{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.10506408909434756,
  "eval_steps": 500,
  "global_step": 1000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.0000000000000002e-07,
      "loss": 0.5147,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.0000000000000003e-07,
      "loss": 1.0359,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.000000000000001e-07,
      "loss": 1.2442,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-07,
      "loss": 0.8008,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 1.4923,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.2000000000000002e-06,
      "loss": 1.5128,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.4000000000000001e-06,
      "loss": 1.8479,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6000000000000001e-06,
      "loss": 1.3834,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.8e-06,
      "loss": 1.3908,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 1.2109,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.2e-06,
      "loss": 1.6184,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 0.8305,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.6e-06,
      "loss": 0.7623,
      "step": 13
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.8000000000000003e-06,
      "loss": 1.4248,
      "step": 14
    },
    {
      "epoch": 0.0,
      "learning_rate": 3e-06,
      "loss": 0.4646,
      "step": 15
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.2000000000000003e-06,
      "loss": 0.6675,
      "step": 16
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.4000000000000005e-06,
      "loss": 1.2929,
      "step": 17
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.6e-06,
      "loss": 0.9805,
      "step": 18
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.8e-06,
      "loss": 0.8487,
      "step": 19
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.9878,
      "step": 20
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.2000000000000004e-06,
      "loss": 1.4433,
      "step": 21
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.4e-06,
      "loss": 0.7954,
      "step": 22
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.6e-06,
      "loss": 1.2045,
      "step": 23
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.800000000000001e-06,
      "loss": 1.0289,
      "step": 24
    },
    {
      "epoch": 0.0,
      "learning_rate": 5e-06,
      "loss": 0.7529,
      "step": 25
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.2e-06,
      "loss": 1.2135,
      "step": 26
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.4e-06,
      "loss": 0.9878,
      "step": 27
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.600000000000001e-06,
      "loss": 1.192,
      "step": 28
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.8e-06,
      "loss": 1.3389,
      "step": 29
    },
    {
      "epoch": 0.0,
      "learning_rate": 6e-06,
      "loss": 0.9383,
      "step": 30
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.2e-06,
      "loss": 1.1623,
      "step": 31
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.4000000000000006e-06,
      "loss": 0.969,
      "step": 32
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.6e-06,
      "loss": 1.0833,
      "step": 33
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.800000000000001e-06,
      "loss": 1.2367,
      "step": 34
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.000000000000001e-06,
      "loss": 0.6725,
      "step": 35
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.2e-06,
      "loss": 1.2712,
      "step": 36
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.4e-06,
      "loss": 1.2394,
      "step": 37
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.6e-06,
      "loss": 1.1948,
      "step": 38
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.8e-06,
      "loss": 0.7856,
      "step": 39
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-06,
      "loss": 1.4131,
      "step": 40
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.200000000000001e-06,
      "loss": 0.6076,
      "step": 41
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.400000000000001e-06,
      "loss": 0.934,
      "step": 42
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.599999999999999e-06,
      "loss": 1.4066,
      "step": 43
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.8e-06,
      "loss": 0.5145,
      "step": 44
    },
    {
      "epoch": 0.0,
      "learning_rate": 9e-06,
      "loss": 0.7606,
      "step": 45
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.2e-06,
      "loss": 1.2383,
      "step": 46
    },
    {
      "epoch": 0.0,
      "learning_rate": 9.4e-06,
      "loss": 1.0737,
      "step": 47
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.600000000000001e-06,
      "loss": 0.7418,
      "step": 48
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.800000000000001e-06,
      "loss": 1.2458,
      "step": 49
    },
    {
      "epoch": 0.01,
      "learning_rate": 1e-05,
      "loss": 1.0782,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.02e-05,
      "loss": 1.496,
      "step": 51
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.04e-05,
      "loss": 1.0337,
      "step": 52
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.06e-05,
      "loss": 0.6506,
      "step": 53
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.08e-05,
      "loss": 1.3999,
      "step": 54
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 1.1575,
      "step": 55
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1200000000000001e-05,
      "loss": 0.8913,
      "step": 56
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1400000000000001e-05,
      "loss": 1.3943,
      "step": 57
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.16e-05,
      "loss": 0.7901,
      "step": 58
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.18e-05,
      "loss": 0.5152,
      "step": 59
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2e-05,
      "loss": 1.3073,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.22e-05,
      "loss": 0.9406,
      "step": 61
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.24e-05,
      "loss": 1.8643,
      "step": 62
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2600000000000001e-05,
      "loss": 1.5102,
      "step": 63
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2800000000000001e-05,
      "loss": 1.1567,
      "step": 64
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.9563,
      "step": 65
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.32e-05,
      "loss": 1.3525,
      "step": 66
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3400000000000002e-05,
      "loss": 0.7416,
      "step": 67
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3600000000000002e-05,
      "loss": 1.299,
      "step": 68
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.3800000000000002e-05,
      "loss": 0.9279,
      "step": 69
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 0.7411,
      "step": 70
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.42e-05,
      "loss": 0.7108,
      "step": 71
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.44e-05,
      "loss": 1.1978,
      "step": 72
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4599999999999999e-05,
      "loss": 0.9814,
      "step": 73
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.48e-05,
      "loss": 1.3739,
      "step": 74
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5e-05,
      "loss": 0.6597,
      "step": 75
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.52e-05,
      "loss": 0.857,
      "step": 76
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.54e-05,
      "loss": 1.0145,
      "step": 77
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.56e-05,
      "loss": 0.9558,
      "step": 78
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.58e-05,
      "loss": 0.9009,
      "step": 79
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 1.187,
      "step": 80
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.62e-05,
      "loss": 1.4067,
      "step": 81
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6400000000000002e-05,
      "loss": 1.0384,
      "step": 82
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.66e-05,
      "loss": 1.0523,
      "step": 83
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6800000000000002e-05,
      "loss": 0.8132,
      "step": 84
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 0.4641,
      "step": 85
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7199999999999998e-05,
      "loss": 1.427,
      "step": 86
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.74e-05,
      "loss": 0.8204,
      "step": 87
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.76e-05,
      "loss": 0.8096,
      "step": 88
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.78e-05,
      "loss": 1.4244,
      "step": 89
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8e-05,
      "loss": 1.0537,
      "step": 90
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8200000000000002e-05,
      "loss": 0.9706,
      "step": 91
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.84e-05,
      "loss": 1.0231,
      "step": 92
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.86e-05,
      "loss": 1.4637,
      "step": 93
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.88e-05,
      "loss": 1.2839,
      "step": 94
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9e-05,
      "loss": 1.0809,
      "step": 95
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9200000000000003e-05,
      "loss": 0.6215,
      "step": 96
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.94e-05,
      "loss": 1.0836,
      "step": 97
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9600000000000002e-05,
      "loss": 1.3463,
      "step": 98
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9800000000000004e-05,
      "loss": 1.5543,
      "step": 99
    },
    {
      "epoch": 0.01,
      "learning_rate": 2e-05,
      "loss": 0.9702,
      "step": 100
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.0200000000000003e-05,
      "loss": 0.7183,
      "step": 101
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.04e-05,
      "loss": 1.9201,
      "step": 102
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.06e-05,
      "loss": 1.7262,
      "step": 103
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.08e-05,
      "loss": 0.8847,
      "step": 104
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.1e-05,
      "loss": 1.2476,
      "step": 105
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.12e-05,
      "loss": 1.3319,
      "step": 106
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.1400000000000002e-05,
      "loss": 1.5576,
      "step": 107
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.16e-05,
      "loss": 1.1616,
      "step": 108
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.18e-05,
      "loss": 0.7254,
      "step": 109
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 1.4791,
      "step": 110
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.22e-05,
      "loss": 0.9266,
      "step": 111
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2400000000000002e-05,
      "loss": 1.4224,
      "step": 112
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.26e-05,
      "loss": 1.2939,
      "step": 113
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2800000000000002e-05,
      "loss": 0.9885,
      "step": 114
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 1.1177,
      "step": 115
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.32e-05,
      "loss": 0.9837,
      "step": 116
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.3400000000000003e-05,
      "loss": 0.9161,
      "step": 117
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.36e-05,
      "loss": 1.1841,
      "step": 118
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.38e-05,
      "loss": 1.1398,
      "step": 119
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4e-05,
      "loss": 1.4146,
      "step": 120
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4200000000000002e-05,
      "loss": 1.1263,
      "step": 121
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.44e-05,
      "loss": 0.6171,
      "step": 122
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.46e-05,
      "loss": 1.3551,
      "step": 123
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.48e-05,
      "loss": 0.7872,
      "step": 124
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.5e-05,
      "loss": 1.3004,
      "step": 125
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.5200000000000003e-05,
      "loss": 1.6624,
      "step": 126
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.54e-05,
      "loss": 0.7562,
      "step": 127
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.5600000000000002e-05,
      "loss": 0.7161,
      "step": 128
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.58e-05,
      "loss": 0.9615,
      "step": 129
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 0.9556,
      "step": 130
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6200000000000003e-05,
      "loss": 0.6519,
      "step": 131
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.64e-05,
      "loss": 0.4542,
      "step": 132
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6600000000000003e-05,
      "loss": 1.1401,
      "step": 133
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6800000000000004e-05,
      "loss": 1.4224,
      "step": 134
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7000000000000002e-05,
      "loss": 1.3867,
      "step": 135
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7200000000000004e-05,
      "loss": 0.8578,
      "step": 136
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7400000000000002e-05,
      "loss": 1.3377,
      "step": 137
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7600000000000003e-05,
      "loss": 0.9124,
      "step": 138
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7800000000000005e-05,
      "loss": 0.9877,
      "step": 139
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 1.4228,
      "step": 140
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8199999999999998e-05,
      "loss": 1.0643,
      "step": 141
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.84e-05,
      "loss": 1.3295,
      "step": 142
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.86e-05,
      "loss": 1.1489,
      "step": 143
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.88e-05,
      "loss": 1.0845,
      "step": 144
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.9e-05,
      "loss": 0.3899,
      "step": 145
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.9199999999999998e-05,
      "loss": 0.5314,
      "step": 146
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.94e-05,
      "loss": 1.0999,
      "step": 147
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.96e-05,
      "loss": 1.0364,
      "step": 148
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.98e-05,
      "loss": 1.0974,
      "step": 149
    },
    {
      "epoch": 0.02,
      "learning_rate": 3e-05,
      "loss": 1.0941,
      "step": 150
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.02e-05,
      "loss": 1.4155,
      "step": 151
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.04e-05,
      "loss": 1.2109,
      "step": 152
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.06e-05,
      "loss": 1.1806,
      "step": 153
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.08e-05,
      "loss": 0.9307,
      "step": 154
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.1e-05,
      "loss": 0.7737,
      "step": 155
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.12e-05,
      "loss": 1.0836,
      "step": 156
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.1400000000000004e-05,
      "loss": 0.5999,
      "step": 157
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.16e-05,
      "loss": 1.0072,
      "step": 158
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.18e-05,
      "loss": 1.2559,
      "step": 159
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 1.4475,
      "step": 160
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.2200000000000003e-05,
      "loss": 1.0085,
      "step": 161
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.24e-05,
      "loss": 1.3529,
      "step": 162
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.26e-05,
      "loss": 0.4488,
      "step": 163
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.2800000000000004e-05,
      "loss": 0.9551,
      "step": 164
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.3e-05,
      "loss": 1.5824,
      "step": 165
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.32e-05,
      "loss": 1.0307,
      "step": 166
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.3400000000000005e-05,
      "loss": 0.7178,
      "step": 167
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.3600000000000004e-05,
      "loss": 1.337,
      "step": 168
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.38e-05,
      "loss": 1.2228,
      "step": 169
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 1.1614,
      "step": 170
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.4200000000000005e-05,
      "loss": 1.2836,
      "step": 171
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.4399999999999996e-05,
      "loss": 1.2975,
      "step": 172
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.46e-05,
      "loss": 1.147,
      "step": 173
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.48e-05,
      "loss": 1.0146,
      "step": 174
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.5e-05,
      "loss": 1.1378,
      "step": 175
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.52e-05,
      "loss": 1.1565,
      "step": 176
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.54e-05,
      "loss": 0.6013,
      "step": 177
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.56e-05,
      "loss": 1.2615,
      "step": 178
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.58e-05,
      "loss": 1.2798,
      "step": 179
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.6e-05,
      "loss": 0.9095,
      "step": 180
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.62e-05,
      "loss": 1.3384,
      "step": 181
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.6400000000000004e-05,
      "loss": 1.1076,
      "step": 182
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.66e-05,
      "loss": 1.3861,
      "step": 183
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.68e-05,
      "loss": 1.3285,
      "step": 184
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.7e-05,
      "loss": 0.4222,
      "step": 185
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.72e-05,
      "loss": 1.1018,
      "step": 186
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.74e-05,
      "loss": 1.1473,
      "step": 187
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.76e-05,
      "loss": 1.1341,
      "step": 188
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.7800000000000004e-05,
      "loss": 1.3934,
      "step": 189
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8e-05,
      "loss": 1.5238,
      "step": 190
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.82e-05,
      "loss": 1.664,
      "step": 191
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8400000000000005e-05,
      "loss": 1.2187,
      "step": 192
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.86e-05,
      "loss": 1.2789,
      "step": 193
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.88e-05,
      "loss": 0.9779,
      "step": 194
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.9000000000000006e-05,
      "loss": 1.3136,
      "step": 195
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.9200000000000004e-05,
      "loss": 0.9833,
      "step": 196
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.94e-05,
      "loss": 0.5529,
      "step": 197
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.960000000000001e-05,
      "loss": 0.7882,
      "step": 198
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.9800000000000005e-05,
      "loss": 1.2528,
      "step": 199
    },
    {
      "epoch": 0.02,
      "learning_rate": 4e-05,
      "loss": 0.8247,
      "step": 200
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.02e-05,
      "loss": 1.5192,
      "step": 201
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.0400000000000006e-05,
      "loss": 1.2239,
      "step": 202
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.0600000000000004e-05,
      "loss": 1.3499,
      "step": 203
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.08e-05,
      "loss": 1.3491,
      "step": 204
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.1e-05,
      "loss": 0.7321,
      "step": 205
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.12e-05,
      "loss": 1.2544,
      "step": 206
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.14e-05,
      "loss": 0.8021,
      "step": 207
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.16e-05,
      "loss": 0.7792,
      "step": 208
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.18e-05,
      "loss": 0.8939,
      "step": 209
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.2e-05,
      "loss": 1.2585,
      "step": 210
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.22e-05,
      "loss": 0.5714,
      "step": 211
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.24e-05,
      "loss": 1.5001,
      "step": 212
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.26e-05,
      "loss": 1.1382,
      "step": 213
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.2800000000000004e-05,
      "loss": 1.0615,
      "step": 214
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.3e-05,
      "loss": 1.0139,
      "step": 215
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.32e-05,
      "loss": 1.3761,
      "step": 216
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.3400000000000005e-05,
      "loss": 1.0633,
      "step": 217
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.36e-05,
      "loss": 1.3294,
      "step": 218
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.38e-05,
      "loss": 1.3122,
      "step": 219
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 0.4348,
      "step": 220
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.4200000000000004e-05,
      "loss": 1.1262,
      "step": 221
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.44e-05,
      "loss": 1.197,
      "step": 222
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.46e-05,
      "loss": 0.8763,
      "step": 223
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.4800000000000005e-05,
      "loss": 0.8927,
      "step": 224
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.5e-05,
      "loss": 0.5612,
      "step": 225
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.52e-05,
      "loss": 1.5945,
      "step": 226
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.5400000000000006e-05,
      "loss": 0.6654,
      "step": 227
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.5600000000000004e-05,
      "loss": 0.8835,
      "step": 228
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.58e-05,
      "loss": 1.1516,
      "step": 229
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.600000000000001e-05,
      "loss": 0.947,
      "step": 230
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.6200000000000005e-05,
      "loss": 1.1408,
      "step": 231
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.64e-05,
      "loss": 1.1119,
      "step": 232
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.660000000000001e-05,
      "loss": 0.7469,
      "step": 233
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.6800000000000006e-05,
      "loss": 0.9026,
      "step": 234
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.7e-05,
      "loss": 1.5774,
      "step": 235
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.72e-05,
      "loss": 1.2098,
      "step": 236
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.74e-05,
      "loss": 0.6599,
      "step": 237
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.76e-05,
      "loss": 0.5967,
      "step": 238
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.78e-05,
      "loss": 1.3286,
      "step": 239
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.8e-05,
      "loss": 0.7375,
      "step": 240
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.82e-05,
      "loss": 1.577,
      "step": 241
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.8400000000000004e-05,
      "loss": 1.201,
      "step": 242
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.86e-05,
      "loss": 1.0599,
      "step": 243
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.88e-05,
      "loss": 0.9123,
      "step": 244
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9e-05,
      "loss": 1.3004,
      "step": 245
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.92e-05,
      "loss": 0.8621,
      "step": 246
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.94e-05,
      "loss": 0.686,
      "step": 247
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.96e-05,
      "loss": 1.2314,
      "step": 248
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9800000000000004e-05,
      "loss": 1.2022,
      "step": 249
    },
    {
      "epoch": 0.03,
      "learning_rate": 5e-05,
      "loss": 1.3039,
      "step": 250
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.02e-05,
      "loss": 1.0916,
      "step": 251
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.0400000000000005e-05,
      "loss": 1.0735,
      "step": 252
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.0600000000000003e-05,
      "loss": 1.1206,
      "step": 253
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.08e-05,
      "loss": 0.74,
      "step": 254
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.1000000000000006e-05,
      "loss": 0.8031,
      "step": 255
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.1200000000000004e-05,
      "loss": 0.9419,
      "step": 256
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.14e-05,
      "loss": 1.1662,
      "step": 257
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.16e-05,
      "loss": 1.2288,
      "step": 258
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.1800000000000005e-05,
      "loss": 1.2055,
      "step": 259
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.2000000000000004e-05,
      "loss": 0.7164,
      "step": 260
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.22e-05,
      "loss": 0.7293,
      "step": 261
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.2400000000000007e-05,
      "loss": 0.4783,
      "step": 262
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.2600000000000005e-05,
      "loss": 0.835,
      "step": 263
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.28e-05,
      "loss": 1.1513,
      "step": 264
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.300000000000001e-05,
      "loss": 1.0213,
      "step": 265
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.3200000000000006e-05,
      "loss": 1.5211,
      "step": 266
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.3400000000000004e-05,
      "loss": 1.2385,
      "step": 267
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.360000000000001e-05,
      "loss": 0.6553,
      "step": 268
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.380000000000001e-05,
      "loss": 0.7151,
      "step": 269
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.4000000000000005e-05,
      "loss": 1.2801,
      "step": 270
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.420000000000001e-05,
      "loss": 1.6315,
      "step": 271
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.440000000000001e-05,
      "loss": 0.7671,
      "step": 272
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.4600000000000006e-05,
      "loss": 0.8112,
      "step": 273
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.4800000000000004e-05,
      "loss": 1.3151,
      "step": 274
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.500000000000001e-05,
      "loss": 1.125,
      "step": 275
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.520000000000001e-05,
      "loss": 0.5224,
      "step": 276
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.5400000000000005e-05,
      "loss": 1.321,
      "step": 277
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.560000000000001e-05,
      "loss": 0.7001,
      "step": 278
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.580000000000001e-05,
      "loss": 0.5794,
      "step": 279
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.6000000000000006e-05,
      "loss": 0.6504,
      "step": 280
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.620000000000001e-05,
      "loss": 1.1106,
      "step": 281
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.6399999999999995e-05,
      "loss": 1.5404,
      "step": 282
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.66e-05,
      "loss": 1.3011,
      "step": 283
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.68e-05,
      "loss": 0.6942,
      "step": 284
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.6999999999999996e-05,
      "loss": 0.887,
      "step": 285
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.72e-05,
      "loss": 1.5039,
      "step": 286
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.74e-05,
      "loss": 1.3413,
      "step": 287
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.76e-05,
      "loss": 1.2419,
      "step": 288
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.7799999999999995e-05,
      "loss": 1.1649,
      "step": 289
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.8e-05,
      "loss": 1.2427,
      "step": 290
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.82e-05,
      "loss": 1.6185,
      "step": 291
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.8399999999999997e-05,
      "loss": 0.7895,
      "step": 292
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.86e-05,
      "loss": 0.6541,
      "step": 293
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.88e-05,
      "loss": 1.4914,
      "step": 294
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.9e-05,
      "loss": 1.5453,
      "step": 295
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.92e-05,
      "loss": 1.5056,
      "step": 296
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.94e-05,
      "loss": 1.3657,
      "step": 297
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.96e-05,
      "loss": 1.5399,
      "step": 298
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.9800000000000003e-05,
      "loss": 1.0309,
      "step": 299
    },
    {
      "epoch": 0.03,
      "learning_rate": 6e-05,
      "loss": 1.9957,
      "step": 300
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.02e-05,
      "loss": 0.667,
      "step": 301
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.04e-05,
      "loss": 1.2875,
      "step": 302
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.06e-05,
      "loss": 1.8473,
      "step": 303
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.08e-05,
      "loss": 1.2755,
      "step": 304
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.1e-05,
      "loss": 0.6024,
      "step": 305
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.12e-05,
      "loss": 1.6293,
      "step": 306
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.14e-05,
      "loss": 0.8741,
      "step": 307
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.16e-05,
      "loss": 0.7479,
      "step": 308
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.18e-05,
      "loss": 0.8534,
      "step": 309
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.2e-05,
      "loss": 0.7573,
      "step": 310
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.220000000000001e-05,
      "loss": 0.7727,
      "step": 311
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.24e-05,
      "loss": 1.2795,
      "step": 312
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.26e-05,
      "loss": 1.3966,
      "step": 313
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.280000000000001e-05,
      "loss": 1.2597,
      "step": 314
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.3e-05,
      "loss": 1.2662,
      "step": 315
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.32e-05,
      "loss": 1.2394,
      "step": 316
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.340000000000001e-05,
      "loss": 0.7125,
      "step": 317
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.36e-05,
      "loss": 1.2991,
      "step": 318
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.38e-05,
      "loss": 1.0087,
      "step": 319
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.400000000000001e-05,
      "loss": 1.2316,
      "step": 320
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.42e-05,
      "loss": 1.2705,
      "step": 321
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.440000000000001e-05,
      "loss": 1.4093,
      "step": 322
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.460000000000001e-05,
      "loss": 0.9789,
      "step": 323
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.48e-05,
      "loss": 1.1718,
      "step": 324
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.500000000000001e-05,
      "loss": 1.454,
      "step": 325
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.52e-05,
      "loss": 0.9867,
      "step": 326
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.54e-05,
      "loss": 0.7592,
      "step": 327
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.560000000000001e-05,
      "loss": 1.2597,
      "step": 328
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.58e-05,
      "loss": 0.7866,
      "step": 329
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.6e-05,
      "loss": 1.118,
      "step": 330
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.620000000000001e-05,
      "loss": 1.0312,
      "step": 331
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.64e-05,
      "loss": 1.4859,
      "step": 332
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.66e-05,
      "loss": 1.0265,
      "step": 333
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.680000000000001e-05,
      "loss": 1.0986,
      "step": 334
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.7e-05,
      "loss": 1.437,
      "step": 335
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.720000000000001e-05,
      "loss": 1.2803,
      "step": 336
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.740000000000001e-05,
      "loss": 1.1289,
      "step": 337
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.76e-05,
      "loss": 1.4711,
      "step": 338
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.780000000000001e-05,
      "loss": 1.0132,
      "step": 339
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.800000000000001e-05,
      "loss": 1.1157,
      "step": 340
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.82e-05,
      "loss": 1.0209,
      "step": 341
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.840000000000001e-05,
      "loss": 0.7813,
      "step": 342
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.860000000000001e-05,
      "loss": 1.4696,
      "step": 343
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.879999999999999e-05,
      "loss": 1.3438,
      "step": 344
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.9e-05,
      "loss": 1.1524,
      "step": 345
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.92e-05,
      "loss": 1.5856,
      "step": 346
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.939999999999999e-05,
      "loss": 0.5561,
      "step": 347
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.96e-05,
      "loss": 1.1503,
      "step": 348
    },
    {
      "epoch": 0.04,
      "learning_rate": 6.98e-05,
      "loss": 0.5888,
      "step": 349
    },
    {
      "epoch": 0.04,
      "learning_rate": 7e-05,
      "loss": 0.7322,
      "step": 350
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.02e-05,
      "loss": 0.977,
      "step": 351
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.04e-05,
      "loss": 0.9464,
      "step": 352
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.06e-05,
      "loss": 1.0909,
      "step": 353
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.08e-05,
      "loss": 1.1843,
      "step": 354
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.1e-05,
      "loss": 0.5669,
      "step": 355
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.12e-05,
      "loss": 1.2487,
      "step": 356
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.14e-05,
      "loss": 1.2812,
      "step": 357
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.16e-05,
      "loss": 1.3443,
      "step": 358
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.18e-05,
      "loss": 0.8615,
      "step": 359
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.2e-05,
      "loss": 1.1377,
      "step": 360
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.22e-05,
      "loss": 1.3459,
      "step": 361
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.24e-05,
      "loss": 0.6404,
      "step": 362
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.26e-05,
      "loss": 1.0849,
      "step": 363
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.280000000000001e-05,
      "loss": 0.9881,
      "step": 364
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.3e-05,
      "loss": 1.1572,
      "step": 365
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.32e-05,
      "loss": 0.5688,
      "step": 366
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.340000000000001e-05,
      "loss": 0.7055,
      "step": 367
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.36e-05,
      "loss": 1.1941,
      "step": 368
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.38e-05,
      "loss": 1.0964,
      "step": 369
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.4e-05,
      "loss": 0.7178,
      "step": 370
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.42e-05,
      "loss": 1.2805,
      "step": 371
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.44e-05,
      "loss": 1.2366,
      "step": 372
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.46e-05,
      "loss": 1.0908,
      "step": 373
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.48e-05,
      "loss": 0.7966,
      "step": 374
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.500000000000001e-05,
      "loss": 1.5489,
      "step": 375
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.52e-05,
      "loss": 0.9741,
      "step": 376
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.54e-05,
      "loss": 0.5604,
      "step": 377
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.560000000000001e-05,
      "loss": 0.6851,
      "step": 378
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.58e-05,
      "loss": 1.0263,
      "step": 379
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.6e-05,
      "loss": 1.5648,
      "step": 380
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.620000000000001e-05,
      "loss": 0.632,
      "step": 381
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.64e-05,
      "loss": 0.9892,
      "step": 382
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.66e-05,
      "loss": 1.1832,
      "step": 383
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.680000000000001e-05,
      "loss": 1.0887,
      "step": 384
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.7e-05,
      "loss": 0.6765,
      "step": 385
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.72e-05,
      "loss": 1.1768,
      "step": 386
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.740000000000001e-05,
      "loss": 0.9968,
      "step": 387
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.76e-05,
      "loss": 1.1145,
      "step": 388
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.780000000000001e-05,
      "loss": 1.3173,
      "step": 389
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.800000000000001e-05,
      "loss": 1.4605,
      "step": 390
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.82e-05,
      "loss": 1.3185,
      "step": 391
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.840000000000001e-05,
      "loss": 1.5281,
      "step": 392
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.860000000000001e-05,
      "loss": 1.3191,
      "step": 393
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.88e-05,
      "loss": 0.9964,
      "step": 394
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.900000000000001e-05,
      "loss": 1.0821,
      "step": 395
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.920000000000001e-05,
      "loss": 1.1694,
      "step": 396
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.94e-05,
      "loss": 1.2506,
      "step": 397
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.960000000000001e-05,
      "loss": 1.0815,
      "step": 398
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.98e-05,
      "loss": 1.1066,
      "step": 399
    },
    {
      "epoch": 0.04,
      "learning_rate": 8e-05,
      "loss": 1.2227,
      "step": 400
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.020000000000001e-05,
      "loss": 1.3202,
      "step": 401
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.04e-05,
      "loss": 1.0453,
      "step": 402
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.060000000000001e-05,
      "loss": 0.7302,
      "step": 403
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.080000000000001e-05,
      "loss": 0.6449,
      "step": 404
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.1e-05,
      "loss": 1.229,
      "step": 405
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.120000000000001e-05,
      "loss": 0.8553,
      "step": 406
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.14e-05,
      "loss": 1.1301,
      "step": 407
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.16e-05,
      "loss": 1.1121,
      "step": 408
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.18e-05,
      "loss": 0.8195,
      "step": 409
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.2e-05,
      "loss": 0.7269,
      "step": 410
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.22e-05,
      "loss": 1.5477,
      "step": 411
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.24e-05,
      "loss": 1.2745,
      "step": 412
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.26e-05,
      "loss": 0.804,
      "step": 413
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.28e-05,
      "loss": 1.0162,
      "step": 414
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.3e-05,
      "loss": 1.2422,
      "step": 415
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.32e-05,
      "loss": 0.8496,
      "step": 416
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.34e-05,
      "loss": 1.3877,
      "step": 417
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.36e-05,
      "loss": 0.9494,
      "step": 418
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.38e-05,
      "loss": 0.9406,
      "step": 419
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.4e-05,
      "loss": 1.2808,
      "step": 420
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.42e-05,
      "loss": 1.0722,
      "step": 421
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.44e-05,
      "loss": 1.0401,
      "step": 422
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.46e-05,
      "loss": 1.6585,
      "step": 423
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.48e-05,
      "loss": 1.0063,
      "step": 424
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.5e-05,
      "loss": 1.2152,
      "step": 425
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.52e-05,
      "loss": 0.5433,
      "step": 426
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.54e-05,
      "loss": 1.063,
      "step": 427
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.560000000000001e-05,
      "loss": 0.7376,
      "step": 428
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.58e-05,
      "loss": 0.8316,
      "step": 429
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.6e-05,
      "loss": 1.3727,
      "step": 430
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.620000000000001e-05,
      "loss": 1.1344,
      "step": 431
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.64e-05,
      "loss": 1.2763,
      "step": 432
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.66e-05,
      "loss": 0.6324,
      "step": 433
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.680000000000001e-05,
      "loss": 1.0709,
      "step": 434
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.7e-05,
      "loss": 0.4515,
      "step": 435
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.72e-05,
      "loss": 1.0522,
      "step": 436
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.740000000000001e-05,
      "loss": 0.939,
      "step": 437
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.76e-05,
      "loss": 1.0412,
      "step": 438
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.78e-05,
      "loss": 1.0124,
      "step": 439
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.800000000000001e-05,
      "loss": 1.5215,
      "step": 440
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.82e-05,
      "loss": 0.7356,
      "step": 441
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.840000000000001e-05,
      "loss": 1.0388,
      "step": 442
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.86e-05,
      "loss": 0.8027,
      "step": 443
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.88e-05,
      "loss": 0.8105,
      "step": 444
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.900000000000001e-05,
      "loss": 1.4067,
      "step": 445
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.92e-05,
      "loss": 0.9554,
      "step": 446
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.94e-05,
      "loss": 1.3724,
      "step": 447
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.960000000000001e-05,
      "loss": 1.2238,
      "step": 448
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.98e-05,
      "loss": 0.4791,
      "step": 449
    },
    {
      "epoch": 0.05,
      "learning_rate": 9e-05,
      "loss": 0.8392,
      "step": 450
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.020000000000001e-05,
      "loss": 1.0729,
      "step": 451
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.04e-05,
      "loss": 1.2355,
      "step": 452
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.06e-05,
      "loss": 1.2601,
      "step": 453
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.080000000000001e-05,
      "loss": 0.9389,
      "step": 454
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.1e-05,
      "loss": 0.6139,
      "step": 455
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.120000000000001e-05,
      "loss": 1.6682,
      "step": 456
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.140000000000001e-05,
      "loss": 0.6317,
      "step": 457
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.16e-05,
      "loss": 0.6849,
      "step": 458
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.180000000000001e-05,
      "loss": 0.7398,
      "step": 459
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.200000000000001e-05,
      "loss": 1.1224,
      "step": 460
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.22e-05,
      "loss": 0.5923,
      "step": 461
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.240000000000001e-05,
      "loss": 1.1985,
      "step": 462
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.260000000000001e-05,
      "loss": 1.2804,
      "step": 463
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.28e-05,
      "loss": 1.0047,
      "step": 464
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.300000000000001e-05,
      "loss": 0.587,
      "step": 465
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.320000000000002e-05,
      "loss": 1.1529,
      "step": 466
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.340000000000001e-05,
      "loss": 1.3616,
      "step": 467
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.360000000000001e-05,
      "loss": 1.4938,
      "step": 468
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.38e-05,
      "loss": 1.0333,
      "step": 469
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.4e-05,
      "loss": 1.286,
      "step": 470
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.42e-05,
      "loss": 1.1995,
      "step": 471
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.44e-05,
      "loss": 0.5875,
      "step": 472
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.46e-05,
      "loss": 0.8804,
      "step": 473
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.48e-05,
      "loss": 0.7862,
      "step": 474
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.5e-05,
      "loss": 1.03,
      "step": 475
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.52e-05,
      "loss": 0.6998,
      "step": 476
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.54e-05,
      "loss": 0.9633,
      "step": 477
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.56e-05,
      "loss": 1.3805,
      "step": 478
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.58e-05,
      "loss": 1.3668,
      "step": 479
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.6e-05,
      "loss": 1.6565,
      "step": 480
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.620000000000001e-05,
      "loss": 1.4485,
      "step": 481
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.64e-05,
      "loss": 1.0282,
      "step": 482
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.66e-05,
      "loss": 0.53,
      "step": 483
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.680000000000001e-05,
      "loss": 0.9214,
      "step": 484
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.7e-05,
      "loss": 1.4821,
      "step": 485
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.72e-05,
      "loss": 1.1442,
      "step": 486
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.74e-05,
      "loss": 0.596,
      "step": 487
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.76e-05,
      "loss": 1.6132,
      "step": 488
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.78e-05,
      "loss": 1.0643,
      "step": 489
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.8e-05,
      "loss": 1.2359,
      "step": 490
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.82e-05,
      "loss": 0.7979,
      "step": 491
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.84e-05,
      "loss": 1.5329,
      "step": 492
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.86e-05,
      "loss": 0.7858,
      "step": 493
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.88e-05,
      "loss": 1.1644,
      "step": 494
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.900000000000001e-05,
      "loss": 0.8736,
      "step": 495
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.92e-05,
      "loss": 0.8781,
      "step": 496
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.94e-05,
      "loss": 1.2642,
      "step": 497
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.960000000000001e-05,
      "loss": 0.9982,
      "step": 498
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.98e-05,
      "loss": 0.617,
      "step": 499
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001,
      "loss": 0.7142,
      "step": 500
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999999696597408e-05,
      "loss": 0.7028,
      "step": 501
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999998786389666e-05,
      "loss": 0.7734,
      "step": 502
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999997269376886e-05,
      "loss": 1.0463,
      "step": 503
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999995145559254e-05,
      "loss": 0.4696,
      "step": 504
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999992414937023e-05,
      "loss": 1.0795,
      "step": 505
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999989077510529e-05,
      "loss": 0.5835,
      "step": 506
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999985133280174e-05,
      "loss": 0.8951,
      "step": 507
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999980582246438e-05,
      "loss": 0.7244,
      "step": 508
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999975424409873e-05,
      "loss": 1.3119,
      "step": 509
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999969659771106e-05,
      "loss": 0.7291,
      "step": 510
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999963288330834e-05,
      "loss": 1.0204,
      "step": 511
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999956310089834e-05,
      "loss": 1.5515,
      "step": 512
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999948725048948e-05,
      "loss": 0.5514,
      "step": 513
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999940533209103e-05,
      "loss": 0.9635,
      "step": 514
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999931734571286e-05,
      "loss": 1.0969,
      "step": 515
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.99992232913657e-05,
      "loss": 1.1407,
      "step": 516
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999912316906094e-05,
      "loss": 1.4571,
      "step": 517
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999901697881076e-05,
      "loss": 1.3362,
      "step": 518
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.9998904720628e-05,
      "loss": 0.5591,
      "step": 519
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999878639452632e-05,
      "loss": 1.0973,
      "step": 520
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999866200052008e-05,
      "loss": 1.1315,
      "step": 521
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.999853153862435e-05,
      "loss": 1.5116,
      "step": 522
    },
    {
      "epoch": 0.05,
      "learning_rate": 9.9998395008855e-05,
      "loss": 1.3071,
      "step": 523
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999825241122856e-05,
      "loss": 0.6677,
      "step": 524
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999810374576235e-05,
      "loss": 0.7311,
      "step": 525
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999794901247444e-05,
      "loss": 1.0712,
      "step": 526
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999778821138357e-05,
      "loss": 0.5546,
      "step": 527
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999762134250926e-05,
      "loss": 0.7659,
      "step": 528
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99974484058718e-05,
      "loss": 1.0204,
      "step": 529
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999726940149212e-05,
      "loss": 1.2139,
      "step": 530
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999708432939199e-05,
      "loss": 0.8289,
      "step": 531
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999689318959386e-05,
      "loss": 1.4426,
      "step": 532
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999669598212092e-05,
      "loss": 0.5919,
      "step": 533
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999649270699709e-05,
      "loss": 0.5368,
      "step": 534
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999628336424706e-05,
      "loss": 1.3905,
      "step": 535
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999606795389622e-05,
      "loss": 1.3093,
      "step": 536
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999584647597075e-05,
      "loss": 1.3036,
      "step": 537
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999561893049748e-05,
      "loss": 1.2168,
      "step": 538
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999538531750405e-05,
      "loss": 1.3502,
      "step": 539
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999514563701882e-05,
      "loss": 1.2013,
      "step": 540
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999489988907084e-05,
      "loss": 1.1542,
      "step": 541
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999464807368999e-05,
      "loss": 1.0601,
      "step": 542
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99943901909068e-05,
      "loss": 1.2476,
      "step": 543
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999412624075254e-05,
      "loss": 1.0611,
      "step": 544
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99938562232593e-05,
      "loss": 1.3745,
      "step": 545
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99935801384598e-05,
      "loss": 0.7918,
      "step": 546
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999329798638756e-05,
      "loss": 0.6255,
      "step": 547
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999300976707687e-05,
      "loss": 1.5143,
      "step": 548
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999271548056263e-05,
      "loss": 0.6207,
      "step": 549
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99924151268806e-05,
      "loss": 1.348,
      "step": 550
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999210870606723e-05,
      "loss": 0.9408,
      "step": 551
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99917962181597e-05,
      "loss": 0.9255,
      "step": 552
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999147766319592e-05,
      "loss": 0.8947,
      "step": 553
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999115304121457e-05,
      "loss": 0.8175,
      "step": 554
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999082235225507e-05,
      "loss": 0.8416,
      "step": 555
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99904855963575e-05,
      "loss": 1.1571,
      "step": 556
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.999014277356276e-05,
      "loss": 1.2331,
      "step": 557
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998979388391245e-05,
      "loss": 1.4191,
      "step": 558
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998943892744889e-05,
      "loss": 1.3521,
      "step": 559
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998907790421522e-05,
      "loss": 1.572,
      "step": 560
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998871081425517e-05,
      "loss": 1.5227,
      "step": 561
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998833765761336e-05,
      "loss": 1.6336,
      "step": 562
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998795843433503e-05,
      "loss": 0.6834,
      "step": 563
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998757314446623e-05,
      "loss": 0.9479,
      "step": 564
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99871817880537e-05,
      "loss": 0.7884,
      "step": 565
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998678436514497e-05,
      "loss": 1.2059,
      "step": 566
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998638087578824e-05,
      "loss": 0.6003,
      "step": 567
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998597132003249e-05,
      "loss": 0.5168,
      "step": 568
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998555569792741e-05,
      "loss": 1.2163,
      "step": 569
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998513400952346e-05,
      "loss": 0.9292,
      "step": 570
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99847062548718e-05,
      "loss": 0.8582,
      "step": 571
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998427243402437e-05,
      "loss": 0.9034,
      "step": 572
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998383254703378e-05,
      "loss": 1.4545,
      "step": 573
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998338659395343e-05,
      "loss": 1.3954,
      "step": 574
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998293457483745e-05,
      "loss": 1.4965,
      "step": 575
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998247648974071e-05,
      "loss": 0.9723,
      "step": 576
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998201233871877e-05,
      "loss": 0.8398,
      "step": 577
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998154212182797e-05,
      "loss": 1.3046,
      "step": 578
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998106583912542e-05,
      "loss": 1.0802,
      "step": 579
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998058349066884e-05,
      "loss": 1.0747,
      "step": 580
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.998009507651684e-05,
      "loss": 1.3065,
      "step": 581
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997960059672865e-05,
      "loss": 1.365,
      "step": 582
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99791000513643e-05,
      "loss": 1.5488,
      "step": 583
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997859344048455e-05,
      "loss": 0.7475,
      "step": 584
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997808076415087e-05,
      "loss": 0.6491,
      "step": 585
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997756202242545e-05,
      "loss": 0.5388,
      "step": 586
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997703721537131e-05,
      "loss": 1.3344,
      "step": 587
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997650634305207e-05,
      "loss": 1.3829,
      "step": 588
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99759694055322e-05,
      "loss": 0.6821,
      "step": 589
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997542640287687e-05,
      "loss": 1.3633,
      "step": 590
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997487733515194e-05,
      "loss": 1.053,
      "step": 591
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997432220242406e-05,
      "loss": 1.1946,
      "step": 592
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997376100476063e-05,
      "loss": 1.2636,
      "step": 593
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997319374222974e-05,
      "loss": 0.3606,
      "step": 594
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997262041490022e-05,
      "loss": 0.6299,
      "step": 595
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997204102284167e-05,
      "loss": 1.1486,
      "step": 596
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99714555661244e-05,
      "loss": 0.6155,
      "step": 597
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997086404481944e-05,
      "loss": 0.7203,
      "step": 598
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.997026645899859e-05,
      "loss": 0.5927,
      "step": 599
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99696628087344e-05,
      "loss": 1.1108,
      "step": 600
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99690530941001e-05,
      "loss": 1.1673,
      "step": 601
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996843731516969e-05,
      "loss": 1.3048,
      "step": 602
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996781547201793e-05,
      "loss": 1.1153,
      "step": 603
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996718756472024e-05,
      "loss": 1.3135,
      "step": 604
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996655359335282e-05,
      "loss": 1.2304,
      "step": 605
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996591355799269e-05,
      "loss": 0.6948,
      "step": 606
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996526745871744e-05,
      "loss": 1.5169,
      "step": 607
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996461529560553e-05,
      "loss": 1.1867,
      "step": 608
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996395706873608e-05,
      "loss": 0.6065,
      "step": 609
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996329277818897e-05,
      "loss": 1.258,
      "step": 610
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996262242404484e-05,
      "loss": 1.3904,
      "step": 611
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996194600638504e-05,
      "loss": 0.5293,
      "step": 612
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996126352529168e-05,
      "loss": 1.4051,
      "step": 613
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.996057498084753e-05,
      "loss": 0.7495,
      "step": 614
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99598803731362e-05,
      "loss": 1.5208,
      "step": 615
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.995917970224197e-05,
      "loss": 0.803,
      "step": 616
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.99584729682499e-05,
      "loss": 1.5495,
      "step": 617
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.995776017124571e-05,
      "loss": 0.9484,
      "step": 618
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995704131131596e-05,
      "loss": 0.7077,
      "step": 619
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995631638854785e-05,
      "loss": 1.3109,
      "step": 620
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995558540302938e-05,
      "loss": 1.1369,
      "step": 621
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995484835484926e-05,
      "loss": 1.4499,
      "step": 622
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995410524409692e-05,
      "loss": 1.2865,
      "step": 623
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995335607086258e-05,
      "loss": 1.2899,
      "step": 624
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995260083523715e-05,
      "loss": 1.1732,
      "step": 625
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995183953731225e-05,
      "loss": 1.0758,
      "step": 626
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995107217718032e-05,
      "loss": 0.6523,
      "step": 627
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.995029875493445e-05,
      "loss": 1.3869,
      "step": 628
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994951927066853e-05,
      "loss": 0.7453,
      "step": 629
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994873372447716e-05,
      "loss": 0.5927,
      "step": 630
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994794211645563e-05,
      "loss": 1.215,
      "step": 631
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994714444670007e-05,
      "loss": 1.572,
      "step": 632
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994634071530725e-05,
      "loss": 1.2814,
      "step": 633
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994553092237473e-05,
      "loss": 0.596,
      "step": 634
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994471506800079e-05,
      "loss": 1.4704,
      "step": 635
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99438931522844e-05,
      "loss": 1.5697,
      "step": 636
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994306517532537e-05,
      "loss": 1.235,
      "step": 637
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994223113722415e-05,
      "loss": 0.7273,
      "step": 638
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994139103808196e-05,
      "loss": 0.7634,
      "step": 639
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.994054487800075e-05,
      "loss": 1.128,
      "step": 640
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993969265708323e-05,
      "loss": 1.1782,
      "step": 641
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99388343754328e-05,
      "loss": 1.4394,
      "step": 642
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993797003315366e-05,
      "loss": 1.3981,
      "step": 643
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99370996303507e-05,
      "loss": 1.2736,
      "step": 644
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993622316712951e-05,
      "loss": 1.4892,
      "step": 645
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993534064359651e-05,
      "loss": 1.5376,
      "step": 646
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993445205985877e-05,
      "loss": 1.6445,
      "step": 647
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993355741602414e-05,
      "loss": 1.3461,
      "step": 648
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99326567122012e-05,
      "loss": 1.1006,
      "step": 649
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993174994849926e-05,
      "loss": 0.7889,
      "step": 650
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.993083712502835e-05,
      "loss": 1.215,
      "step": 651
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992991824189928e-05,
      "loss": 1.408,
      "step": 652
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992899329922354e-05,
      "loss": 0.4755,
      "step": 653
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99280622971134e-05,
      "loss": 0.5892,
      "step": 654
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992712523568183e-05,
      "loss": 1.1628,
      "step": 655
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992618211504256e-05,
      "loss": 0.603,
      "step": 656
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992523293531007e-05,
      "loss": 1.0192,
      "step": 657
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992427769659952e-05,
      "loss": 0.7802,
      "step": 658
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992331639902685e-05,
      "loss": 0.9805,
      "step": 659
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992234904270874e-05,
      "loss": 1.1351,
      "step": 660
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992137562776256e-05,
      "loss": 0.94,
      "step": 661
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.992039615430648e-05,
      "loss": 0.9847,
      "step": 662
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991941062245934e-05,
      "loss": 0.45,
      "step": 663
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991841903234076e-05,
      "loss": 1.3327,
      "step": 664
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991742138407107e-05,
      "loss": 0.8068,
      "step": 665
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991641767777134e-05,
      "loss": 0.8112,
      "step": 666
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991540791356342e-05,
      "loss": 1.2377,
      "step": 667
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99143920915698e-05,
      "loss": 1.0605,
      "step": 668
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991337021191381e-05,
      "loss": 0.9956,
      "step": 669
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991234227471945e-05,
      "loss": 1.4656,
      "step": 670
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991130828011145e-05,
      "loss": 1.6875,
      "step": 671
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.991026822821534e-05,
      "loss": 0.9998,
      "step": 672
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99092221191573e-05,
      "loss": 1.1447,
      "step": 673
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99081699530643e-05,
      "loss": 1.3256,
      "step": 674
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.990711173006403e-05,
      "loss": 0.8783,
      "step": 675
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.990604745028494e-05,
      "loss": 1.0185,
      "step": 676
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.990497711385617e-05,
      "loss": 0.8896,
      "step": 677
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.990390072090762e-05,
      "loss": 1.3446,
      "step": 678
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.990281827156992e-05,
      "loss": 1.0091,
      "step": 679
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.990172976597445e-05,
      "loss": 1.3831,
      "step": 680
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.99006352042533e-05,
      "loss": 1.3698,
      "step": 681
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989953458653929e-05,
      "loss": 1.2736,
      "step": 682
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989842791296603e-05,
      "loss": 0.9642,
      "step": 683
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.98973151836678e-05,
      "loss": 0.9409,
      "step": 684
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989619639877968e-05,
      "loss": 1.2176,
      "step": 685
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989507155843738e-05,
      "loss": 1.0084,
      "step": 686
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989394066277746e-05,
      "loss": 0.7551,
      "step": 687
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989280371193718e-05,
      "loss": 1.6169,
      "step": 688
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.989166070605447e-05,
      "loss": 0.6749,
      "step": 689
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.98905116452681e-05,
      "loss": 1.3913,
      "step": 690
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988935652971747e-05,
      "loss": 1.4098,
      "step": 691
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988819535954281e-05,
      "loss": 1.5099,
      "step": 692
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988702813488502e-05,
      "loss": 1.6457,
      "step": 693
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988585485588576e-05,
      "loss": 0.9918,
      "step": 694
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988467552268741e-05,
      "loss": 1.5477,
      "step": 695
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.98834901354331e-05,
      "loss": 1.2559,
      "step": 696
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988229869426671e-05,
      "loss": 1.1911,
      "step": 697
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.988110119933281e-05,
      "loss": 0.6814,
      "step": 698
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987989765077672e-05,
      "loss": 1.116,
      "step": 699
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987868804874453e-05,
      "loss": 1.4332,
      "step": 700
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987747239338306e-05,
      "loss": 0.6897,
      "step": 701
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987625068483978e-05,
      "loss": 1.3695,
      "step": 702
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987502292326301e-05,
      "loss": 1.0693,
      "step": 703
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987378910880172e-05,
      "loss": 1.4126,
      "step": 704
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987254924160566e-05,
      "loss": 1.1757,
      "step": 705
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.98713033218253e-05,
      "loss": 1.4369,
      "step": 706
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.987005134961185e-05,
      "loss": 1.3517,
      "step": 707
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.986879332511726e-05,
      "loss": 1.3974,
      "step": 708
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.986752924849419e-05,
      "loss": 0.7276,
      "step": 709
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.986625911989604e-05,
      "loss": 0.8544,
      "step": 710
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.986498293947697e-05,
      "loss": 1.0697,
      "step": 711
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.986370070739187e-05,
      "loss": 0.7162,
      "step": 712
    },
    {
      "epoch": 0.07,
      "learning_rate": 9.986241242379633e-05,
      "loss": 0.8902,
      "step": 713
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.986111808884671e-05,
      "loss": 0.7228,
      "step": 714
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985981770270007e-05,
      "loss": 1.2249,
      "step": 715
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985851126551428e-05,
      "loss": 1.3135,
      "step": 716
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985719877744783e-05,
      "loss": 1.2724,
      "step": 717
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985588023866004e-05,
      "loss": 1.1748,
      "step": 718
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985455564931092e-05,
      "loss": 1.5257,
      "step": 719
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.98532250095612e-05,
      "loss": 1.2931,
      "step": 720
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985188831957242e-05,
      "loss": 1.2338,
      "step": 721
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.985054557950674e-05,
      "loss": 1.0715,
      "step": 722
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.984919678952718e-05,
      "loss": 1.1845,
      "step": 723
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.984784194979739e-05,
      "loss": 1.228,
      "step": 724
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.98464810604818e-05,
      "loss": 1.5172,
      "step": 725
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.984511412174557e-05,
      "loss": 0.5897,
      "step": 726
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.984374113375461e-05,
      "loss": 0.9165,
      "step": 727
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.984236209667553e-05,
      "loss": 1.1547,
      "step": 728
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.98409770106757e-05,
      "loss": 0.8435,
      "step": 729
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983958587592319e-05,
      "loss": 0.7764,
      "step": 730
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983818869258687e-05,
      "loss": 1.2112,
      "step": 731
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983678546083628e-05,
      "loss": 0.5342,
      "step": 732
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983537618084171e-05,
      "loss": 1.3425,
      "step": 733
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983396085277421e-05,
      "loss": 0.8257,
      "step": 734
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983253947680553e-05,
      "loss": 0.8778,
      "step": 735
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.983111205310819e-05,
      "loss": 1.2013,
      "step": 736
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982967858185542e-05,
      "loss": 1.4641,
      "step": 737
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982823906322116e-05,
      "loss": 1.1029,
      "step": 738
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982679349738014e-05,
      "loss": 1.3809,
      "step": 739
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982534188450778e-05,
      "loss": 0.7902,
      "step": 740
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982388422478026e-05,
      "loss": 1.0384,
      "step": 741
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982242051837448e-05,
      "loss": 0.7375,
      "step": 742
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.982095076546807e-05,
      "loss": 0.6584,
      "step": 743
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981947496623942e-05,
      "loss": 1.2841,
      "step": 744
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981799312086762e-05,
      "loss": 1.3534,
      "step": 745
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981650522953248e-05,
      "loss": 0.5626,
      "step": 746
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981501129241463e-05,
      "loss": 1.329,
      "step": 747
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981351130969534e-05,
      "loss": 1.1135,
      "step": 748
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981200528155666e-05,
      "loss": 1.3702,
      "step": 749
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.981049320818135e-05,
      "loss": 1.4757,
      "step": 750
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.980897508975294e-05,
      "loss": 0.4567,
      "step": 751
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.980745092645564e-05,
      "loss": 1.3336,
      "step": 752
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.980592071847445e-05,
      "loss": 1.7605,
      "step": 753
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.980438446599507e-05,
      "loss": 1.0117,
      "step": 754
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.980284216920393e-05,
      "loss": 0.9213,
      "step": 755
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.980129382828823e-05,
      "loss": 0.4212,
      "step": 756
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.979973944343584e-05,
      "loss": 0.9445,
      "step": 757
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.979817901483544e-05,
      "loss": 1.248,
      "step": 758
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.979661254267638e-05,
      "loss": 1.0449,
      "step": 759
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97950400271488e-05,
      "loss": 1.2656,
      "step": 760
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.979346146844351e-05,
      "loss": 1.1661,
      "step": 761
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.979187686675208e-05,
      "loss": 1.0254,
      "step": 762
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.979028622226685e-05,
      "loss": 1.1144,
      "step": 763
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.978868953518084e-05,
      "loss": 1.1096,
      "step": 764
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.978708680568784e-05,
      "loss": 1.5102,
      "step": 765
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.978547803398235e-05,
      "loss": 1.3489,
      "step": 766
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.978386322025961e-05,
      "loss": 0.8688,
      "step": 767
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97822423647156e-05,
      "loss": 1.5165,
      "step": 768
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.978061546754703e-05,
      "loss": 1.1871,
      "step": 769
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.977898252895134e-05,
      "loss": 1.3936,
      "step": 770
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97773435491267e-05,
      "loss": 1.3528,
      "step": 771
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.977569852827202e-05,
      "loss": 0.9926,
      "step": 772
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.977404746658696e-05,
      "loss": 0.7894,
      "step": 773
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.977239036427188e-05,
      "loss": 0.8583,
      "step": 774
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.977072722152787e-05,
      "loss": 1.5327,
      "step": 775
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.976905803855679e-05,
      "loss": 0.5102,
      "step": 776
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.976738281556121e-05,
      "loss": 1.2363,
      "step": 777
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.976570155274446e-05,
      "loss": 1.2401,
      "step": 778
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.976401425031054e-05,
      "loss": 0.9295,
      "step": 779
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.976232090846424e-05,
      "loss": 0.5442,
      "step": 780
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.976062152741107e-05,
      "loss": 1.0803,
      "step": 781
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.975891610735728e-05,
      "loss": 0.8488,
      "step": 782
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.975720464850981e-05,
      "loss": 0.792,
      "step": 783
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.975548715107639e-05,
      "loss": 0.4693,
      "step": 784
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.975376361526543e-05,
      "loss": 1.3893,
      "step": 785
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.975203404128614e-05,
      "loss": 0.9634,
      "step": 786
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.975029842934838e-05,
      "loss": 1.2634,
      "step": 787
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.974855677966283e-05,
      "loss": 0.8203,
      "step": 788
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.974680909244084e-05,
      "loss": 1.0295,
      "step": 789
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.974505536789447e-05,
      "loss": 1.333,
      "step": 790
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97432956062366e-05,
      "loss": 0.6689,
      "step": 791
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97415298076808e-05,
      "loss": 1.3343,
      "step": 792
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.973975797244138e-05,
      "loss": 1.1755,
      "step": 793
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.973798010073332e-05,
      "loss": 1.1036,
      "step": 794
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.973619619277242e-05,
      "loss": 1.2033,
      "step": 795
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.973440624877514e-05,
      "loss": 0.9428,
      "step": 796
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.973261026895877e-05,
      "loss": 0.9121,
      "step": 797
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.973080825354124e-05,
      "loss": 1.2227,
      "step": 798
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.972900020274123e-05,
      "loss": 0.8702,
      "step": 799
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.97271861167782e-05,
      "loss": 0.5274,
      "step": 800
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.972536599587228e-05,
      "loss": 1.2051,
      "step": 801
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.972353984024437e-05,
      "loss": 1.0447,
      "step": 802
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.972170765011611e-05,
      "loss": 1.021,
      "step": 803
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.971986942570984e-05,
      "loss": 1.1914,
      "step": 804
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.971802516724864e-05,
      "loss": 1.19,
      "step": 805
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.971617487495635e-05,
      "loss": 0.6571,
      "step": 806
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.971431854905752e-05,
      "loss": 0.6127,
      "step": 807
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.971245618977742e-05,
      "loss": 0.8383,
      "step": 808
    },
    {
      "epoch": 0.08,
      "learning_rate": 9.971058779734211e-05,
      "loss": 0.6268,
      "step": 809
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.970871337197828e-05,
      "loss": 1.0983,
      "step": 810
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.970683291391346e-05,
      "loss": 0.7968,
      "step": 811
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.970494642337585e-05,
      "loss": 0.8836,
      "step": 812
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.970305390059437e-05,
      "loss": 1.2473,
      "step": 813
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.970115534579875e-05,
      "loss": 1.0361,
      "step": 814
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.969925075921936e-05,
      "loss": 1.2961,
      "step": 815
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.969734014108737e-05,
      "loss": 1.135,
      "step": 816
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.969542349163461e-05,
      "loss": 1.044,
      "step": 817
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.969350081109375e-05,
      "loss": 1.4854,
      "step": 818
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.969157209969806e-05,
      "loss": 1.1936,
      "step": 819
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.968963735768166e-05,
      "loss": 0.771,
      "step": 820
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.968769658527935e-05,
      "loss": 1.1106,
      "step": 821
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.968574978272665e-05,
      "loss": 1.4841,
      "step": 822
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.968379695025981e-05,
      "loss": 1.099,
      "step": 823
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.968183808811586e-05,
      "loss": 1.3412,
      "step": 824
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.967987319653251e-05,
      "loss": 1.4232,
      "step": 825
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.967790227574823e-05,
      "loss": 1.4605,
      "step": 826
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96759253260022e-05,
      "loss": 1.3547,
      "step": 827
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.967394234753435e-05,
      "loss": 1.525,
      "step": 828
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.967195334058536e-05,
      "loss": 0.9738,
      "step": 829
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.966995830539658e-05,
      "loss": 1.2094,
      "step": 830
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.966795724221016e-05,
      "loss": 1.4562,
      "step": 831
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.966595015126894e-05,
      "loss": 1.4202,
      "step": 832
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96639370328165e-05,
      "loss": 0.6975,
      "step": 833
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.966191788709716e-05,
      "loss": 0.6287,
      "step": 834
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.965989271435595e-05,
      "loss": 1.3874,
      "step": 835
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.965786151483867e-05,
      "loss": 0.9942,
      "step": 836
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96558242887918e-05,
      "loss": 1.5124,
      "step": 837
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.965378103646261e-05,
      "loss": 1.1854,
      "step": 838
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.965173175809906e-05,
      "loss": 1.1436,
      "step": 839
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.964967645394984e-05,
      "loss": 0.6441,
      "step": 840
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96476151242644e-05,
      "loss": 1.1908,
      "step": 841
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96455477692929e-05,
      "loss": 0.9733,
      "step": 842
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.964347438928623e-05,
      "loss": 1.3604,
      "step": 843
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.964139498449603e-05,
      "loss": 1.0567,
      "step": 844
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.963930955517464e-05,
      "loss": 0.5014,
      "step": 845
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.963721810157518e-05,
      "loss": 0.6732,
      "step": 846
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.963512062395145e-05,
      "loss": 0.6192,
      "step": 847
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.9633017122558e-05,
      "loss": 0.6476,
      "step": 848
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.963090759765012e-05,
      "loss": 0.8187,
      "step": 849
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.962879204948381e-05,
      "loss": 1.4724,
      "step": 850
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.962667047831584e-05,
      "loss": 1.6065,
      "step": 851
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.962454288440368e-05,
      "loss": 1.2907,
      "step": 852
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.962240926800553e-05,
      "loss": 1.1274,
      "step": 853
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.962026962938032e-05,
      "loss": 0.5885,
      "step": 854
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.961812396878772e-05,
      "loss": 1.1867,
      "step": 855
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.961597228648814e-05,
      "loss": 1.2307,
      "step": 856
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96138145827427e-05,
      "loss": 1.1636,
      "step": 857
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.961165085781329e-05,
      "loss": 1.4505,
      "step": 858
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.960948111196247e-05,
      "loss": 1.2584,
      "step": 859
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.960730534545358e-05,
      "loss": 1.426,
      "step": 860
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.960512355855064e-05,
      "loss": 1.6087,
      "step": 861
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96029357515185e-05,
      "loss": 1.3281,
      "step": 862
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.96007419246226e-05,
      "loss": 0.7987,
      "step": 863
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.959854207812923e-05,
      "loss": 1.0634,
      "step": 864
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.959633621230535e-05,
      "loss": 1.099,
      "step": 865
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.959412432741869e-05,
      "loss": 1.1635,
      "step": 866
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.959190642373765e-05,
      "loss": 0.8464,
      "step": 867
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.958968250153142e-05,
      "loss": 1.3227,
      "step": 868
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.958745256106991e-05,
      "loss": 1.0479,
      "step": 869
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.958521660262371e-05,
      "loss": 0.8466,
      "step": 870
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.958297462646419e-05,
      "loss": 0.9906,
      "step": 871
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.958072663286348e-05,
      "loss": 1.3412,
      "step": 872
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.957847262209435e-05,
      "loss": 1.1449,
      "step": 873
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.957621259443035e-05,
      "loss": 0.6198,
      "step": 874
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.957394655014579e-05,
      "loss": 0.8211,
      "step": 875
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.957167448951567e-05,
      "loss": 0.7339,
      "step": 876
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.956939641281571e-05,
      "loss": 1.1439,
      "step": 877
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.95671123203224e-05,
      "loss": 1.4849,
      "step": 878
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.956482221231293e-05,
      "loss": 1.5012,
      "step": 879
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.956252608906524e-05,
      "loss": 1.3153,
      "step": 880
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.956022395085798e-05,
      "loss": 1.623,
      "step": 881
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.955791579797053e-05,
      "loss": 1.4195,
      "step": 882
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.955560163068304e-05,
      "loss": 0.7595,
      "step": 883
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.955328144927633e-05,
      "loss": 1.2983,
      "step": 884
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.9550955254032e-05,
      "loss": 0.5835,
      "step": 885
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.954862304523235e-05,
      "loss": 1.1641,
      "step": 886
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.954628482316042e-05,
      "loss": 0.5821,
      "step": 887
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.954394058809998e-05,
      "loss": 1.4094,
      "step": 888
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.954159034033554e-05,
      "loss": 0.8763,
      "step": 889
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.95392340801523e-05,
      "loss": 0.7776,
      "step": 890
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.953687180783624e-05,
      "loss": 1.5091,
      "step": 891
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.953450352367404e-05,
      "loss": 1.2801,
      "step": 892
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.953212922795314e-05,
      "loss": 0.6603,
      "step": 893
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.952974892096165e-05,
      "loss": 1.3199,
      "step": 894
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.952736260298846e-05,
      "loss": 1.493,
      "step": 895
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.95249702743232e-05,
      "loss": 1.889,
      "step": 896
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.952257193525615e-05,
      "loss": 0.8567,
      "step": 897
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.952016758607845e-05,
      "loss": 1.1683,
      "step": 898
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.951775722708184e-05,
      "loss": 0.8459,
      "step": 899
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.951534085855885e-05,
      "loss": 1.1369,
      "step": 900
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.951291848080276e-05,
      "loss": 1.366,
      "step": 901
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.951049009410751e-05,
      "loss": 1.5734,
      "step": 902
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.950805569876786e-05,
      "loss": 1.3372,
      "step": 903
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.95056152950792e-05,
      "loss": 0.4965,
      "step": 904
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.950316888333775e-05,
      "loss": 1.3115,
      "step": 905
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.950071646384037e-05,
      "loss": 1.0258,
      "step": 906
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.949825803688471e-05,
      "loss": 0.8791,
      "step": 907
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.949579360276912e-05,
      "loss": 1.2303,
      "step": 908
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.949332316179269e-05,
      "loss": 0.9167,
      "step": 909
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.949084671425523e-05,
      "loss": 1.3614,
      "step": 910
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.948836426045728e-05,
      "loss": 1.0742,
      "step": 911
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.948587580070012e-05,
      "loss": 0.5246,
      "step": 912
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.948338133528576e-05,
      "loss": 1.0793,
      "step": 913
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.948088086451691e-05,
      "loss": 1.6566,
      "step": 914
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.947837438869704e-05,
      "loss": 0.4854,
      "step": 915
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.947586190813036e-05,
      "loss": 0.8178,
      "step": 916
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.947334342312176e-05,
      "loss": 1.0783,
      "step": 917
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.94708189339769e-05,
      "loss": 0.6771,
      "step": 918
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.946828844100214e-05,
      "loss": 1.1437,
      "step": 919
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.946575194450458e-05,
      "loss": 0.6977,
      "step": 920
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.946320944479208e-05,
      "loss": 1.1065,
      "step": 921
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.946066094217318e-05,
      "loss": 0.5999,
      "step": 922
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.945810643695717e-05,
      "loss": 1.0021,
      "step": 923
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.945554592945407e-05,
      "loss": 1.3826,
      "step": 924
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.945297941997464e-05,
      "loss": 1.4021,
      "step": 925
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.945040690883033e-05,
      "loss": 0.5645,
      "step": 926
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.944782839633335e-05,
      "loss": 0.5668,
      "step": 927
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.944524388279663e-05,
      "loss": 1.5026,
      "step": 928
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.944265336853385e-05,
      "loss": 0.4877,
      "step": 929
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.944005685385937e-05,
      "loss": 1.5571,
      "step": 930
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.943745433908831e-05,
      "loss": 1.5987,
      "step": 931
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.943484582453653e-05,
      "loss": 1.1794,
      "step": 932
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.94322313105206e-05,
      "loss": 1.3225,
      "step": 933
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.94296107973578e-05,
      "loss": 0.8015,
      "step": 934
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.942698428536616e-05,
      "loss": 0.4423,
      "step": 935
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.942435177486446e-05,
      "loss": 0.5739,
      "step": 936
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.942171326617218e-05,
      "loss": 1.3807,
      "step": 937
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.941906875960952e-05,
      "loss": 1.4181,
      "step": 938
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.941641825549741e-05,
      "loss": 1.258,
      "step": 939
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.941376175415755e-05,
      "loss": 1.1324,
      "step": 940
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.94110992559123e-05,
      "loss": 0.878,
      "step": 941
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.940843076108481e-05,
      "loss": 1.0451,
      "step": 942
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.940575626999891e-05,
      "loss": 1.4167,
      "step": 943
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.94030757829792e-05,
      "loss": 1.1862,
      "step": 944
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.940038930035098e-05,
      "loss": 0.8138,
      "step": 945
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.939769682244028e-05,
      "loss": 1.3366,
      "step": 946
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.939499834957386e-05,
      "loss": 0.5101,
      "step": 947
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.939229388207921e-05,
      "loss": 0.9917,
      "step": 948
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.938958342028456e-05,
      "loss": 1.2722,
      "step": 949
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.938686696451884e-05,
      "loss": 1.3982,
      "step": 950
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.938414451511173e-05,
      "loss": 1.3981,
      "step": 951
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.938141607239362e-05,
      "loss": 1.8356,
      "step": 952
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.937868163669565e-05,
      "loss": 1.2166,
      "step": 953
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.937594120834965e-05,
      "loss": 0.9584,
      "step": 954
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.937319478768824e-05,
      "loss": 0.7718,
      "step": 955
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.93704423750447e-05,
      "loss": 1.0461,
      "step": 956
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.936768397075307e-05,
      "loss": 0.8388,
      "step": 957
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.93649195751481e-05,
      "loss": 1.4428,
      "step": 958
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.93621491885653e-05,
      "loss": 1.1652,
      "step": 959
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.935937281134089e-05,
      "loss": 1.0133,
      "step": 960
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.93565904438118e-05,
      "loss": 1.1574,
      "step": 961
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.935380208631572e-05,
      "loss": 1.057,
      "step": 962
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.935100773919102e-05,
      "loss": 0.8511,
      "step": 963
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.934820740277685e-05,
      "loss": 1.1405,
      "step": 964
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.934540107741304e-05,
      "loss": 0.7235,
      "step": 965
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.934258876344018e-05,
      "loss": 0.5497,
      "step": 966
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.933977046119959e-05,
      "loss": 1.1873,
      "step": 967
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.933694617103327e-05,
      "loss": 1.8115,
      "step": 968
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.9334115893284e-05,
      "loss": 1.0816,
      "step": 969
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.933127962829527e-05,
      "loss": 0.827,
      "step": 970
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.932843737641127e-05,
      "loss": 1.4088,
      "step": 971
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.932558913797696e-05,
      "loss": 1.2906,
      "step": 972
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.9322734913338e-05,
      "loss": 1.3112,
      "step": 973
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.931987470284077e-05,
      "loss": 1.3161,
      "step": 974
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.931700850683242e-05,
      "loss": 0.8538,
      "step": 975
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.931413632566074e-05,
      "loss": 1.2561,
      "step": 976
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.931125815967434e-05,
      "loss": 1.0723,
      "step": 977
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.930837400922252e-05,
      "loss": 1.1425,
      "step": 978
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.93054838746553e-05,
      "loss": 0.6706,
      "step": 979
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.93025877563234e-05,
      "loss": 0.5838,
      "step": 980
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.929968565457832e-05,
      "loss": 1.0826,
      "step": 981
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.929677756977228e-05,
      "loss": 0.9103,
      "step": 982
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.929386350225818e-05,
      "loss": 0.6866,
      "step": 983
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.929094345238968e-05,
      "loss": 1.0696,
      "step": 984
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.928801742052116e-05,
      "loss": 0.7385,
      "step": 985
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.928508540700774e-05,
      "loss": 1.234,
      "step": 986
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.928214741220523e-05,
      "loss": 1.2262,
      "step": 987
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.927920343647021e-05,
      "loss": 1.4375,
      "step": 988
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.927625348015996e-05,
      "loss": 1.4102,
      "step": 989
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.927329754363248e-05,
      "loss": 1.2538,
      "step": 990
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.92703356272465e-05,
      "loss": 1.4239,
      "step": 991
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.92673677313615e-05,
      "loss": 1.2304,
      "step": 992
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.926439385633766e-05,
      "loss": 1.2144,
      "step": 993
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.926141400253588e-05,
      "loss": 1.1922,
      "step": 994
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.925842817031781e-05,
      "loss": 1.3711,
      "step": 995
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.92554363600458e-05,
      "loss": 1.0155,
      "step": 996
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.925243857208298e-05,
      "loss": 1.3198,
      "step": 997
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.924943480679311e-05,
      "loss": 0.7377,
      "step": 998
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.924642506454075e-05,
      "loss": 1.3732,
      "step": 999
    },
    {
      "epoch": 0.11,
      "learning_rate": 9.924340934569121e-05,
      "loss": 0.7496,
      "step": 1000
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 9518,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 200,
  "total_flos": 1.690338267758592e+18,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}