      "loss": 0.1703,
      "step": 2245
    },
    {
      "epoch": 1.49,
      "learning_rate": 3.159404744386014e-06,
      "loss": 0.1535,
      "step": 2246
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.1515420575543587e-06,
      "loss": 0.1737,
      "step": 2247
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.1436873361949204e-06,
      "loss": 0.1833,
      "step": 2248
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.135840589443555e-06,
      "loss": 0.1722,
      "step": 2249
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.1280018264268576e-06,
      "loss": 0.1676,
      "step": 2250
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.120171056262119e-06,
      "loss": 0.1742,
      "step": 2251
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.1123482880573484e-06,
      "loss": 0.1677,
      "step": 2252
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.104533530911247e-06,
      "loss": 0.2191,
      "step": 2253
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.096726793913184e-06,
      "loss": 0.2178,
      "step": 2254
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0889280861432204e-06,
      "loss": 0.1583,
      "step": 2255
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0811374166720586e-06,
      "loss": 0.1687,
      "step": 2256
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.073354794561071e-06,
      "loss": 0.1974,
      "step": 2257
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0655802288622505e-06,
      "loss": 0.1551,
      "step": 2258
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0578137286182383e-06,
      "loss": 0.2179,
      "step": 2259
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.050055302862279e-06,
      "loss": 0.1846,
      "step": 2260
    },
    {
      "epoch": 1.5,
      "learning_rate": 3.0423049606182386e-06,
      "loss": 0.1799,
      "step": 2261
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.0345627109005703e-06,
      "loss": 0.1981,
      "step": 2262
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.026828562714319e-06,
      "loss": 0.1916,
      "step": 2263
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.019102525055112e-06,
      "loss": 0.2,
      "step": 2264
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.0113846069091323e-06,
      "loss": 0.1543,
      "step": 2265
    },
    {
      "epoch": 1.51,
      "learning_rate": 3.0036748172531317e-06,
      "loss": 0.177,
      "step": 2266
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.9959731650543945e-06,
      "loss": 0.2269,
      "step": 2267
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.988279659270753e-06,
      "loss": 0.1608,
      "step": 2268
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.980594308850554e-06,
      "loss": 0.1726,
      "step": 2269
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.972917122732667e-06,
      "loss": 0.1927,
      "step": 2270
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.965248109846457e-06,
      "loss": 0.1948,
      "step": 2271
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.957587279111793e-06,
      "loss": 0.1813,
      "step": 2272
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.949934639439015e-06,
      "loss": 0.1857,
      "step": 2273
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.94229019972895e-06,
      "loss": 0.136,
      "step": 2274
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.9346539688728737e-06,
      "loss": 0.1641,
      "step": 2275
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.9270259557525247e-06,
      "loss": 0.159,
      "step": 2276
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.9194061692400764e-06,
      "loss": 0.2048,
      "step": 2277
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.9117946181981405e-06,
      "loss": 0.1804,
      "step": 2278
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.9041913114797404e-06,
      "loss": 0.1635,
      "step": 2279
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8965962579283204e-06,
      "loss": 0.2113,
      "step": 2280
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8890094663777246e-06,
      "loss": 0.1781,
      "step": 2281
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.881430945652176e-06,
      "loss": 0.1915,
      "step": 2282
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8738607045662948e-06,
      "loss": 0.1864,
      "step": 2283
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8662987519250563e-06,
      "loss": 0.2063,
      "step": 2284
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.858745096523805e-06,
      "loss": 0.2715,
      "step": 2285
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8511997471482277e-06,
      "loss": 0.175,
      "step": 2286
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8436627125743597e-06,
      "loss": 0.1568,
      "step": 2287
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8361340015685533e-06,
      "loss": 0.1534,
      "step": 2288
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.828613622887494e-06,
      "loss": 0.169,
      "step": 2289
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.8211015852781585e-06,
      "loss": 0.1911,
      "step": 2290
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.813597897477841e-06,
      "loss": 0.1695,
      "step": 2291
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.8061025682141076e-06,
      "loss": 0.183,
      "step": 2292
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7986156062048133e-06,
      "loss": 0.1835,
      "step": 2293
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7911370201580736e-06,
      "loss": 0.1656,
      "step": 2294
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7836668187722703e-06,
      "loss": 0.1519,
      "step": 2295
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7762050107360237e-06,
      "loss": 0.1792,
      "step": 2296
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7687516047281983e-06,
      "loss": 0.1704,
      "step": 2297
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7613066094178865e-06,
      "loss": 0.1914,
      "step": 2298
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.753870033464391e-06,
      "loss": 0.2362,
      "step": 2299
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7464418855172316e-06,
      "loss": 0.1798,
      "step": 2300
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.739022174216115e-06,
      "loss": 0.1585,
      "step": 2301
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.731610908190948e-06,
      "loss": 0.2097,
      "step": 2302
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7242080960618e-06,
      "loss": 0.1919,
      "step": 2303
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.716813746438921e-06,
      "loss": 0.1791,
      "step": 2304
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.709427867922706e-06,
      "loss": 0.1947,
      "step": 2305
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.702050469103711e-06,
      "loss": 0.1327,
      "step": 2306
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.694681558562615e-06,
      "loss": 0.1441,
      "step": 2307
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6873211448702363e-06,
      "loss": 0.2057,
      "step": 2308
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6799692365875007e-06,
      "loss": 0.245,
      "step": 2309
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6726258422654495e-06,
      "loss": 0.219,
      "step": 2310
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6652909704452147e-06,
      "loss": 0.2056,
      "step": 2311
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6579646296580206e-06,
      "loss": 0.2132,
      "step": 2312
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6506468284251653e-06,
      "loss": 0.2114,
      "step": 2313
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.643337575258017e-06,
      "loss": 0.1682,
      "step": 2314
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6360368786580036e-06,
      "loss": 0.1582,
      "step": 2315
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.628744747116594e-06,
      "loss": 0.1857,
      "step": 2316
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.621461189115303e-06,
      "loss": 0.1586,
      "step": 2317
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6141862131256647e-06,
      "loss": 0.2043,
      "step": 2318
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6069198276092433e-06,
      "loss": 0.1674,
      "step": 2319
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.5996620410175964e-06,
      "loss": 0.2031,
      "step": 2320
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.5924128617922993e-06,
      "loss": 0.143,
      "step": 2321
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5851722983648907e-06,
      "loss": 0.1954,
      "step": 2322
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5779403591569106e-06,
      "loss": 0.2173,
      "step": 2323
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.570717052579862e-06,
      "loss": 0.2409,
      "step": 2324
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5635023870351993e-06,
      "loss": 0.1726,
      "step": 2325
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5562963709143396e-06,
      "loss": 0.1945,
      "step": 2326
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5490990125986237e-06,
      "loss": 0.2013,
      "step": 2327
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.541910320459341e-06,
      "loss": 0.1786,
      "step": 2328
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.534730302857682e-06,
      "loss": 0.189,
      "step": 2329
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5275589681447664e-06,
      "loss": 0.2064,
      "step": 2330
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.520396324661599e-06,
      "loss": 0.179,
      "step": 2331
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5132423807390883e-06,
      "loss": 0.2365,
      "step": 2332
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.506097144698012e-06,
      "loss": 0.2011,
      "step": 2333
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.498960624849034e-06,
      "loss": 0.1885,
      "step": 2334
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.491832829492664e-06,
      "loss": 0.234,
      "step": 2335
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.484713766919282e-06,
      "loss": 0.2235,
      "step": 2336
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.4776034454090935e-06,
      "loss": 0.2138,
      "step": 2337
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.4705018732321516e-06,
      "loss": 0.1764,
      "step": 2338
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.4634090586483216e-06,
      "loss": 0.1613,
      "step": 2339
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.4563250099072943e-06,
      "loss": 0.1831,
      "step": 2340
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.449249735248551e-06,
      "loss": 0.192,
      "step": 2341
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.442183242901379e-06,
      "loss": 0.2278,
      "step": 2342
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.435125541084852e-06,
      "loss": 0.1647,
      "step": 2343
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.428076638007807e-06,
      "loss": 0.1764,
      "step": 2344
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.4210365418688606e-06,
      "loss": 0.1763,
      "step": 2345
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.414005260856376e-06,
      "loss": 0.1834,
      "step": 2346
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.406982803148472e-06,
      "loss": 0.1624,
      "step": 2347
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.3999691769129964e-06,
      "loss": 0.1696,
      "step": 2348
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.392964390307534e-06,
      "loss": 0.1686,
      "step": 2349
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.3859684514793767e-06,
      "loss": 0.1306,
      "step": 2350
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.3789813685655416e-06,
      "loss": 0.1965,
      "step": 2351
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.3720031496927275e-06,
      "loss": 0.1747,
      "step": 2352
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.365033802977339e-06,
      "loss": 0.1729,
      "step": 2353
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.358073336525448e-06,
      "loss": 0.1589,
      "step": 2354
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.3511217584328116e-06,
      "loss": 0.1767,
      "step": 2355
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.3441790767848347e-06,
      "loss": 0.1746,
      "step": 2356
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.3372452996565877e-06,
      "loss": 0.1627,
      "step": 2357
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.330320435112772e-06,
      "loss": 0.1811,
      "step": 2358
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.3234044912077314e-06,
      "loss": 0.206,
      "step": 2359
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.3164974759854343e-06,
      "loss": 0.1759,
      "step": 2360
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.309599397479455e-06,
      "loss": 0.1547,
      "step": 2361
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.302710263712987e-06,
      "loss": 0.1792,
      "step": 2362
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.2958300826988057e-06,
      "loss": 0.2009,
      "step": 2363
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.2889588624392853e-06,
      "loss": 0.1928,
      "step": 2364
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.2820966109263698e-06,
      "loss": 0.1604,
      "step": 2365
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.275243336141578e-06,
      "loss": 0.1649,
      "step": 2366
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2683990460559803e-06,
      "loss": 0.1357,
      "step": 2367
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2615637486302068e-06,
      "loss": 0.1904,
      "step": 2368
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2547374518144173e-06,
      "loss": 0.2124,
      "step": 2369
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.247920163548313e-06,
      "loss": 0.169,
      "step": 2370
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2411118917611074e-06,
      "loss": 0.1998,
      "step": 2371
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2343126443715367e-06,
      "loss": 0.162,
      "step": 2372
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2275224292878326e-06,
      "loss": 0.2349,
      "step": 2373
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2207412544077287e-06,
      "loss": 0.1877,
      "step": 2374
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.213969127618436e-06,
      "loss": 0.1919,
      "step": 2375
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2072060567966457e-06,
      "loss": 0.1901,
      "step": 2376
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.2004520498085213e-06,
      "loss": 0.1959,
      "step": 2377
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.193707114509672e-06,
      "loss": 0.1689,
      "step": 2378
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.1869712587451676e-06,
      "loss": 0.2026,
      "step": 2379
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.18024449034951e-06,
      "loss": 0.1716,
      "step": 2380
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.1735268171466326e-06,
      "loss": 0.2106,
      "step": 2381
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.166818246949891e-06,
      "loss": 0.2005,
      "step": 2382
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1601187875620567e-06,
      "loss": 0.1578,
      "step": 2383
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1534284467752965e-06,
      "loss": 0.183,
      "step": 2384
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.146747232371178e-06,
      "loss": 0.1984,
      "step": 2385
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.140075152120654e-06,
      "loss": 0.1487,
      "step": 2386
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1334122137840475e-06,
      "loss": 0.1783,
      "step": 2387
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.126758425111056e-06,
      "loss": 0.2053,
      "step": 2388
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1201137938407257e-06,
      "loss": 0.1457,
      "step": 2389
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1134783277014627e-06,
      "loss": 0.1968,
      "step": 2390
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1068520344110024e-06,
      "loss": 0.1673,
      "step": 2391
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.1002349216764208e-06,
      "loss": 0.1538,
      "step": 2392
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.093626997194107e-06,
      "loss": 0.2117,
      "step": 2393
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.0870282686497722e-06,
      "loss": 0.172,
      "step": 2394
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.0804387437184225e-06,
      "loss": 0.1561,
      "step": 2395
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.0738584300643694e-06,
      "loss": 0.1954,
      "step": 2396
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.0672873353411993e-06,
      "loss": 0.1809,
      "step": 2397
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.0607254671917888e-06,
      "loss": 0.1702,
      "step": 2398
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.054172833248269e-06,
      "loss": 0.2249,
      "step": 2399
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.047629441132044e-06,
      "loss": 0.1864,
      "step": 2400
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.041095298453758e-06,
      "loss": 0.2023,
      "step": 2401
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.034570412813306e-06,
      "loss": 0.2052,
      "step": 2402
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.0280547917998063e-06,
      "loss": 0.1596,
      "step": 2403
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.021548442991611e-06,
      "loss": 0.1786,
      "step": 2404
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.015051373956285e-06,
      "loss": 0.1756,
      "step": 2405
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.008563592250594e-06,
      "loss": 0.157,
      "step": 2406
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.0020851054205103e-06,
      "loss": 0.1481,
      "step": 2407
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.995615921001187e-06,
      "loss": 0.1977,
      "step": 2408
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.9891560465169666e-06,
      "loss": 0.2188,
      "step": 2409
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.9827054894813536e-06,
      "loss": 0.2019,
      "step": 2410
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.9762642573970224e-06,
      "loss": 0.1898,
      "step": 2411
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.9698323577557966e-06,
      "loss": 0.1621,
      "step": 2412
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9634097980386525e-06,
      "loss": 0.1684,
      "step": 2413
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.95699658571569e-06,
      "loss": 0.186,
      "step": 2414
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9505927282461544e-06,
      "loss": 0.2177,
      "step": 2415
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9441982330783935e-06,
      "loss": 0.1474,
      "step": 2416
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.93781310764988e-06,
      "loss": 0.1616,
      "step": 2417
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.931437359387175e-06,
      "loss": 0.1565,
      "step": 2418
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9250709957059464e-06,
      "loss": 0.1698,
      "step": 2419
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9187140240109348e-06,
      "loss": 0.2128,
      "step": 2420
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.912366451695965e-06,
      "loss": 0.1986,
      "step": 2421
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9060282861439305e-06,
      "loss": 0.2256,
      "step": 2422
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.8996995347267744e-06,
      "loss": 0.1404,
      "step": 2423
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.8933802048055005e-06,
      "loss": 0.1749,
      "step": 2424
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.8870703037301452e-06,
      "loss": 0.1679,
      "step": 2425
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.8807698388397878e-06,
      "loss": 0.1584,
      "step": 2426
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.8744788174625216e-06,
      "loss": 0.1914,
      "step": 2427
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8681972469154685e-06,
      "loss": 0.1797,
      "step": 2428
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.861925134504745e-06,
      "loss": 0.1624,
      "step": 2429
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8556624875254792e-06,
      "loss": 0.1612,
      "step": 2430
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8494093132617775e-06,
      "loss": 0.1914,
      "step": 2431
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.843165618986742e-06,
      "loss": 0.2055,
      "step": 2432
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.836931411962435e-06,
      "loss": 0.1302,
      "step": 2433
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.830706699439897e-06,
      "loss": 0.203,
      "step": 2434
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8244914886591124e-06,
      "loss": 0.1779,
      "step": 2435
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8182857868490268e-06,
      "loss": 0.173,
      "step": 2436
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8120896012275159e-06,
      "loss": 0.1991,
      "step": 2437
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.8059029390013916e-06,
      "loss": 0.1655,
      "step": 2438
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.7997258073663926e-06,
      "loss": 0.1843,
      "step": 2439
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.793558213507164e-06,
      "loss": 0.1342,
      "step": 2440
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.787400164597265e-06,
      "loss": 0.2076,
      "step": 2441
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.7812516677991444e-06,
      "loss": 0.191,
      "step": 2442
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7751127302641525e-06,
      "loss": 0.1938,
      "step": 2443
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7689833591325111e-06,
      "loss": 0.1799,
      "step": 2444
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7628635615333234e-06,
      "loss": 0.1965,
      "step": 2445
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7567533445845497e-06,
      "loss": 0.1499,
      "step": 2446
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.750652715393011e-06,
      "loss": 0.1471,
      "step": 2447
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7445616810543808e-06,
      "loss": 0.152,
      "step": 2448
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7384802486531627e-06,
      "loss": 0.1537,
      "step": 2449
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7324084252627038e-06,
      "loss": 0.1854,
      "step": 2450
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.726346217945164e-06,
      "loss": 0.1411,
      "step": 2451
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7202936337515287e-06,
      "loss": 0.1424,
      "step": 2452
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7142506797215808e-06,
      "loss": 0.1682,
      "step": 2453
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7082173628839116e-06,
      "loss": 0.1972,
      "step": 2454
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.7021936902558956e-06,
      "loss": 0.1861,
      "step": 2455
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.696179668843696e-06,
      "loss": 0.1484,
      "step": 2456
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.6901753056422465e-06,
      "loss": 0.1962,
      "step": 2457
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6841806076352507e-06,
      "loss": 0.1805,
      "step": 2458
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6781955817951668e-06,
      "loss": 0.1985,
      "step": 2459
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6722202350832073e-06,
      "loss": 0.1626,
      "step": 2460
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.666254574449322e-06,
      "loss": 0.2196,
      "step": 2461
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6602986068322014e-06,
      "loss": 0.1731,
      "step": 2462
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6543523391592543e-06,
      "loss": 0.2029,
      "step": 2463
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6484157783466148e-06,
      "loss": 0.1749,
      "step": 2464
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6424889312991199e-06,
      "loss": 0.1622,
      "step": 2465
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.636571804910314e-06,
      "loss": 0.1603,
      "step": 2466
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6306644060624354e-06,
      "loss": 0.194,
      "step": 2467
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6247667416264023e-06,
      "loss": 0.199,
      "step": 2468
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6188788184618188e-06,
      "loss": 0.1819,
      "step": 2469
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6130006434169509e-06,
      "loss": 0.1739,
      "step": 2470
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6071322233287334e-06,
      "loss": 0.1947,
      "step": 2471
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6012735650227496e-06,
      "loss": 0.1895,
      "step": 2472
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.595424675313234e-06,
      "loss": 0.1955,
      "step": 2473
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5895855610030532e-06,
      "loss": 0.1778,
      "step": 2474
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5837562288837116e-06,
      "loss": 0.1987,
      "step": 2475
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5779366857353262e-06,
      "loss": 0.1888,
      "step": 2476
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5721269383266403e-06,
      "loss": 0.1741,
      "step": 2477
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5663269934149906e-06,
      "loss": 0.1844,
      "step": 2478
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5605368577463242e-06,
      "loss": 0.1975,
      "step": 2479
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5547565380551698e-06,
      "loss": 0.1542,
      "step": 2480
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5489860410646472e-06,
      "loss": 0.216,
      "step": 2481
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5432253734864444e-06,
      "loss": 0.1395,
      "step": 2482
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5374745420208204e-06,
      "loss": 0.1741,
      "step": 2483
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5317335533565947e-06,
      "loss": 0.1275,
      "step": 2484
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.526002414171135e-06,
      "loss": 0.1616,
      "step": 2485
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5202811311303579e-06,
      "loss": 0.2111,
      "step": 2486
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.5145697108887092e-06,
      "loss": 0.1539,
      "step": 2487
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.508868160089172e-06,
      "loss": 0.1497,
      "step": 2488
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.5031764853632403e-06,
      "loss": 0.1743,
      "step": 2489
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4974946933309332e-06,
      "loss": 0.1926,
      "step": 2490
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.491822790600761e-06,
      "loss": 0.1354,
      "step": 2491
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4861607837697467e-06,
      "loss": 0.1868,
      "step": 2492
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4805086794233904e-06,
      "loss": 0.1752,
      "step": 2493
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4748664841356841e-06,
      "loss": 0.1381,
      "step": 2494
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4692342044690854e-06,
      "loss": 0.1833,
      "step": 2495
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.463611846974531e-06,
      "loss": 0.2345,
      "step": 2496
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4579994181914037e-06,
      "loss": 0.1592,
      "step": 2497
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.452396924647549e-06,
      "loss": 0.1939,
      "step": 2498
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4468043728592506e-06,
      "loss": 0.1581,
      "step": 2499
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4412217693312292e-06,
      "loss": 0.2009,
      "step": 2500
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.435649120556639e-06,
      "loss": 0.1853,
      "step": 2501
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.4300864330170494e-06,
      "loss": 0.17,
      "step": 2502
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.424533713182451e-06,
      "loss": 0.2144,
      "step": 2503
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4189909675112333e-06,
      "loss": 0.1568,
      "step": 2504
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.413458202450194e-06,
      "loss": 0.1656,
      "step": 2505
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4079354244345112e-06,
      "loss": 0.2016,
      "step": 2506
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.4024226398877572e-06,
      "loss": 0.1688,
      "step": 2507
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3969198552218731e-06,
      "loss": 0.205,
      "step": 2508
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.391427076837174e-06,
      "loss": 0.225,
      "step": 2509
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.385944311122337e-06,
      "loss": 0.1455,
      "step": 2510
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3804715644543886e-06,
      "loss": 0.1794,
      "step": 2511
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3750088431987073e-06,
      "loss": 0.255,
      "step": 2512
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3695561537090062e-06,
      "loss": 0.1835,
      "step": 2513
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3641135023273366e-06,
      "loss": 0.1435,
      "step": 2514
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3586808953840668e-06,
      "loss": 0.1779,
      "step": 2515
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3532583391978904e-06,
      "loss": 0.1963,
      "step": 2516
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.3478458400758021e-06,
      "loss": 0.193,
      "step": 2517
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3424434043131084e-06,
      "loss": 0.1555,
      "step": 2518
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3370510381934032e-06,
      "loss": 0.1645,
      "step": 2519
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3316687479885748e-06,
      "loss": 0.1922,
      "step": 2520
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3262965399587858e-06,
      "loss": 0.1563,
      "step": 2521
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.32093442035248e-06,
      "loss": 0.1925,
      "step": 2522
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3155823954063585e-06,
      "loss": 0.1531,
      "step": 2523
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3102404713453876e-06,
      "loss": 0.178,
      "step": 2524
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.3049086543827827e-06,
      "loss": 0.1886,
      "step": 2525
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.299586950720003e-06,
      "loss": 0.1544,
      "step": 2526
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.2942753665467488e-06,
      "loss": 0.1659,
      "step": 2527
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.2889739080409446e-06,
      "loss": 0.2179,
      "step": 2528
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.283682581368745e-06,
      "loss": 0.2181,
      "step": 2529
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.2784013926845084e-06,
      "loss": 0.2074,
      "step": 2530
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.2731303481308178e-06,
      "loss": 0.1881,
      "step": 2531
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.2678694538384407e-06,
      "loss": 0.1315,
      "step": 2532
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2626187159263536e-06,
      "loss": 0.1498,
      "step": 2533
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2573781405017083e-06,
      "loss": 0.1768,
      "step": 2534
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2521477336598453e-06,
      "loss": 0.2021,
      "step": 2535
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2469275014842708e-06,
      "loss": 0.1541,
      "step": 2536
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2417174500466644e-06,
      "loss": 0.1878,
      "step": 2537
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2365175854068556e-06,
      "loss": 0.1834,
      "step": 2538
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2313279136128365e-06,
      "loss": 0.1896,
      "step": 2539
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2261484407007306e-06,
      "loss": 0.1904,
      "step": 2540
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2209791726948127e-06,
      "loss": 0.1987,
      "step": 2541
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2158201156074756e-06,
      "loss": 0.2205,
      "step": 2542
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.210671275439248e-06,
      "loss": 0.1948,
      "step": 2543
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.205532658178763e-06,
      "loss": 0.2015,
      "step": 2544
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.2004042698027728e-06,
      "loss": 0.1629,
      "step": 2545
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.1952861162761298e-06,
      "loss": 0.1712,
      "step": 2546
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.1901782035517783e-06,
      "loss": 0.2045,
      "step": 2547
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1850805375707587e-06,
      "loss": 0.1756,
      "step": 2548
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1799931242621842e-06,
      "loss": 0.159,
      "step": 2549
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1749159695432522e-06,
      "loss": 0.1803,
      "step": 2550
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1698490793192218e-06,
      "loss": 0.1604,
      "step": 2551
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1647924594834181e-06,
      "loss": 0.1781,
      "step": 2552
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1597461159172151e-06,
      "loss": 0.1907,
      "step": 2553
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1547100544900414e-06,
      "loss": 0.2205,
      "step": 2554
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1496842810593578e-06,
      "loss": 0.1197,
      "step": 2555
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1446688014706698e-06,
      "loss": 0.1608,
      "step": 2556
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1396636215574997e-06,
      "loss": 0.2051,
      "step": 2557
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1346687471413975e-06,
      "loss": 0.1682,
      "step": 2558
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.129684184031924e-06,
      "loss": 0.1577,
      "step": 2559
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1247099380266435e-06,
      "loss": 0.1739,
      "step": 2560
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1197460149111282e-06,
      "loss": 0.2007,
      "step": 2561
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.1147924204589355e-06,
      "loss": 0.172,
      "step": 2562
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.1098491604316185e-06,
      "loss": 0.1555,
      "step": 2563
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.1049162405786985e-06,
      "loss": 0.2006,
      "step": 2564
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0999936666376832e-06,
      "loss": 0.1674,
      "step": 2565
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0950814443340352e-06,
      "loss": 0.1761,
      "step": 2566
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0901795793811876e-06,
      "loss": 0.19,
      "step": 2567
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0852880774805163e-06,
      "loss": 0.2172,
      "step": 2568
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0804069443213527e-06,
      "loss": 0.1983,
      "step": 2569
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.075536185580962e-06,
      "loss": 0.1599,
      "step": 2570
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0706758069245482e-06,
      "loss": 0.158,
      "step": 2571
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0658258140052392e-06,
      "loss": 0.1654,
      "step": 2572
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0609862124640814e-06,
      "loss": 0.1796,
      "step": 2573
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0561570079300409e-06,
      "loss": 0.176,
      "step": 2574
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0513382060199816e-06,
      "loss": 0.1912,
      "step": 2575
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0465298123386802e-06,
      "loss": 0.2205,
      "step": 2576
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.0417318324787961e-06,
      "loss": 0.1592,
      "step": 2577
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0369442720208832e-06,
      "loss": 0.1648,
      "step": 2578
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0321671365333742e-06,
      "loss": 0.1806,
      "step": 2579
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0274004315725772e-06,
      "loss": 0.2191,
      "step": 2580
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0226441626826667e-06,
      "loss": 0.1773,
      "step": 2581
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0178983353956838e-06,
      "loss": 0.1733,
      "step": 2582
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.013162955231517e-06,
      "loss": 0.1564,
      "step": 2583
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0084380276979134e-06,
      "loss": 0.2025,
      "step": 2584
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.0037235582904513e-06,
      "loss": 0.1915,
      "step": 2585
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.990195524925584e-07,
      "loss": 0.1731,
      "step": 2586
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.943260157754775e-07,
      "loss": 0.1955,
      "step": 2587
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.896429535982866e-07,
      "loss": 0.1549,
      "step": 2588
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.849703714078784e-07,
      "loss": 0.183,
      "step": 2589
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.803082746389493e-07,
      "loss": 0.2093,
      "step": 2590
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.75656668714009e-07,
      "loss": 0.1834,
      "step": 2591
    },
    {
      "epoch": 1.72,
      "learning_rate": 9.710155590433568e-07,
      "loss": 0.1788,
      "step": 2592
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.663849510250945e-07,
      "loss": 0.186,
      "step": 2593
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.61764850045096e-07,
      "loss": 0.1761,
      "step": 2594
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.571552614770286e-07,
      "loss": 0.1748,
      "step": 2595
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.525561906823188e-07,
      "loss": 0.1498,
      "step": 2596
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.479676430101726e-07,
      "loss": 0.1801,
      "step": 2597
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.433896237975448e-07,
      "loss": 0.2041,
      "step": 2598
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.388221383691553e-07,
      "loss": 0.1794,
      "step": 2599
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.342651920374624e-07,
      "loss": 0.1794,
      "step": 2600
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.297187901026749e-07,
      "loss": 0.151,
      "step": 2601
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.251829378527288e-07,
      "loss": 0.1839,
      "step": 2602
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.206576405632995e-07,
      "loss": 0.1424,
      "step": 2603
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.161429034977753e-07,
      "loss": 0.2076,
      "step": 2604
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.116387319072717e-07,
      "loss": 0.2276,
      "step": 2605
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.071451310306078e-07,
      "loss": 0.208,
      "step": 2606
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.02662106094312e-07,
      "loss": 0.1521,
      "step": 2607
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.981896623126118e-07,
      "loss": 0.1503,
      "step": 2608
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.937278048874254e-07,
      "loss": 0.163,
      "step": 2609
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.892765390083602e-07,
      "loss": 0.1962,
      "step": 2610
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.848358698527026e-07,
      "loss": 0.1726,
      "step": 2611
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.804058025854157e-07,
      "loss": 0.162,
      "step": 2612
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.759863423591286e-07,
      "loss": 0.1938,
      "step": 2613
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.715774943141398e-07,
      "loss": 0.1696,
      "step": 2614
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.671792635783948e-07,
      "loss": 0.2065,
      "step": 2615
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.627916552674997e-07,
      "loss": 0.1665,
      "step": 2616
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.584146744846978e-07,
      "loss": 0.1306,
      "step": 2617
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.540483263208799e-07,
      "loss": 0.1669,
      "step": 2618
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.496926158545626e-07,
      "loss": 0.2206,
      "step": 2619
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.453475481518891e-07,
      "loss": 0.2028,
      "step": 2620
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.410131282666334e-07,
      "loss": 0.1503,
      "step": 2621
    },
    {
      "epoch": 1.74,
      "learning_rate": 8.366893612401739e-07,
      "loss": 0.2198,
      "step": 2622
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.323762521015056e-07,
      "loss": 0.1582,
      "step": 2623
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.280738058672244e-07,
      "loss": 0.1555,
      "step": 2624
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.237820275415265e-07,
      "loss": 0.1851,
      "step": 2625
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.195009221161954e-07,
      "loss": 0.1544,
      "step": 2626
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.152304945706068e-07,
      "loss": 0.2038,
      "step": 2627
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.109707498717123e-07,
      "loss": 0.1583,
      "step": 2628
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.06721692974043e-07,
      "loss": 0.1505,
      "step": 2629
    },
    {
      "epoch": 1.75,
      "learning_rate": 8.02483328819692e-07,
      "loss": 0.1582,
      "step": 2630
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.982556623383242e-07,
      "loss": 0.1602,
      "step": 2631
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.940386984471538e-07,
      "loss": 0.1842,
      "step": 2632
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.898324420509507e-07,
      "loss": 0.1841,
      "step": 2633
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.856368980420336e-07,
      "loss": 0.1724,
      "step": 2634
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.814520713002549e-07,
      "loss": 0.1956,
      "step": 2635
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.772779666930097e-07,
      "loss": 0.1858,
      "step": 2636
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.731145890752134e-07,
      "loss": 0.1629,
      "step": 2637
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.689619432893126e-07,
      "loss": 0.1701,
      "step": 2638
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.648200341652645e-07,
      "loss": 0.1831,
      "step": 2639
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.60688866520547e-07,
      "loss": 0.1563,
      "step": 2640
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.565684451601363e-07,
      "loss": 0.2179,
      "step": 2641
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.524587748765144e-07,
      "loss": 0.1918,
      "step": 2642
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.483598604496556e-07,
      "loss": 0.1645,
      "step": 2643
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.442717066470295e-07,
      "loss": 0.1427,
      "step": 2644
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.401943182235804e-07,
      "loss": 0.1443,
      "step": 2645
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.361276999217437e-07,
      "loss": 0.1795,
      "step": 2646
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.320718564714158e-07,
      "loss": 0.1966,
      "step": 2647
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.280267925899698e-07,
      "loss": 0.2755,
      "step": 2648
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.239925129822356e-07,
      "loss": 0.1794,
      "step": 2649
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.19969022340502e-07,
      "loss": 0.2099,
      "step": 2650
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.159563253445123e-07,
      "loss": 0.1412,
      "step": 2651
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.119544266614486e-07,
      "loss": 0.1465,
      "step": 2652
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.079633309459411e-07,
      "loss": 0.2086,
      "step": 2653
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.039830428400463e-07,
      "loss": 0.2032,
      "step": 2654
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.000135669732588e-07,
      "loss": 0.1569,
      "step": 2655
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.960549079624923e-07,
      "loss": 0.1779,
      "step": 2656
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.921070704120825e-07,
      "loss": 0.1502,
      "step": 2657
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.881700589137741e-07,
      "loss": 0.1638,
      "step": 2658
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.842438780467264e-07,
      "loss": 0.2324,
      "step": 2659
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.803285323774944e-07,
      "loss": 0.1523,
      "step": 2660
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.764240264600375e-07,
      "loss": 0.1568,
      "step": 2661
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.725303648357018e-07,
      "loss": 0.1845,
      "step": 2662
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.686475520332237e-07,
      "loss": 0.1917,
      "step": 2663
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.647755925687194e-07,
      "loss": 0.1785,
      "step": 2664
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.609144909456844e-07,
      "loss": 0.1733,
      "step": 2665
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.570642516549808e-07,
      "loss": 0.2003,
      "step": 2666
    },
    {
      "epoch": 1.77,
      "learning_rate": 6.532248791748408e-07,
      "loss": 0.1749,
      "step": 2667
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.493963779708545e-07,
      "loss": 0.1605,
      "step": 2668
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.4557875249597e-07,
      "loss": 0.1808,
      "step": 2669
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.417720071904854e-07,
      "loss": 0.1857,
      "step": 2670
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.379761464820399e-07,
      "loss": 0.1862,
      "step": 2671
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.341911747856211e-07,
      "loss": 0.2183,
      "step": 2672
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.304170965035428e-07,
      "loss": 0.1369,
      "step": 2673
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.266539160254536e-07,
      "loss": 0.2117,
      "step": 2674
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.229016377283248e-07,
      "loss": 0.1901,
      "step": 2675
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.191602659764506e-07,
      "loss": 0.1961,
      "step": 2676
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.154298051214336e-07,
      "loss": 0.1802,
      "step": 2677
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.11710259502194e-07,
      "loss": 0.1643,
      "step": 2678
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.08001633444949e-07,
      "loss": 0.1647,
      "step": 2679
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.043039312632182e-07,
      "loss": 0.2143,
      "step": 2680
    },
    {
      "epoch": 1.78,
      "learning_rate": 6.006171572578168e-07,
      "loss": 0.1887,
      "step": 2681
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.969413157168468e-07,
      "loss": 0.1423,
      "step": 2682
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.932764109156985e-07,
      "loss": 0.1624,
      "step": 2683
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.896224471170353e-07,
      "loss": 0.1675,
      "step": 2684
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.859794285708031e-07,
      "loss": 0.2221,
      "step": 2685
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.823473595142104e-07,
      "loss": 0.1352,
      "step": 2686
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.787262441717356e-07,
      "loss": 0.1608,
      "step": 2687
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.751160867551098e-07,
      "loss": 0.1588,
      "step": 2688
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.715168914633284e-07,
      "loss": 0.2142,
      "step": 2689
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.679286624826275e-07,
      "loss": 0.1613,
      "step": 2690
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.643514039864961e-07,
      "loss": 0.2188,
      "step": 2691
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.607851201356563e-07,
      "loss": 0.1767,
      "step": 2692
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.572298150780708e-07,
      "loss": 0.1808,
      "step": 2693
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.536854929489288e-07,
      "loss": 0.1717,
      "step": 2694
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.501521578706492e-07,
      "loss": 0.1369,
      "step": 2695
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.466298139528714e-07,
      "loss": 0.1609,
      "step": 2696
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.43118465292446e-07,
      "loss": 0.1711,
      "step": 2697
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.396181159734414e-07,
      "loss": 0.1478,
      "step": 2698
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.361287700671269e-07,
      "loss": 0.1764,
      "step": 2699
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.326504316319792e-07,
      "loss": 0.146,
      "step": 2700
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.291831047136664e-07,
      "loss": 0.1955,
      "step": 2701
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.25726793345055e-07,
      "loss": 0.1485,
      "step": 2702
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.222815015461935e-07,
      "loss": 0.1486,
      "step": 2703
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.188472333243189e-07,
      "loss": 0.1562,
      "step": 2704
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.154239926738402e-07,
      "loss": 0.1278,
      "step": 2705
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.120117835763494e-07,
      "loss": 0.166,
      "step": 2706
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.086106100005972e-07,
      "loss": 0.1759,
      "step": 2707
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.052204759025081e-07,
      "loss": 0.2212,
      "step": 2708
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.01841385225158e-07,
      "loss": 0.1468,
      "step": 2709
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.984733418987864e-07,
      "loss": 0.2167,
      "step": 2710
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.951163498407763e-07,
      "loss": 0.1707,
      "step": 2711
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.917704129556622e-07,
      "loss": 0.1503,
      "step": 2712
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.884355351351189e-07,
      "loss": 0.1768,
      "step": 2713
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.851117202579558e-07,
      "loss": 0.2251,
      "step": 2714
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.817989721901195e-07,
      "loss": 0.1448,
      "step": 2715
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.784972947846789e-07,
      "loss": 0.155,
      "step": 2716
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.7520669188183324e-07,
      "loss": 0.1583,
      "step": 2717
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.719271673088954e-07,
      "loss": 0.1836,
      "step": 2718
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.686587248802954e-07,
      "loss": 0.1644,
      "step": 2719
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.6540136839757335e-07,
      "loss": 0.1358,
      "step": 2720
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.621551016493764e-07,
      "loss": 0.1557,
      "step": 2721
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.5891992841144873e-07,
      "loss": 0.1499,
      "step": 2722
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.556958524466393e-07,
      "loss": 0.1563,
      "step": 2723
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.524828775048828e-07,
      "loss": 0.1734,
      "step": 2724
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.4928100732320565e-07,
      "loss": 0.1565,
      "step": 2725
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.4609024562571545e-07,
      "loss": 0.1706,
      "step": 2726
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.4291059612360464e-07,
      "loss": 0.164,
      "step": 2727
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.3974206251513495e-07,
      "loss": 0.1926,
      "step": 2728
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.365846484856451e-07,
      "loss": 0.1497,
      "step": 2729
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.3343835770753405e-07,
      "loss": 0.1734,
      "step": 2730
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.30303193840268e-07,
      "loss": 0.1492,
      "step": 2731
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.2717916053037324e-07,
      "loss": 0.1793,
      "step": 2732
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.2406626141142217e-07,
      "loss": 0.1576,
      "step": 2733
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.2096450010404524e-07,
      "loss": 0.1804,
      "step": 2734
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.1787388021591327e-07,
      "loss": 0.175,
      "step": 2735
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.1479440534174075e-07,
      "loss": 0.1754,
      "step": 2736
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.117260790632793e-07,
      "loss": 0.1599,
      "step": 2737
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.08668904949312e-07,
      "loss": 0.1431,
      "step": 2738
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.0562288655565106e-07,
      "loss": 0.196,
      "step": 2739
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.0258802742513483e-07,
      "loss": 0.1599,
      "step": 2740
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.995643310876229e-07,
      "loss": 0.1603,
      "step": 2741
    },
    {
      "epoch": 1.82,
      "learning_rate": 3.965518010599889e-07,
      "loss": 0.1634,
      "step": 2742
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.9355044084612105e-07,
      "loss": 0.1721,
      "step": 2743
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.9056025393691245e-07,
      "loss": 0.1888,
      "step": 2744
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.875812438102655e-07,
      "loss": 0.1845,
      "step": 2745
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.8461341393107867e-07,
      "loss": 0.1869,
      "step": 2746
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.816567677512506e-07,
      "loss": 0.1812,
      "step": 2747
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.787113087096661e-07,
      "loss": 0.202,
      "step": 2748
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.7577704023220583e-07,
      "loss": 0.1945,
      "step": 2749
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.728539657317265e-07,
      "loss": 0.149,
      "step": 2750
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.6994208860807293e-07,
      "loss": 0.1874,
      "step": 2751
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.6704141224806034e-07,
      "loss": 0.1455,
      "step": 2752
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.6415194002547895e-07,
      "loss": 0.1438,
      "step": 2753
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.612736753010859e-07,
      "loss": 0.1749,
      "step": 2754
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.584066214226067e-07,
      "loss": 0.1403,
      "step": 2755
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.5555078172472166e-07,
      "loss": 0.1826,
      "step": 2756
    },
    {
      "epoch": 1.83,
      "learning_rate": 3.527061595290715e-07,
      "loss": 0.1881,
      "step": 2757
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.498727581442507e-07,
      "loss": 0.1617,
      "step": 2758
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.4705058086579866e-07,
      "loss": 0.1434,
      "step": 2759
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.4423963097620415e-07,
      "loss": 0.2372,
      "step": 2760
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.4143991174489524e-07,
      "loss": 0.1533,
      "step": 2761
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.386514264282381e-07,
      "loss": 0.1686,
      "step": 2762
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.3587417826953183e-07,
      "loss": 0.1722,
      "step": 2763
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.331081704990069e-07,
      "loss": 0.1423,
      "step": 2764
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.303534063338187e-07,
      "loss": 0.174,
      "step": 2765
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.276098889780477e-07,
      "loss": 0.1433,
      "step": 2766
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.248776216226879e-07,
      "loss": 0.2232,
      "step": 2767
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.22156607445655e-07,
      "loss": 0.1972,
      "step": 2768
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.1944684961177084e-07,
      "loss": 0.1798,
      "step": 2769
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.167483512727687e-07,
      "loss": 0.1886,
      "step": 2770
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.140611155672824e-07,
      "loss": 0.2015,
      "step": 2771
    },
    {
      "epoch": 1.84,
      "learning_rate": 3.113851456208494e-07,
      "loss": 0.1726,
      "step": 2772
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.0872044454590024e-07,
      "loss": 0.1416,
      "step": 2773
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.060670154417622e-07,
      "loss": 0.1636,
      "step": 2774
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.0342486139465113e-07,
      "loss": 0.1815,
      "step": 2775
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.007939854776665e-07,
      "loss": 0.1849,
      "step": 2776
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.9817439075079393e-07,
      "loss": 0.2087,
      "step": 2777
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.9556608026089394e-07,
      "loss": 0.2023,
      "step": 2778
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.9296905704170544e-07,
      "loss": 0.2182,
      "step": 2779
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.9038332411383765e-07,
      "loss": 0.1597,
      "step": 2780
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.878088844847682e-07,
      "loss": 0.1615,
      "step": 2781
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.852457411488385e-07,
      "loss": 0.191,
      "step": 2782
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.826938970872528e-07,
      "loss": 0.174,
      "step": 2783
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.801533552680724e-07,
      "loss": 0.1777,
      "step": 2784
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.7762411864621364e-07,
      "loss": 0.1692,
      "step": 2785
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.7510619016344107e-07,
      "loss": 0.1749,
      "step": 2786
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.72599572748371e-07,
      "loss": 0.1811,
      "step": 2787
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.7010426931645796e-07,
      "loss": 0.153,
      "step": 2788
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.6762028277000475e-07,
      "loss": 0.1642,
      "step": 2789
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.651476159981448e-07,
      "loss": 0.1717,
      "step": 2790
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.6268627187684745e-07,
      "loss": 0.1839,
      "step": 2791
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.602362532689162e-07,
      "loss": 0.1707,
      "step": 2792
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.5779756302397484e-07,
      "loss": 0.15,
      "step": 2793
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.553702039784789e-07,
      "loss": 0.1727,
      "step": 2794
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.5295417895569665e-07,
      "loss": 0.1548,
      "step": 2795
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.505494907657213e-07,
      "loss": 0.1703,
      "step": 2796
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.4815614220545234e-07,
      "loss": 0.1942,
      "step": 2797
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.4577413605860857e-07,
      "loss": 0.1766,
      "step": 2798
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.434034750957082e-07,
      "loss": 0.1498,
      "step": 2799
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.410441620740811e-07,
      "loss": 0.1864,
      "step": 2800
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.386961997378512e-07,
      "loss": 0.1808,
      "step": 2801
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.3635959081794724e-07,
      "loss": 0.1788,
      "step": 2802
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.3403433803208862e-07,
      "loss": 0.1974,
      "step": 2803
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.3172044408478645e-07,
      "loss": 0.1467,
      "step": 2804
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.2941791166734228e-07,
      "loss": 0.1973,
      "step": 2805
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.271267434578417e-07,
      "loss": 0.1763,
      "step": 2806
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.2484694212115187e-07,
      "loss": 0.1393,
      "step": 2807
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.2257851030892064e-07,
      "loss": 0.196,
      "step": 2808
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.203214506595719e-07,
      "loss": 0.1507,
      "step": 2809
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.1807576579830126e-07,
      "loss": 0.2193,
      "step": 2810
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.1584145833707382e-07,
      "loss": 0.1436,
      "step": 2811
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.1361853087462414e-07,
      "loss": 0.1701,
      "step": 2812
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.114069859964496e-07,
      "loss": 0.1843,
      "step": 2813
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.092068262748048e-07,
      "loss": 0.2435,
      "step": 2814
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0701805426870947e-07,
      "loss": 0.1714,
      "step": 2815
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0484067252393157e-07,
      "loss": 0.1758,
      "step": 2816
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0267468357299425e-07,
      "loss": 0.2323,
      "step": 2817
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0052008993516893e-07,
      "loss": 0.1436,
      "step": 2818
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9837689411647321e-07,
      "loss": 0.1577,
      "step": 2819
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9624509860966867e-07,
      "loss": 0.1925,
      "step": 2820
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9412470589425415e-07,
      "loss": 0.1739,
      "step": 2821
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9201571843647015e-07,
      "loss": 0.1693,
      "step": 2822
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.89918138689289e-07,
      "loss": 0.2188,
      "step": 2823
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.8783196909241352e-07,
      "loss": 0.153,
      "step": 2824
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.8575721207227837e-07,
      "loss": 0.1937,
      "step": 2825
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.836938700420432e-07,
      "loss": 0.2035,
      "step": 2826
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.816419454015894e-07,
      "loss": 0.1524,
      "step": 2827
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.7960144053752015e-07,
      "loss": 0.1839,
      "step": 2828
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.7757235782315585e-07,
      "loss": 0.1633,
      "step": 2829
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.7555469961853089e-07,
      "loss": 0.1869,
      "step": 2830
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.7354846827039363e-07,
      "loss": 0.1535,
      "step": 2831
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.7155366611220082e-07,
      "loss": 0.1847,
      "step": 2832
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.695702954641143e-07,
      "loss": 0.1753,
      "step": 2833
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6759835863300434e-07,
      "loss": 0.2071,
      "step": 2834
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6563785791243625e-07,
      "loss": 0.1576,
      "step": 2835
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6368879558267825e-07,
      "loss": 0.1667,
      "step": 2836
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6175117391069583e-07,
      "loss": 0.1372,
      "step": 2837
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.5982499515014183e-07,
      "loss": 0.1844,
      "step": 2838
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.5791026154136635e-07,
      "loss": 0.1874,
      "step": 2839
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.5600697531140353e-07,
      "loss": 0.1678,
      "step": 2840
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.5411513867397366e-07,
      "loss": 0.145,
      "step": 2841
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.5223475382947995e-07,
      "loss": 0.1893,
      "step": 2842
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.5036582296500846e-07,
      "loss": 0.1835,
      "step": 2843
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.4850834825432036e-07,
      "loss": 0.1983,
      "step": 2844
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.4666233185785194e-07,
      "loss": 0.1729,
      "step": 2845
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.4482777592271347e-07,
      "loss": 0.1734,
      "step": 2846
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.4300468258268586e-07,
      "loss": 0.1964,
      "step": 2847
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.4119305395821626e-07,
      "loss": 0.2009,
      "step": 2848
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3939289215641915e-07,
      "loss": 0.1696,
      "step": 2849
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.376041992710697e-07,
      "loss": 0.1782,
      "step": 2850
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3582697738260487e-07,
      "loss": 0.1428,
      "step": 2851
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3406122855812e-07,
      "loss": 0.2289,
      "step": 2852
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.323069548513656e-07,
      "loss": 0.158,
      "step": 2853
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.3056415830274505e-07,
      "loss": 0.1583,
      "step": 2854
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2883284093931137e-07,
      "loss": 0.1337,
      "step": 2855
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.271130047747715e-07,
      "loss": 0.1562,
      "step": 2856
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2540465180946982e-07,
      "loss": 0.1801,
      "step": 2857
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.237077840304024e-07,
      "loss": 0.1586,
      "step": 2858
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.2202240341120163e-07,
      "loss": 0.1725,
      "step": 2859
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.203485119121439e-07,
      "loss": 0.1786,
      "step": 2860
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1868611148013742e-07,
      "loss": 0.1756,
      "step": 2861
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.1703520404872993e-07,
      "loss": 0.1891,
      "step": 2862
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1539579153809766e-07,
      "loss": 0.1964,
      "step": 2863
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1376787585504867e-07,
      "loss": 0.1477,
      "step": 2864
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1215145889302059e-07,
      "loss": 0.15,
      "step": 2865
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.1054654253207397e-07,
      "loss": 0.217,
      "step": 2866
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.0895312863889452e-07,
      "loss": 0.1271,
      "step": 2867
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.0737121906678749e-07,
      "loss": 0.161,
      "step": 2868
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.0580081565568224e-07,
      "loss": 0.1626,
      "step": 2869
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.0424192023211988e-07,
      "loss": 0.1856,
      "step": 2870
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.0269453460925894e-07,
      "loss": 0.164,
      "step": 2871
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.0115866058686974e-07,
      "loss": 0.1772,
      "step": 2872
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.963429995133667e-08,
      "loss": 0.1759,
      "step": 2873
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.812145447564925e-08,
      "loss": 0.1809,
      "step": 2874
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.66201259194044e-08,
      "loss": 0.1906,
      "step": 2875
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.513031602880418e-08,
      "loss": 0.1611,
      "step": 2876
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.365202653665362e-08,
      "loss": 0.182,
      "step": 2877
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.21852591623562e-08,
      "loss": 0.1763,
      "step": 2878
    },
    {
      "epoch": 1.92,
      "learning_rate": 9.07300156119184e-08,
      "loss": 0.2007,
      "step": 2879
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.928629757793961e-08,
      "loss": 0.1989,
      "step": 2880
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.785410673961325e-08,
      "loss": 0.2208,
      "step": 2881
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.643344476272908e-08,
      "loss": 0.1723,
      "step": 2882
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.50243132996631e-08,
      "loss": 0.1796,
      "step": 2883
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.362671398938537e-08,
      "loss": 0.1878,
      "step": 2884
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.224064845744673e-08,
      "loss": 0.2257,
      "step": 2885
    },
    {
      "epoch": 1.92,
      "learning_rate": 8.086611831598867e-08,
      "loss": 0.1631,
      "step": 2886
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.950312516373238e-08,
      "loss": 0.1909,
      "step": 2887
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.815167058598084e-08,
      "loss": 0.1692,
      "step": 2888
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.681175615461889e-08,
      "loss": 0.1999,
      "step": 2889
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.548338342810435e-08,
      "loss": 0.1541,
      "step": 2890
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.416655395147576e-08,
      "loss": 0.1893,
      "step": 2891
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.28612692563424e-08,
      "loss": 0.1974,
      "step": 2892
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.156753086088542e-08,
      "loss": 0.191,
      "step": 2893
    },
    {
      "epoch": 1.93,
      "learning_rate": 7.028534026986111e-08,
      "loss": 0.1918,
      "step": 2894
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.901469897458768e-08,
      "loss": 0.1666,
      "step": 2895
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.775560845295625e-08,
      "loss": 0.1862,
      "step": 2896
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.650807016941873e-08,
      "loss": 0.1953,
      "step": 2897
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.527208557499443e-08,
      "loss": 0.1705,
      "step": 2898
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.404765610726116e-08,
      "loss": 0.1917,
      "step": 2899
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.283478319035862e-08,
      "loss": 0.1974,
      "step": 2900
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.163346823498506e-08,
      "loss": 0.1849,
      "step": 2901
    },
    {
      "epoch": 1.93,
      "learning_rate": 6.044371263839499e-08,
      "loss": 0.1522,
      "step": 2902
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9265517784399266e-08,
      "loss": 0.1838,
      "step": 2903
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.8098885043360585e-08,
      "loss": 0.1572,
      "step": 2904
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.694381577219577e-08,
      "loss": 0.1708,
      "step": 2905
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.5800311314369026e-08,
      "loss": 0.1963,
      "step": 2906
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.466837299989758e-08,
      "loss": 0.1674,
      "step": 2907
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.354800214534384e-08,
      "loss": 0.2044,
      "step": 2908
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.243920005381542e-08,
      "loss": 0.1563,
      "step": 2909
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.134196801496516e-08,
      "loss": 0.1937,
      "step": 2910
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.025630730498887e-08,
      "loss": 0.1746,
      "step": 2911
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.918221918662314e-08,
      "loss": 0.1898,
      "step": 2912
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.8119704909145306e-08,
      "loss": 0.1681,
      "step": 2913
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.7068765708369045e-08,
      "loss": 0.1883,
      "step": 2914
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.602940280664769e-08,
      "loss": 0.1793,
      "step": 2915
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.500161741286868e-08,
      "loss": 0.146,
      "step": 2916
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.398541072245355e-08,
      "loss": 0.1695,
      "step": 2917
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.298078391735794e-08,
      "loss": 0.1765,
      "step": 2918
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.198773816606605e-08,
      "loss": 0.1784,
      "step": 2919
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.100627462359508e-08,
      "loss": 0.2176,
      "step": 2920
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.0036394431489656e-08,
      "loss": 0.1481,
      "step": 2921
    },
    {
      "epoch": 1.94,
      "learning_rate": 3.907809871782076e-08,
      "loss": 0.1783,
      "step": 2922
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.8131388597187904e-08,
      "loss": 0.1899,
      "step": 2923
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.719626517071251e-08,
      "loss": 0.1752,
      "step": 2924
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.627272952604233e-08,
      "loss": 0.1784,
      "step": 2925
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.536078273734367e-08,
      "loss": 0.1606,
      "step": 2926
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.4460425865308066e-08,
      "loss": 0.1525,
      "step": 2927
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.357165995714451e-08,
      "loss": 0.1573,
      "step": 2928
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.269448604657832e-08,
      "loss": 0.1522,
      "step": 2929
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.1828905153856725e-08,
      "loss": 0.1797,
      "step": 2930
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.0974918285741064e-08,
      "loss": 0.1628,
      "step": 2931
    },
    {
      "epoch": 1.95,
      "learning_rate": 3.013252643550568e-08,
      "loss": 0.1992,
      "step": 2932
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.9301730582940168e-08,
      "loss": 0.1866,
      "step": 2933
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.8482531694348225e-08,
      "loss": 0.1694,
      "step": 2934
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.7674930722543237e-08,
      "loss": 0.1884,
      "step": 2935
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.6878928606849376e-08,
      "loss": 0.1418,
      "step": 2936
    },
    {
      "epoch": 1.95,
      "learning_rate": 2.6094526273100496e-08,
      "loss": 0.1472,
      "step": 2937
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.5321724633639022e-08,
      "loss": 0.1792,
      "step": 2938
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.456052458731262e-08,
      "loss": 0.1633,
      "step": 2939
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.3810927019477513e-08,
      "loss": 0.1499,
      "step": 2940
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.3072932801994074e-08,
      "loss": 0.1908,
      "step": 2941
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.234654279322568e-08,
      "loss": 0.1934,
      "step": 2942
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.1631757838039836e-08,
      "loss": 0.1755,
      "step": 2943
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.0928578767808182e-08,
      "loss": 0.1609,
      "step": 2944
    },
    {
      "epoch": 1.96,
      "learning_rate": 2.023700640039983e-08,
      "loss": 0.1699,
      "step": 2945
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9557041540185783e-08,
      "loss": 0.1518,
      "step": 2946
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8888684978035642e-08,
      "loss": 0.1823,
      "step": 2947
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8231937491319794e-08,
      "loss": 0.2003,
      "step": 2948
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.758679984390277e-08,
      "loss": 0.1861,
      "step": 2949
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.6953272786147667e-08,
      "loss": 0.1857,
      "step": 2950
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.6331357054911734e-08,
      "loss": 0.1613,
      "step": 2951
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.5721053373548567e-08,
      "loss": 0.1653,
      "step": 2952
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.5122362451904794e-08,
      "loss": 0.2039,
      "step": 2953
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.4535284986322285e-08,
      "loss": 0.2104,
      "step": 2954
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.395982165963039e-08,
      "loss": 0.1463,
      "step": 2955
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.3395973141155927e-08,
      "loss": 0.1923,
      "step": 2956
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.2843740086712075e-08,
      "loss": 0.1691,
      "step": 2957
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.2303123138603934e-08,
      "loss": 0.1837,
      "step": 2958
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1774122925625186e-08,
      "loss": 0.1487,
      "step": 2959
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.1256740063059213e-08,
      "loss": 0.1843,
      "step": 2960
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0750975152675758e-08,
      "loss": 0.1747,
      "step": 2961
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.0256828782733153e-08,
      "loss": 0.1701,
      "step": 2962
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.774301527972762e-09,
      "loss": 0.1434,
      "step": 2963
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.30339394962676e-09,
      "loss": 0.1835,
      "step": 2964
    },
    {
      "epoch": 1.97,
      "learning_rate": 8.84410659540924e-09,
      "loss": 0.1917,
      "step": 2965
    },
    {
      "epoch": 1.97,
      "learning_rate": 8.396439999519556e-09,
      "loss": 0.1787,
      "step": 2966
    },
    {
      "epoch": 1.97,
      "learning_rate": 7.960394682638983e-09,
      "loss": 0.1511,
      "step": 2967
    },
    {
      "epoch": 1.97,
      "learning_rate": 7.535971151936271e-09,
      "loss": 0.1734,
      "step": 2968
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.123169901058769e-09,
      "loss": 0.1939,
      "step": 2969
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.721991410137962e-09,
      "loss": 0.1673,
      "step": 2970
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.3324361457872685e-09,
      "loss": 0.1912,
      "step": 2971
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.954504561097585e-09,
      "loss": 0.1937,
      "step": 2972
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.5881970956450646e-09,
      "loss": 0.2231,
      "step": 2973
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.233514175483345e-09,
      "loss": 0.1822,
      "step": 2974
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.890456213144656e-09,
      "loss": 0.1432,
      "step": 2975
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.5590236076431536e-09,
      "loss": 0.1559,
      "step": 2976
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.239216744467145e-09,
      "loss": 0.1465,
      "step": 2977
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.931035995586863e-09,
      "loss": 0.1493,
      "step": 2978
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.6344817194489123e-09,
      "loss": 0.1516,
      "step": 2979
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.349554260976273e-09,
      "loss": 0.1746,
      "step": 2980
    },
    {
      "epoch": 1.98,
      "learning_rate": 3.0762539515694077e-09,
      "loss": 0.1848,
      "step": 2981
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.8145811091062626e-09,
      "loss": 0.1587,
      "step": 2982
    },
    {
      "epoch": 1.98,
      "learning_rate": 2.564536037937826e-09,
      "loss": 0.1632,
      "step": 2983
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.326119028893681e-09,
      "loss": 0.1727,
      "step": 2984
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.0993303592775628e-09,
      "loss": 0.1854,
      "step": 2985
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.88417029286847e-09,
      "loss": 0.1441,
      "step": 2986
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.680639079920665e-09,
      "loss": 0.2128,
      "step": 2987
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.4887369571603416e-09,
      "loss": 0.1573,
      "step": 2988
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.3084641477922878e-09,
      "loss": 0.1623,
      "step": 2989
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.1398208614910034e-09,
      "loss": 0.1556,
      "step": 2990
    },
    {
      "epoch": 1.99,
      "learning_rate": 9.828072944051414e-10,
      "loss": 0.1457,
      "step": 2991
    },
    {
      "epoch": 1.99,
      "learning_rate": 8.374236291608384e-10,
      "loss": 0.1563,
      "step": 2992
    },
    {
      "epoch": 1.99,
      "learning_rate": 7.036700348528324e-10,
      "loss": 0.1816,
      "step": 2993
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.815466670500147e-10,
      "loss": 0.174,
      "step": 2994
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.710536677954292e-10,
      "loss": 0.141,
      "step": 2995
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.72191165602942e-10,
      "loss": 0.1747,
      "step": 2996
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.849592754616826e-10,
      "loss": 0.1844,
      "step": 2997
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.093580988304922e-10,
      "loss": 0.179,
      "step": 2998
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.4538772364236508e-10,
      "loss": 0.1688,
      "step": 2999
    },
    {
      "epoch": 2.0,
      "learning_rate": 9.304822430000749e-11,
      "loss": 0.2051,
      "step": 3000
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.233966168027849e-11,
      "loss": 0.1558,
      "step": 3001
    },
    {
      "epoch": 2.0,
      "learning_rate": 2.3262083130859426e-11,
      "loss": 0.123,
      "step": 3002
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.815522473584523e-12,
      "loss": 0.1744,
      "step": 3003
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0,
      "loss": 0.1928,
      "step": 3004
    },
    {
      "epoch": 2.0,
      "step": 3004,
      "total_flos": 1.5919114231845224e+19,
      "train_loss": 0.5608125850855907,
      "train_runtime": 132555.8326,
      "train_samples_per_second": 1.451,
      "train_steps_per_second": 0.023
    }
  ],
  "max_steps": 3004,
  "num_train_epochs": 2,
  "total_flos": 1.5919114231845224e+19,
  "trial_name": null,
  "trial_params": null
}