{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.99958796868562,
  "global_step": 7280,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000001e-07,
      "loss": 11.954,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0000000000000002e-07,
      "loss": 11.7574,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.0000000000000003e-07,
      "loss": 11.7662,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.000000000000001e-07,
      "loss": 11.6897,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-07,
      "loss": 11.7292,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 11.7205,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2000000000000002e-06,
      "loss": 11.4608,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4000000000000001e-06,
      "loss": 11.282,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.6000000000000001e-06,
      "loss": 11.0321,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8e-06,
      "loss": 10.9982,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 10.7355,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.2e-06,
      "loss": 10.4186,
      "step": 24
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 10.2501,
      "step": 26
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.6e-06,
      "loss": 10.1457,
      "step": 28
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8000000000000003e-06,
      "loss": 9.7192,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 3e-06,
      "loss": 9.4468,
      "step": 32
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.2000000000000003e-06,
      "loss": 9.3485,
      "step": 34
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.4000000000000005e-06,
      "loss": 9.1247,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.6e-06,
      "loss": 8.8148,
      "step": 38
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.8e-06,
      "loss": 8.5495,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.000000000000001e-06,
      "loss": 8.4185,
      "step": 42
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.2000000000000004e-06,
      "loss": 8.2214,
      "step": 44
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.4e-06,
      "loss": 8.0612,
      "step": 46
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.6e-06,
      "loss": 7.8571,
      "step": 48
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.800000000000001e-06,
      "loss": 7.6998,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 5e-06,
      "loss": 7.6994,
      "step": 52
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.2e-06,
      "loss": 7.4386,
      "step": 54
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.4e-06,
      "loss": 7.4094,
      "step": 56
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.600000000000001e-06,
      "loss": 7.3206,
      "step": 58
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.8e-06,
      "loss": 7.2811,
      "step": 60
    },
    {
      "epoch": 0.03,
      "learning_rate": 6e-06,
      "loss": 7.2123,
      "step": 62
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.2e-06,
      "loss": 7.1106,
      "step": 64
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.4000000000000006e-06,
      "loss": 7.182,
      "step": 66
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.6e-06,
      "loss": 7.0434,
      "step": 68
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.800000000000001e-06,
      "loss": 7.0735,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.000000000000001e-06,
      "loss": 6.9208,
      "step": 72
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.2e-06,
      "loss": 6.9198,
      "step": 74
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.4e-06,
      "loss": 6.8729,
      "step": 76
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.6e-06,
      "loss": 6.7778,
      "step": 78
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.8e-06,
      "loss": 6.7507,
      "step": 80
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.000000000000001e-06,
      "loss": 6.6864,
      "step": 82
    },
    {
      "epoch": 0.03,
      "learning_rate": 8.200000000000001e-06,
      "loss": 6.7855,
      "step": 84
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.400000000000001e-06,
      "loss": 6.6503,
      "step": 86
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.599999999999999e-06,
      "loss": 6.6798,
      "step": 88
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.8e-06,
      "loss": 6.5844,
      "step": 90
    },
    {
      "epoch": 0.04,
      "learning_rate": 9e-06,
      "loss": 6.6041,
      "step": 92
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.2e-06,
      "loss": 6.539,
      "step": 94
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.4e-06,
      "loss": 6.5601,
      "step": 96
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.600000000000001e-06,
      "loss": 6.5664,
      "step": 98
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.800000000000001e-06,
      "loss": 6.5179,
      "step": 100
    },
    {
      "epoch": 0.04,
      "learning_rate": 1e-05,
      "loss": 6.5246,
      "step": 102
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.02e-05,
      "loss": 6.5961,
      "step": 104
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.04e-05,
      "loss": 6.435,
      "step": 106
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.06e-05,
      "loss": 6.4144,
      "step": 108
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.08e-05,
      "loss": 6.426,
      "step": 110
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 6.4593,
      "step": 112
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1200000000000001e-05,
      "loss": 6.44,
      "step": 114
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1400000000000001e-05,
      "loss": 6.2686,
      "step": 116
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.16e-05,
      "loss": 6.482,
      "step": 118
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.18e-05,
      "loss": 6.3319,
      "step": 120
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.2e-05,
      "loss": 6.325,
      "step": 122
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.22e-05,
      "loss": 6.3218,
      "step": 124
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.24e-05,
      "loss": 6.3082,
      "step": 126
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.2600000000000001e-05,
      "loss": 6.3179,
      "step": 128
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.2800000000000001e-05,
      "loss": 6.3464,
      "step": 130
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 6.3182,
      "step": 132
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.32e-05,
      "loss": 6.2068,
      "step": 134
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3400000000000002e-05,
      "loss": 6.0798,
      "step": 136
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3600000000000002e-05,
      "loss": 6.2293,
      "step": 138
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3800000000000002e-05,
      "loss": 6.1612,
      "step": 140
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 6.1521,
      "step": 142
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.42e-05,
      "loss": 6.168,
      "step": 144
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.44e-05,
      "loss": 6.0825,
      "step": 146
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4599999999999999e-05,
      "loss": 6.0993,
      "step": 148
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.48e-05,
      "loss": 6.1154,
      "step": 150
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.5e-05,
      "loss": 6.0066,
      "step": 152
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.52e-05,
      "loss": 5.9245,
      "step": 154
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.54e-05,
      "loss": 5.9375,
      "step": 156
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.56e-05,
      "loss": 5.9601,
      "step": 158
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.58e-05,
      "loss": 5.9027,
      "step": 160
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 5.8572,
      "step": 162
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.62e-05,
      "loss": 5.7952,
      "step": 164
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6400000000000002e-05,
      "loss": 5.8793,
      "step": 166
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.66e-05,
      "loss": 5.7643,
      "step": 168
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6800000000000002e-05,
      "loss": 5.7912,
      "step": 170
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.7000000000000003e-05,
      "loss": 5.7259,
      "step": 172
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.7199999999999998e-05,
      "loss": 5.7517,
      "step": 174
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.74e-05,
      "loss": 5.7468,
      "step": 176
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.76e-05,
      "loss": 5.6949,
      "step": 178
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.78e-05,
      "loss": 5.7367,
      "step": 180
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.8e-05,
      "loss": 5.6554,
      "step": 182
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8200000000000002e-05,
      "loss": 5.6252,
      "step": 184
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.84e-05,
      "loss": 5.6907,
      "step": 186
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.86e-05,
      "loss": 5.5224,
      "step": 188
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.88e-05,
      "loss": 5.6275,
      "step": 190
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9e-05,
      "loss": 5.4918,
      "step": 192
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9200000000000003e-05,
      "loss": 5.4904,
      "step": 194
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.94e-05,
      "loss": 5.5485,
      "step": 196
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9600000000000002e-05,
      "loss": 5.4543,
      "step": 198
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9800000000000004e-05,
      "loss": 5.4329,
      "step": 200
    },
    {
      "epoch": 0.08,
      "learning_rate": 2e-05,
      "loss": 5.6126,
      "step": 202
    },
    {
      "epoch": 0.08,
      "learning_rate": 2.0200000000000003e-05,
      "loss": 5.3985,
      "step": 204
    },
    {
      "epoch": 0.08,
      "learning_rate": 2.04e-05,
      "loss": 5.3746,
      "step": 206
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.06e-05,
      "loss": 5.4144,
      "step": 208
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.08e-05,
      "loss": 5.3626,
      "step": 210
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.1e-05,
      "loss": 5.2793,
      "step": 212
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.12e-05,
      "loss": 5.3459,
      "step": 214
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.1400000000000002e-05,
      "loss": 5.4056,
      "step": 216
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.16e-05,
      "loss": 5.2179,
      "step": 218
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.18e-05,
      "loss": 5.2413,
      "step": 220
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 5.1814,
      "step": 222
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.22e-05,
      "loss": 4.9489,
      "step": 224
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.2400000000000002e-05,
      "loss": 4.975,
      "step": 226
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.26e-05,
      "loss": 4.9943,
      "step": 228
    },
    {
      "epoch": 0.09,
      "learning_rate": 2.2800000000000002e-05,
      "loss": 4.7421,
      "step": 230
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.3000000000000003e-05,
      "loss": 4.4864,
      "step": 232
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.32e-05,
      "loss": 4.4499,
      "step": 234
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.3400000000000003e-05,
      "loss": 4.1762,
      "step": 236
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.36e-05,
      "loss": 3.9751,
      "step": 238
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.37e-05,
      "loss": 3.9673,
      "step": 240
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.39e-05,
      "loss": 3.413,
      "step": 242
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.41e-05,
      "loss": 2.7623,
      "step": 244
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.43e-05,
      "loss": 2.5487,
      "step": 246
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.45e-05,
      "loss": 1.8896,
      "step": 248
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.47e-05,
      "loss": 2.0058,
      "step": 250
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.4900000000000002e-05,
      "loss": 1.6642,
      "step": 252
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.51e-05,
      "loss": 1.3299,
      "step": 254
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.5300000000000002e-05,
      "loss": 1.2625,
      "step": 256
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.5500000000000003e-05,
      "loss": 1.2002,
      "step": 258
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.57e-05,
      "loss": 1.1715,
      "step": 260
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.5900000000000003e-05,
      "loss": 1.1752,
      "step": 262
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.61e-05,
      "loss": 1.062,
      "step": 264
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.6300000000000002e-05,
      "loss": 1.0381,
      "step": 266
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.6500000000000004e-05,
      "loss": 0.9979,
      "step": 268
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.6700000000000002e-05,
      "loss": 1.0363,
      "step": 270
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.6900000000000003e-05,
      "loss": 0.9941,
      "step": 272
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.7100000000000005e-05,
      "loss": 0.9799,
      "step": 274
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.7300000000000003e-05,
      "loss": 0.9831,
      "step": 276
    },
    {
      "epoch": 0.11,
      "learning_rate": 2.7500000000000004e-05,
      "loss": 0.9996,
      "step": 278
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.7700000000000002e-05,
      "loss": 0.9783,
      "step": 280
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.7900000000000004e-05,
      "loss": 0.945,
      "step": 282
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.8100000000000005e-05,
      "loss": 0.9243,
      "step": 284
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.83e-05,
      "loss": 0.9685,
      "step": 286
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.8499999999999998e-05,
      "loss": 0.9208,
      "step": 288
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.87e-05,
      "loss": 0.9502,
      "step": 290
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.8899999999999998e-05,
      "loss": 0.975,
      "step": 292
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.91e-05,
      "loss": 0.9952,
      "step": 294
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.93e-05,
      "loss": 0.9405,
      "step": 296
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.95e-05,
      "loss": 0.8896,
      "step": 298
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.97e-05,
      "loss": 1.0118,
      "step": 300
    },
    {
      "epoch": 0.12,
      "learning_rate": 2.9900000000000002e-05,
      "loss": 0.9056,
      "step": 302
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.01e-05,
      "loss": 0.9318,
      "step": 304
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.03e-05,
      "loss": 0.8904,
      "step": 306
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.05e-05,
      "loss": 0.9237,
      "step": 308
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.07e-05,
      "loss": 0.8975,
      "step": 310
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.09e-05,
      "loss": 0.9402,
      "step": 312
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.1100000000000004e-05,
      "loss": 0.9393,
      "step": 314
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.13e-05,
      "loss": 0.8767,
      "step": 316
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.15e-05,
      "loss": 0.8804,
      "step": 318
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.1700000000000005e-05,
      "loss": 0.8926,
      "step": 320
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.19e-05,
      "loss": 0.9633,
      "step": 322
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.21e-05,
      "loss": 0.8566,
      "step": 324
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.2300000000000006e-05,
      "loss": 0.8893,
      "step": 326
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.2500000000000004e-05,
      "loss": 0.8954,
      "step": 328
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.27e-05,
      "loss": 0.943,
      "step": 330
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.29e-05,
      "loss": 0.9046,
      "step": 332
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.3100000000000005e-05,
      "loss": 0.8691,
      "step": 334
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.33e-05,
      "loss": 0.8766,
      "step": 336
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.35e-05,
      "loss": 0.8979,
      "step": 338
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.3700000000000006e-05,
      "loss": 0.9405,
      "step": 340
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.3900000000000004e-05,
      "loss": 0.937,
      "step": 342
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.41e-05,
      "loss": 0.9517,
      "step": 344
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.430000000000001e-05,
      "loss": 0.8925,
      "step": 346
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.45e-05,
      "loss": 0.9017,
      "step": 348
    },
    {
      "epoch": 0.14,
      "learning_rate": 3.4699999999999996e-05,
      "loss": 0.9481,
      "step": 350
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.49e-05,
      "loss": 0.9065,
      "step": 352
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.51e-05,
      "loss": 0.8895,
      "step": 354
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.53e-05,
      "loss": 0.9089,
      "step": 356
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.55e-05,
      "loss": 0.9032,
      "step": 358
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.57e-05,
      "loss": 0.8839,
      "step": 360
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.59e-05,
      "loss": 0.9038,
      "step": 362
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.61e-05,
      "loss": 0.8902,
      "step": 364
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.63e-05,
      "loss": 0.909,
      "step": 366
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.65e-05,
      "loss": 0.9007,
      "step": 368
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.6700000000000004e-05,
      "loss": 0.8603,
      "step": 370
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.69e-05,
      "loss": 0.9076,
      "step": 372
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.71e-05,
      "loss": 0.9497,
      "step": 374
    },
    {
      "epoch": 0.15,
      "learning_rate": 3.73e-05,
      "loss": 0.861,
      "step": 376
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.7500000000000003e-05,
      "loss": 0.8768,
      "step": 378
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.77e-05,
      "loss": 0.8686,
      "step": 380
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.79e-05,
      "loss": 0.8976,
      "step": 382
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.8100000000000005e-05,
      "loss": 0.9487,
      "step": 384
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.83e-05,
      "loss": 0.8876,
      "step": 386
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.85e-05,
      "loss": 0.8372,
      "step": 388
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.8700000000000006e-05,
      "loss": 0.8659,
      "step": 390
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.8900000000000004e-05,
      "loss": 0.8629,
      "step": 392
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.91e-05,
      "loss": 0.8987,
      "step": 394
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.9300000000000007e-05,
      "loss": 0.8044,
      "step": 396
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 0.8774,
      "step": 398
    },
    {
      "epoch": 0.16,
      "learning_rate": 3.97e-05,
      "loss": 0.8448,
      "step": 400
    },
    {
      "epoch": 0.17,
      "learning_rate": 3.99e-05,
      "loss": 0.8648,
      "step": 402
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.0100000000000006e-05,
      "loss": 0.9247,
      "step": 404
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.0300000000000004e-05,
      "loss": 0.9516,
      "step": 406
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.05e-05,
      "loss": 0.9199,
      "step": 408
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.07e-05,
      "loss": 0.8626,
      "step": 410
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.09e-05,
      "loss": 0.8827,
      "step": 412
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.11e-05,
      "loss": 0.8758,
      "step": 414
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.13e-05,
      "loss": 0.8837,
      "step": 416
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.15e-05,
      "loss": 0.8269,
      "step": 418
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.17e-05,
      "loss": 0.8785,
      "step": 420
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.19e-05,
      "loss": 0.8848,
      "step": 422
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.21e-05,
      "loss": 0.8468,
      "step": 424
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.23e-05,
      "loss": 0.8769,
      "step": 426
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.25e-05,
      "loss": 0.8842,
      "step": 428
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.27e-05,
      "loss": 0.8689,
      "step": 430
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.29e-05,
      "loss": 0.8532,
      "step": 432
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.3100000000000004e-05,
      "loss": 0.8785,
      "step": 434
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.33e-05,
      "loss": 0.8459,
      "step": 436
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.35e-05,
      "loss": 0.9716,
      "step": 438
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.3700000000000005e-05,
      "loss": 0.8705,
      "step": 440
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.39e-05,
      "loss": 0.8851,
      "step": 442
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.41e-05,
      "loss": 0.8803,
      "step": 444
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.43e-05,
      "loss": 0.8405,
      "step": 446
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.4500000000000004e-05,
      "loss": 0.8476,
      "step": 448
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.47e-05,
      "loss": 0.9052,
      "step": 450
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.49e-05,
      "loss": 0.8506,
      "step": 452
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5100000000000005e-05,
      "loss": 0.8367,
      "step": 454
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.53e-05,
      "loss": 0.8569,
      "step": 456
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.55e-05,
      "loss": 0.8898,
      "step": 458
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5700000000000006e-05,
      "loss": 0.9018,
      "step": 460
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5900000000000004e-05,
      "loss": 0.8873,
      "step": 462
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.61e-05,
      "loss": 0.8827,
      "step": 464
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.630000000000001e-05,
      "loss": 0.8778,
      "step": 466
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.6500000000000005e-05,
      "loss": 0.8425,
      "step": 468
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.6700000000000003e-05,
      "loss": 0.8578,
      "step": 470
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.69e-05,
      "loss": 0.9361,
      "step": 472
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.71e-05,
      "loss": 0.8809,
      "step": 474
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.73e-05,
      "loss": 0.8798,
      "step": 476
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.75e-05,
      "loss": 0.8578,
      "step": 478
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.77e-05,
      "loss": 0.8839,
      "step": 480
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.79e-05,
      "loss": 0.8434,
      "step": 482
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.8100000000000004e-05,
      "loss": 0.8752,
      "step": 484
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.83e-05,
      "loss": 0.9135,
      "step": 486
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.85e-05,
      "loss": 0.8467,
      "step": 488
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.87e-05,
      "loss": 0.8543,
      "step": 490
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.89e-05,
      "loss": 0.8767,
      "step": 492
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.91e-05,
      "loss": 0.9209,
      "step": 494
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.93e-05,
      "loss": 0.8609,
      "step": 496
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 0.8317,
      "step": 498
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.97e-05,
      "loss": 0.8808,
      "step": 500
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.99e-05,
      "loss": 0.8848,
      "step": 502
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.999262645627489e-05,
      "loss": 0.8544,
      "step": 504
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.997787936882466e-05,
      "loss": 0.8409,
      "step": 506
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.996313228137443e-05,
      "loss": 0.8588,
      "step": 508
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.99483851939242e-05,
      "loss": 0.8377,
      "step": 510
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9933638106473976e-05,
      "loss": 0.9157,
      "step": 512
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9918891019023746e-05,
      "loss": 0.8672,
      "step": 514
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9904143931573517e-05,
      "loss": 0.8357,
      "step": 516
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.988939684412329e-05,
      "loss": 0.8829,
      "step": 518
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.987464975667306e-05,
      "loss": 0.9034,
      "step": 520
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.985990266922283e-05,
      "loss": 0.8809,
      "step": 522
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.98451555817726e-05,
      "loss": 0.899,
      "step": 524
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9830408494322375e-05,
      "loss": 0.8546,
      "step": 526
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9815661406872146e-05,
      "loss": 0.8629,
      "step": 528
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9800914319421916e-05,
      "loss": 0.8685,
      "step": 530
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9786167231971686e-05,
      "loss": 0.8077,
      "step": 532
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.977142014452146e-05,
      "loss": 0.878,
      "step": 534
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.975667305707123e-05,
      "loss": 0.8549,
      "step": 536
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9741925969621e-05,
      "loss": 0.8689,
      "step": 538
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9727178882170774e-05,
      "loss": 0.8742,
      "step": 540
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9712431794720545e-05,
      "loss": 0.849,
      "step": 542
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9697684707270315e-05,
      "loss": 0.8723,
      "step": 544
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9682937619820086e-05,
      "loss": 0.8409,
      "step": 546
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.9668190532369856e-05,
      "loss": 0.8383,
      "step": 548
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.9653443444919626e-05,
      "loss": 0.8507,
      "step": 550
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.96386963574694e-05,
      "loss": 0.8375,
      "step": 552
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.9623949270019174e-05,
      "loss": 0.8783,
      "step": 554
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.9609202182568944e-05,
      "loss": 0.8638,
      "step": 556
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.9594455095118715e-05,
      "loss": 0.8609,
      "step": 558
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.957970800766849e-05,
      "loss": 0.8278,
      "step": 560
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.956496092021826e-05,
      "loss": 0.8456,
      "step": 562
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.955021383276803e-05,
      "loss": 0.8514,
      "step": 564
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.95354667453178e-05,
      "loss": 0.8789,
      "step": 566
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.952071965786757e-05,
      "loss": 0.8477,
      "step": 568
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.950597257041735e-05,
      "loss": 0.8417,
      "step": 570
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.949122548296712e-05,
      "loss": 0.8501,
      "step": 572
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.947647839551689e-05,
      "loss": 0.8384,
      "step": 574
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.946173130806666e-05,
      "loss": 0.8437,
      "step": 576
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.944698422061643e-05,
      "loss": 0.8619,
      "step": 578
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.94322371331662e-05,
      "loss": 0.8724,
      "step": 580
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.941749004571597e-05,
      "loss": 0.7811,
      "step": 582
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.940274295826575e-05,
      "loss": 0.8787,
      "step": 584
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.938799587081552e-05,
      "loss": 0.8281,
      "step": 586
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.937324878336529e-05,
      "loss": 0.7954,
      "step": 588
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.935850169591506e-05,
      "loss": 0.8124,
      "step": 590
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.934375460846483e-05,
      "loss": 0.8741,
      "step": 592
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.93290075210146e-05,
      "loss": 0.8196,
      "step": 594
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.931426043356437e-05,
      "loss": 0.8318,
      "step": 596
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.929951334611415e-05,
      "loss": 0.9004,
      "step": 598
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.928476625866392e-05,
      "loss": 0.8497,
      "step": 600
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.927001917121369e-05,
      "loss": 0.8404,
      "step": 602
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.925527208376346e-05,
      "loss": 0.8338,
      "step": 604
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.924052499631323e-05,
      "loss": 0.8533,
      "step": 606
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.9225777908863e-05,
      "loss": 0.8912,
      "step": 608
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.921103082141277e-05,
      "loss": 0.8374,
      "step": 610
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.919628373396254e-05,
      "loss": 0.8796,
      "step": 612
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.918153664651232e-05,
      "loss": 0.8042,
      "step": 614
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.916678955906209e-05,
      "loss": 0.7941,
      "step": 616
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.915204247161186e-05,
      "loss": 0.933,
      "step": 618
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.913729538416163e-05,
      "loss": 0.8431,
      "step": 620
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.91225482967114e-05,
      "loss": 0.8813,
      "step": 622
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.910780120926117e-05,
      "loss": 0.8526,
      "step": 624
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.909305412181094e-05,
      "loss": 0.8631,
      "step": 626
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.907830703436072e-05,
      "loss": 0.8055,
      "step": 628
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.906355994691049e-05,
      "loss": 0.8616,
      "step": 630
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.904881285946026e-05,
      "loss": 0.9097,
      "step": 632
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.903406577201003e-05,
      "loss": 0.8544,
      "step": 634
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.90193186845598e-05,
      "loss": 0.8254,
      "step": 636
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.900457159710957e-05,
      "loss": 0.8253,
      "step": 638
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.898982450965934e-05,
      "loss": 0.7993,
      "step": 640
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.897507742220912e-05,
      "loss": 0.879,
      "step": 642
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.896033033475889e-05,
      "loss": 0.8417,
      "step": 644
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.894558324730866e-05,
      "loss": 0.8124,
      "step": 646
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.893083615985843e-05,
      "loss": 0.8684,
      "step": 648
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.89160890724082e-05,
      "loss": 0.8399,
      "step": 650
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.890134198495797e-05,
      "loss": 0.8246,
      "step": 652
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.888659489750774e-05,
      "loss": 0.8645,
      "step": 654
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.887184781005752e-05,
      "loss": 0.8278,
      "step": 656
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.885710072260729e-05,
      "loss": 0.9073,
      "step": 658
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.884235363515706e-05,
      "loss": 0.8387,
      "step": 660
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.882760654770683e-05,
      "loss": 0.8482,
      "step": 662
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.88128594602566e-05,
      "loss": 0.8576,
      "step": 664
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.879811237280637e-05,
      "loss": 0.8412,
      "step": 666
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.878336528535614e-05,
      "loss": 0.8005,
      "step": 668
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8768618197905916e-05,
      "loss": 0.8123,
      "step": 670
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.875387111045569e-05,
      "loss": 0.8396,
      "step": 672
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8739124023005464e-05,
      "loss": 0.8386,
      "step": 674
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8724376935555234e-05,
      "loss": 0.9202,
      "step": 676
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8709629848105004e-05,
      "loss": 0.7992,
      "step": 678
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8694882760654775e-05,
      "loss": 0.876,
      "step": 680
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8680135673204545e-05,
      "loss": 0.8453,
      "step": 682
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.8665388585754315e-05,
      "loss": 0.8252,
      "step": 684
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.865064149830409e-05,
      "loss": 0.8319,
      "step": 686
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.863589441085386e-05,
      "loss": 0.8373,
      "step": 688
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.862114732340363e-05,
      "loss": 0.8236,
      "step": 690
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8606400235953404e-05,
      "loss": 0.7889,
      "step": 692
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8591653148503174e-05,
      "loss": 0.8248,
      "step": 694
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8576906061052944e-05,
      "loss": 0.8454,
      "step": 696
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8562158973602715e-05,
      "loss": 0.8636,
      "step": 698
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8547411886152485e-05,
      "loss": 0.884,
      "step": 700
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.853266479870226e-05,
      "loss": 0.8469,
      "step": 702
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.851791771125203e-05,
      "loss": 0.8405,
      "step": 704
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.85031706238018e-05,
      "loss": 0.7826,
      "step": 706
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.848842353635157e-05,
      "loss": 0.8442,
      "step": 708
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8473676448901344e-05,
      "loss": 0.8202,
      "step": 710
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8458929361451114e-05,
      "loss": 0.8535,
      "step": 712
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.8444182274000884e-05,
      "loss": 0.8496,
      "step": 714
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.842943518655066e-05,
      "loss": 0.8235,
      "step": 716
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.841468809910043e-05,
      "loss": 0.8485,
      "step": 718
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.83999410116502e-05,
      "loss": 0.8674,
      "step": 720
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.838519392419997e-05,
      "loss": 0.8795,
      "step": 722
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.837044683674974e-05,
      "loss": 0.8539,
      "step": 724
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8355699749299513e-05,
      "loss": 0.8626,
      "step": 726
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8340952661849284e-05,
      "loss": 0.8122,
      "step": 728
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.832620557439906e-05,
      "loss": 0.8793,
      "step": 730
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.831145848694883e-05,
      "loss": 0.81,
      "step": 732
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.82967113994986e-05,
      "loss": 0.9482,
      "step": 734
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.828196431204837e-05,
      "loss": 0.8222,
      "step": 736
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.826721722459814e-05,
      "loss": 0.8519,
      "step": 738
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.825247013714791e-05,
      "loss": 0.8499,
      "step": 740
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.823772304969768e-05,
      "loss": 0.851,
      "step": 742
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.822297596224746e-05,
      "loss": 0.8685,
      "step": 744
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.820822887479723e-05,
      "loss": 0.8846,
      "step": 746
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.8193481787347e-05,
      "loss": 0.8518,
      "step": 748
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.817873469989677e-05,
      "loss": 0.8471,
      "step": 750
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.816398761244654e-05,
      "loss": 0.8832,
      "step": 752
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.814924052499631e-05,
      "loss": 0.8588,
      "step": 754
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.813449343754608e-05,
      "loss": 0.8577,
      "step": 756
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.811974635009586e-05,
      "loss": 0.8579,
      "step": 758
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.810499926264563e-05,
      "loss": 0.8324,
      "step": 760
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.80902521751954e-05,
      "loss": 0.8126,
      "step": 762
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.807550508774517e-05,
      "loss": 0.832,
      "step": 764
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.806075800029494e-05,
      "loss": 0.8345,
      "step": 766
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.804601091284471e-05,
      "loss": 0.8391,
      "step": 768
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.803126382539448e-05,
      "loss": 0.8035,
      "step": 770
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.801651673794426e-05,
      "loss": 0.8089,
      "step": 772
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.800176965049403e-05,
      "loss": 0.8706,
      "step": 774
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.79870225630438e-05,
      "loss": 0.8626,
      "step": 776
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.797227547559357e-05,
      "loss": 0.8638,
      "step": 778
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.795752838814334e-05,
      "loss": 0.9007,
      "step": 780
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.794278130069312e-05,
      "loss": 0.8611,
      "step": 782
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.792803421324289e-05,
      "loss": 0.8556,
      "step": 784
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.791328712579266e-05,
      "loss": 0.8462,
      "step": 786
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.7898540038342435e-05,
      "loss": 0.8515,
      "step": 788
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7883792950892206e-05,
      "loss": 0.8073,
      "step": 790
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7869045863441976e-05,
      "loss": 0.8134,
      "step": 792
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7854298775991747e-05,
      "loss": 0.851,
      "step": 794
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.783955168854152e-05,
      "loss": 0.848,
      "step": 796
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.782480460109129e-05,
      "loss": 0.8346,
      "step": 798
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.781005751364106e-05,
      "loss": 0.8662,
      "step": 800
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.779531042619083e-05,
      "loss": 0.8581,
      "step": 802
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7780563338740605e-05,
      "loss": 0.8505,
      "step": 804
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7765816251290376e-05,
      "loss": 0.8554,
      "step": 806
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7751069163840146e-05,
      "loss": 0.9062,
      "step": 808
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7736322076389916e-05,
      "loss": 0.8006,
      "step": 810
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.7721574988939687e-05,
      "loss": 0.8087,
      "step": 812
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.770682790148946e-05,
      "loss": 0.8345,
      "step": 814
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.769208081403923e-05,
      "loss": 0.8532,
      "step": 816
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7677333726589004e-05,
      "loss": 0.8842,
      "step": 818
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7662586639138775e-05,
      "loss": 0.8885,
      "step": 820
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7647839551688545e-05,
      "loss": 0.856,
      "step": 822
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7633092464238316e-05,
      "loss": 0.8625,
      "step": 824
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7618345376788086e-05,
      "loss": 0.8237,
      "step": 826
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7603598289337856e-05,
      "loss": 0.7523,
      "step": 828
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.758885120188763e-05,
      "loss": 0.847,
      "step": 830
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7574104114437404e-05,
      "loss": 0.8751,
      "step": 832
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7559357026987174e-05,
      "loss": 0.8276,
      "step": 834
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.7544609939536945e-05,
      "loss": 0.8154,
      "step": 836
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7529862852086715e-05,
      "loss": 0.8765,
      "step": 838
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7515115764636485e-05,
      "loss": 0.8684,
      "step": 840
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7500368677186256e-05,
      "loss": 0.8158,
      "step": 842
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7485621589736026e-05,
      "loss": 0.8219,
      "step": 844
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.74708745022858e-05,
      "loss": 0.8181,
      "step": 846
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7456127414835574e-05,
      "loss": 0.8315,
      "step": 848
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7441380327385344e-05,
      "loss": 0.8109,
      "step": 850
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7426633239935114e-05,
      "loss": 0.8479,
      "step": 852
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7411886152484885e-05,
      "loss": 0.8438,
      "step": 854
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7397139065034655e-05,
      "loss": 0.8652,
      "step": 856
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.7382391977584425e-05,
      "loss": 0.8379,
      "step": 858
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.73676448901342e-05,
      "loss": 0.8482,
      "step": 860
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.735289780268397e-05,
      "loss": 0.8513,
      "step": 862
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.733815071523374e-05,
      "loss": 0.8628,
      "step": 864
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.7323403627783514e-05,
      "loss": 0.8107,
      "step": 866
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.7308656540333284e-05,
      "loss": 0.8356,
      "step": 868
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.7293909452883054e-05,
      "loss": 0.8361,
      "step": 870
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.7279162365432825e-05,
      "loss": 0.8223,
      "step": 872
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.72644152779826e-05,
      "loss": 0.8262,
      "step": 874
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.724966819053237e-05,
      "loss": 0.8054,
      "step": 876
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.723492110308214e-05,
      "loss": 0.7922,
      "step": 878
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.722017401563191e-05,
      "loss": 0.8221,
      "step": 880
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.720542692818168e-05,
      "loss": 0.826,
      "step": 882
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.7190679840731454e-05,
      "loss": 0.8521,
      "step": 884
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7175932753281224e-05,
      "loss": 0.8767,
      "step": 886
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.7161185665831e-05,
      "loss": 0.8165,
      "step": 888
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.714643857838077e-05,
      "loss": 0.8441,
      "step": 890
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.713169149093054e-05,
      "loss": 0.7979,
      "step": 892
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.711694440348032e-05,
      "loss": 0.8034,
      "step": 894
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.710219731603009e-05,
      "loss": 0.8003,
      "step": 896
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.708745022857986e-05,
      "loss": 0.8737,
      "step": 898
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.707270314112963e-05,
      "loss": 0.8413,
      "step": 900
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.70579560536794e-05,
      "loss": 0.8639,
      "step": 902
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.704320896622917e-05,
      "loss": 0.841,
      "step": 904
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.702846187877895e-05,
      "loss": 0.828,
      "step": 906
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.701371479132872e-05,
      "loss": 0.8221,
      "step": 908
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.699896770387849e-05,
      "loss": 0.8361,
      "step": 910
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.698422061642826e-05,
      "loss": 0.8356,
      "step": 912
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.696947352897803e-05,
      "loss": 0.8892,
      "step": 914
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.69547264415278e-05,
      "loss": 0.8679,
      "step": 916
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.693997935407757e-05,
      "loss": 0.857,
      "step": 918
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.692523226662735e-05,
      "loss": 0.8376,
      "step": 920
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.691048517917712e-05,
      "loss": 0.7994,
      "step": 922
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.689573809172689e-05,
      "loss": 0.8265,
      "step": 924
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.688099100427666e-05,
      "loss": 0.8302,
      "step": 926
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.686624391682643e-05,
      "loss": 0.8605,
      "step": 928
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.68514968293762e-05,
      "loss": 0.8348,
      "step": 930
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.683674974192597e-05,
      "loss": 0.9294,
      "step": 932
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.682200265447575e-05,
      "loss": 0.8049,
      "step": 934
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.680725556702552e-05,
      "loss": 0.8367,
      "step": 936
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.679250847957529e-05,
      "loss": 0.8214,
      "step": 938
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.677776139212506e-05,
      "loss": 0.8354,
      "step": 940
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.676301430467483e-05,
      "loss": 0.7735,
      "step": 942
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.67482672172246e-05,
      "loss": 0.8547,
      "step": 944
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.673352012977437e-05,
      "loss": 0.9028,
      "step": 946
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.6718773042324146e-05,
      "loss": 0.8421,
      "step": 948
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.6704025954873916e-05,
      "loss": 0.7982,
      "step": 950
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.668927886742369e-05,
      "loss": 0.8219,
      "step": 952
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.667453177997346e-05,
      "loss": 0.7998,
      "step": 954
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.665978469252323e-05,
      "loss": 0.8252,
      "step": 956
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.6645037605073e-05,
      "loss": 0.8232,
      "step": 958
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.663029051762277e-05,
      "loss": 0.8422,
      "step": 960
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6615543430172545e-05,
      "loss": 0.8727,
      "step": 962
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6600796342722316e-05,
      "loss": 0.8578,
      "step": 964
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6586049255272086e-05,
      "loss": 0.837,
      "step": 966
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6571302167821856e-05,
      "loss": 0.9421,
      "step": 968
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.655655508037163e-05,
      "loss": 0.8091,
      "step": 970
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.65418079929214e-05,
      "loss": 0.8594,
      "step": 972
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.652706090547117e-05,
      "loss": 0.8507,
      "step": 974
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6512313818020945e-05,
      "loss": 0.8609,
      "step": 976
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6497566730570715e-05,
      "loss": 0.8813,
      "step": 978
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6482819643120485e-05,
      "loss": 0.8057,
      "step": 980
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.6468072555670256e-05,
      "loss": 0.8062,
      "step": 982
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6453325468220026e-05,
      "loss": 0.8508,
      "step": 984
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6438578380769797e-05,
      "loss": 0.8332,
      "step": 986
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.642383129331957e-05,
      "loss": 0.8184,
      "step": 988
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6409084205869344e-05,
      "loss": 0.8128,
      "step": 990
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6394337118419114e-05,
      "loss": 0.8545,
      "step": 992
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6379590030968885e-05,
      "loss": 0.8364,
      "step": 994
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6364842943518655e-05,
      "loss": 0.8842,
      "step": 996
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6350095856068426e-05,
      "loss": 0.8796,
      "step": 998
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6335348768618196e-05,
      "loss": 0.8242,
      "step": 1000
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6320601681167966e-05,
      "loss": 0.8147,
      "step": 1002
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6305854593717743e-05,
      "loss": 0.8647,
      "step": 1004
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.6291107506267514e-05,
      "loss": 0.8221,
      "step": 1006
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.627636041881729e-05,
      "loss": 0.846,
      "step": 1008
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.626161333136706e-05,
      "loss": 0.8636,
      "step": 1010
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.624686624391683e-05,
      "loss": 0.8473,
      "step": 1012
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.62321191564666e-05,
      "loss": 0.8087,
      "step": 1014
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.621737206901637e-05,
      "loss": 0.8669,
      "step": 1016
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.620262498156614e-05,
      "loss": 0.8327,
      "step": 1018
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.618787789411591e-05,
      "loss": 0.8566,
      "step": 1020
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.617313080666569e-05,
      "loss": 0.808,
      "step": 1022
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.615838371921546e-05,
      "loss": 0.8282,
      "step": 1024
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.614363663176523e-05,
      "loss": 0.8135,
      "step": 1026
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.6128889544315e-05,
      "loss": 0.8184,
      "step": 1028
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.611414245686477e-05,
      "loss": 0.8127,
      "step": 1030
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.609939536941454e-05,
      "loss": 0.8339,
      "step": 1032
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.608464828196431e-05,
      "loss": 0.8238,
      "step": 1034
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.606990119451409e-05,
      "loss": 0.8214,
      "step": 1036
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.605515410706386e-05,
      "loss": 0.8232,
      "step": 1038
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.604040701961363e-05,
      "loss": 0.7824,
      "step": 1040
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.60256599321634e-05,
      "loss": 0.8186,
      "step": 1042
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.601091284471317e-05,
      "loss": 0.8536,
      "step": 1044
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.599616575726294e-05,
      "loss": 0.8062,
      "step": 1046
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.598141866981271e-05,
      "loss": 0.8714,
      "step": 1048
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.596667158236249e-05,
      "loss": 0.8435,
      "step": 1050
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.595192449491226e-05,
      "loss": 0.8228,
      "step": 1052
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.593717740746203e-05,
      "loss": 0.7986,
      "step": 1054
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.59224303200118e-05,
      "loss": 0.8247,
      "step": 1056
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.590768323256157e-05,
      "loss": 0.8276,
      "step": 1058
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.589293614511134e-05,
      "loss": 0.8052,
      "step": 1060
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.587818905766111e-05,
      "loss": 0.8156,
      "step": 1062
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.586344197021089e-05,
      "loss": 0.8256,
      "step": 1064
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.584869488276066e-05,
      "loss": 0.8305,
      "step": 1066
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.583394779531043e-05,
      "loss": 0.7752,
      "step": 1068
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.58192007078602e-05,
      "loss": 0.8572,
      "step": 1070
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.580445362040997e-05,
      "loss": 0.8221,
      "step": 1072
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.578970653295974e-05,
      "loss": 0.8453,
      "step": 1074
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.577495944550951e-05,
      "loss": 0.8213,
      "step": 1076
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.576021235805929e-05,
      "loss": 0.8088,
      "step": 1078
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.574546527060906e-05,
      "loss": 0.8339,
      "step": 1080
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.573071818315883e-05,
      "loss": 0.8472,
      "step": 1082
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.57159710957086e-05,
      "loss": 0.8698,
      "step": 1084
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.570122400825837e-05,
      "loss": 0.7912,
      "step": 1086
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.568647692080814e-05,
      "loss": 0.8813,
      "step": 1088
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.567172983335791e-05,
      "loss": 0.7854,
      "step": 1090
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.565698274590769e-05,
      "loss": 0.842,
      "step": 1092
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.564223565845746e-05,
      "loss": 0.8375,
      "step": 1094
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.562748857100723e-05,
      "loss": 0.8403,
      "step": 1096
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.5612741483557e-05,
      "loss": 0.8051,
      "step": 1098
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.559799439610677e-05,
      "loss": 0.9037,
      "step": 1100
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.558324730865654e-05,
      "loss": 0.8582,
      "step": 1102
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.556850022120631e-05,
      "loss": 0.8324,
      "step": 1104
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5553753133756086e-05,
      "loss": 0.84,
      "step": 1106
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.553900604630586e-05,
      "loss": 0.8334,
      "step": 1108
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.552425895885563e-05,
      "loss": 0.8086,
      "step": 1110
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.55095118714054e-05,
      "loss": 0.8272,
      "step": 1112
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.549476478395517e-05,
      "loss": 0.8429,
      "step": 1114
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5480017696504945e-05,
      "loss": 0.7907,
      "step": 1116
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5465270609054715e-05,
      "loss": 0.9171,
      "step": 1118
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5450523521604486e-05,
      "loss": 0.8435,
      "step": 1120
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5435776434154256e-05,
      "loss": 0.8191,
      "step": 1122
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.542102934670403e-05,
      "loss": 0.8398,
      "step": 1124
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5406282259253803e-05,
      "loss": 0.8147,
      "step": 1126
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5391535171803574e-05,
      "loss": 0.8307,
      "step": 1128
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5376788084353344e-05,
      "loss": 0.7734,
      "step": 1130
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5362040996903115e-05,
      "loss": 0.8031,
      "step": 1132
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5347293909452885e-05,
      "loss": 0.8328,
      "step": 1134
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5332546822002655e-05,
      "loss": 0.8652,
      "step": 1136
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.531779973455243e-05,
      "loss": 0.88,
      "step": 1138
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.53030526471022e-05,
      "loss": 0.8037,
      "step": 1140
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.528830555965197e-05,
      "loss": 0.8142,
      "step": 1142
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5273558472201744e-05,
      "loss": 0.8702,
      "step": 1144
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5258811384751514e-05,
      "loss": 0.8482,
      "step": 1146
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5244064297301284e-05,
      "loss": 0.7982,
      "step": 1148
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.5229317209851055e-05,
      "loss": 0.807,
      "step": 1150
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.521457012240083e-05,
      "loss": 0.834,
      "step": 1152
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.51998230349506e-05,
      "loss": 0.7921,
      "step": 1154
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.518507594750037e-05,
      "loss": 0.8493,
      "step": 1156
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.517032886005014e-05,
      "loss": 0.8312,
      "step": 1158
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.515558177259991e-05,
      "loss": 0.8723,
      "step": 1160
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.5140834685149684e-05,
      "loss": 0.8187,
      "step": 1162
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.5126087597699454e-05,
      "loss": 0.8091,
      "step": 1164
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.511134051024923e-05,
      "loss": 0.8081,
      "step": 1166
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.5096593422799e-05,
      "loss": 0.8407,
      "step": 1168
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.508184633534877e-05,
      "loss": 0.8222,
      "step": 1170
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.506709924789854e-05,
      "loss": 0.7976,
      "step": 1172
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.505235216044831e-05,
      "loss": 0.8155,
      "step": 1174
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.503760507299808e-05,
      "loss": 0.8629,
      "step": 1176
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.502285798554785e-05,
      "loss": 0.8829,
      "step": 1178
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.500811089809763e-05,
      "loss": 0.8536,
      "step": 1180
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.49933638106474e-05,
      "loss": 0.8669,
      "step": 1182
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.497861672319717e-05,
      "loss": 0.8298,
      "step": 1184
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.496386963574694e-05,
      "loss": 0.8221,
      "step": 1186
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.494912254829671e-05,
      "loss": 0.8372,
      "step": 1188
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.493437546084648e-05,
      "loss": 0.8382,
      "step": 1190
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.491962837339625e-05,
      "loss": 0.8355,
      "step": 1192
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.490488128594603e-05,
      "loss": 0.8213,
      "step": 1194
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.48901341984958e-05,
      "loss": 0.8561,
      "step": 1196
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.487538711104557e-05,
      "loss": 0.7984,
      "step": 1198
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.486064002359534e-05,
      "loss": 0.826,
      "step": 1200
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.484589293614511e-05,
      "loss": 0.8236,
      "step": 1202
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.483114584869488e-05,
      "loss": 0.8138,
      "step": 1204
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.481639876124465e-05,
      "loss": 0.8331,
      "step": 1206
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.480165167379443e-05,
      "loss": 0.873,
      "step": 1208
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.47869045863442e-05,
      "loss": 0.8731,
      "step": 1210
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.477215749889397e-05,
      "loss": 0.8127,
      "step": 1212
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.475741041144374e-05,
      "loss": 0.8225,
      "step": 1214
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.474266332399351e-05,
      "loss": 0.8282,
      "step": 1216
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.472791623654328e-05,
      "loss": 0.8676,
      "step": 1218
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.471316914909305e-05,
      "loss": 0.822,
      "step": 1220
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.469842206164282e-05,
      "loss": 0.8263,
      "step": 1222
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.46836749741926e-05,
      "loss": 0.8309,
      "step": 1224
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4668927886742376e-05,
      "loss": 0.8127,
      "step": 1226
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4654180799292146e-05,
      "loss": 0.8313,
      "step": 1228
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.463943371184192e-05,
      "loss": 0.8109,
      "step": 1230
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.462468662439169e-05,
      "loss": 0.8252,
      "step": 1232
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.460993953694146e-05,
      "loss": 0.8146,
      "step": 1234
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.459519244949123e-05,
      "loss": 0.8171,
      "step": 1236
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4580445362041e-05,
      "loss": 0.8425,
      "step": 1238
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4565698274590775e-05,
      "loss": 0.8072,
      "step": 1240
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4550951187140546e-05,
      "loss": 0.8385,
      "step": 1242
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4536204099690316e-05,
      "loss": 0.8169,
      "step": 1244
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.4521457012240086e-05,
      "loss": 0.8715,
      "step": 1246
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.450670992478986e-05,
      "loss": 0.8152,
      "step": 1248
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.449196283733963e-05,
      "loss": 0.7933,
      "step": 1250
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.44772157498894e-05,
      "loss": 0.8323,
      "step": 1252
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4462468662439175e-05,
      "loss": 0.8309,
      "step": 1254
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4447721574988945e-05,
      "loss": 0.8761,
      "step": 1256
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4432974487538715e-05,
      "loss": 0.8615,
      "step": 1258
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4418227400088486e-05,
      "loss": 0.8257,
      "step": 1260
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4403480312638256e-05,
      "loss": 0.8479,
      "step": 1262
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4388733225188027e-05,
      "loss": 0.8621,
      "step": 1264
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.43739861377378e-05,
      "loss": 0.8419,
      "step": 1266
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4359239050287574e-05,
      "loss": 0.8843,
      "step": 1268
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4344491962837344e-05,
      "loss": 0.8091,
      "step": 1270
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4329744875387115e-05,
      "loss": 0.8366,
      "step": 1272
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.4314997787936885e-05,
      "loss": 0.8674,
      "step": 1274
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4300250700486655e-05,
      "loss": 0.7962,
      "step": 1276
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4285503613036426e-05,
      "loss": 0.8376,
      "step": 1278
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4270756525586196e-05,
      "loss": 0.807,
      "step": 1280
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.425600943813597e-05,
      "loss": 0.8511,
      "step": 1282
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4241262350685744e-05,
      "loss": 0.8021,
      "step": 1284
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4226515263235514e-05,
      "loss": 0.8676,
      "step": 1286
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4211768175785284e-05,
      "loss": 0.826,
      "step": 1288
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4197021088335055e-05,
      "loss": 0.7824,
      "step": 1290
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4182274000884825e-05,
      "loss": 0.8464,
      "step": 1292
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.4167526913434596e-05,
      "loss": 0.8568,
      "step": 1294
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.415277982598437e-05,
      "loss": 0.8431,
      "step": 1296
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.413803273853414e-05,
      "loss": 0.8588,
      "step": 1298
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.4123285651083913e-05,
      "loss": 0.7948,
      "step": 1300
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.4108538563633684e-05,
      "loss": 0.8336,
      "step": 1302
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.4093791476183454e-05,
      "loss": 0.8257,
      "step": 1304
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.4079044388733225e-05,
      "loss": 0.8376,
      "step": 1306
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.4064297301282995e-05,
      "loss": 0.8444,
      "step": 1308
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.404955021383277e-05,
      "loss": 0.787,
      "step": 1310
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.403480312638254e-05,
      "loss": 0.8364,
      "step": 1312
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.402005603893231e-05,
      "loss": 0.8233,
      "step": 1314
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.400530895148208e-05,
      "loss": 0.8191,
      "step": 1316
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.3990561864031854e-05,
      "loss": 0.8113,
      "step": 1318
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.3975814776581624e-05,
      "loss": 0.7833,
      "step": 1320
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.3961067689131394e-05,
      "loss": 0.7891,
      "step": 1322
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.3946320601681165e-05,
      "loss": 0.8593,
      "step": 1324
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.393157351423094e-05,
      "loss": 0.8615,
      "step": 1326
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.391682642678071e-05,
      "loss": 0.772,
      "step": 1328
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.390207933933048e-05,
      "loss": 0.7948,
      "step": 1330
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.388733225188025e-05,
      "loss": 0.8139,
      "step": 1332
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.387258516443002e-05,
      "loss": 0.8183,
      "step": 1334
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.3857838076979794e-05,
      "loss": 0.8156,
      "step": 1336
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.384309098952957e-05,
      "loss": 0.8836,
      "step": 1338
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.382834390207934e-05,
      "loss": 0.8565,
      "step": 1340
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.381359681462912e-05,
      "loss": 0.8458,
      "step": 1342
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.379884972717889e-05,
      "loss": 0.8234,
      "step": 1344
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.378410263972866e-05,
      "loss": 0.8144,
      "step": 1346
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.376935555227843e-05,
      "loss": 0.852,
      "step": 1348
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.37546084648282e-05,
      "loss": 0.8161,
      "step": 1350
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.373986137737797e-05,
      "loss": 0.838,
      "step": 1352
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.372511428992774e-05,
      "loss": 0.8771,
      "step": 1354
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.371036720247752e-05,
      "loss": 0.8506,
      "step": 1356
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.369562011502729e-05,
      "loss": 0.8154,
      "step": 1358
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.368087302757706e-05,
      "loss": 0.8098,
      "step": 1360
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.366612594012683e-05,
      "loss": 0.8262,
      "step": 1362
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.36513788526766e-05,
      "loss": 0.7663,
      "step": 1364
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.363663176522637e-05,
      "loss": 0.8635,
      "step": 1366
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.362188467777614e-05,
      "loss": 0.8576,
      "step": 1368
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.360713759032592e-05,
      "loss": 0.7925,
      "step": 1370
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.359239050287569e-05,
      "loss": 0.7896,
      "step": 1372
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.357764341542546e-05,
      "loss": 0.8537,
      "step": 1374
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.356289632797523e-05,
      "loss": 0.8585,
      "step": 1376
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.3548149240525e-05,
      "loss": 0.8166,
      "step": 1378
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.353340215307477e-05,
      "loss": 0.9279,
      "step": 1380
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.351865506562454e-05,
      "loss": 0.8345,
      "step": 1382
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.3503907978174316e-05,
      "loss": 0.8519,
      "step": 1384
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.348916089072409e-05,
      "loss": 0.8501,
      "step": 1386
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.347441380327386e-05,
      "loss": 0.8254,
      "step": 1388
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.345966671582363e-05,
      "loss": 0.8241,
      "step": 1390
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.34449196283734e-05,
      "loss": 0.7967,
      "step": 1392
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.343017254092317e-05,
      "loss": 0.8526,
      "step": 1394
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.341542545347294e-05,
      "loss": 0.8325,
      "step": 1396
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.3400678366022716e-05,
      "loss": 0.8454,
      "step": 1398
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.3385931278572486e-05,
      "loss": 0.7675,
      "step": 1400
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.3371184191122256e-05,
      "loss": 0.8005,
      "step": 1402
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.335643710367203e-05,
      "loss": 0.806,
      "step": 1404
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.33416900162218e-05,
      "loss": 0.7897,
      "step": 1406
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.332694292877157e-05,
      "loss": 0.8342,
      "step": 1408
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.331219584132134e-05,
      "loss": 0.8208,
      "step": 1410
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.329744875387111e-05,
      "loss": 0.8533,
      "step": 1412
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.3282701666420885e-05,
      "loss": 0.802,
      "step": 1414
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.3267954578970656e-05,
      "loss": 0.8007,
      "step": 1416
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.3253207491520426e-05,
      "loss": 0.8142,
      "step": 1418
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3238460404070196e-05,
      "loss": 0.8091,
      "step": 1420
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.322371331661997e-05,
      "loss": 0.8296,
      "step": 1422
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.320896622916974e-05,
      "loss": 0.8258,
      "step": 1424
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.319421914171951e-05,
      "loss": 0.841,
      "step": 1426
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3179472054269285e-05,
      "loss": 0.8215,
      "step": 1428
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3164724966819055e-05,
      "loss": 0.8278,
      "step": 1430
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3149977879368825e-05,
      "loss": 0.85,
      "step": 1432
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3135230791918596e-05,
      "loss": 0.8618,
      "step": 1434
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3120483704468366e-05,
      "loss": 0.788,
      "step": 1436
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3105736617018136e-05,
      "loss": 0.793,
      "step": 1438
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.309098952956791e-05,
      "loss": 0.8573,
      "step": 1440
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3076242442117684e-05,
      "loss": 0.8381,
      "step": 1442
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.3061495354667454e-05,
      "loss": 0.8146,
      "step": 1444
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.3046748267217225e-05,
      "loss": 0.7729,
      "step": 1446
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.3032001179767e-05,
      "loss": 0.8063,
      "step": 1448
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.301725409231677e-05,
      "loss": 0.8212,
      "step": 1450
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.300250700486654e-05,
      "loss": 0.7833,
      "step": 1452
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.298775991741631e-05,
      "loss": 0.8,
      "step": 1454
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.297301282996608e-05,
      "loss": 0.8289,
      "step": 1456
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.295826574251586e-05,
      "loss": 0.8185,
      "step": 1458
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.294351865506563e-05,
      "loss": 0.7993,
      "step": 1460
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.29287715676154e-05,
      "loss": 0.8207,
      "step": 1462
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.291402448016517e-05,
      "loss": 0.8709,
      "step": 1464
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.289927739271494e-05,
      "loss": 0.7985,
      "step": 1466
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.288453030526471e-05,
      "loss": 0.8283,
      "step": 1468
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.286978321781448e-05,
      "loss": 0.8379,
      "step": 1470
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.285503613036426e-05,
      "loss": 0.8803,
      "step": 1472
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.284028904291403e-05,
      "loss": 0.8251,
      "step": 1474
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.28255419554638e-05,
      "loss": 0.8228,
      "step": 1476
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.281079486801357e-05,
      "loss": 0.7585,
      "step": 1478
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.279604778056334e-05,
      "loss": 0.8203,
      "step": 1480
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.278130069311311e-05,
      "loss": 0.8617,
      "step": 1482
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.276655360566288e-05,
      "loss": 0.8306,
      "step": 1484
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.275180651821266e-05,
      "loss": 0.827,
      "step": 1486
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.273705943076243e-05,
      "loss": 0.8528,
      "step": 1488
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.27223123433122e-05,
      "loss": 0.8362,
      "step": 1490
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.270756525586197e-05,
      "loss": 0.8442,
      "step": 1492
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.269281816841174e-05,
      "loss": 0.8021,
      "step": 1494
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.267807108096151e-05,
      "loss": 0.8326,
      "step": 1496
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.266332399351128e-05,
      "loss": 0.8301,
      "step": 1498
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.264857690606106e-05,
      "loss": 0.8409,
      "step": 1500
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.263382981861083e-05,
      "loss": 0.7758,
      "step": 1502
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.26190827311606e-05,
      "loss": 0.8291,
      "step": 1504
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.260433564371037e-05,
      "loss": 0.8028,
      "step": 1506
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.258958855626014e-05,
      "loss": 0.8122,
      "step": 1508
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.257484146880991e-05,
      "loss": 0.8234,
      "step": 1510
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.256009438135968e-05,
      "loss": 0.8213,
      "step": 1512
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.254534729390945e-05,
      "loss": 0.8344,
      "step": 1514
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.253060020645923e-05,
      "loss": 0.8247,
      "step": 1516
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.2515853119009e-05,
      "loss": 0.8342,
      "step": 1518
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.250110603155877e-05,
      "loss": 0.8142,
      "step": 1520
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.248635894410854e-05,
      "loss": 0.785,
      "step": 1522
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.247161185665831e-05,
      "loss": 0.7957,
      "step": 1524
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.245686476920808e-05,
      "loss": 0.8078,
      "step": 1526
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.244211768175785e-05,
      "loss": 0.7706,
      "step": 1528
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.242737059430763e-05,
      "loss": 0.7852,
      "step": 1530
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.24126235068574e-05,
      "loss": 0.8091,
      "step": 1532
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.239787641940717e-05,
      "loss": 0.7481,
      "step": 1534
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.238312933195694e-05,
      "loss": 0.7796,
      "step": 1536
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.236838224450671e-05,
      "loss": 0.7684,
      "step": 1538
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.235363515705648e-05,
      "loss": 0.801,
      "step": 1540
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.233888806960625e-05,
      "loss": 0.7736,
      "step": 1542
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.232414098215603e-05,
      "loss": 0.8029,
      "step": 1544
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.23093938947058e-05,
      "loss": 0.8057,
      "step": 1546
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.229464680725557e-05,
      "loss": 0.7621,
      "step": 1548
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.227989971980534e-05,
      "loss": 0.8406,
      "step": 1550
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.226515263235511e-05,
      "loss": 0.7833,
      "step": 1552
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.225040554490488e-05,
      "loss": 0.7774,
      "step": 1554
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.223565845745465e-05,
      "loss": 0.733,
      "step": 1556
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.2220911370004426e-05,
      "loss": 0.8172,
      "step": 1558
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.22061642825542e-05,
      "loss": 0.7769,
      "step": 1560
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.2191417195103974e-05,
      "loss": 0.7938,
      "step": 1562
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.2176670107653744e-05,
      "loss": 0.8023,
      "step": 1564
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2161923020203514e-05,
      "loss": 0.7959,
      "step": 1566
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2147175932753285e-05,
      "loss": 0.8241,
      "step": 1568
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2132428845303055e-05,
      "loss": 0.7865,
      "step": 1570
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2117681757852826e-05,
      "loss": 0.715,
      "step": 1572
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.21029346704026e-05,
      "loss": 0.7328,
      "step": 1574
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.208818758295237e-05,
      "loss": 0.8034,
      "step": 1576
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2073440495502143e-05,
      "loss": 0.7578,
      "step": 1578
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2058693408051914e-05,
      "loss": 0.8372,
      "step": 1580
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2043946320601684e-05,
      "loss": 0.7244,
      "step": 1582
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2029199233151455e-05,
      "loss": 0.7408,
      "step": 1584
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.2014452145701225e-05,
      "loss": 0.7625,
      "step": 1586
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.1999705058251e-05,
      "loss": 0.7836,
      "step": 1588
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.198495797080077e-05,
      "loss": 0.748,
      "step": 1590
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.197021088335054e-05,
      "loss": 0.7324,
      "step": 1592
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.195546379590031e-05,
      "loss": 0.7164,
      "step": 1594
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.1940716708450083e-05,
      "loss": 0.7824,
      "step": 1596
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.1925969620999854e-05,
      "loss": 0.7625,
      "step": 1598
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.1911222533549624e-05,
      "loss": 0.787,
      "step": 1600
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.18964754460994e-05,
      "loss": 0.761,
      "step": 1602
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.188172835864917e-05,
      "loss": 0.7429,
      "step": 1604
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.186698127119894e-05,
      "loss": 0.7039,
      "step": 1606
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.185223418374871e-05,
      "loss": 0.7159,
      "step": 1608
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.183748709629848e-05,
      "loss": 0.7142,
      "step": 1610
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.182274000884825e-05,
      "loss": 0.74,
      "step": 1612
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.1807992921398024e-05,
      "loss": 0.6937,
      "step": 1614
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.1793245833947794e-05,
      "loss": 0.7695,
      "step": 1616
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.177849874649757e-05,
      "loss": 0.7314,
      "step": 1618
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.176375165904734e-05,
      "loss": 0.6929,
      "step": 1620
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.174900457159711e-05,
      "loss": 0.7062,
      "step": 1622
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.173425748414688e-05,
      "loss": 0.6904,
      "step": 1624
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.171951039669665e-05,
      "loss": 0.6807,
      "step": 1626
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.170476330924642e-05,
      "loss": 0.675,
      "step": 1628
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.169001622179619e-05,
      "loss": 0.6715,
      "step": 1630
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.167526913434597e-05,
      "loss": 0.6523,
      "step": 1632
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.166052204689574e-05,
      "loss": 0.7049,
      "step": 1634
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.164577495944551e-05,
      "loss": 0.6815,
      "step": 1636
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.163102787199528e-05,
      "loss": 0.6788,
      "step": 1638
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.161628078454505e-05,
      "loss": 0.7023,
      "step": 1640
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.160153369709482e-05,
      "loss": 0.6764,
      "step": 1642
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.158678660964459e-05,
      "loss": 0.6692,
      "step": 1644
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.157203952219437e-05,
      "loss": 0.6781,
      "step": 1646
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.155729243474414e-05,
      "loss": 0.6691,
      "step": 1648
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.154254534729391e-05,
      "loss": 0.685,
      "step": 1650
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.152779825984368e-05,
      "loss": 0.8321,
      "step": 1652
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.151305117239345e-05,
      "loss": 0.6421,
      "step": 1654
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.149830408494322e-05,
      "loss": 0.7624,
      "step": 1656
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.148355699749299e-05,
      "loss": 0.7034,
      "step": 1658
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.146880991004277e-05,
      "loss": 0.6854,
      "step": 1660
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.145406282259254e-05,
      "loss": 0.6872,
      "step": 1662
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.143931573514231e-05,
      "loss": 0.6933,
      "step": 1664
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.142456864769208e-05,
      "loss": 0.6613,
      "step": 1666
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.140982156024185e-05,
      "loss": 0.6774,
      "step": 1668
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.139507447279162e-05,
      "loss": 0.6366,
      "step": 1670
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.13803273853414e-05,
      "loss": 0.6133,
      "step": 1672
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.136558029789117e-05,
      "loss": 0.6792,
      "step": 1674
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.1350833210440946e-05,
      "loss": 0.6475,
      "step": 1676
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.1336086122990716e-05,
      "loss": 0.6826,
      "step": 1678
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.1321339035540486e-05,
      "loss": 0.6444,
      "step": 1680
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.130659194809026e-05,
      "loss": 0.671,
      "step": 1682
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.129184486064003e-05,
      "loss": 0.7377,
      "step": 1684
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.12770977731898e-05,
      "loss": 0.6623,
      "step": 1686
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.126235068573957e-05,
      "loss": 0.6232,
      "step": 1688
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1247603598289345e-05,
      "loss": 0.6682,
      "step": 1690
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1232856510839115e-05,
      "loss": 0.6431,
      "step": 1692
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1218109423388886e-05,
      "loss": 0.5756,
      "step": 1694
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1203362335938656e-05,
      "loss": 0.681,
      "step": 1696
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1188615248488426e-05,
      "loss": 0.6087,
      "step": 1698
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.11738681610382e-05,
      "loss": 0.6113,
      "step": 1700
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.115912107358797e-05,
      "loss": 0.626,
      "step": 1702
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.114437398613774e-05,
      "loss": 0.6438,
      "step": 1704
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1129626898687515e-05,
      "loss": 0.7078,
      "step": 1706
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1114879811237285e-05,
      "loss": 0.6205,
      "step": 1708
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1100132723787055e-05,
      "loss": 0.6851,
      "step": 1710
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1085385636336826e-05,
      "loss": 0.6878,
      "step": 1712
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1070638548886596e-05,
      "loss": 0.659,
      "step": 1714
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1055891461436366e-05,
      "loss": 0.6709,
      "step": 1716
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.104114437398614e-05,
      "loss": 0.6479,
      "step": 1718
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1026397286535914e-05,
      "loss": 0.6662,
      "step": 1720
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1011650199085684e-05,
      "loss": 0.6182,
      "step": 1722
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0996903111635455e-05,
      "loss": 0.6007,
      "step": 1724
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0982156024185225e-05,
      "loss": 0.6237,
      "step": 1726
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0967408936734995e-05,
      "loss": 0.6529,
      "step": 1728
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0952661849284766e-05,
      "loss": 0.6719,
      "step": 1730
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.0937914761834536e-05,
      "loss": 0.6873,
      "step": 1732
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.092316767438431e-05,
      "loss": 0.6082,
      "step": 1734
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0908420586934084e-05,
      "loss": 0.6193,
      "step": 1736
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0893673499483854e-05,
      "loss": 0.6521,
      "step": 1738
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0878926412033624e-05,
      "loss": 0.5972,
      "step": 1740
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0864179324583395e-05,
      "loss": 0.6502,
      "step": 1742
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0849432237133165e-05,
      "loss": 0.6213,
      "step": 1744
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0834685149682935e-05,
      "loss": 0.6415,
      "step": 1746
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.081993806223271e-05,
      "loss": 0.6002,
      "step": 1748
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.080519097478248e-05,
      "loss": 0.6413,
      "step": 1750
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.079044388733225e-05,
      "loss": 0.6248,
      "step": 1752
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0775696799882024e-05,
      "loss": 0.6517,
      "step": 1754
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0760949712431794e-05,
      "loss": 0.5927,
      "step": 1756
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.0746202624981564e-05,
      "loss": 0.585,
      "step": 1758
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.0731455537531335e-05,
      "loss": 0.6427,
      "step": 1760
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.071670845008111e-05,
      "loss": 0.6499,
      "step": 1762
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.070196136263088e-05,
      "loss": 0.6628,
      "step": 1764
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.068721427518065e-05,
      "loss": 0.5609,
      "step": 1766
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.067246718773042e-05,
      "loss": 0.5767,
      "step": 1768
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.0657720100280193e-05,
      "loss": 0.6732,
      "step": 1770
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.0642973012829964e-05,
      "loss": 0.6323,
      "step": 1772
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.0628225925379734e-05,
      "loss": 0.612,
      "step": 1774
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.061347883792951e-05,
      "loss": 0.5862,
      "step": 1776
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.059873175047928e-05,
      "loss": 0.5813,
      "step": 1778
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.058398466302905e-05,
      "loss": 0.6015,
      "step": 1780
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.056923757557883e-05,
      "loss": 0.5844,
      "step": 1782
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.05544904881286e-05,
      "loss": 0.64,
      "step": 1784
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.053974340067837e-05,
      "loss": 0.5815,
      "step": 1786
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.052499631322814e-05,
      "loss": 0.6386,
      "step": 1788
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.051024922577791e-05,
      "loss": 0.5999,
      "step": 1790
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.049550213832769e-05,
      "loss": 0.6016,
      "step": 1792
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.048075505087746e-05,
      "loss": 0.6203,
      "step": 1794
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.046600796342723e-05,
      "loss": 0.6047,
      "step": 1796
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.0451260875977e-05,
      "loss": 0.6276,
      "step": 1798
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.043651378852677e-05,
      "loss": 0.6086,
      "step": 1800
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.042176670107654e-05,
      "loss": 0.6421,
      "step": 1802
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.040701961362631e-05,
      "loss": 0.6232,
      "step": 1804
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.039227252617608e-05,
      "loss": 0.5741,
      "step": 1806
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.037752543872586e-05,
      "loss": 0.6024,
      "step": 1808
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.036277835127563e-05,
      "loss": 0.5927,
      "step": 1810
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.03480312638254e-05,
      "loss": 0.5929,
      "step": 1812
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.033328417637517e-05,
      "loss": 0.6248,
      "step": 1814
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.031853708892494e-05,
      "loss": 0.5759,
      "step": 1816
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.030379000147471e-05,
      "loss": 0.5706,
      "step": 1818
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.028904291402448e-05,
      "loss": 0.6462,
      "step": 1820
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.027429582657426e-05,
      "loss": 0.6058,
      "step": 1822
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.025954873912403e-05,
      "loss": 0.5881,
      "step": 1824
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.02448016516738e-05,
      "loss": 0.6122,
      "step": 1826
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.023005456422357e-05,
      "loss": 0.5945,
      "step": 1828
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.021530747677334e-05,
      "loss": 0.597,
      "step": 1830
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.020056038932311e-05,
      "loss": 0.5553,
      "step": 1832
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.018581330187288e-05,
      "loss": 0.6057,
      "step": 1834
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.0171066214422656e-05,
      "loss": 0.575,
      "step": 1836
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.0156319126972427e-05,
      "loss": 0.6124,
      "step": 1838
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.01415720395222e-05,
      "loss": 0.6177,
      "step": 1840
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.012682495207197e-05,
      "loss": 0.5781,
      "step": 1842
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.011207786462174e-05,
      "loss": 0.6067,
      "step": 1844
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.009733077717151e-05,
      "loss": 0.559,
      "step": 1846
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.008258368972128e-05,
      "loss": 0.6018,
      "step": 1848
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.0067836602271056e-05,
      "loss": 0.6155,
      "step": 1850
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.0053089514820826e-05,
      "loss": 0.6113,
      "step": 1852
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.0038342427370596e-05,
      "loss": 0.5407,
      "step": 1854
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.002359533992037e-05,
      "loss": 0.5407,
      "step": 1856
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.000884825247014e-05,
      "loss": 0.5448,
      "step": 1858
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.999410116501991e-05,
      "loss": 0.569,
      "step": 1860
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.997935407756968e-05,
      "loss": 0.572,
      "step": 1862
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9964606990119455e-05,
      "loss": 0.5465,
      "step": 1864
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9949859902669225e-05,
      "loss": 0.6186,
      "step": 1866
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9935112815218996e-05,
      "loss": 0.5391,
      "step": 1868
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9920365727768766e-05,
      "loss": 0.5822,
      "step": 1870
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9905618640318536e-05,
      "loss": 0.5548,
      "step": 1872
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.989087155286831e-05,
      "loss": 0.6418,
      "step": 1874
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.987612446541808e-05,
      "loss": 0.5635,
      "step": 1876
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9861377377967854e-05,
      "loss": 0.5891,
      "step": 1878
    },
    {
      "epoch": 0.77,
      "learning_rate": 3.9846630290517625e-05,
      "loss": 0.5836,
      "step": 1880
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.9831883203067395e-05,
      "loss": 0.5769,
      "step": 1882
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.9817136115617165e-05,
      "loss": 0.5997,
      "step": 1884
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.9802389028166936e-05,
      "loss": 0.5533,
      "step": 1886
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.9787641940716706e-05,
      "loss": 0.5398,
      "step": 1888
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.9772894853266476e-05,
      "loss": 0.5413,
      "step": 1890
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.9758147765816254e-05,
      "loss": 0.523,
      "step": 1892
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.974340067836603e-05,
      "loss": 0.5979,
      "step": 1894
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.97286535909158e-05,
      "loss": 0.5861,
      "step": 1896
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.971390650346557e-05,
      "loss": 0.5185,
      "step": 1898
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.969915941601534e-05,
      "loss": 0.6024,
      "step": 1900
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.968441232856511e-05,
      "loss": 0.5562,
      "step": 1902
    },
    {
      "epoch": 0.78,
      "learning_rate": 3.966966524111488e-05,
      "loss": 0.5882,
      "step": 1904
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.965491815366465e-05,
      "loss": 0.5644,
      "step": 1906
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.964017106621442e-05,
      "loss": 0.6252,
      "step": 1908
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.96254239787642e-05,
      "loss": 0.5662,
      "step": 1910
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.961067689131397e-05,
      "loss": 0.5367,
      "step": 1912
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.959592980386374e-05,
      "loss": 0.5991,
      "step": 1914
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.958118271641351e-05,
      "loss": 0.5349,
      "step": 1916
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.956643562896328e-05,
      "loss": 0.5764,
      "step": 1918
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.955168854151305e-05,
      "loss": 0.5159,
      "step": 1920
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.953694145406282e-05,
      "loss": 0.5909,
      "step": 1922
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.95221943666126e-05,
      "loss": 0.5929,
      "step": 1924
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.950744727916237e-05,
      "loss": 0.5818,
      "step": 1926
    },
    {
      "epoch": 0.79,
      "learning_rate": 3.949270019171214e-05,
      "loss": 0.5754,
      "step": 1928
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.947795310426191e-05,
      "loss": 0.5308,
      "step": 1930
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.946320601681168e-05,
      "loss": 0.5257,
      "step": 1932
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.944845892936145e-05,
      "loss": 0.5067,
      "step": 1934
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.943371184191122e-05,
      "loss": 0.6011,
      "step": 1936
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.9418964754461e-05,
      "loss": 0.5923,
      "step": 1938
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.940421766701077e-05,
      "loss": 0.5545,
      "step": 1940
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.938947057956054e-05,
      "loss": 0.5329,
      "step": 1942
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.937472349211031e-05,
      "loss": 0.5169,
      "step": 1944
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.935997640466008e-05,
      "loss": 0.5292,
      "step": 1946
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.934522931720985e-05,
      "loss": 0.5997,
      "step": 1948
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.933048222975962e-05,
      "loss": 0.5849,
      "step": 1950
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.93157351423094e-05,
      "loss": 0.5549,
      "step": 1952
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.930098805485917e-05,
      "loss": 0.5789,
      "step": 1954
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.928624096740894e-05,
      "loss": 0.5595,
      "step": 1956
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.927149387995871e-05,
      "loss": 0.5548,
      "step": 1958
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.925674679250848e-05,
      "loss": 0.4898,
      "step": 1960
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.924199970505825e-05,
      "loss": 0.5542,
      "step": 1962
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.922725261760802e-05,
      "loss": 0.562,
      "step": 1964
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.92125055301578e-05,
      "loss": 0.5437,
      "step": 1966
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.919775844270757e-05,
      "loss": 0.5792,
      "step": 1968
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.918301135525734e-05,
      "loss": 0.5387,
      "step": 1970
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.916826426780711e-05,
      "loss": 0.5311,
      "step": 1972
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.915351718035688e-05,
      "loss": 0.5059,
      "step": 1974
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.913877009290665e-05,
      "loss": 0.5655,
      "step": 1976
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.912402300545642e-05,
      "loss": 0.5516,
      "step": 1978
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.91092759180062e-05,
      "loss": 0.5494,
      "step": 1980
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.909452883055597e-05,
      "loss": 0.521,
      "step": 1982
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.907978174310574e-05,
      "loss": 0.5534,
      "step": 1984
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.906503465565551e-05,
      "loss": 0.575,
      "step": 1986
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.905028756820528e-05,
      "loss": 0.5384,
      "step": 1988
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.903554048075505e-05,
      "loss": 0.5486,
      "step": 1990
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.902079339330482e-05,
      "loss": 0.4873,
      "step": 1992
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.9006046305854596e-05,
      "loss": 0.3682,
      "step": 1994
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.899129921840437e-05,
      "loss": 0.2667,
      "step": 1996
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.897655213095414e-05,
      "loss": 0.1665,
      "step": 1998
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.896180504350391e-05,
      "loss": 0.1086,
      "step": 2000
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.894705795605368e-05,
      "loss": 0.058,
      "step": 2002
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8932310868603455e-05,
      "loss": 0.0663,
      "step": 2004
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8917563781153225e-05,
      "loss": 0.0442,
      "step": 2006
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8902816693702996e-05,
      "loss": 0.0255,
      "step": 2008
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8888069606252766e-05,
      "loss": 0.0213,
      "step": 2010
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.887332251880254e-05,
      "loss": 0.0197,
      "step": 2012
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8858575431352314e-05,
      "loss": 0.0262,
      "step": 2014
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8843828343902084e-05,
      "loss": 0.0388,
      "step": 2016
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8829081256451854e-05,
      "loss": 0.0271,
      "step": 2018
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8814334169001625e-05,
      "loss": 0.0324,
      "step": 2020
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8799587081551395e-05,
      "loss": 0.018,
      "step": 2022
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.8784839994101165e-05,
      "loss": 0.0144,
      "step": 2024
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.877009290665094e-05,
      "loss": 0.0143,
      "step": 2026
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.875534581920071e-05,
      "loss": 0.0095,
      "step": 2028
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.874059873175048e-05,
      "loss": 0.0504,
      "step": 2030
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.8725851644300254e-05,
      "loss": 0.0278,
      "step": 2032
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.8711104556850024e-05,
      "loss": 0.0121,
      "step": 2034
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.8696357469399794e-05,
      "loss": 0.0253,
      "step": 2036
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.8681610381949565e-05,
      "loss": 0.0143,
      "step": 2038
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.866686329449934e-05,
      "loss": 0.0168,
      "step": 2040
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.865211620704911e-05,
      "loss": 0.0136,
      "step": 2042
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.863736911959888e-05,
      "loss": 0.0125,
      "step": 2044
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.862262203214865e-05,
      "loss": 0.0109,
      "step": 2046
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.8607874944698423e-05,
      "loss": 0.0422,
      "step": 2048
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.8593127857248194e-05,
      "loss": 0.0185,
      "step": 2050
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.8578380769797964e-05,
      "loss": 0.0116,
      "step": 2052
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.856363368234774e-05,
      "loss": 0.0068,
      "step": 2054
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.854888659489751e-05,
      "loss": 0.0077,
      "step": 2056
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.853413950744728e-05,
      "loss": 0.0064,
      "step": 2058
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.851939241999705e-05,
      "loss": 0.0041,
      "step": 2060
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.850464533254682e-05,
      "loss": 0.0048,
      "step": 2062
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.848989824509659e-05,
      "loss": 0.0103,
      "step": 2064
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.8475151157646363e-05,
      "loss": 0.016,
      "step": 2066
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.846040407019614e-05,
      "loss": 0.0041,
      "step": 2068
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.844565698274591e-05,
      "loss": 0.0032,
      "step": 2070
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.843090989529568e-05,
      "loss": 0.0282,
      "step": 2072
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.841616280784545e-05,
      "loss": 0.0019,
      "step": 2074
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.840141572039522e-05,
      "loss": 0.0041,
      "step": 2076
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.838666863294499e-05,
      "loss": 0.0037,
      "step": 2078
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.837192154549476e-05,
      "loss": 0.0057,
      "step": 2080
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.835717445804454e-05,
      "loss": 0.0104,
      "step": 2082
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.834242737059431e-05,
      "loss": 0.0041,
      "step": 2084
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.832768028314408e-05,
      "loss": 0.0053,
      "step": 2086
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.831293319569385e-05,
      "loss": 0.0035,
      "step": 2088
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.829818610824362e-05,
      "loss": 0.0169,
      "step": 2090
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.828343902079339e-05,
      "loss": 0.0127,
      "step": 2092
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.826869193334316e-05,
      "loss": 0.0265,
      "step": 2094
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.825394484589294e-05,
      "loss": 0.0078,
      "step": 2096
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.823919775844271e-05,
      "loss": 0.0057,
      "step": 2098
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.822445067099248e-05,
      "loss": 0.0038,
      "step": 2100
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.820970358354225e-05,
      "loss": 0.0041,
      "step": 2102
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.819495649609202e-05,
      "loss": 0.0023,
      "step": 2104
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.818020940864179e-05,
      "loss": 0.0024,
      "step": 2106
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.816546232119156e-05,
      "loss": 0.0017,
      "step": 2108
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.815071523374134e-05,
      "loss": 0.0026,
      "step": 2110
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.813596814629111e-05,
      "loss": 0.003,
      "step": 2112
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.812122105884088e-05,
      "loss": 0.0018,
      "step": 2114
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.8106473971390657e-05,
      "loss": 0.0026,
      "step": 2116
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.809172688394043e-05,
      "loss": 0.0013,
      "step": 2118
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.80769797964902e-05,
      "loss": 0.0008,
      "step": 2120
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.806223270903997e-05,
      "loss": 0.0019,
      "step": 2122
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.804748562158974e-05,
      "loss": 0.0035,
      "step": 2124
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.803273853413951e-05,
      "loss": 0.0048,
      "step": 2126
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.8017991446689285e-05,
      "loss": 0.0014,
      "step": 2128
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.8003244359239056e-05,
      "loss": 0.0016,
      "step": 2130
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.7988497271788826e-05,
      "loss": 0.0014,
      "step": 2132
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.7973750184338597e-05,
      "loss": 0.0029,
      "step": 2134
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.795900309688837e-05,
      "loss": 0.002,
      "step": 2136
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.794425600943814e-05,
      "loss": 0.0009,
      "step": 2138
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.792950892198791e-05,
      "loss": 0.001,
      "step": 2140
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.7914761834537685e-05,
      "loss": 0.0008,
      "step": 2142
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.7900014747087455e-05,
      "loss": 0.0008,
      "step": 2144
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.7885267659637226e-05,
      "loss": 0.0021,
      "step": 2146
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7870520572186996e-05,
      "loss": 0.001,
      "step": 2148
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7855773484736766e-05,
      "loss": 0.0011,
      "step": 2150
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.784102639728654e-05,
      "loss": 0.0014,
      "step": 2152
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.782627930983631e-05,
      "loss": 0.0008,
      "step": 2154
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7811532222386084e-05,
      "loss": 0.0005,
      "step": 2156
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7796785134935855e-05,
      "loss": 0.0054,
      "step": 2158
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7782038047485625e-05,
      "loss": 0.0007,
      "step": 2160
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7767290960035395e-05,
      "loss": 0.0065,
      "step": 2162
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7752543872585166e-05,
      "loss": 0.0014,
      "step": 2164
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7737796785134936e-05,
      "loss": 0.0004,
      "step": 2166
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7723049697684706e-05,
      "loss": 0.0009,
      "step": 2168
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7708302610234484e-05,
      "loss": 0.0009,
      "step": 2170
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.7693555522784254e-05,
      "loss": 0.001,
      "step": 2172
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7678808435334024e-05,
      "loss": 0.0079,
      "step": 2174
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7664061347883795e-05,
      "loss": 0.0018,
      "step": 2176
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7649314260433565e-05,
      "loss": 0.0017,
      "step": 2178
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7634567172983335e-05,
      "loss": 0.0024,
      "step": 2180
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7619820085533106e-05,
      "loss": 0.0013,
      "step": 2182
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.760507299808288e-05,
      "loss": 0.0009,
      "step": 2184
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.759032591063265e-05,
      "loss": 0.0012,
      "step": 2186
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7575578823182424e-05,
      "loss": 0.0006,
      "step": 2188
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7560831735732194e-05,
      "loss": 0.0031,
      "step": 2190
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7546084648281964e-05,
      "loss": 0.0007,
      "step": 2192
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7531337560831735e-05,
      "loss": 0.0009,
      "step": 2194
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.7516590473381505e-05,
      "loss": 0.0011,
      "step": 2196
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.750184338593128e-05,
      "loss": 0.0017,
      "step": 2198
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.748709629848105e-05,
      "loss": 0.0038,
      "step": 2200
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.747234921103082e-05,
      "loss": 0.0005,
      "step": 2202
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.745760212358059e-05,
      "loss": 0.0008,
      "step": 2204
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7442855036130364e-05,
      "loss": 0.0006,
      "step": 2206
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7428107948680134e-05,
      "loss": 0.0011,
      "step": 2208
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.7413360861229904e-05,
      "loss": 0.0009,
      "step": 2210
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.739861377377968e-05,
      "loss": 0.0005,
      "step": 2212
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.738386668632945e-05,
      "loss": 0.0011,
      "step": 2214
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.736911959887922e-05,
      "loss": 0.004,
      "step": 2216
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.735437251142899e-05,
      "loss": 0.0004,
      "step": 2218
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.733962542397876e-05,
      "loss": 0.0015,
      "step": 2220
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.732487833652853e-05,
      "loss": 0.0014,
      "step": 2222
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7310131249078304e-05,
      "loss": 0.001,
      "step": 2224
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.7295384161628074e-05,
      "loss": 0.001,
      "step": 2226
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.728063707417785e-05,
      "loss": 0.0007,
      "step": 2228
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.726588998672763e-05,
      "loss": 0.0009,
      "step": 2230
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.72511428992774e-05,
      "loss": 0.0006,
      "step": 2232
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.723639581182717e-05,
      "loss": 0.0023,
      "step": 2234
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.722164872437694e-05,
      "loss": 0.0003,
      "step": 2236
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.720690163692671e-05,
      "loss": 0.0007,
      "step": 2238
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.719215454947648e-05,
      "loss": 0.0016,
      "step": 2240
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.717740746202625e-05,
      "loss": 0.0007,
      "step": 2242
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.716266037457603e-05,
      "loss": 0.0004,
      "step": 2244
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.71479132871258e-05,
      "loss": 0.0004,
      "step": 2246
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.713316619967557e-05,
      "loss": 0.0006,
      "step": 2248
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.711841911222534e-05,
      "loss": 0.0015,
      "step": 2250
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.710367202477511e-05,
      "loss": 0.0004,
      "step": 2252
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.708892493732488e-05,
      "loss": 0.0021,
      "step": 2254
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.707417784987465e-05,
      "loss": 0.0006,
      "step": 2256
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.705943076242443e-05,
      "loss": 0.0005,
      "step": 2258
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.70446836749742e-05,
      "loss": 0.0213,
      "step": 2260
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.702993658752397e-05,
      "loss": 0.0002,
      "step": 2262
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.701518950007374e-05,
      "loss": 0.0004,
      "step": 2264
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.700044241262351e-05,
      "loss": 0.0009,
      "step": 2266
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.698569532517328e-05,
      "loss": 0.0063,
      "step": 2268
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.697094823772305e-05,
      "loss": 0.0005,
      "step": 2270
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.6956201150272826e-05,
      "loss": 0.0007,
      "step": 2272
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.69414540628226e-05,
      "loss": 0.0005,
      "step": 2274
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.692670697537237e-05,
      "loss": 0.0045,
      "step": 2276
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.691195988792214e-05,
      "loss": 0.0004,
      "step": 2278
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.689721280047191e-05,
      "loss": 0.0016,
      "step": 2280
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.688246571302168e-05,
      "loss": 0.001,
      "step": 2282
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.686771862557145e-05,
      "loss": 0.0004,
      "step": 2284
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.6852971538121226e-05,
      "loss": 0.001,
      "step": 2286
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.6838224450670996e-05,
      "loss": 0.002,
      "step": 2288
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.6823477363220766e-05,
      "loss": 0.0006,
      "step": 2290
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.680873027577054e-05,
      "loss": 0.0006,
      "step": 2292
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.679398318832031e-05,
      "loss": 0.0008,
      "step": 2294
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.677923610087008e-05,
      "loss": 0.0004,
      "step": 2296
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.676448901341985e-05,
      "loss": 0.0005,
      "step": 2298
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6749741925969625e-05,
      "loss": 0.0005,
      "step": 2300
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6734994838519395e-05,
      "loss": 0.0006,
      "step": 2302
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6720247751069166e-05,
      "loss": 0.0003,
      "step": 2304
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6705500663618936e-05,
      "loss": 0.0004,
      "step": 2306
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6690753576168707e-05,
      "loss": 0.0003,
      "step": 2308
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.667600648871848e-05,
      "loss": 0.0011,
      "step": 2310
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.666125940126825e-05,
      "loss": 0.0005,
      "step": 2312
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6646512313818024e-05,
      "loss": 0.0003,
      "step": 2314
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.6631765226367795e-05,
      "loss": 0.0006,
      "step": 2316
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6617018138917565e-05,
      "loss": 0.0005,
      "step": 2318
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6602271051467336e-05,
      "loss": 0.0004,
      "step": 2320
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6587523964017106e-05,
      "loss": 0.0021,
      "step": 2322
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6572776876566876e-05,
      "loss": 0.0003,
      "step": 2324
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.655802978911665e-05,
      "loss": 0.0003,
      "step": 2326
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.654328270166642e-05,
      "loss": 0.0005,
      "step": 2328
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6528535614216194e-05,
      "loss": 0.0005,
      "step": 2330
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6513788526765964e-05,
      "loss": 0.0002,
      "step": 2332
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6499041439315735e-05,
      "loss": 0.0042,
      "step": 2334
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.6484294351865505e-05,
      "loss": 0.0004,
      "step": 2336
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.646954726441528e-05,
      "loss": 0.0003,
      "step": 2338
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.645480017696505e-05,
      "loss": 0.0021,
      "step": 2340
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.644005308951482e-05,
      "loss": 0.0014,
      "step": 2342
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.6425306002064593e-05,
      "loss": 0.0007,
      "step": 2344
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.641055891461437e-05,
      "loss": 0.0007,
      "step": 2346
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.639581182716414e-05,
      "loss": 0.0074,
      "step": 2348
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.638106473971391e-05,
      "loss": 0.0003,
      "step": 2350
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.636631765226368e-05,
      "loss": 0.0059,
      "step": 2352
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.635157056481345e-05,
      "loss": 0.0009,
      "step": 2354
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.633682347736322e-05,
      "loss": 0.0014,
      "step": 2356
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.632207638991299e-05,
      "loss": 0.0029,
      "step": 2358
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.630732930246277e-05,
      "loss": 0.0005,
      "step": 2360
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.629258221501254e-05,
      "loss": 0.0024,
      "step": 2362
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.627783512756231e-05,
      "loss": 0.0003,
      "step": 2364
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.626308804011208e-05,
      "loss": 0.0005,
      "step": 2366
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.624834095266185e-05,
      "loss": 0.0006,
      "step": 2368
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.623359386521162e-05,
      "loss": 0.0006,
      "step": 2370
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.621884677776139e-05,
      "loss": 0.0003,
      "step": 2372
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.620409969031117e-05,
      "loss": 0.0004,
      "step": 2374
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.618935260286094e-05,
      "loss": 0.0009,
      "step": 2376
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.617460551541071e-05,
      "loss": 0.0006,
      "step": 2378
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.615985842796048e-05,
      "loss": 0.0003,
      "step": 2380
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.614511134051025e-05,
      "loss": 0.0004,
      "step": 2382
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.613036425306002e-05,
      "loss": 0.0004,
      "step": 2384
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.611561716560979e-05,
      "loss": 0.0012,
      "step": 2386
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.610087007815957e-05,
      "loss": 0.0003,
      "step": 2388
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.608612299070934e-05,
      "loss": 0.0002,
      "step": 2390
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.607137590325911e-05,
      "loss": 0.0002,
      "step": 2392
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.605662881580888e-05,
      "loss": 0.0003,
      "step": 2394
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.604188172835865e-05,
      "loss": 0.0005,
      "step": 2396
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.602713464090842e-05,
      "loss": 0.0005,
      "step": 2398
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.601238755345819e-05,
      "loss": 0.0003,
      "step": 2400
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.599764046600797e-05,
      "loss": 0.0004,
      "step": 2402
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.598289337855774e-05,
      "loss": 0.0012,
      "step": 2404
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.596814629110751e-05,
      "loss": 0.0002,
      "step": 2406
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.595339920365728e-05,
      "loss": 0.0003,
      "step": 2408
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.593865211620705e-05,
      "loss": 0.0002,
      "step": 2410
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.592390502875682e-05,
      "loss": 0.0001,
      "step": 2412
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.590915794130659e-05,
      "loss": 0.001,
      "step": 2414
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.589441085385637e-05,
      "loss": 0.0002,
      "step": 2416
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.587966376640614e-05,
      "loss": 0.0003,
      "step": 2418
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.586491667895591e-05,
      "loss": 0.0003,
      "step": 2420
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.585016959150568e-05,
      "loss": 0.0004,
      "step": 2422
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.583542250405545e-05,
      "loss": 0.0002,
      "step": 2424
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.582067541660522e-05,
      "loss": 0.0001,
      "step": 2426
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.580592832915499e-05,
      "loss": 0.0002,
      "step": 2428
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.579118124170476e-05,
      "loss": 0.0003,
      "step": 2430
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.577643415425454e-05,
      "loss": 0.0002,
      "step": 2432
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.576168706680431e-05,
      "loss": 0.0003,
      "step": 2434
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.574693997935408e-05,
      "loss": 0.0004,
      "step": 2436
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.573219289190385e-05,
      "loss": 0.0002,
      "step": 2438
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.571744580445362e-05,
      "loss": 0.0002,
      "step": 2440
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.570269871700339e-05,
      "loss": 0.0002,
      "step": 2442
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.568795162955316e-05,
      "loss": 0.0003,
      "step": 2444
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5673204542102936e-05,
      "loss": 0.0001,
      "step": 2446
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.565845745465271e-05,
      "loss": 0.0002,
      "step": 2448
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5643710367202484e-05,
      "loss": 0.0002,
      "step": 2450
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5628963279752254e-05,
      "loss": 0.0003,
      "step": 2452
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5614216192302025e-05,
      "loss": 0.001,
      "step": 2454
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5599469104851795e-05,
      "loss": 0.0014,
      "step": 2456
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5584722017401565e-05,
      "loss": 0.0002,
      "step": 2458
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.5569974929951336e-05,
      "loss": 0.0014,
      "step": 2460
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.555522784250111e-05,
      "loss": 0.0002,
      "step": 2462
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.554048075505088e-05,
      "loss": 0.0001,
      "step": 2464
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5525733667600654e-05,
      "loss": 0.0006,
      "step": 2466
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5510986580150424e-05,
      "loss": 0.0004,
      "step": 2468
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5496239492700194e-05,
      "loss": 0.0006,
      "step": 2470
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5481492405249965e-05,
      "loss": 0.0003,
      "step": 2472
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5466745317799735e-05,
      "loss": 0.0018,
      "step": 2474
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.545199823034951e-05,
      "loss": 0.0006,
      "step": 2476
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.543725114289928e-05,
      "loss": 0.0031,
      "step": 2478
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.542250405544905e-05,
      "loss": 0.0003,
      "step": 2480
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.540775696799882e-05,
      "loss": 0.0002,
      "step": 2482
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5393009880548594e-05,
      "loss": 0.0003,
      "step": 2484
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.5378262793098364e-05,
      "loss": 0.0004,
      "step": 2486
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.5363515705648134e-05,
      "loss": 0.0004,
      "step": 2488
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.534876861819791e-05,
      "loss": 0.0006,
      "step": 2490
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.533402153074768e-05,
      "loss": 0.0007,
      "step": 2492
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.531927444329745e-05,
      "loss": 0.0005,
      "step": 2494
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.530452735584722e-05,
      "loss": 0.0007,
      "step": 2496
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.528978026839699e-05,
      "loss": 0.001,
      "step": 2498
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.527503318094676e-05,
      "loss": 0.0003,
      "step": 2500
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.5260286093496534e-05,
      "loss": 0.0002,
      "step": 2502
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.524553900604631e-05,
      "loss": 0.0002,
      "step": 2504
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.523079191859608e-05,
      "loss": 0.0009,
      "step": 2506
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.521604483114585e-05,
      "loss": 0.0002,
      "step": 2508
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.520129774369562e-05,
      "loss": 0.0003,
      "step": 2510
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.518655065624539e-05,
      "loss": 0.0028,
      "step": 2512
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.517180356879516e-05,
      "loss": 0.0004,
      "step": 2514
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.515705648134493e-05,
      "loss": 0.0002,
      "step": 2516
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.5142309393894703e-05,
      "loss": 0.0006,
      "step": 2518
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.512756230644448e-05,
      "loss": 0.0117,
      "step": 2520
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.511281521899425e-05,
      "loss": 0.0003,
      "step": 2522
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.509806813154402e-05,
      "loss": 0.0012,
      "step": 2524
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.508332104409379e-05,
      "loss": 0.0008,
      "step": 2526
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.506857395664356e-05,
      "loss": 0.0003,
      "step": 2528
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.505382686919333e-05,
      "loss": 0.0002,
      "step": 2530
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.50390797817431e-05,
      "loss": 0.0002,
      "step": 2532
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.502433269429288e-05,
      "loss": 0.0004,
      "step": 2534
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.500958560684265e-05,
      "loss": 0.0002,
      "step": 2536
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.499483851939242e-05,
      "loss": 0.0003,
      "step": 2538
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.498009143194219e-05,
      "loss": 0.0004,
      "step": 2540
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.496534434449196e-05,
      "loss": 0.0006,
      "step": 2542
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.495059725704173e-05,
      "loss": 0.0003,
      "step": 2544
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.49358501695915e-05,
      "loss": 0.0003,
      "step": 2546
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.492110308214128e-05,
      "loss": 0.0002,
      "step": 2548
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.490635599469105e-05,
      "loss": 0.0002,
      "step": 2550
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.489160890724082e-05,
      "loss": 0.0003,
      "step": 2552
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.487686181979059e-05,
      "loss": 0.0003,
      "step": 2554
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.486211473234036e-05,
      "loss": 0.0007,
      "step": 2556
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.484736764489013e-05,
      "loss": 0.0002,
      "step": 2558
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.483262055743991e-05,
      "loss": 0.0003,
      "step": 2560
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.481787346998968e-05,
      "loss": 0.0002,
      "step": 2562
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4803126382539456e-05,
      "loss": 0.0004,
      "step": 2564
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4788379295089226e-05,
      "loss": 0.0003,
      "step": 2566
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4773632207638996e-05,
      "loss": 0.0002,
      "step": 2568
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.475888512018877e-05,
      "loss": 0.0008,
      "step": 2570
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.474413803273854e-05,
      "loss": 0.0002,
      "step": 2572
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.472939094528831e-05,
      "loss": 0.0007,
      "step": 2574
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.471464385783808e-05,
      "loss": 0.0001,
      "step": 2576
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4699896770387855e-05,
      "loss": 0.0002,
      "step": 2578
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4685149682937625e-05,
      "loss": 0.0005,
      "step": 2580
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4670402595487396e-05,
      "loss": 0.0004,
      "step": 2582
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.4655655508037166e-05,
      "loss": 0.0001,
      "step": 2584
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4640908420586937e-05,
      "loss": 0.0001,
      "step": 2586
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.462616133313671e-05,
      "loss": 0.0004,
      "step": 2588
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.461141424568648e-05,
      "loss": 0.0002,
      "step": 2590
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4596667158236254e-05,
      "loss": 0.0003,
      "step": 2592
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4581920070786025e-05,
      "loss": 0.0016,
      "step": 2594
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4567172983335795e-05,
      "loss": 0.0002,
      "step": 2596
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4552425895885565e-05,
      "loss": 0.0001,
      "step": 2598
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4537678808435336e-05,
      "loss": 0.0003,
      "step": 2600
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4522931720985106e-05,
      "loss": 0.0004,
      "step": 2602
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4508184633534877e-05,
      "loss": 0.0004,
      "step": 2604
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4493437546084654e-05,
      "loss": 0.0009,
      "step": 2606
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.4478690458634424e-05,
      "loss": 0.0002,
      "step": 2608
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4463943371184194e-05,
      "loss": 0.0002,
      "step": 2610
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4449196283733965e-05,
      "loss": 0.0004,
      "step": 2612
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4434449196283735e-05,
      "loss": 0.0004,
      "step": 2614
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4419702108833506e-05,
      "loss": 0.0004,
      "step": 2616
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4404955021383276e-05,
      "loss": 0.0002,
      "step": 2618
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4390207933933046e-05,
      "loss": 0.0002,
      "step": 2620
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4375460846482823e-05,
      "loss": 0.0002,
      "step": 2622
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4360713759032594e-05,
      "loss": 0.0004,
      "step": 2624
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4345966671582364e-05,
      "loss": 0.0008,
      "step": 2626
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4331219584132135e-05,
      "loss": 0.0002,
      "step": 2628
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4316472496681905e-05,
      "loss": 0.0002,
      "step": 2630
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.4301725409231675e-05,
      "loss": 0.0006,
      "step": 2632
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.4286978321781446e-05,
      "loss": 0.0033,
      "step": 2634
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.427223123433122e-05,
      "loss": 0.0003,
      "step": 2636
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.425748414688099e-05,
      "loss": 0.0003,
      "step": 2638
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.4242737059430764e-05,
      "loss": 0.0004,
      "step": 2640
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.4227989971980534e-05,
      "loss": 0.001,
      "step": 2642
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.4213242884530304e-05,
      "loss": 0.0006,
      "step": 2644
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.4198495797080075e-05,
      "loss": 0.0004,
      "step": 2646
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.4183748709629845e-05,
      "loss": 0.0002,
      "step": 2648
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.416900162217962e-05,
      "loss": 0.0002,
      "step": 2650
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.415425453472939e-05,
      "loss": 0.0002,
      "step": 2652
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.413950744727916e-05,
      "loss": 0.0001,
      "step": 2654
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.412476035982893e-05,
      "loss": 0.0002,
      "step": 2656
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.4110013272378704e-05,
      "loss": 0.0002,
      "step": 2658
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.4095266184928474e-05,
      "loss": 0.0001,
      "step": 2660
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.4080519097478244e-05,
      "loss": 0.0002,
      "step": 2662
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.406577201002802e-05,
      "loss": 0.0002,
      "step": 2664
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.405102492257779e-05,
      "loss": 0.0001,
      "step": 2666
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.403627783512756e-05,
      "loss": 0.0002,
      "step": 2668
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.402153074767733e-05,
      "loss": 0.0002,
      "step": 2670
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.400678366022711e-05,
      "loss": 0.0002,
      "step": 2672
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.399203657277688e-05,
      "loss": 0.0001,
      "step": 2674
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.397728948532665e-05,
      "loss": 0.0001,
      "step": 2676
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.396254239787642e-05,
      "loss": 0.0003,
      "step": 2678
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.39477953104262e-05,
      "loss": 0.0007,
      "step": 2680
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.393304822297597e-05,
      "loss": 0.0003,
      "step": 2682
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.391830113552574e-05,
      "loss": 0.0001,
      "step": 2684
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.390355404807551e-05,
      "loss": 0.0001,
      "step": 2686
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.388880696062528e-05,
      "loss": 0.0001,
      "step": 2688
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.387405987317505e-05,
      "loss": 0.0002,
      "step": 2690
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.385931278572482e-05,
      "loss": 0.0008,
      "step": 2692
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.38445656982746e-05,
      "loss": 0.0001,
      "step": 2694
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.382981861082437e-05,
      "loss": 0.0002,
      "step": 2696
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.381507152337414e-05,
      "loss": 0.0001,
      "step": 2698
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.380032443592391e-05,
      "loss": 0.0002,
      "step": 2700
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.378557734847368e-05,
      "loss": 0.0001,
      "step": 2702
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.377083026102345e-05,
      "loss": 0.0001,
      "step": 2704
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.375608317357322e-05,
      "loss": 0.0002,
      "step": 2706
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.3741336086123e-05,
      "loss": 0.0001,
      "step": 2708
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.372658899867277e-05,
      "loss": 0.0001,
      "step": 2710
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.371184191122254e-05,
      "loss": 0.0001,
      "step": 2712
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.369709482377231e-05,
      "loss": 0.0002,
      "step": 2714
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.368234773632208e-05,
      "loss": 0.0001,
      "step": 2716
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.366760064887185e-05,
      "loss": 0.0001,
      "step": 2718
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.365285356142162e-05,
      "loss": 0.0001,
      "step": 2720
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.363810647397139e-05,
      "loss": 0.0001,
      "step": 2722
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.3623359386521166e-05,
      "loss": 0.0001,
      "step": 2724
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.360861229907094e-05,
      "loss": 0.0002,
      "step": 2726
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.359386521162071e-05,
      "loss": 0.0002,
      "step": 2728
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.357911812417048e-05,
      "loss": 0.0008,
      "step": 2730
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.356437103672025e-05,
      "loss": 0.0001,
      "step": 2732
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.354962394927002e-05,
      "loss": 0.0001,
      "step": 2734
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.353487686181979e-05,
      "loss": 0.0002,
      "step": 2736
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.3520129774369566e-05,
      "loss": 0.0002,
      "step": 2738
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.3505382686919336e-05,
      "loss": 0.0001,
      "step": 2740
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.3490635599469106e-05,
      "loss": 0.0001,
      "step": 2742
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.347588851201888e-05,
      "loss": 0.0001,
      "step": 2744
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.346114142456865e-05,
      "loss": 0.0001,
      "step": 2746
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.344639433711842e-05,
      "loss": 0.0001,
      "step": 2748
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.343164724966819e-05,
      "loss": 0.0001,
      "step": 2750
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.3416900162217965e-05,
      "loss": 0.0002,
      "step": 2752
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.3402153074767735e-05,
      "loss": 0.0001,
      "step": 2754
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3387405987317506e-05,
      "loss": 0.0001,
      "step": 2756
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3372658899867276e-05,
      "loss": 0.0001,
      "step": 2758
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3357911812417046e-05,
      "loss": 0.0001,
      "step": 2760
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.334316472496682e-05,
      "loss": 0.0001,
      "step": 2762
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.332841763751659e-05,
      "loss": 0.0001,
      "step": 2764
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3313670550066364e-05,
      "loss": 0.0004,
      "step": 2766
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3298923462616135e-05,
      "loss": 0.0001,
      "step": 2768
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3284176375165905e-05,
      "loss": 0.0001,
      "step": 2770
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3269429287715675e-05,
      "loss": 0.0001,
      "step": 2772
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3254682200265446e-05,
      "loss": 0.0001,
      "step": 2774
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3239935112815216e-05,
      "loss": 0.0001,
      "step": 2776
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.3225188025364987e-05,
      "loss": 0.0001,
      "step": 2778
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.3210440937914764e-05,
      "loss": 0.0001,
      "step": 2780
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.319569385046454e-05,
      "loss": 0.0001,
      "step": 2782
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.318094676301431e-05,
      "loss": 0.0001,
      "step": 2784
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.316619967556408e-05,
      "loss": 0.0001,
      "step": 2786
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.315145258811385e-05,
      "loss": 0.0001,
      "step": 2788
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.313670550066362e-05,
      "loss": 0.0001,
      "step": 2790
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.312195841321339e-05,
      "loss": 0.0001,
      "step": 2792
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.310721132576316e-05,
      "loss": 0.0001,
      "step": 2794
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.309246423831294e-05,
      "loss": 0.0001,
      "step": 2796
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.307771715086271e-05,
      "loss": 0.0001,
      "step": 2798
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.306297006341248e-05,
      "loss": 0.0001,
      "step": 2800
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.304822297596225e-05,
      "loss": 0.0001,
      "step": 2802
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.303347588851202e-05,
      "loss": 0.0011,
      "step": 2804
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.301872880106179e-05,
      "loss": 0.0001,
      "step": 2806
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.300398171361156e-05,
      "loss": 0.0001,
      "step": 2808
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.298923462616133e-05,
      "loss": 0.0001,
      "step": 2810
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.297448753871111e-05,
      "loss": 0.0002,
      "step": 2812
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.295974045126088e-05,
      "loss": 0.0001,
      "step": 2814
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.294499336381065e-05,
      "loss": 0.0004,
      "step": 2816
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.293024627636042e-05,
      "loss": 0.0001,
      "step": 2818
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.291549918891019e-05,
      "loss": 0.0001,
      "step": 2820
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.290075210145996e-05,
      "loss": 0.0001,
      "step": 2822
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.288600501400973e-05,
      "loss": 0.0001,
      "step": 2824
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.287125792655951e-05,
      "loss": 0.0001,
      "step": 2826
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.285651083910928e-05,
      "loss": 0.0001,
      "step": 2828
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.284176375165905e-05,
      "loss": 0.0001,
      "step": 2830
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.282701666420882e-05,
      "loss": 0.0001,
      "step": 2832
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.281226957675859e-05,
      "loss": 0.0002,
      "step": 2834
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.279752248930836e-05,
      "loss": 0.0001,
      "step": 2836
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.278277540185813e-05,
      "loss": 0.0001,
      "step": 2838
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.276802831440791e-05,
      "loss": 0.0001,
      "step": 2840
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.275328122695768e-05,
      "loss": 0.0004,
      "step": 2842
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.273853413950745e-05,
      "loss": 0.0002,
      "step": 2844
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.272378705205722e-05,
      "loss": 0.0001,
      "step": 2846
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.270903996460699e-05,
      "loss": 0.0002,
      "step": 2848
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.269429287715676e-05,
      "loss": 0.0001,
      "step": 2850
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.267954578970653e-05,
      "loss": 0.0001,
      "step": 2852
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.266479870225631e-05,
      "loss": 0.0001,
      "step": 2854
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.265005161480608e-05,
      "loss": 0.0006,
      "step": 2856
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.263530452735585e-05,
      "loss": 0.0004,
      "step": 2858
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.262055743990562e-05,
      "loss": 0.0006,
      "step": 2860
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.260581035245539e-05,
      "loss": 0.0001,
      "step": 2862
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.259106326500516e-05,
      "loss": 0.0002,
      "step": 2864
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.257631617755493e-05,
      "loss": 0.0001,
      "step": 2866
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.256156909010471e-05,
      "loss": 0.0001,
      "step": 2868
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.254682200265448e-05,
      "loss": 0.0004,
      "step": 2870
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.253207491520425e-05,
      "loss": 0.0001,
      "step": 2872
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.251732782775402e-05,
      "loss": 0.0001,
      "step": 2874
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.250258074030379e-05,
      "loss": 0.0007,
      "step": 2876
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.248783365285356e-05,
      "loss": 0.0002,
      "step": 2878
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.247308656540333e-05,
      "loss": 0.0002,
      "step": 2880
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.2458339477953107e-05,
      "loss": 0.0002,
      "step": 2882
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.244359239050288e-05,
      "loss": 0.0004,
      "step": 2884
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.242884530305265e-05,
      "loss": 0.0002,
      "step": 2886
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.241409821560242e-05,
      "loss": 0.0003,
      "step": 2888
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.239935112815219e-05,
      "loss": 0.0002,
      "step": 2890
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.238460404070196e-05,
      "loss": 0.0003,
      "step": 2892
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.2369856953251736e-05,
      "loss": 0.0002,
      "step": 2894
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.2355109865801506e-05,
      "loss": 0.0002,
      "step": 2896
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.234036277835128e-05,
      "loss": 0.0002,
      "step": 2898
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.2325615690901053e-05,
      "loss": 0.0002,
      "step": 2900
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2310868603450824e-05,
      "loss": 0.0001,
      "step": 2902
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2296121516000594e-05,
      "loss": 0.0002,
      "step": 2904
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2281374428550365e-05,
      "loss": 0.0001,
      "step": 2906
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2266627341100135e-05,
      "loss": 0.0002,
      "step": 2908
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2251880253649905e-05,
      "loss": 0.0002,
      "step": 2910
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2237133166199676e-05,
      "loss": 0.0001,
      "step": 2912
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.222238607874945e-05,
      "loss": 0.0001,
      "step": 2914
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.220763899129922e-05,
      "loss": 0.0001,
      "step": 2916
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2192891903848993e-05,
      "loss": 0.0001,
      "step": 2918
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2178144816398764e-05,
      "loss": 0.0002,
      "step": 2920
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2163397728948534e-05,
      "loss": 0.0001,
      "step": 2922
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.2148650641498305e-05,
      "loss": 0.0001,
      "step": 2924
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.2133903554048075e-05,
      "loss": 0.0001,
      "step": 2926
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.211915646659785e-05,
      "loss": 0.0002,
      "step": 2928
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.210440937914762e-05,
      "loss": 0.0001,
      "step": 2930
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.208966229169739e-05,
      "loss": 0.0001,
      "step": 2932
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.207491520424716e-05,
      "loss": 0.0001,
      "step": 2934
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.2060168116796934e-05,
      "loss": 0.0001,
      "step": 2936
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.2045421029346704e-05,
      "loss": 0.0002,
      "step": 2938
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.2030673941896474e-05,
      "loss": 0.0001,
      "step": 2940
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.201592685444625e-05,
      "loss": 0.0001,
      "step": 2942
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.200117976699602e-05,
      "loss": 0.0001,
      "step": 2944
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.198643267954579e-05,
      "loss": 0.0003,
      "step": 2946
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.197168559209556e-05,
      "loss": 0.0001,
      "step": 2948
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.195693850464533e-05,
      "loss": 0.0003,
      "step": 2950
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.19421914171951e-05,
      "loss": 0.0008,
      "step": 2952
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.1927444329744874e-05,
      "loss": 0.0001,
      "step": 2954
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.191269724229465e-05,
      "loss": 0.0001,
      "step": 2956
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.189795015484442e-05,
      "loss": 0.0001,
      "step": 2958
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.188320306739419e-05,
      "loss": 0.0001,
      "step": 2960
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.186845597994396e-05,
      "loss": 0.0002,
      "step": 2962
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.185370889249373e-05,
      "loss": 0.0001,
      "step": 2964
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.18389618050435e-05,
      "loss": 0.0001,
      "step": 2966
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.182421471759327e-05,
      "loss": 0.0001,
      "step": 2968
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.180946763014305e-05,
      "loss": 0.0001,
      "step": 2970
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.179472054269282e-05,
      "loss": 0.0001,
      "step": 2972
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.177997345524259e-05,
      "loss": 0.0001,
      "step": 2974
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.176522636779236e-05,
      "loss": 0.0001,
      "step": 2976
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.175047928034213e-05,
      "loss": 0.0001,
      "step": 2978
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.17357321928919e-05,
      "loss": 0.0002,
      "step": 2980
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.172098510544167e-05,
      "loss": 0.0001,
      "step": 2982
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.170623801799145e-05,
      "loss": 0.0001,
      "step": 2984
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.169149093054122e-05,
      "loss": 0.0001,
      "step": 2986
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.167674384309099e-05,
      "loss": 0.0001,
      "step": 2988
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.166199675564076e-05,
      "loss": 0.0001,
      "step": 2990
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.164724966819053e-05,
      "loss": 0.0001,
      "step": 2992
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.16325025807403e-05,
      "loss": 0.0001,
      "step": 2994
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.161775549329007e-05,
      "loss": 0.0001,
      "step": 2996
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.160300840583985e-05,
      "loss": 0.0001,
      "step": 2998
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.158826131838962e-05,
      "loss": 0.0001,
      "step": 3000
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.157351423093939e-05,
      "loss": 0.0001,
      "step": 3002
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.155876714348916e-05,
      "loss": 0.0001,
      "step": 3004
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.154402005603894e-05,
      "loss": 0.0001,
      "step": 3006
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.152927296858871e-05,
      "loss": 0.0001,
      "step": 3008
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.151452588113848e-05,
      "loss": 0.0002,
      "step": 3010
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.149977879368825e-05,
      "loss": 0.0001,
      "step": 3012
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.148503170623802e-05,
      "loss": 0.0001,
      "step": 3014
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.1470284618787796e-05,
      "loss": 0.0001,
      "step": 3016
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.1455537531337566e-05,
      "loss": 0.0001,
      "step": 3018
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.1440790443887336e-05,
      "loss": 0.0001,
      "step": 3020
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.142604335643711e-05,
      "loss": 0.0001,
      "step": 3022
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.141129626898688e-05,
      "loss": 0.0001,
      "step": 3024
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.139654918153665e-05,
      "loss": 0.0001,
      "step": 3026
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.138180209408642e-05,
      "loss": 0.0001,
      "step": 3028
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.1367055006636195e-05,
      "loss": 0.0001,
      "step": 3030
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.1352307919185965e-05,
      "loss": 0.0001,
      "step": 3032
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.1337560831735736e-05,
      "loss": 0.002,
      "step": 3034
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.1322813744285506e-05,
      "loss": 0.0001,
      "step": 3036
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.1308066656835276e-05,
      "loss": 0.0001,
      "step": 3038
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.129331956938505e-05,
      "loss": 0.0001,
      "step": 3040
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.127857248193482e-05,
      "loss": 0.0001,
      "step": 3042
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.1263825394484594e-05,
      "loss": 0.0001,
      "step": 3044
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1249078307034365e-05,
      "loss": 0.0001,
      "step": 3046
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1234331219584135e-05,
      "loss": 0.0001,
      "step": 3048
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1219584132133905e-05,
      "loss": 0.0001,
      "step": 3050
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1204837044683676e-05,
      "loss": 0.0001,
      "step": 3052
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1190089957233446e-05,
      "loss": 0.0001,
      "step": 3054
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1175342869783217e-05,
      "loss": 0.0001,
      "step": 3056
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1160595782332994e-05,
      "loss": 0.0001,
      "step": 3058
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1145848694882764e-05,
      "loss": 0.0001,
      "step": 3060
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1131101607432534e-05,
      "loss": 0.0001,
      "step": 3062
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1116354519982305e-05,
      "loss": 0.0003,
      "step": 3064
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1101607432532075e-05,
      "loss": 0.0004,
      "step": 3066
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1086860345081845e-05,
      "loss": 0.0001,
      "step": 3068
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.1072113257631616e-05,
      "loss": 0.0,
      "step": 3070
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.105736617018139e-05,
      "loss": 0.0001,
      "step": 3072
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.104261908273116e-05,
      "loss": 0.0001,
      "step": 3074
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.1027871995280934e-05,
      "loss": 0.0001,
      "step": 3076
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.1013124907830704e-05,
      "loss": 0.0001,
      "step": 3078
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.0998377820380474e-05,
      "loss": 0.0001,
      "step": 3080
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.0983630732930245e-05,
      "loss": 0.0005,
      "step": 3082
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.0968883645480015e-05,
      "loss": 0.0001,
      "step": 3084
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.095413655802979e-05,
      "loss": 0.0001,
      "step": 3086
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.093938947057956e-05,
      "loss": 0.0003,
      "step": 3088
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.092464238312933e-05,
      "loss": 0.0001,
      "step": 3090
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.0909895295679103e-05,
      "loss": 0.0008,
      "step": 3092
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.0895148208228874e-05,
      "loss": 0.0001,
      "step": 3094
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.0880401120778644e-05,
      "loss": 0.0001,
      "step": 3096
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.0865654033328415e-05,
      "loss": 0.0001,
      "step": 3098
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.085090694587819e-05,
      "loss": 0.0001,
      "step": 3100
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.083615985842796e-05,
      "loss": 0.0001,
      "step": 3102
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.082141277097773e-05,
      "loss": 0.0001,
      "step": 3104
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.08066656835275e-05,
      "loss": 0.0001,
      "step": 3106
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.079191859607727e-05,
      "loss": 0.0002,
      "step": 3108
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.0777171508627044e-05,
      "loss": 0.0005,
      "step": 3110
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.0762424421176814e-05,
      "loss": 0.0001,
      "step": 3112
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.074767733372659e-05,
      "loss": 0.0001,
      "step": 3114
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.073293024627636e-05,
      "loss": 0.0001,
      "step": 3116
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.071818315882614e-05,
      "loss": 0.0,
      "step": 3118
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.070343607137591e-05,
      "loss": 0.0002,
      "step": 3120
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.068868898392568e-05,
      "loss": 0.0001,
      "step": 3122
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.067394189647545e-05,
      "loss": 0.0001,
      "step": 3124
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.065919480902522e-05,
      "loss": 0.0001,
      "step": 3126
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.064444772157499e-05,
      "loss": 0.0001,
      "step": 3128
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.062970063412476e-05,
      "loss": 0.0001,
      "step": 3130
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.061495354667454e-05,
      "loss": 0.0001,
      "step": 3132
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.060020645922431e-05,
      "loss": 0.0001,
      "step": 3134
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.058545937177408e-05,
      "loss": 0.0001,
      "step": 3136
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.057071228432385e-05,
      "loss": 0.0001,
      "step": 3138
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.055596519687362e-05,
      "loss": 0.0001,
      "step": 3140
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.054121810942339e-05,
      "loss": 0.0001,
      "step": 3142
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.052647102197316e-05,
      "loss": 0.0012,
      "step": 3144
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0511723934522934e-05,
      "loss": 0.0002,
      "step": 3146
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0496976847072704e-05,
      "loss": 0.0001,
      "step": 3148
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0482229759622478e-05,
      "loss": 0.0001,
      "step": 3150
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0467482672172248e-05,
      "loss": 0.0001,
      "step": 3152
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.045273558472202e-05,
      "loss": 0.0,
      "step": 3154
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0437988497271792e-05,
      "loss": 0.0001,
      "step": 3156
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0423241409821563e-05,
      "loss": 0.0001,
      "step": 3158
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0408494322371333e-05,
      "loss": 0.0001,
      "step": 3160
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0393747234921104e-05,
      "loss": 0.0002,
      "step": 3162
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0379000147470877e-05,
      "loss": 0.0001,
      "step": 3164
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.0364253060020648e-05,
      "loss": 0.0002,
      "step": 3166
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0349505972570418e-05,
      "loss": 0.0001,
      "step": 3168
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0334758885120192e-05,
      "loss": 0.0001,
      "step": 3170
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0320011797669962e-05,
      "loss": 0.0001,
      "step": 3172
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0305264710219732e-05,
      "loss": 0.0001,
      "step": 3174
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0290517622769503e-05,
      "loss": 0.0001,
      "step": 3176
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0275770535319277e-05,
      "loss": 0.0001,
      "step": 3178
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0261023447869047e-05,
      "loss": 0.0001,
      "step": 3180
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0246276360418817e-05,
      "loss": 0.0001,
      "step": 3182
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.023152927296859e-05,
      "loss": 0.0003,
      "step": 3184
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.021678218551836e-05,
      "loss": 0.0017,
      "step": 3186
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0202035098068132e-05,
      "loss": 0.0001,
      "step": 3188
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.0187288010617902e-05,
      "loss": 0.0001,
      "step": 3190
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0172540923167676e-05,
      "loss": 0.001,
      "step": 3192
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0157793835717446e-05,
      "loss": 0.0001,
      "step": 3194
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0143046748267217e-05,
      "loss": 0.0001,
      "step": 3196
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0128299660816987e-05,
      "loss": 0.0001,
      "step": 3198
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.011355257336676e-05,
      "loss": 0.0001,
      "step": 3200
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.009880548591653e-05,
      "loss": 0.0004,
      "step": 3202
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.00840583984663e-05,
      "loss": 0.0002,
      "step": 3204
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0069311311016075e-05,
      "loss": 0.0001,
      "step": 3206
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0054564223565846e-05,
      "loss": 0.0001,
      "step": 3208
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0039817136115616e-05,
      "loss": 0.0023,
      "step": 3210
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.0025070048665386e-05,
      "loss": 0.0001,
      "step": 3212
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.001032296121516e-05,
      "loss": 0.0015,
      "step": 3214
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.999557587376493e-05,
      "loss": 0.0003,
      "step": 3216
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.99808287863147e-05,
      "loss": 0.0002,
      "step": 3218
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9966081698864475e-05,
      "loss": 0.0001,
      "step": 3220
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9951334611414245e-05,
      "loss": 0.0001,
      "step": 3222
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9936587523964015e-05,
      "loss": 0.0002,
      "step": 3224
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9921840436513786e-05,
      "loss": 0.0001,
      "step": 3226
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9907093349063563e-05,
      "loss": 0.0002,
      "step": 3228
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9892346261613337e-05,
      "loss": 0.0001,
      "step": 3230
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9877599174163107e-05,
      "loss": 0.0002,
      "step": 3232
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9862852086712877e-05,
      "loss": 0.0001,
      "step": 3234
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9848104999262648e-05,
      "loss": 0.0001,
      "step": 3236
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.983335791181242e-05,
      "loss": 0.0001,
      "step": 3238
    },
    {
      "epoch": 1.33,
      "learning_rate": 2.9818610824362192e-05,
      "loss": 0.0001,
      "step": 3240
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9803863736911962e-05,
      "loss": 0.0001,
      "step": 3242
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9789116649461736e-05,
      "loss": 0.0001,
      "step": 3244
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9774369562011506e-05,
      "loss": 0.0001,
      "step": 3246
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9759622474561277e-05,
      "loss": 0.0001,
      "step": 3248
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9744875387111047e-05,
      "loss": 0.0001,
      "step": 3250
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.973012829966082e-05,
      "loss": 0.0002,
      "step": 3252
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.971538121221059e-05,
      "loss": 0.0001,
      "step": 3254
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.970063412476036e-05,
      "loss": 0.0002,
      "step": 3256
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9685887037310135e-05,
      "loss": 0.0001,
      "step": 3258
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9671139949859906e-05,
      "loss": 0.0001,
      "step": 3260
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9656392862409676e-05,
      "loss": 0.0001,
      "step": 3262
    },
    {
      "epoch": 1.34,
      "learning_rate": 2.9641645774959446e-05,
      "loss": 0.0003,
      "step": 3264
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.962689868750922e-05,
      "loss": 0.0001,
      "step": 3266
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.961215160005899e-05,
      "loss": 0.0001,
      "step": 3268
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.959740451260876e-05,
      "loss": 0.0001,
      "step": 3270
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.9582657425158535e-05,
      "loss": 0.0016,
      "step": 3272
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.9567910337708305e-05,
      "loss": 0.0001,
      "step": 3274
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.9553163250258075e-05,
      "loss": 0.0001,
      "step": 3276
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.9538416162807846e-05,
      "loss": 0.0001,
      "step": 3278
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.952366907535762e-05,
      "loss": 0.0001,
      "step": 3280
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.950892198790739e-05,
      "loss": 0.0001,
      "step": 3282
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.949417490045716e-05,
      "loss": 0.0001,
      "step": 3284
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.9479427813006934e-05,
      "loss": 0.0001,
      "step": 3286
    },
    {
      "epoch": 1.35,
      "learning_rate": 2.9464680725556704e-05,
      "loss": 0.0001,
      "step": 3288
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.9449933638106475e-05,
      "loss": 0.0001,
      "step": 3290
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.9435186550656245e-05,
      "loss": 0.0001,
      "step": 3292
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.942043946320602e-05,
      "loss": 0.0001,
      "step": 3294
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.940569237575579e-05,
      "loss": 0.0001,
      "step": 3296
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.939094528830556e-05,
      "loss": 0.0001,
      "step": 3298
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.937619820085533e-05,
      "loss": 0.0001,
      "step": 3300
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.9361451113405104e-05,
      "loss": 0.0001,
      "step": 3302
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.9346704025954874e-05,
      "loss": 0.0001,
      "step": 3304
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.9331956938504644e-05,
      "loss": 0.0001,
      "step": 3306
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.9317209851054418e-05,
      "loss": 0.0001,
      "step": 3308
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.930246276360419e-05,
      "loss": 0.0002,
      "step": 3310
    },
    {
      "epoch": 1.36,
      "learning_rate": 2.928771567615396e-05,
      "loss": 0.0001,
      "step": 3312
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.927296858870373e-05,
      "loss": 0.0001,
      "step": 3314
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9258221501253503e-05,
      "loss": 0.0001,
      "step": 3316
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9243474413803273e-05,
      "loss": 0.0001,
      "step": 3318
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9228727326353044e-05,
      "loss": 0.0001,
      "step": 3320
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9213980238902818e-05,
      "loss": 0.0001,
      "step": 3322
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9199233151452588e-05,
      "loss": 0.0001,
      "step": 3324
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9184486064002358e-05,
      "loss": 0.0001,
      "step": 3326
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.916973897655213e-05,
      "loss": 0.0001,
      "step": 3328
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9154991889101902e-05,
      "loss": 0.0001,
      "step": 3330
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9140244801651673e-05,
      "loss": 0.0001,
      "step": 3332
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9125497714201443e-05,
      "loss": 0.0001,
      "step": 3334
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.9110750626751217e-05,
      "loss": 0.0001,
      "step": 3336
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.909600353930099e-05,
      "loss": 0.0001,
      "step": 3338
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.9081256451850764e-05,
      "loss": 0.0001,
      "step": 3340
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.9066509364400535e-05,
      "loss": 0.0001,
      "step": 3342
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.9051762276950305e-05,
      "loss": 0.0001,
      "step": 3344
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.903701518950008e-05,
      "loss": 0.0001,
      "step": 3346
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.902226810204985e-05,
      "loss": 0.0001,
      "step": 3348
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.900752101459962e-05,
      "loss": 0.0001,
      "step": 3350
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.899277392714939e-05,
      "loss": 0.0002,
      "step": 3352
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.8978026839699164e-05,
      "loss": 0.0001,
      "step": 3354
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.8963279752248934e-05,
      "loss": 0.0001,
      "step": 3356
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.8948532664798704e-05,
      "loss": 0.0001,
      "step": 3358
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.8933785577348478e-05,
      "loss": 0.0001,
      "step": 3360
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.891903848989825e-05,
      "loss": 0.0001,
      "step": 3362
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.890429140244802e-05,
      "loss": 0.0001,
      "step": 3364
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.888954431499779e-05,
      "loss": 0.0001,
      "step": 3366
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8874797227547563e-05,
      "loss": 0.0001,
      "step": 3368
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8860050140097333e-05,
      "loss": 0.0,
      "step": 3370
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8845303052647104e-05,
      "loss": 0.0001,
      "step": 3372
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8830555965196878e-05,
      "loss": 0.0001,
      "step": 3374
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8815808877746648e-05,
      "loss": 0.0,
      "step": 3376
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8801061790296418e-05,
      "loss": 0.0001,
      "step": 3378
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.878631470284619e-05,
      "loss": 0.0001,
      "step": 3380
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8771567615395962e-05,
      "loss": 0.0001,
      "step": 3382
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.8756820527945733e-05,
      "loss": 0.0,
      "step": 3384
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8742073440495503e-05,
      "loss": 0.0001,
      "step": 3386
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8727326353045277e-05,
      "loss": 0.0,
      "step": 3388
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8712579265595047e-05,
      "loss": 0.0,
      "step": 3390
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8697832178144818e-05,
      "loss": 0.0001,
      "step": 3392
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8683085090694588e-05,
      "loss": 0.0001,
      "step": 3394
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8668338003244362e-05,
      "loss": 0.0001,
      "step": 3396
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8653590915794132e-05,
      "loss": 0.0001,
      "step": 3398
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8638843828343902e-05,
      "loss": 0.0002,
      "step": 3400
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8624096740893673e-05,
      "loss": 0.0,
      "step": 3402
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8609349653443447e-05,
      "loss": 0.0001,
      "step": 3404
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8594602565993217e-05,
      "loss": 0.0,
      "step": 3406
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.8579855478542987e-05,
      "loss": 0.0001,
      "step": 3408
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.856510839109276e-05,
      "loss": 0.0001,
      "step": 3410
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.855036130364253e-05,
      "loss": 0.0001,
      "step": 3412
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8535614216192302e-05,
      "loss": 0.0,
      "step": 3414
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8520867128742072e-05,
      "loss": 0.0001,
      "step": 3416
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8506120041291846e-05,
      "loss": 0.0001,
      "step": 3418
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8491372953841616e-05,
      "loss": 0.0001,
      "step": 3420
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8476625866391387e-05,
      "loss": 0.0,
      "step": 3422
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.846187877894116e-05,
      "loss": 0.0,
      "step": 3424
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.844713169149093e-05,
      "loss": 0.0001,
      "step": 3426
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.84323846040407e-05,
      "loss": 0.0001,
      "step": 3428
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.841763751659047e-05,
      "loss": 0.0001,
      "step": 3430
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8402890429140245e-05,
      "loss": 0.0001,
      "step": 3432
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.8388143341690016e-05,
      "loss": 0.0001,
      "step": 3434
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8373396254239786e-05,
      "loss": 0.0,
      "step": 3436
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.835864916678956e-05,
      "loss": 0.0001,
      "step": 3438
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.834390207933933e-05,
      "loss": 0.0001,
      "step": 3440
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.83291549918891e-05,
      "loss": 0.0,
      "step": 3442
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.831440790443887e-05,
      "loss": 0.0001,
      "step": 3444
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8299660816988645e-05,
      "loss": 0.0,
      "step": 3446
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8284913729538415e-05,
      "loss": 0.0,
      "step": 3448
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8270166642088192e-05,
      "loss": 0.0001,
      "step": 3450
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8255419554637962e-05,
      "loss": 0.0,
      "step": 3452
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8240672467187733e-05,
      "loss": 0.0001,
      "step": 3454
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8225925379737507e-05,
      "loss": 0.0001,
      "step": 3456
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.8211178292287277e-05,
      "loss": 0.0001,
      "step": 3458
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.8196431204837047e-05,
      "loss": 0.0001,
      "step": 3460
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.818168411738682e-05,
      "loss": 0.0,
      "step": 3462
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.816693702993659e-05,
      "loss": 0.0,
      "step": 3464
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.8152189942486362e-05,
      "loss": 0.0,
      "step": 3466
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.8137442855036132e-05,
      "loss": 0.0001,
      "step": 3468
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.8122695767585906e-05,
      "loss": 0.0,
      "step": 3470
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.8107948680135676e-05,
      "loss": 0.0,
      "step": 3472
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.8093201592685447e-05,
      "loss": 0.0001,
      "step": 3474
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.807845450523522e-05,
      "loss": 0.0,
      "step": 3476
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.806370741778499e-05,
      "loss": 0.0,
      "step": 3478
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.804896033033476e-05,
      "loss": 0.0001,
      "step": 3480
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.803421324288453e-05,
      "loss": 0.0,
      "step": 3482
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.8019466155434305e-05,
      "loss": 0.0001,
      "step": 3484
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.8004719067984076e-05,
      "loss": 0.0001,
      "step": 3486
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.7989971980533846e-05,
      "loss": 0.0,
      "step": 3488
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.7975224893083616e-05,
      "loss": 0.0,
      "step": 3490
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.796047780563339e-05,
      "loss": 0.0001,
      "step": 3492
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.794573071818316e-05,
      "loss": 0.0,
      "step": 3494
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.793098363073293e-05,
      "loss": 0.0001,
      "step": 3496
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.7916236543282705e-05,
      "loss": 0.0,
      "step": 3498
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.7901489455832475e-05,
      "loss": 0.0001,
      "step": 3500
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.7886742368382245e-05,
      "loss": 0.0001,
      "step": 3502
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.7871995280932016e-05,
      "loss": 0.0001,
      "step": 3504
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.785724819348179e-05,
      "loss": 0.0,
      "step": 3506
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.784250110603156e-05,
      "loss": 0.0001,
      "step": 3508
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.782775401858133e-05,
      "loss": 0.0,
      "step": 3510
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7813006931131104e-05,
      "loss": 0.0,
      "step": 3512
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7798259843680874e-05,
      "loss": 0.0001,
      "step": 3514
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7783512756230645e-05,
      "loss": 0.0002,
      "step": 3516
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7768765668780415e-05,
      "loss": 0.0,
      "step": 3518
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.775401858133019e-05,
      "loss": 0.0003,
      "step": 3520
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.773927149387996e-05,
      "loss": 0.0,
      "step": 3522
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.772452440642973e-05,
      "loss": 0.0,
      "step": 3524
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7709777318979503e-05,
      "loss": 0.0,
      "step": 3526
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7695030231529274e-05,
      "loss": 0.0,
      "step": 3528
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.7680283144079044e-05,
      "loss": 0.0,
      "step": 3530
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7665536056628814e-05,
      "loss": 0.0,
      "step": 3532
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7650788969178588e-05,
      "loss": 0.0,
      "step": 3534
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.763604188172836e-05,
      "loss": 0.0001,
      "step": 3536
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.762129479427813e-05,
      "loss": 0.0,
      "step": 3538
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7606547706827903e-05,
      "loss": 0.0001,
      "step": 3540
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7591800619377673e-05,
      "loss": 0.0001,
      "step": 3542
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7577053531927443e-05,
      "loss": 0.0001,
      "step": 3544
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7562306444477214e-05,
      "loss": 0.0,
      "step": 3546
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7547559357026987e-05,
      "loss": 0.0001,
      "step": 3548
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7532812269576758e-05,
      "loss": 0.0,
      "step": 3550
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.7518065182126528e-05,
      "loss": 0.0001,
      "step": 3552
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.75033180946763e-05,
      "loss": 0.0,
      "step": 3554
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7488571007226072e-05,
      "loss": 0.0001,
      "step": 3556
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7473823919775843e-05,
      "loss": 0.0001,
      "step": 3558
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7459076832325613e-05,
      "loss": 0.0001,
      "step": 3560
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.744432974487539e-05,
      "loss": 0.0,
      "step": 3562
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7429582657425164e-05,
      "loss": 0.0,
      "step": 3564
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7414835569974934e-05,
      "loss": 0.0001,
      "step": 3566
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7400088482524705e-05,
      "loss": 0.0,
      "step": 3568
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7385341395074475e-05,
      "loss": 0.0,
      "step": 3570
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.737059430762425e-05,
      "loss": 0.0,
      "step": 3572
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.735584722017402e-05,
      "loss": 0.0,
      "step": 3574
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.734110013272379e-05,
      "loss": 0.0001,
      "step": 3576
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.7326353045273563e-05,
      "loss": 0.0007,
      "step": 3578
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7311605957823334e-05,
      "loss": 0.0,
      "step": 3580
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7296858870373104e-05,
      "loss": 0.0,
      "step": 3582
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7282111782922874e-05,
      "loss": 0.0001,
      "step": 3584
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7267364695472648e-05,
      "loss": 0.0,
      "step": 3586
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.725261760802242e-05,
      "loss": 0.0001,
      "step": 3588
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.723787052057219e-05,
      "loss": 0.0001,
      "step": 3590
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.722312343312196e-05,
      "loss": 0.0001,
      "step": 3592
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7208376345671733e-05,
      "loss": 0.0001,
      "step": 3594
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7193629258221503e-05,
      "loss": 0.0003,
      "step": 3596
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7178882170771274e-05,
      "loss": 0.0001,
      "step": 3598
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7164135083321047e-05,
      "loss": 0.0001,
      "step": 3600
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7149387995870818e-05,
      "loss": 0.0,
      "step": 3602
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.7134640908420588e-05,
      "loss": 0.0,
      "step": 3604
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.711989382097036e-05,
      "loss": 0.0001,
      "step": 3606
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7105146733520132e-05,
      "loss": 0.0,
      "step": 3608
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7090399646069903e-05,
      "loss": 0.0,
      "step": 3610
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7075652558619673e-05,
      "loss": 0.0,
      "step": 3612
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7060905471169447e-05,
      "loss": 0.0,
      "step": 3614
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7046158383719217e-05,
      "loss": 0.0,
      "step": 3616
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7031411296268988e-05,
      "loss": 0.0,
      "step": 3618
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.7016664208818758e-05,
      "loss": 0.0001,
      "step": 3620
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.700191712136853e-05,
      "loss": 0.0001,
      "step": 3622
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.6987170033918302e-05,
      "loss": 0.0001,
      "step": 3624
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.6972422946468072e-05,
      "loss": 0.0,
      "step": 3626
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.6957675859017846e-05,
      "loss": 0.0001,
      "step": 3628
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6942928771567617e-05,
      "loss": 0.0001,
      "step": 3630
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6928181684117387e-05,
      "loss": 0.0001,
      "step": 3632
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6913434596667157e-05,
      "loss": 0.0001,
      "step": 3634
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.689868750921693e-05,
      "loss": 0.0,
      "step": 3636
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.68839404217667e-05,
      "loss": 0.0001,
      "step": 3638
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6869193334316472e-05,
      "loss": 0.0,
      "step": 3640
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6854446246866246e-05,
      "loss": 0.0001,
      "step": 3642
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6839699159416016e-05,
      "loss": 0.0,
      "step": 3644
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6824952071965786e-05,
      "loss": 0.0001,
      "step": 3646
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.6810204984515557e-05,
      "loss": 0.0,
      "step": 3648
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.679545789706533e-05,
      "loss": 0.0001,
      "step": 3650
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.67807108096151e-05,
      "loss": 0.0001,
      "step": 3652
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.676596372216487e-05,
      "loss": 0.0001,
      "step": 3654
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.675121663471464e-05,
      "loss": 0.0,
      "step": 3656
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6736469547264415e-05,
      "loss": 0.0001,
      "step": 3658
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6721722459814186e-05,
      "loss": 0.0001,
      "step": 3660
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6706975372363956e-05,
      "loss": 0.0001,
      "step": 3662
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.669222828491373e-05,
      "loss": 0.0004,
      "step": 3664
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.66774811974635e-05,
      "loss": 0.0,
      "step": 3666
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.666273411001327e-05,
      "loss": 0.0003,
      "step": 3668
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.664798702256304e-05,
      "loss": 0.0001,
      "step": 3670
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6633239935112818e-05,
      "loss": 0.0,
      "step": 3672
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6618492847662592e-05,
      "loss": 0.0,
      "step": 3674
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.6603745760212362e-05,
      "loss": 0.0,
      "step": 3676
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6588998672762132e-05,
      "loss": 0.0001,
      "step": 3678
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6574251585311906e-05,
      "loss": 0.0001,
      "step": 3680
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6559504497861677e-05,
      "loss": 0.0,
      "step": 3682
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6544757410411447e-05,
      "loss": 0.0001,
      "step": 3684
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6530010322961217e-05,
      "loss": 0.0001,
      "step": 3686
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.651526323551099e-05,
      "loss": 0.0001,
      "step": 3688
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.650051614806076e-05,
      "loss": 0.0,
      "step": 3690
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6485769060610532e-05,
      "loss": 0.0001,
      "step": 3692
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6471021973160302e-05,
      "loss": 0.0001,
      "step": 3694
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6456274885710076e-05,
      "loss": 0.0001,
      "step": 3696
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6441527798259846e-05,
      "loss": 0.0001,
      "step": 3698
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.6426780710809617e-05,
      "loss": 0.0,
      "step": 3700
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.641203362335939e-05,
      "loss": 0.0,
      "step": 3702
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.639728653590916e-05,
      "loss": 0.0001,
      "step": 3704
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.638253944845893e-05,
      "loss": 0.0,
      "step": 3706
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.63677923610087e-05,
      "loss": 0.0,
      "step": 3708
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.6353045273558475e-05,
      "loss": 0.0001,
      "step": 3710
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.6338298186108246e-05,
      "loss": 0.0001,
      "step": 3712
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.6323551098658016e-05,
      "loss": 0.0001,
      "step": 3714
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.630880401120779e-05,
      "loss": 0.0,
      "step": 3716
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.629405692375756e-05,
      "loss": 0.0,
      "step": 3718
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.627930983630733e-05,
      "loss": 0.0,
      "step": 3720
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.62645627488571e-05,
      "loss": 0.0001,
      "step": 3722
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.6249815661406875e-05,
      "loss": 0.0,
      "step": 3724
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6235068573956645e-05,
      "loss": 0.0003,
      "step": 3726
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6220321486506415e-05,
      "loss": 0.0,
      "step": 3728
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.620557439905619e-05,
      "loss": 0.0,
      "step": 3730
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.619082731160596e-05,
      "loss": 0.0001,
      "step": 3732
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.617608022415573e-05,
      "loss": 0.0,
      "step": 3734
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.61613331367055e-05,
      "loss": 0.0,
      "step": 3736
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6146586049255274e-05,
      "loss": 0.0001,
      "step": 3738
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6131838961805044e-05,
      "loss": 0.0001,
      "step": 3740
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.6117091874354815e-05,
      "loss": 0.0001,
      "step": 3742
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.610234478690459e-05,
      "loss": 0.0001,
      "step": 3744
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.608759769945436e-05,
      "loss": 0.0,
      "step": 3746
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.607285061200413e-05,
      "loss": 0.0001,
      "step": 3748
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.60581035245539e-05,
      "loss": 0.0001,
      "step": 3750
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.6043356437103673e-05,
      "loss": 0.0,
      "step": 3752
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.6028609349653444e-05,
      "loss": 0.0,
      "step": 3754
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.6013862262203214e-05,
      "loss": 0.0001,
      "step": 3756
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5999115174752984e-05,
      "loss": 0.0,
      "step": 3758
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5984368087302758e-05,
      "loss": 0.0,
      "step": 3760
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.596962099985253e-05,
      "loss": 0.0001,
      "step": 3762
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.59548739124023e-05,
      "loss": 0.0,
      "step": 3764
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5940126824952073e-05,
      "loss": 0.0001,
      "step": 3766
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5925379737501843e-05,
      "loss": 0.0,
      "step": 3768
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5910632650051613e-05,
      "loss": 0.0001,
      "step": 3770
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.5895885562601384e-05,
      "loss": 0.0,
      "step": 3772
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5881138475151157e-05,
      "loss": 0.0001,
      "step": 3774
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5866391387700928e-05,
      "loss": 0.0001,
      "step": 3776
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5851644300250698e-05,
      "loss": 0.0,
      "step": 3778
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5836897212800472e-05,
      "loss": 0.0001,
      "step": 3780
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5822150125350242e-05,
      "loss": 0.0001,
      "step": 3782
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.580740303790002e-05,
      "loss": 0.0,
      "step": 3784
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.579265595044979e-05,
      "loss": 0.0,
      "step": 3786
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.577790886299956e-05,
      "loss": 0.0,
      "step": 3788
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5763161775549334e-05,
      "loss": 0.0,
      "step": 3790
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5748414688099104e-05,
      "loss": 0.0001,
      "step": 3792
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5733667600648875e-05,
      "loss": 0.0001,
      "step": 3794
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.5718920513198645e-05,
      "loss": 0.0,
      "step": 3796
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.570417342574842e-05,
      "loss": 0.0002,
      "step": 3798
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.568942633829819e-05,
      "loss": 0.0,
      "step": 3800
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.567467925084796e-05,
      "loss": 0.0,
      "step": 3802
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5659932163397733e-05,
      "loss": 0.0001,
      "step": 3804
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5645185075947504e-05,
      "loss": 0.0,
      "step": 3806
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5630437988497274e-05,
      "loss": 0.0,
      "step": 3808
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5615690901047044e-05,
      "loss": 0.0,
      "step": 3810
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5600943813596818e-05,
      "loss": 0.0,
      "step": 3812
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.558619672614659e-05,
      "loss": 0.0,
      "step": 3814
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.557144963869636e-05,
      "loss": 0.0,
      "step": 3816
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5556702551246133e-05,
      "loss": 0.0,
      "step": 3818
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5541955463795903e-05,
      "loss": 0.0,
      "step": 3820
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.5527208376345673e-05,
      "loss": 0.0001,
      "step": 3822
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5512461288895444e-05,
      "loss": 0.0,
      "step": 3824
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5497714201445217e-05,
      "loss": 0.0,
      "step": 3826
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5482967113994988e-05,
      "loss": 0.0,
      "step": 3828
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5468220026544758e-05,
      "loss": 0.0,
      "step": 3830
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5453472939094532e-05,
      "loss": 0.0,
      "step": 3832
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5438725851644302e-05,
      "loss": 0.0,
      "step": 3834
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5423978764194073e-05,
      "loss": 0.0,
      "step": 3836
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5409231676743843e-05,
      "loss": 0.0001,
      "step": 3838
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5394484589293617e-05,
      "loss": 0.0,
      "step": 3840
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5379737501843387e-05,
      "loss": 0.0,
      "step": 3842
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5364990414393158e-05,
      "loss": 0.0001,
      "step": 3844
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.5350243326942928e-05,
      "loss": 0.0001,
      "step": 3846
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.53354962394927e-05,
      "loss": 0.0003,
      "step": 3848
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5320749152042472e-05,
      "loss": 0.0,
      "step": 3850
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5306002064592242e-05,
      "loss": 0.0,
      "step": 3852
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5291254977142016e-05,
      "loss": 0.0,
      "step": 3854
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5276507889691787e-05,
      "loss": 0.0,
      "step": 3856
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5261760802241557e-05,
      "loss": 0.0001,
      "step": 3858
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5247013714791327e-05,
      "loss": 0.0001,
      "step": 3860
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.52322666273411e-05,
      "loss": 0.0,
      "step": 3862
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.521751953989087e-05,
      "loss": 0.0001,
      "step": 3864
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5202772452440642e-05,
      "loss": 0.0006,
      "step": 3866
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5188025364990415e-05,
      "loss": 0.0,
      "step": 3868
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.5173278277540186e-05,
      "loss": 0.0,
      "step": 3870
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.5158531190089956e-05,
      "loss": 0.0,
      "step": 3872
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.5143784102639727e-05,
      "loss": 0.0001,
      "step": 3874
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.51290370151895e-05,
      "loss": 0.0,
      "step": 3876
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.511428992773927e-05,
      "loss": 0.0,
      "step": 3878
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.509954284028904e-05,
      "loss": 0.0,
      "step": 3880
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.5084795752838815e-05,
      "loss": 0.0,
      "step": 3882
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.5070048665388585e-05,
      "loss": 0.0,
      "step": 3884
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.5055301577938356e-05,
      "loss": 0.0001,
      "step": 3886
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.5040554490488126e-05,
      "loss": 0.0001,
      "step": 3888
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.50258074030379e-05,
      "loss": 0.0001,
      "step": 3890
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.501106031558767e-05,
      "loss": 0.0001,
      "step": 3892
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.4996313228137444e-05,
      "loss": 0.0001,
      "step": 3894
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4981566140687214e-05,
      "loss": 0.0,
      "step": 3896
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4966819053236988e-05,
      "loss": 0.0,
      "step": 3898
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4952071965786758e-05,
      "loss": 0.0,
      "step": 3900
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.493732487833653e-05,
      "loss": 0.0001,
      "step": 3902
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.49225777908863e-05,
      "loss": 0.0001,
      "step": 3904
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4907830703436073e-05,
      "loss": 0.0001,
      "step": 3906
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4893083615985843e-05,
      "loss": 0.0,
      "step": 3908
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4878336528535613e-05,
      "loss": 0.0001,
      "step": 3910
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4863589441085387e-05,
      "loss": 0.0,
      "step": 3912
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4848842353635158e-05,
      "loss": 0.0001,
      "step": 3914
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.4834095266184928e-05,
      "loss": 0.0001,
      "step": 3916
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.48193481787347e-05,
      "loss": 0.0,
      "step": 3918
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4804601091284472e-05,
      "loss": 0.0,
      "step": 3920
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4789854003834246e-05,
      "loss": 0.0001,
      "step": 3922
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4775106916384016e-05,
      "loss": 0.0,
      "step": 3924
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4760359828933787e-05,
      "loss": 0.0001,
      "step": 3926
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.474561274148356e-05,
      "loss": 0.0001,
      "step": 3928
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.473086565403333e-05,
      "loss": 0.0001,
      "step": 3930
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.47161185665831e-05,
      "loss": 0.0004,
      "step": 3932
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4701371479132875e-05,
      "loss": 0.0,
      "step": 3934
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4686624391682645e-05,
      "loss": 0.0,
      "step": 3936
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4671877304232416e-05,
      "loss": 0.0,
      "step": 3938
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.4657130216782186e-05,
      "loss": 0.0002,
      "step": 3940
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.464238312933196e-05,
      "loss": 0.0,
      "step": 3942
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.462763604188173e-05,
      "loss": 0.0001,
      "step": 3944
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.46128889544315e-05,
      "loss": 0.0001,
      "step": 3946
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.459814186698127e-05,
      "loss": 0.0,
      "step": 3948
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.4583394779531045e-05,
      "loss": 0.0001,
      "step": 3950
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.4568647692080815e-05,
      "loss": 0.0001,
      "step": 3952
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.4553900604630585e-05,
      "loss": 0.0,
      "step": 3954
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.453915351718036e-05,
      "loss": 0.0,
      "step": 3956
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.452440642973013e-05,
      "loss": 0.0,
      "step": 3958
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.45096593422799e-05,
      "loss": 0.0,
      "step": 3960
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.449491225482967e-05,
      "loss": 0.0,
      "step": 3962
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.4480165167379444e-05,
      "loss": 0.0,
      "step": 3964
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.4465418079929214e-05,
      "loss": 0.0,
      "step": 3966
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.4450670992478985e-05,
      "loss": 0.0009,
      "step": 3968
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.443592390502876e-05,
      "loss": 0.0001,
      "step": 3970
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.442117681757853e-05,
      "loss": 0.0001,
      "step": 3972
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.44064297301283e-05,
      "loss": 0.0001,
      "step": 3974
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.439168264267807e-05,
      "loss": 0.0001,
      "step": 3976
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4376935555227847e-05,
      "loss": 0.0001,
      "step": 3978
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4362188467777617e-05,
      "loss": 0.0003,
      "step": 3980
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4347441380327387e-05,
      "loss": 0.0001,
      "step": 3982
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4332694292877158e-05,
      "loss": 0.0001,
      "step": 3984
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.431794720542693e-05,
      "loss": 0.0001,
      "step": 3986
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4303200117976702e-05,
      "loss": 0.0001,
      "step": 3988
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4288453030526472e-05,
      "loss": 0.0001,
      "step": 3990
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.4273705943076243e-05,
      "loss": 0.0001,
      "step": 3992
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4258958855626016e-05,
      "loss": 0.0001,
      "step": 3994
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4244211768175787e-05,
      "loss": 0.0002,
      "step": 3996
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4229464680725557e-05,
      "loss": 0.0001,
      "step": 3998
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.421471759327533e-05,
      "loss": 0.0001,
      "step": 4000
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.41999705058251e-05,
      "loss": 0.0002,
      "step": 4002
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.418522341837487e-05,
      "loss": 0.0002,
      "step": 4004
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4170476330924642e-05,
      "loss": 0.0001,
      "step": 4006
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4155729243474416e-05,
      "loss": 0.0001,
      "step": 4008
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4140982156024186e-05,
      "loss": 0.0001,
      "step": 4010
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.4126235068573956e-05,
      "loss": 0.0001,
      "step": 4012
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.411148798112373e-05,
      "loss": 0.0001,
      "step": 4014
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.40967408936735e-05,
      "loss": 0.0001,
      "step": 4016
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.408199380622327e-05,
      "loss": 0.0003,
      "step": 4018
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.406724671877304e-05,
      "loss": 0.0001,
      "step": 4020
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.4052499631322815e-05,
      "loss": 0.0001,
      "step": 4022
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.4037752543872585e-05,
      "loss": 0.0001,
      "step": 4024
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.4023005456422356e-05,
      "loss": 0.0001,
      "step": 4026
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.400825836897213e-05,
      "loss": 0.0,
      "step": 4028
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.39935112815219e-05,
      "loss": 0.0001,
      "step": 4030
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.397876419407167e-05,
      "loss": 0.0,
      "step": 4032
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.3964017106621444e-05,
      "loss": 0.0001,
      "step": 4034
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.3949270019171218e-05,
      "loss": 0.0001,
      "step": 4036
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.3934522931720988e-05,
      "loss": 0.0001,
      "step": 4038
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.391977584427076e-05,
      "loss": 0.0001,
      "step": 4040
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.390502875682053e-05,
      "loss": 0.0,
      "step": 4042
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3890281669370303e-05,
      "loss": 0.0001,
      "step": 4044
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3875534581920073e-05,
      "loss": 0.0001,
      "step": 4046
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3860787494469843e-05,
      "loss": 0.0,
      "step": 4048
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3846040407019614e-05,
      "loss": 0.0001,
      "step": 4050
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3831293319569387e-05,
      "loss": 0.0,
      "step": 4052
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3816546232119158e-05,
      "loss": 0.0001,
      "step": 4054
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3801799144668928e-05,
      "loss": 0.002,
      "step": 4056
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3787052057218702e-05,
      "loss": 0.0001,
      "step": 4058
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3772304969768472e-05,
      "loss": 0.0001,
      "step": 4060
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3757557882318243e-05,
      "loss": 0.0,
      "step": 4062
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.3742810794868013e-05,
      "loss": 0.0001,
      "step": 4064
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3728063707417787e-05,
      "loss": 0.0001,
      "step": 4066
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3713316619967557e-05,
      "loss": 0.0001,
      "step": 4068
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3698569532517327e-05,
      "loss": 0.0001,
      "step": 4070
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.36838224450671e-05,
      "loss": 0.0001,
      "step": 4072
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.366907535761687e-05,
      "loss": 0.0001,
      "step": 4074
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3654328270166642e-05,
      "loss": 0.0001,
      "step": 4076
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3639581182716412e-05,
      "loss": 0.0001,
      "step": 4078
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3624834095266186e-05,
      "loss": 0.0001,
      "step": 4080
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3610087007815956e-05,
      "loss": 0.0001,
      "step": 4082
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.3595339920365727e-05,
      "loss": 0.0002,
      "step": 4084
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.35805928329155e-05,
      "loss": 0.0001,
      "step": 4086
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.356584574546527e-05,
      "loss": 0.0001,
      "step": 4088
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3551098658015045e-05,
      "loss": 0.0025,
      "step": 4090
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3536351570564815e-05,
      "loss": 0.0001,
      "step": 4092
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3521604483114585e-05,
      "loss": 0.0001,
      "step": 4094
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.350685739566436e-05,
      "loss": 0.0001,
      "step": 4096
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.349211030821413e-05,
      "loss": 0.0001,
      "step": 4098
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.34773632207639e-05,
      "loss": 0.0001,
      "step": 4100
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3462616133313674e-05,
      "loss": 0.0001,
      "step": 4102
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3447869045863444e-05,
      "loss": 0.0002,
      "step": 4104
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3433121958413214e-05,
      "loss": 0.0001,
      "step": 4106
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.3418374870962985e-05,
      "loss": 0.0002,
      "step": 4108
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.340362778351276e-05,
      "loss": 0.0004,
      "step": 4110
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.338888069606253e-05,
      "loss": 0.0002,
      "step": 4112
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.33741336086123e-05,
      "loss": 0.0001,
      "step": 4114
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3359386521162073e-05,
      "loss": 0.0001,
      "step": 4116
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3344639433711843e-05,
      "loss": 0.0001,
      "step": 4118
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3329892346261614e-05,
      "loss": 0.0001,
      "step": 4120
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3315145258811384e-05,
      "loss": 0.0001,
      "step": 4122
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3300398171361158e-05,
      "loss": 0.0001,
      "step": 4124
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3285651083910928e-05,
      "loss": 0.0001,
      "step": 4126
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.32709039964607e-05,
      "loss": 0.0001,
      "step": 4128
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3256156909010472e-05,
      "loss": 0.0,
      "step": 4130
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3241409821560243e-05,
      "loss": 0.0001,
      "step": 4132
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3226662734110013e-05,
      "loss": 0.0001,
      "step": 4134
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3211915646659783e-05,
      "loss": 0.0003,
      "step": 4136
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.3197168559209557e-05,
      "loss": 0.0001,
      "step": 4138
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3182421471759328e-05,
      "loss": 0.0001,
      "step": 4140
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3167674384309098e-05,
      "loss": 0.0,
      "step": 4142
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3152927296858872e-05,
      "loss": 0.0001,
      "step": 4144
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3138180209408645e-05,
      "loss": 0.0001,
      "step": 4146
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3123433121958416e-05,
      "loss": 0.0,
      "step": 4148
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3108686034508186e-05,
      "loss": 0.0001,
      "step": 4150
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3093938947057957e-05,
      "loss": 0.0001,
      "step": 4152
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.307919185960773e-05,
      "loss": 0.0001,
      "step": 4154
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.30644447721575e-05,
      "loss": 0.0002,
      "step": 4156
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.304969768470727e-05,
      "loss": 0.0002,
      "step": 4158
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3034950597257045e-05,
      "loss": 0.0,
      "step": 4160
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3020203509806815e-05,
      "loss": 0.0001,
      "step": 4162
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.3005456422356586e-05,
      "loss": 0.0001,
      "step": 4164
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.2990709334906356e-05,
      "loss": 0.0,
      "step": 4166
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.297596224745613e-05,
      "loss": 0.0001,
      "step": 4168
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.29612151600059e-05,
      "loss": 0.0001,
      "step": 4170
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.294646807255567e-05,
      "loss": 0.0001,
      "step": 4172
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.2931720985105444e-05,
      "loss": 0.0001,
      "step": 4174
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.2916973897655214e-05,
      "loss": 0.0001,
      "step": 4176
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.2902226810204985e-05,
      "loss": 0.0,
      "step": 4178
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.2887479722754755e-05,
      "loss": 0.0001,
      "step": 4180
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.287273263530453e-05,
      "loss": 0.0001,
      "step": 4182
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.28579855478543e-05,
      "loss": 0.0001,
      "step": 4184
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.284323846040407e-05,
      "loss": 0.0001,
      "step": 4186
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2828491372953843e-05,
      "loss": 0.0001,
      "step": 4188
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2813744285503614e-05,
      "loss": 0.0,
      "step": 4190
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2798997198053384e-05,
      "loss": 0.0001,
      "step": 4192
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2784250110603155e-05,
      "loss": 0.0,
      "step": 4194
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.276950302315293e-05,
      "loss": 0.0001,
      "step": 4196
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.27547559357027e-05,
      "loss": 0.0001,
      "step": 4198
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2740008848252472e-05,
      "loss": 0.0001,
      "step": 4200
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2725261760802243e-05,
      "loss": 0.0,
      "step": 4202
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2710514673352017e-05,
      "loss": 0.0001,
      "step": 4204
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2695767585901787e-05,
      "loss": 0.0001,
      "step": 4206
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2681020498451557e-05,
      "loss": 0.0001,
      "step": 4208
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.2666273411001328e-05,
      "loss": 0.0,
      "step": 4210
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.26515263235511e-05,
      "loss": 0.0,
      "step": 4212
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2636779236100872e-05,
      "loss": 0.0,
      "step": 4214
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2622032148650642e-05,
      "loss": 0.0,
      "step": 4216
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2607285061200416e-05,
      "loss": 0.0,
      "step": 4218
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2592537973750186e-05,
      "loss": 0.0001,
      "step": 4220
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2577790886299957e-05,
      "loss": 0.0,
      "step": 4222
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2563043798849727e-05,
      "loss": 0.0001,
      "step": 4224
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.25482967113995e-05,
      "loss": 0.0,
      "step": 4226
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.253354962394927e-05,
      "loss": 0.0001,
      "step": 4228
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.251880253649904e-05,
      "loss": 0.0,
      "step": 4230
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2504055449048815e-05,
      "loss": 0.0001,
      "step": 4232
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.2489308361598586e-05,
      "loss": 0.0,
      "step": 4234
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2474561274148356e-05,
      "loss": 0.0,
      "step": 4236
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2459814186698126e-05,
      "loss": 0.0,
      "step": 4238
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.24450670992479e-05,
      "loss": 0.0001,
      "step": 4240
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.243032001179767e-05,
      "loss": 0.0,
      "step": 4242
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.241557292434744e-05,
      "loss": 0.0,
      "step": 4244
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2400825836897215e-05,
      "loss": 0.0001,
      "step": 4246
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2386078749446985e-05,
      "loss": 0.0,
      "step": 4248
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2371331661996755e-05,
      "loss": 0.0,
      "step": 4250
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2356584574546526e-05,
      "loss": 0.0,
      "step": 4252
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.23418374870963e-05,
      "loss": 0.0,
      "step": 4254
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2327090399646073e-05,
      "loss": 0.0,
      "step": 4256
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.2312343312195844e-05,
      "loss": 0.0,
      "step": 4258
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2297596224745614e-05,
      "loss": 0.0001,
      "step": 4260
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2282849137295388e-05,
      "loss": 0.0001,
      "step": 4262
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2268102049845158e-05,
      "loss": 0.0,
      "step": 4264
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.225335496239493e-05,
      "loss": 0.0005,
      "step": 4266
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.22386078749447e-05,
      "loss": 0.0,
      "step": 4268
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2223860787494473e-05,
      "loss": 0.0001,
      "step": 4270
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2209113700044243e-05,
      "loss": 0.0001,
      "step": 4272
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2194366612594013e-05,
      "loss": 0.0007,
      "step": 4274
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2179619525143787e-05,
      "loss": 0.0,
      "step": 4276
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2164872437693557e-05,
      "loss": 0.0001,
      "step": 4278
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2150125350243328e-05,
      "loss": 0.0001,
      "step": 4280
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2135378262793098e-05,
      "loss": 0.0001,
      "step": 4282
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2120631175342872e-05,
      "loss": 0.0002,
      "step": 4284
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2105884087892642e-05,
      "loss": 0.0002,
      "step": 4286
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2091137000442413e-05,
      "loss": 0.0003,
      "step": 4288
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2076389912992186e-05,
      "loss": 0.0001,
      "step": 4290
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2061642825541957e-05,
      "loss": 0.0002,
      "step": 4292
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2046895738091727e-05,
      "loss": 0.0002,
      "step": 4294
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2032148650641497e-05,
      "loss": 0.0002,
      "step": 4296
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.201740156319127e-05,
      "loss": 0.0004,
      "step": 4298
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.200265447574104e-05,
      "loss": 0.0002,
      "step": 4300
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.1987907388290812e-05,
      "loss": 0.0001,
      "step": 4302
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.1973160300840582e-05,
      "loss": 0.0002,
      "step": 4304
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.1958413213390356e-05,
      "loss": 0.0003,
      "step": 4306
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1943666125940126e-05,
      "loss": 0.0001,
      "step": 4308
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1928919038489897e-05,
      "loss": 0.0002,
      "step": 4310
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.191417195103967e-05,
      "loss": 0.0001,
      "step": 4312
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1899424863589444e-05,
      "loss": 0.0001,
      "step": 4314
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1884677776139215e-05,
      "loss": 0.0001,
      "step": 4316
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1869930688688985e-05,
      "loss": 0.0005,
      "step": 4318
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.185518360123876e-05,
      "loss": 0.0002,
      "step": 4320
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.184043651378853e-05,
      "loss": 0.0001,
      "step": 4322
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.18256894263383e-05,
      "loss": 0.0001,
      "step": 4324
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.181094233888807e-05,
      "loss": 0.0001,
      "step": 4326
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1796195251437844e-05,
      "loss": 0.0002,
      "step": 4328
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1781448163987614e-05,
      "loss": 0.0002,
      "step": 4330
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.1766701076537384e-05,
      "loss": 0.0001,
      "step": 4332
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1751953989087158e-05,
      "loss": 0.0001,
      "step": 4334
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.173720690163693e-05,
      "loss": 0.0003,
      "step": 4336
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.17224598141867e-05,
      "loss": 0.0001,
      "step": 4338
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.170771272673647e-05,
      "loss": 0.0001,
      "step": 4340
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1692965639286243e-05,
      "loss": 0.0001,
      "step": 4342
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1678218551836013e-05,
      "loss": 0.0001,
      "step": 4344
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1663471464385784e-05,
      "loss": 0.0003,
      "step": 4346
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1648724376935554e-05,
      "loss": 0.0001,
      "step": 4348
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1633977289485328e-05,
      "loss": 0.0001,
      "step": 4350
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1619230202035098e-05,
      "loss": 0.0001,
      "step": 4352
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.160448311458487e-05,
      "loss": 0.0002,
      "step": 4354
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.1589736027134642e-05,
      "loss": 0.0001,
      "step": 4356
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1574988939684413e-05,
      "loss": 0.0001,
      "step": 4358
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1560241852234183e-05,
      "loss": 0.0,
      "step": 4360
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1545494764783953e-05,
      "loss": 0.0001,
      "step": 4362
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1530747677333727e-05,
      "loss": 0.0001,
      "step": 4364
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.15160005898835e-05,
      "loss": 0.0001,
      "step": 4366
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.150125350243327e-05,
      "loss": 0.0001,
      "step": 4368
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.148650641498304e-05,
      "loss": 0.0001,
      "step": 4370
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1471759327532815e-05,
      "loss": 0.0001,
      "step": 4372
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1457012240082586e-05,
      "loss": 0.0001,
      "step": 4374
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.1442265152632356e-05,
      "loss": 0.0001,
      "step": 4376
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.142751806518213e-05,
      "loss": 0.0001,
      "step": 4378
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.14127709777319e-05,
      "loss": 0.0001,
      "step": 4380
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.139802389028167e-05,
      "loss": 0.0001,
      "step": 4382
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.138327680283144e-05,
      "loss": 0.0002,
      "step": 4384
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.1368529715381215e-05,
      "loss": 0.0002,
      "step": 4386
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.1353782627930985e-05,
      "loss": 0.0001,
      "step": 4388
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.1339035540480755e-05,
      "loss": 0.0001,
      "step": 4390
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.132428845303053e-05,
      "loss": 0.0001,
      "step": 4392
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.13095413655803e-05,
      "loss": 0.0001,
      "step": 4394
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.129479427813007e-05,
      "loss": 0.0,
      "step": 4396
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.128004719067984e-05,
      "loss": 0.0001,
      "step": 4398
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.1265300103229614e-05,
      "loss": 0.0,
      "step": 4400
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.1250553015779384e-05,
      "loss": 0.0001,
      "step": 4402
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.1235805928329155e-05,
      "loss": 0.0,
      "step": 4404
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1221058840878925e-05,
      "loss": 0.0,
      "step": 4406
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.12063117534287e-05,
      "loss": 0.0001,
      "step": 4408
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.119156466597847e-05,
      "loss": 0.0001,
      "step": 4410
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.117681757852824e-05,
      "loss": 0.0001,
      "step": 4412
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1162070491078013e-05,
      "loss": 0.0,
      "step": 4414
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1147323403627784e-05,
      "loss": 0.0,
      "step": 4416
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1132576316177554e-05,
      "loss": 0.0001,
      "step": 4418
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1117829228727325e-05,
      "loss": 0.0,
      "step": 4420
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.11030821412771e-05,
      "loss": 0.0001,
      "step": 4422
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1088335053826872e-05,
      "loss": 0.0001,
      "step": 4424
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1073587966376642e-05,
      "loss": 0.0,
      "step": 4426
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1058840878926413e-05,
      "loss": 0.0,
      "step": 4428
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.1044093791476187e-05,
      "loss": 0.0,
      "step": 4430
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.1029346704025957e-05,
      "loss": 0.0001,
      "step": 4432
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.1014599616575727e-05,
      "loss": 0.0,
      "step": 4434
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.09998525291255e-05,
      "loss": 0.0,
      "step": 4436
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.098510544167527e-05,
      "loss": 0.0,
      "step": 4438
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.0970358354225042e-05,
      "loss": 0.0,
      "step": 4440
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.0955611266774812e-05,
      "loss": 0.0,
      "step": 4442
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.0940864179324586e-05,
      "loss": 0.0001,
      "step": 4444
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.0926117091874356e-05,
      "loss": 0.0001,
      "step": 4446
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.0911370004424127e-05,
      "loss": 0.0,
      "step": 4448
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.0896622916973897e-05,
      "loss": 0.0001,
      "step": 4450
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.088187582952367e-05,
      "loss": 0.0,
      "step": 4452
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.086712874207344e-05,
      "loss": 0.0001,
      "step": 4454
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.085238165462321e-05,
      "loss": 0.0004,
      "step": 4456
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0837634567172985e-05,
      "loss": 0.0,
      "step": 4458
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0822887479722756e-05,
      "loss": 0.0,
      "step": 4460
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0808140392272526e-05,
      "loss": 0.0001,
      "step": 4462
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0793393304822296e-05,
      "loss": 0.0,
      "step": 4464
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.077864621737207e-05,
      "loss": 0.0,
      "step": 4466
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.076389912992184e-05,
      "loss": 0.0001,
      "step": 4468
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.074915204247161e-05,
      "loss": 0.0001,
      "step": 4470
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0734404955021385e-05,
      "loss": 0.0001,
      "step": 4472
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0719657867571155e-05,
      "loss": 0.0001,
      "step": 4474
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.0704910780120925e-05,
      "loss": 0.0001,
      "step": 4476
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.06901636926707e-05,
      "loss": 0.0001,
      "step": 4478
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0675416605220473e-05,
      "loss": 0.0,
      "step": 4480
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0660669517770243e-05,
      "loss": 0.0,
      "step": 4482
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0645922430320014e-05,
      "loss": 0.0002,
      "step": 4484
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0631175342869784e-05,
      "loss": 0.0001,
      "step": 4486
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0616428255419558e-05,
      "loss": 0.0001,
      "step": 4488
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0601681167969328e-05,
      "loss": 0.0001,
      "step": 4490
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.05869340805191e-05,
      "loss": 0.0001,
      "step": 4492
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.057218699306887e-05,
      "loss": 0.0001,
      "step": 4494
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0557439905618642e-05,
      "loss": 0.0001,
      "step": 4496
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0542692818168413e-05,
      "loss": 0.0001,
      "step": 4498
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0527945730718183e-05,
      "loss": 0.0001,
      "step": 4500
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.0513198643267957e-05,
      "loss": 0.0,
      "step": 4502
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0498451555817727e-05,
      "loss": 0.0001,
      "step": 4504
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0483704468367498e-05,
      "loss": 0.0001,
      "step": 4506
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0468957380917268e-05,
      "loss": 0.0001,
      "step": 4508
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0454210293467042e-05,
      "loss": 0.0001,
      "step": 4510
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0439463206016812e-05,
      "loss": 0.0,
      "step": 4512
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0424716118566583e-05,
      "loss": 0.0,
      "step": 4514
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0409969031116356e-05,
      "loss": 0.0001,
      "step": 4516
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0395221943666127e-05,
      "loss": 0.0,
      "step": 4518
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0380474856215897e-05,
      "loss": 0.0001,
      "step": 4520
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.0365727768765667e-05,
      "loss": 0.0003,
      "step": 4522
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.035098068131544e-05,
      "loss": 0.0,
      "step": 4524
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.033623359386521e-05,
      "loss": 0.0001,
      "step": 4526
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0321486506414982e-05,
      "loss": 0.0001,
      "step": 4528
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0306739418964756e-05,
      "loss": 0.0003,
      "step": 4530
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0291992331514526e-05,
      "loss": 0.0,
      "step": 4532
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.02772452440643e-05,
      "loss": 0.0,
      "step": 4534
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.026249815661407e-05,
      "loss": 0.0001,
      "step": 4536
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0247751069163844e-05,
      "loss": 0.0001,
      "step": 4538
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0233003981713614e-05,
      "loss": 0.0,
      "step": 4540
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0218256894263385e-05,
      "loss": 0.0,
      "step": 4542
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.0203509806813155e-05,
      "loss": 0.0003,
      "step": 4544
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.018876271936293e-05,
      "loss": 0.0001,
      "step": 4546
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.01740156319127e-05,
      "loss": 0.0,
      "step": 4548
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.015926854446247e-05,
      "loss": 0.0,
      "step": 4550
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.014452145701224e-05,
      "loss": 0.0,
      "step": 4552
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0129774369562014e-05,
      "loss": 0.0,
      "step": 4554
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0115027282111784e-05,
      "loss": 0.0,
      "step": 4556
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0100280194661554e-05,
      "loss": 0.0001,
      "step": 4558
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0085533107211328e-05,
      "loss": 0.0003,
      "step": 4560
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.00707860197611e-05,
      "loss": 0.0001,
      "step": 4562
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.005603893231087e-05,
      "loss": 0.0001,
      "step": 4564
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.004129184486064e-05,
      "loss": 0.0001,
      "step": 4566
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0026544757410413e-05,
      "loss": 0.0,
      "step": 4568
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0011797669960183e-05,
      "loss": 0.0001,
      "step": 4570
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9997050582509954e-05,
      "loss": 0.0001,
      "step": 4572
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9982303495059727e-05,
      "loss": 0.0001,
      "step": 4574
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9967556407609498e-05,
      "loss": 0.0001,
      "step": 4576
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9952809320159268e-05,
      "loss": 0.0001,
      "step": 4578
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.993806223270904e-05,
      "loss": 0.0001,
      "step": 4580
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9923315145258812e-05,
      "loss": 0.0058,
      "step": 4582
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9908568057808583e-05,
      "loss": 0.0001,
      "step": 4584
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9893820970358353e-05,
      "loss": 0.0001,
      "step": 4586
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9879073882908127e-05,
      "loss": 0.0001,
      "step": 4588
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.98643267954579e-05,
      "loss": 0.0001,
      "step": 4590
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.984957970800767e-05,
      "loss": 0.0001,
      "step": 4592
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.983483262055744e-05,
      "loss": 0.0001,
      "step": 4594
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.982008553310721e-05,
      "loss": 0.0001,
      "step": 4596
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9805338445656985e-05,
      "loss": 0.0001,
      "step": 4598
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9790591358206756e-05,
      "loss": 0.0,
      "step": 4600
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9775844270756526e-05,
      "loss": 0.0001,
      "step": 4602
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.97610971833063e-05,
      "loss": 0.0,
      "step": 4604
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.974635009585607e-05,
      "loss": 0.0001,
      "step": 4606
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.973160300840584e-05,
      "loss": 0.0001,
      "step": 4608
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.971685592095561e-05,
      "loss": 0.0,
      "step": 4610
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9702108833505385e-05,
      "loss": 0.0,
      "step": 4612
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9687361746055155e-05,
      "loss": 0.0001,
      "step": 4614
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9672614658604925e-05,
      "loss": 0.0001,
      "step": 4616
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.96578675711547e-05,
      "loss": 0.0001,
      "step": 4618
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.964312048370447e-05,
      "loss": 0.0,
      "step": 4620
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.962837339625424e-05,
      "loss": 0.0,
      "step": 4622
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.961362630880401e-05,
      "loss": 0.0001,
      "step": 4624
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9598879221353784e-05,
      "loss": 0.0001,
      "step": 4626
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9584132133903554e-05,
      "loss": 0.0001,
      "step": 4628
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9569385046453325e-05,
      "loss": 0.0001,
      "step": 4630
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.95546379590031e-05,
      "loss": 0.0,
      "step": 4632
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.953989087155287e-05,
      "loss": 0.0,
      "step": 4634
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.952514378410264e-05,
      "loss": 0.0001,
      "step": 4636
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.951039669665241e-05,
      "loss": 0.0,
      "step": 4638
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9495649609202183e-05,
      "loss": 0.0001,
      "step": 4640
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9480902521751954e-05,
      "loss": 0.0001,
      "step": 4642
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9466155434301728e-05,
      "loss": 0.0002,
      "step": 4644
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9451408346851498e-05,
      "loss": 0.0,
      "step": 4646
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.943666125940127e-05,
      "loss": 0.0,
      "step": 4648
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9421914171951042e-05,
      "loss": 0.0001,
      "step": 4650
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9407167084500812e-05,
      "loss": 0.0012,
      "step": 4652
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9392419997050583e-05,
      "loss": 0.0,
      "step": 4654
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9377672909600356e-05,
      "loss": 0.0001,
      "step": 4656
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9362925822150127e-05,
      "loss": 0.0,
      "step": 4658
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9348178734699897e-05,
      "loss": 0.0041,
      "step": 4660
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.933343164724967e-05,
      "loss": 0.0001,
      "step": 4662
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.931868455979944e-05,
      "loss": 0.0,
      "step": 4664
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9303937472349212e-05,
      "loss": 0.0001,
      "step": 4666
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9289190384898982e-05,
      "loss": 0.0001,
      "step": 4668
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.9274443297448756e-05,
      "loss": 0.0011,
      "step": 4670
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9259696209998526e-05,
      "loss": 0.0001,
      "step": 4672
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9244949122548297e-05,
      "loss": 0.0001,
      "step": 4674
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.923020203509807e-05,
      "loss": 0.0002,
      "step": 4676
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.921545494764784e-05,
      "loss": 0.0002,
      "step": 4678
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.920070786019761e-05,
      "loss": 0.0001,
      "step": 4680
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.918596077274738e-05,
      "loss": 0.0001,
      "step": 4682
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9171213685297155e-05,
      "loss": 0.0001,
      "step": 4684
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9156466597846926e-05,
      "loss": 0.0002,
      "step": 4686
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9141719510396696e-05,
      "loss": 0.0001,
      "step": 4688
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.912697242294647e-05,
      "loss": 0.0002,
      "step": 4690
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.911222533549624e-05,
      "loss": 0.0001,
      "step": 4692
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.909747824804601e-05,
      "loss": 0.0001,
      "step": 4694
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.908273116059578e-05,
      "loss": 0.0004,
      "step": 4696
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.9067984073145555e-05,
      "loss": 0.0001,
      "step": 4698
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.9053236985695328e-05,
      "loss": 0.0005,
      "step": 4700
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.90384898982451e-05,
      "loss": 0.0001,
      "step": 4702
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.902374281079487e-05,
      "loss": 0.0,
      "step": 4704
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.9008995723344643e-05,
      "loss": 0.0001,
      "step": 4706
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.8994248635894413e-05,
      "loss": 0.0,
      "step": 4708
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.8979501548444183e-05,
      "loss": 0.0001,
      "step": 4710
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.8964754460993954e-05,
      "loss": 0.0,
      "step": 4712
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.8950007373543728e-05,
      "loss": 0.0001,
      "step": 4714
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.8935260286093498e-05,
      "loss": 0.0001,
      "step": 4716
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.892051319864327e-05,
      "loss": 0.0001,
      "step": 4718
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.8905766111193042e-05,
      "loss": 0.0,
      "step": 4720
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8891019023742812e-05,
      "loss": 0.0001,
      "step": 4722
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8876271936292583e-05,
      "loss": 0.0,
      "step": 4724
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8861524848842353e-05,
      "loss": 0.0001,
      "step": 4726
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8846777761392127e-05,
      "loss": 0.0001,
      "step": 4728
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8832030673941897e-05,
      "loss": 0.0001,
      "step": 4730
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8817283586491668e-05,
      "loss": 0.0001,
      "step": 4732
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.880253649904144e-05,
      "loss": 0.0001,
      "step": 4734
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8787789411591212e-05,
      "loss": 0.0,
      "step": 4736
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8773042324140982e-05,
      "loss": 0.0,
      "step": 4738
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8758295236690753e-05,
      "loss": 0.0,
      "step": 4740
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8743548149240526e-05,
      "loss": 0.0,
      "step": 4742
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.8728801061790297e-05,
      "loss": 0.0001,
      "step": 4744
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8714053974340067e-05,
      "loss": 0.0001,
      "step": 4746
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.869930688688984e-05,
      "loss": 0.0001,
      "step": 4748
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.868455979943961e-05,
      "loss": 0.0001,
      "step": 4750
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.866981271198938e-05,
      "loss": 0.0,
      "step": 4752
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8655065624539152e-05,
      "loss": 0.0,
      "step": 4754
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8640318537088926e-05,
      "loss": 0.0001,
      "step": 4756
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.86255714496387e-05,
      "loss": 0.0,
      "step": 4758
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.861082436218847e-05,
      "loss": 0.0001,
      "step": 4760
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.859607727473824e-05,
      "loss": 0.0001,
      "step": 4762
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8581330187288014e-05,
      "loss": 0.0,
      "step": 4764
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8566583099837784e-05,
      "loss": 0.0,
      "step": 4766
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.8551836012387555e-05,
      "loss": 0.0001,
      "step": 4768
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8537088924937325e-05,
      "loss": 0.0,
      "step": 4770
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.85223418374871e-05,
      "loss": 0.0,
      "step": 4772
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.850759475003687e-05,
      "loss": 0.0001,
      "step": 4774
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.849284766258664e-05,
      "loss": 0.0,
      "step": 4776
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8478100575136413e-05,
      "loss": 0.0001,
      "step": 4778
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8463353487686184e-05,
      "loss": 0.0,
      "step": 4780
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8448606400235954e-05,
      "loss": 0.0002,
      "step": 4782
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8433859312785724e-05,
      "loss": 0.0,
      "step": 4784
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8419112225335498e-05,
      "loss": 0.0001,
      "step": 4786
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.840436513788527e-05,
      "loss": 0.0001,
      "step": 4788
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.838961805043504e-05,
      "loss": 0.0,
      "step": 4790
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.8374870962984813e-05,
      "loss": 0.0,
      "step": 4792
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8360123875534583e-05,
      "loss": 0.0,
      "step": 4794
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8345376788084353e-05,
      "loss": 0.0001,
      "step": 4796
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8330629700634124e-05,
      "loss": 0.0,
      "step": 4798
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8315882613183897e-05,
      "loss": 0.0001,
      "step": 4800
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8301135525733668e-05,
      "loss": 0.0001,
      "step": 4802
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8286388438283438e-05,
      "loss": 0.0001,
      "step": 4804
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.827164135083321e-05,
      "loss": 0.0001,
      "step": 4806
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8256894263382982e-05,
      "loss": 0.0001,
      "step": 4808
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8242147175932753e-05,
      "loss": 0.0001,
      "step": 4810
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8227400088482526e-05,
      "loss": 0.0,
      "step": 4812
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8212653001032297e-05,
      "loss": 0.0001,
      "step": 4814
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.819790591358207e-05,
      "loss": 0.0,
      "step": 4816
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.818315882613184e-05,
      "loss": 0.0001,
      "step": 4818
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.816841173868161e-05,
      "loss": 0.0,
      "step": 4820
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8153664651231385e-05,
      "loss": 0.0001,
      "step": 4822
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8138917563781155e-05,
      "loss": 0.0,
      "step": 4824
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8124170476330926e-05,
      "loss": 0.0,
      "step": 4826
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8109423388880696e-05,
      "loss": 0.0001,
      "step": 4828
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.809467630143047e-05,
      "loss": 0.0001,
      "step": 4830
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.807992921398024e-05,
      "loss": 0.0,
      "step": 4832
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.806518212653001e-05,
      "loss": 0.0,
      "step": 4834
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8050435039079784e-05,
      "loss": 0.0,
      "step": 4836
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8035687951629555e-05,
      "loss": 0.0001,
      "step": 4838
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8020940864179325e-05,
      "loss": 0.0001,
      "step": 4840
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.8006193776729095e-05,
      "loss": 0.0,
      "step": 4842
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.799144668927887e-05,
      "loss": 0.0001,
      "step": 4844
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.797669960182864e-05,
      "loss": 0.0001,
      "step": 4846
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.796195251437841e-05,
      "loss": 0.0001,
      "step": 4848
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.7947205426928184e-05,
      "loss": 0.0,
      "step": 4850
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.7932458339477954e-05,
      "loss": 0.0,
      "step": 4852
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.7917711252027724e-05,
      "loss": 0.0001,
      "step": 4854
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.7902964164577495e-05,
      "loss": 0.0,
      "step": 4856
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.788821707712727e-05,
      "loss": 0.0,
      "step": 4858
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.787346998967704e-05,
      "loss": 0.0001,
      "step": 4860
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.785872290222681e-05,
      "loss": 0.0001,
      "step": 4862
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.784397581477658e-05,
      "loss": 0.0,
      "step": 4864
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.7829228727326353e-05,
      "loss": 0.0,
      "step": 4866
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7814481639876127e-05,
      "loss": 0.0001,
      "step": 4868
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7799734552425897e-05,
      "loss": 0.0001,
      "step": 4870
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7784987464975668e-05,
      "loss": 0.0001,
      "step": 4872
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.777024037752544e-05,
      "loss": 0.0,
      "step": 4874
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7755493290075212e-05,
      "loss": 0.0001,
      "step": 4876
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7740746202624982e-05,
      "loss": 0.0001,
      "step": 4878
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7725999115174756e-05,
      "loss": 0.0,
      "step": 4880
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7711252027724526e-05,
      "loss": 0.0001,
      "step": 4882
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7696504940274297e-05,
      "loss": 0.0,
      "step": 4884
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.7681757852824067e-05,
      "loss": 0.0001,
      "step": 4886
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.766701076537384e-05,
      "loss": 0.0,
      "step": 4888
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.765226367792361e-05,
      "loss": 0.0,
      "step": 4890
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.763751659047338e-05,
      "loss": 0.0,
      "step": 4892
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7622769503023155e-05,
      "loss": 0.0,
      "step": 4894
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7608022415572926e-05,
      "loss": 0.0,
      "step": 4896
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7593275328122696e-05,
      "loss": 0.0,
      "step": 4898
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7578528240672467e-05,
      "loss": 0.0,
      "step": 4900
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.756378115322224e-05,
      "loss": 0.0001,
      "step": 4902
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.754903406577201e-05,
      "loss": 0.0,
      "step": 4904
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.753428697832178e-05,
      "loss": 0.0,
      "step": 4906
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.751953989087155e-05,
      "loss": 0.0,
      "step": 4908
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7504792803421325e-05,
      "loss": 0.0,
      "step": 4910
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7490045715971095e-05,
      "loss": 0.0,
      "step": 4912
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.7475298628520866e-05,
      "loss": 0.0,
      "step": 4914
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.746055154107064e-05,
      "loss": 0.0169,
      "step": 4916
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.744580445362041e-05,
      "loss": 0.0001,
      "step": 4918
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.743105736617018e-05,
      "loss": 0.0014,
      "step": 4920
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7416310278719954e-05,
      "loss": 0.0,
      "step": 4922
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7401563191269728e-05,
      "loss": 0.0,
      "step": 4924
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7386816103819498e-05,
      "loss": 0.0,
      "step": 4926
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.737206901636927e-05,
      "loss": 0.0,
      "step": 4928
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.735732192891904e-05,
      "loss": 0.0,
      "step": 4930
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7342574841468813e-05,
      "loss": 0.0001,
      "step": 4932
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7327827754018583e-05,
      "loss": 0.0001,
      "step": 4934
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7313080666568353e-05,
      "loss": 0.0089,
      "step": 4936
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.7298333579118127e-05,
      "loss": 0.0001,
      "step": 4938
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7283586491667898e-05,
      "loss": 0.0026,
      "step": 4940
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7268839404217668e-05,
      "loss": 0.0004,
      "step": 4942
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7254092316767438e-05,
      "loss": 0.0001,
      "step": 4944
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7239345229317212e-05,
      "loss": 0.0002,
      "step": 4946
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7224598141866982e-05,
      "loss": 0.0007,
      "step": 4948
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7209851054416753e-05,
      "loss": 0.0003,
      "step": 4950
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7195103966966523e-05,
      "loss": 0.0008,
      "step": 4952
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7180356879516297e-05,
      "loss": 0.0001,
      "step": 4954
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7165609792066067e-05,
      "loss": 0.0001,
      "step": 4956
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7150862704615838e-05,
      "loss": 0.0001,
      "step": 4958
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.713611561716561e-05,
      "loss": 0.0002,
      "step": 4960
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7121368529715382e-05,
      "loss": 0.0001,
      "step": 4962
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7106621442265152e-05,
      "loss": 0.0002,
      "step": 4964
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7091874354814922e-05,
      "loss": 0.0001,
      "step": 4966
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7077127267364696e-05,
      "loss": 0.0001,
      "step": 4968
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7062380179914467e-05,
      "loss": 0.0001,
      "step": 4970
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7047633092464237e-05,
      "loss": 0.0001,
      "step": 4972
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.703288600501401e-05,
      "loss": 0.0001,
      "step": 4974
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.701813891756378e-05,
      "loss": 0.0001,
      "step": 4976
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7003391830113555e-05,
      "loss": 0.0001,
      "step": 4978
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.6988644742663325e-05,
      "loss": 0.0001,
      "step": 4980
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.69738976552131e-05,
      "loss": 0.0002,
      "step": 4982
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.695915056776287e-05,
      "loss": 0.0001,
      "step": 4984
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.694440348031264e-05,
      "loss": 0.0001,
      "step": 4986
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.692965639286241e-05,
      "loss": 0.0001,
      "step": 4988
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6914909305412184e-05,
      "loss": 0.0002,
      "step": 4990
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6900162217961954e-05,
      "loss": 0.0,
      "step": 4992
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6885415130511725e-05,
      "loss": 0.0001,
      "step": 4994
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.68706680430615e-05,
      "loss": 0.0001,
      "step": 4996
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.685592095561127e-05,
      "loss": 0.0001,
      "step": 4998
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.684117386816104e-05,
      "loss": 0.0001,
      "step": 5000
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.682642678071081e-05,
      "loss": 0.0001,
      "step": 5002
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6811679693260583e-05,
      "loss": 0.0002,
      "step": 5004
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6796932605810354e-05,
      "loss": 0.0001,
      "step": 5006
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6782185518360124e-05,
      "loss": 0.0,
      "step": 5008
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.6767438430909894e-05,
      "loss": 0.0001,
      "step": 5010
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6752691343459668e-05,
      "loss": 0.0,
      "step": 5012
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.673794425600944e-05,
      "loss": 0.0001,
      "step": 5014
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.672319716855921e-05,
      "loss": 0.0001,
      "step": 5016
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6708450081108983e-05,
      "loss": 0.0001,
      "step": 5018
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6693702993658753e-05,
      "loss": 0.0,
      "step": 5020
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6678955906208523e-05,
      "loss": 0.0001,
      "step": 5022
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6664208818758294e-05,
      "loss": 0.0001,
      "step": 5024
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6649461731308067e-05,
      "loss": 0.0,
      "step": 5026
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6634714643857838e-05,
      "loss": 0.0001,
      "step": 5028
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6619967556407608e-05,
      "loss": 0.0001,
      "step": 5030
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6605220468957382e-05,
      "loss": 0.0001,
      "step": 5032
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6590473381507156e-05,
      "loss": 0.0001,
      "step": 5034
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.6575726294056926e-05,
      "loss": 0.0001,
      "step": 5036
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.6560979206606696e-05,
      "loss": 0.0,
      "step": 5038
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.654623211915647e-05,
      "loss": 0.0,
      "step": 5040
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.653148503170624e-05,
      "loss": 0.0001,
      "step": 5042
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.651673794425601e-05,
      "loss": 0.0001,
      "step": 5044
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.650199085680578e-05,
      "loss": 0.0001,
      "step": 5046
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.6487243769355555e-05,
      "loss": 0.0001,
      "step": 5048
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.6472496681905325e-05,
      "loss": 0.0,
      "step": 5050
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.6457749594455096e-05,
      "loss": 0.0001,
      "step": 5052
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.6443002507004866e-05,
      "loss": 0.0,
      "step": 5054
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.642825541955464e-05,
      "loss": 0.0002,
      "step": 5056
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.641350833210441e-05,
      "loss": 0.0001,
      "step": 5058
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.639876124465418e-05,
      "loss": 0.0001,
      "step": 5060
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6384014157203954e-05,
      "loss": 0.0001,
      "step": 5062
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6369267069753725e-05,
      "loss": 0.0001,
      "step": 5064
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6354519982303495e-05,
      "loss": 0.0,
      "step": 5066
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6339772894853265e-05,
      "loss": 0.0,
      "step": 5068
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.632502580740304e-05,
      "loss": 0.0001,
      "step": 5070
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.631027871995281e-05,
      "loss": 0.0,
      "step": 5072
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.629553163250258e-05,
      "loss": 0.0001,
      "step": 5074
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6280784545052354e-05,
      "loss": 0.0001,
      "step": 5076
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6266037457602124e-05,
      "loss": 0.0,
      "step": 5078
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6251290370151894e-05,
      "loss": 0.0,
      "step": 5080
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.6236543282701665e-05,
      "loss": 0.0,
      "step": 5082
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.622179619525144e-05,
      "loss": 0.0,
      "step": 5084
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.620704910780121e-05,
      "loss": 0.0001,
      "step": 5086
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.619230202035098e-05,
      "loss": 0.0001,
      "step": 5088
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6177554932900753e-05,
      "loss": 0.0,
      "step": 5090
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6162807845450527e-05,
      "loss": 0.0001,
      "step": 5092
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6148060758000297e-05,
      "loss": 0.0,
      "step": 5094
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6133313670550067e-05,
      "loss": 0.0,
      "step": 5096
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6118566583099838e-05,
      "loss": 0.0,
      "step": 5098
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.610381949564961e-05,
      "loss": 0.0,
      "step": 5100
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6089072408199382e-05,
      "loss": 0.0001,
      "step": 5102
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6074325320749152e-05,
      "loss": 0.0001,
      "step": 5104
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6059578233298926e-05,
      "loss": 0.0001,
      "step": 5106
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.6044831145848696e-05,
      "loss": 0.0001,
      "step": 5108
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.6030084058398467e-05,
      "loss": 0.0001,
      "step": 5110
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.6015336970948237e-05,
      "loss": 0.0,
      "step": 5112
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.600058988349801e-05,
      "loss": 0.0,
      "step": 5114
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.598584279604778e-05,
      "loss": 0.0001,
      "step": 5116
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.597109570859755e-05,
      "loss": 0.0,
      "step": 5118
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5956348621147325e-05,
      "loss": 0.0,
      "step": 5120
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5941601533697096e-05,
      "loss": 0.0001,
      "step": 5122
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5926854446246866e-05,
      "loss": 0.0001,
      "step": 5124
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5912107358796636e-05,
      "loss": 0.0001,
      "step": 5126
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.589736027134641e-05,
      "loss": 0.0001,
      "step": 5128
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.588261318389618e-05,
      "loss": 0.0001,
      "step": 5130
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.586786609644595e-05,
      "loss": 0.0,
      "step": 5132
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5853119008995725e-05,
      "loss": 0.0001,
      "step": 5134
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5838371921545495e-05,
      "loss": 0.0,
      "step": 5136
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5823624834095265e-05,
      "loss": 0.0,
      "step": 5138
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5808877746645036e-05,
      "loss": 0.0,
      "step": 5140
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.579413065919481e-05,
      "loss": 0.0001,
      "step": 5142
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.577938357174458e-05,
      "loss": 0.0,
      "step": 5144
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5764636484294354e-05,
      "loss": 0.0002,
      "step": 5146
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5749889396844124e-05,
      "loss": 0.0001,
      "step": 5148
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5735142309393898e-05,
      "loss": 0.0,
      "step": 5150
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5720395221943668e-05,
      "loss": 0.0001,
      "step": 5152
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.570564813449344e-05,
      "loss": 0.0001,
      "step": 5154
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.569090104704321e-05,
      "loss": 0.0001,
      "step": 5156
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5676153959592983e-05,
      "loss": 0.0,
      "step": 5158
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5661406872142753e-05,
      "loss": 0.0,
      "step": 5160
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5646659784692523e-05,
      "loss": 0.0,
      "step": 5162
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5631912697242297e-05,
      "loss": 0.0,
      "step": 5164
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5617165609792068e-05,
      "loss": 0.0,
      "step": 5166
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5602418522341838e-05,
      "loss": 0.0001,
      "step": 5168
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5587671434891608e-05,
      "loss": 0.0001,
      "step": 5170
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5572924347441382e-05,
      "loss": 0.0,
      "step": 5172
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5558177259991152e-05,
      "loss": 0.0001,
      "step": 5174
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5543430172540923e-05,
      "loss": 0.0001,
      "step": 5176
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5528683085090696e-05,
      "loss": 0.0,
      "step": 5178
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5513935997640467e-05,
      "loss": 0.0001,
      "step": 5180
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5499188910190237e-05,
      "loss": 0.0,
      "step": 5182
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5484441822740008e-05,
      "loss": 0.0001,
      "step": 5184
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.546969473528978e-05,
      "loss": 0.0001,
      "step": 5186
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5454947647839552e-05,
      "loss": 0.0,
      "step": 5188
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5440200560389322e-05,
      "loss": 0.0,
      "step": 5190
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5425453472939096e-05,
      "loss": 0.0,
      "step": 5192
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5410706385488866e-05,
      "loss": 0.0,
      "step": 5194
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5395959298038637e-05,
      "loss": 0.0,
      "step": 5196
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5381212210588407e-05,
      "loss": 0.0,
      "step": 5198
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.536646512313818e-05,
      "loss": 0.0,
      "step": 5200
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5351718035687954e-05,
      "loss": 0.0,
      "step": 5202
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5336970948237725e-05,
      "loss": 0.0,
      "step": 5204
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5322223860787495e-05,
      "loss": 0.0,
      "step": 5206
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.530747677333727e-05,
      "loss": 0.0,
      "step": 5208
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.529272968588704e-05,
      "loss": 0.0,
      "step": 5210
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.527798259843681e-05,
      "loss": 0.0,
      "step": 5212
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.526323551098658e-05,
      "loss": 0.0,
      "step": 5214
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5248488423536352e-05,
      "loss": 0.0001,
      "step": 5216
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5233741336086124e-05,
      "loss": 0.0001,
      "step": 5218
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5218994248635896e-05,
      "loss": 0.0,
      "step": 5220
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5204247161185667e-05,
      "loss": 0.0,
      "step": 5222
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5189500073735439e-05,
      "loss": 0.0,
      "step": 5224
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5174752986285209e-05,
      "loss": 0.0,
      "step": 5226
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5160005898834981e-05,
      "loss": 0.0,
      "step": 5228
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5145258811384751e-05,
      "loss": 0.0,
      "step": 5230
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5130511723934523e-05,
      "loss": 0.0,
      "step": 5232
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5115764636484296e-05,
      "loss": 0.0001,
      "step": 5234
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5101017549034066e-05,
      "loss": 0.0,
      "step": 5236
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5086270461583838e-05,
      "loss": 0.0,
      "step": 5238
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5071523374133608e-05,
      "loss": 0.0,
      "step": 5240
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.505677628668338e-05,
      "loss": 0.0001,
      "step": 5242
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.504202919923315e-05,
      "loss": 0.0,
      "step": 5244
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5027282111782923e-05,
      "loss": 0.0,
      "step": 5246
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5012535024332693e-05,
      "loss": 0.0,
      "step": 5248
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.4997787936882465e-05,
      "loss": 0.0,
      "step": 5250
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.4983040849432237e-05,
      "loss": 0.0,
      "step": 5252
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.4968293761982008e-05,
      "loss": 0.0,
      "step": 5254
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4953546674531781e-05,
      "loss": 0.0,
      "step": 5256
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4938799587081553e-05,
      "loss": 0.0001,
      "step": 5258
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4924052499631324e-05,
      "loss": 0.0,
      "step": 5260
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4909305412181096e-05,
      "loss": 0.0,
      "step": 5262
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4894558324730868e-05,
      "loss": 0.0,
      "step": 5264
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4879811237280638e-05,
      "loss": 0.0001,
      "step": 5266
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.486506414983041e-05,
      "loss": 0.0,
      "step": 5268
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.485031706238018e-05,
      "loss": 0.0,
      "step": 5270
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4835569974929953e-05,
      "loss": 0.0,
      "step": 5272
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4820822887479723e-05,
      "loss": 0.0,
      "step": 5274
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4806075800029495e-05,
      "loss": 0.0,
      "step": 5276
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.4791328712579267e-05,
      "loss": 0.0,
      "step": 5278
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4776581625129038e-05,
      "loss": 0.0,
      "step": 5280
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.476183453767881e-05,
      "loss": 0.0001,
      "step": 5282
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.474708745022858e-05,
      "loss": 0.0,
      "step": 5284
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4732340362778352e-05,
      "loss": 0.0,
      "step": 5286
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4717593275328123e-05,
      "loss": 0.0001,
      "step": 5288
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4702846187877895e-05,
      "loss": 0.0,
      "step": 5290
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4688099100427665e-05,
      "loss": 0.0,
      "step": 5292
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4673352012977437e-05,
      "loss": 0.0,
      "step": 5294
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4658604925527209e-05,
      "loss": 0.0,
      "step": 5296
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.464385783807698e-05,
      "loss": 0.0,
      "step": 5298
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4629110750626752e-05,
      "loss": 0.0,
      "step": 5300
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.4614363663176522e-05,
      "loss": 0.0,
      "step": 5302
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4599616575726294e-05,
      "loss": 0.0,
      "step": 5304
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4584869488276064e-05,
      "loss": 0.0,
      "step": 5306
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4570122400825836e-05,
      "loss": 0.0001,
      "step": 5308
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4555375313375608e-05,
      "loss": 0.0,
      "step": 5310
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4540628225925382e-05,
      "loss": 0.0,
      "step": 5312
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4525881138475153e-05,
      "loss": 0.0,
      "step": 5314
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4511134051024925e-05,
      "loss": 0.0,
      "step": 5316
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4496386963574695e-05,
      "loss": 0.0,
      "step": 5318
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4481639876124467e-05,
      "loss": 0.0001,
      "step": 5320
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4466892788674239e-05,
      "loss": 0.0,
      "step": 5322
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.445214570122401e-05,
      "loss": 0.0,
      "step": 5324
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.4437398613773782e-05,
      "loss": 0.0,
      "step": 5326
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4422651526323552e-05,
      "loss": 0.0001,
      "step": 5328
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4407904438873324e-05,
      "loss": 0.0001,
      "step": 5330
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4393157351423094e-05,
      "loss": 0.0,
      "step": 5332
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4378410263972866e-05,
      "loss": 0.0,
      "step": 5334
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4363663176522638e-05,
      "loss": 0.0,
      "step": 5336
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4348916089072409e-05,
      "loss": 0.0,
      "step": 5338
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4334169001622181e-05,
      "loss": 0.0,
      "step": 5340
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4319421914171951e-05,
      "loss": 0.0001,
      "step": 5342
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4304674826721723e-05,
      "loss": 0.0,
      "step": 5344
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4289927739271494e-05,
      "loss": 0.0,
      "step": 5346
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4275180651821266e-05,
      "loss": 0.0,
      "step": 5348
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4260433564371036e-05,
      "loss": 0.0,
      "step": 5350
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4245686476920808e-05,
      "loss": 0.0,
      "step": 5352
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.423093938947058e-05,
      "loss": 0.0,
      "step": 5354
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.421619230202035e-05,
      "loss": 0.0,
      "step": 5356
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4201445214570123e-05,
      "loss": 0.0003,
      "step": 5358
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4186698127119893e-05,
      "loss": 0.0,
      "step": 5360
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4171951039669665e-05,
      "loss": 0.0001,
      "step": 5362
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4157203952219435e-05,
      "loss": 0.0,
      "step": 5364
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4142456864769207e-05,
      "loss": 0.0,
      "step": 5366
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4127709777318981e-05,
      "loss": 0.0,
      "step": 5368
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4112962689868753e-05,
      "loss": 0.0001,
      "step": 5370
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4098215602418524e-05,
      "loss": 0.0001,
      "step": 5372
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4083468514968296e-05,
      "loss": 0.0001,
      "step": 5374
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.4068721427518066e-05,
      "loss": 0.0,
      "step": 5376
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.4053974340067838e-05,
      "loss": 0.0,
      "step": 5378
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.403922725261761e-05,
      "loss": 0.0,
      "step": 5380
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.402448016516738e-05,
      "loss": 0.0,
      "step": 5382
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.4009733077717153e-05,
      "loss": 0.0003,
      "step": 5384
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3994985990266923e-05,
      "loss": 0.0,
      "step": 5386
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3980238902816695e-05,
      "loss": 0.0001,
      "step": 5388
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3965491815366465e-05,
      "loss": 0.0,
      "step": 5390
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3950744727916237e-05,
      "loss": 0.0,
      "step": 5392
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3935997640466008e-05,
      "loss": 0.0,
      "step": 5394
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.392125055301578e-05,
      "loss": 0.0,
      "step": 5396
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3906503465565552e-05,
      "loss": 0.0001,
      "step": 5398
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.3891756378115322e-05,
      "loss": 0.0,
      "step": 5400
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3877009290665094e-05,
      "loss": 0.0001,
      "step": 5402
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3862262203214865e-05,
      "loss": 0.0,
      "step": 5404
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3847515115764637e-05,
      "loss": 0.0,
      "step": 5406
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3832768028314407e-05,
      "loss": 0.0,
      "step": 5408
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.381802094086418e-05,
      "loss": 0.0,
      "step": 5410
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3803273853413951e-05,
      "loss": 0.0,
      "step": 5412
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3788526765963722e-05,
      "loss": 0.0,
      "step": 5414
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3773779678513494e-05,
      "loss": 0.0,
      "step": 5416
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3759032591063264e-05,
      "loss": 0.0,
      "step": 5418
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3744285503613036e-05,
      "loss": 0.0,
      "step": 5420
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3729538416162807e-05,
      "loss": 0.0001,
      "step": 5422
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.3714791328712582e-05,
      "loss": 0.0,
      "step": 5424
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3700044241262352e-05,
      "loss": 0.0001,
      "step": 5426
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3685297153812124e-05,
      "loss": 0.0,
      "step": 5428
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3670550066361895e-05,
      "loss": 0.0001,
      "step": 5430
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3655802978911667e-05,
      "loss": 0.0,
      "step": 5432
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3641055891461437e-05,
      "loss": 0.0,
      "step": 5434
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.362630880401121e-05,
      "loss": 0.0,
      "step": 5436
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.361156171656098e-05,
      "loss": 0.0,
      "step": 5438
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3596814629110752e-05,
      "loss": 0.0,
      "step": 5440
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3582067541660524e-05,
      "loss": 0.0,
      "step": 5442
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3567320454210294e-05,
      "loss": 0.0,
      "step": 5444
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3552573366760066e-05,
      "loss": 0.0,
      "step": 5446
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.3537826279309837e-05,
      "loss": 0.0,
      "step": 5448
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3523079191859609e-05,
      "loss": 0.0002,
      "step": 5450
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3508332104409379e-05,
      "loss": 0.0,
      "step": 5452
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3493585016959151e-05,
      "loss": 0.0,
      "step": 5454
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3478837929508923e-05,
      "loss": 0.0001,
      "step": 5456
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3464090842058693e-05,
      "loss": 0.0,
      "step": 5458
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3449343754608466e-05,
      "loss": 0.0,
      "step": 5460
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3434596667158236e-05,
      "loss": 0.0,
      "step": 5462
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3419849579708008e-05,
      "loss": 0.0,
      "step": 5464
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3405102492257778e-05,
      "loss": 0.0,
      "step": 5466
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.339035540480755e-05,
      "loss": 0.0,
      "step": 5468
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.337560831735732e-05,
      "loss": 0.0,
      "step": 5470
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.3360861229907093e-05,
      "loss": 0.0,
      "step": 5472
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3346114142456865e-05,
      "loss": 0.0,
      "step": 5474
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3331367055006635e-05,
      "loss": 0.0,
      "step": 5476
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3316619967556409e-05,
      "loss": 0.0,
      "step": 5478
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3301872880106181e-05,
      "loss": 0.0001,
      "step": 5480
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3287125792655953e-05,
      "loss": 0.0,
      "step": 5482
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3272378705205723e-05,
      "loss": 0.0,
      "step": 5484
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3257631617755496e-05,
      "loss": 0.0,
      "step": 5486
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3242884530305266e-05,
      "loss": 0.0001,
      "step": 5488
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3228137442855038e-05,
      "loss": 0.0,
      "step": 5490
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.3213390355404808e-05,
      "loss": 0.0,
      "step": 5492
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.319864326795458e-05,
      "loss": 0.0,
      "step": 5494
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.318389618050435e-05,
      "loss": 0.0003,
      "step": 5496
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3169149093054123e-05,
      "loss": 0.0,
      "step": 5498
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3154402005603895e-05,
      "loss": 0.0,
      "step": 5500
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3139654918153665e-05,
      "loss": 0.0,
      "step": 5502
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3124907830703437e-05,
      "loss": 0.0,
      "step": 5504
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3110160743253208e-05,
      "loss": 0.0,
      "step": 5506
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.309541365580298e-05,
      "loss": 0.0,
      "step": 5508
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.308066656835275e-05,
      "loss": 0.0,
      "step": 5510
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3065919480902522e-05,
      "loss": 0.0,
      "step": 5512
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3051172393452294e-05,
      "loss": 0.0001,
      "step": 5514
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3036425306002065e-05,
      "loss": 0.0,
      "step": 5516
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3021678218551837e-05,
      "loss": 0.0,
      "step": 5518
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3006931131101607e-05,
      "loss": 0.0001,
      "step": 5520
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2992184043651379e-05,
      "loss": 0.0,
      "step": 5522
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.297743695620115e-05,
      "loss": 0.0,
      "step": 5524
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2962689868750921e-05,
      "loss": 0.0,
      "step": 5526
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2947942781300692e-05,
      "loss": 0.0,
      "step": 5528
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2933195693850464e-05,
      "loss": 0.0,
      "step": 5530
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2918448606400236e-05,
      "loss": 0.0,
      "step": 5532
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.290370151895001e-05,
      "loss": 0.0,
      "step": 5534
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.288895443149978e-05,
      "loss": 0.0,
      "step": 5536
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2874207344049552e-05,
      "loss": 0.0002,
      "step": 5538
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2859460256599323e-05,
      "loss": 0.0,
      "step": 5540
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2844713169149095e-05,
      "loss": 0.0001,
      "step": 5542
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.2829966081698867e-05,
      "loss": 0.0,
      "step": 5544
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2815218994248637e-05,
      "loss": 0.0,
      "step": 5546
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2800471906798409e-05,
      "loss": 0.0,
      "step": 5548
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.278572481934818e-05,
      "loss": 0.0,
      "step": 5550
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2770977731897951e-05,
      "loss": 0.0,
      "step": 5552
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2756230644447722e-05,
      "loss": 0.0,
      "step": 5554
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2741483556997494e-05,
      "loss": 0.0001,
      "step": 5556
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2726736469547266e-05,
      "loss": 0.0,
      "step": 5558
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2711989382097036e-05,
      "loss": 0.0,
      "step": 5560
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2697242294646808e-05,
      "loss": 0.0,
      "step": 5562
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2682495207196579e-05,
      "loss": 0.0,
      "step": 5564
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.266774811974635e-05,
      "loss": 0.0,
      "step": 5566
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.2653001032296121e-05,
      "loss": 0.0,
      "step": 5568
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2638253944845893e-05,
      "loss": 0.0,
      "step": 5570
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2623506857395664e-05,
      "loss": 0.0,
      "step": 5572
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2608759769945436e-05,
      "loss": 0.0,
      "step": 5574
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2594012682495208e-05,
      "loss": 0.0,
      "step": 5576
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2579265595044978e-05,
      "loss": 0.0,
      "step": 5578
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.256451850759475e-05,
      "loss": 0.0,
      "step": 5580
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.254977142014452e-05,
      "loss": 0.0,
      "step": 5582
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2535024332694293e-05,
      "loss": 0.0,
      "step": 5584
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2520277245244063e-05,
      "loss": 0.0,
      "step": 5586
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2505530157793835e-05,
      "loss": 0.0,
      "step": 5588
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2490783070343607e-05,
      "loss": 0.0,
      "step": 5590
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2476035982893379e-05,
      "loss": 0.0,
      "step": 5592
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.246128889544315e-05,
      "loss": 0.0001,
      "step": 5594
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2446541807992922e-05,
      "loss": 0.0,
      "step": 5596
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2431794720542694e-05,
      "loss": 0.0,
      "step": 5598
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2417047633092464e-05,
      "loss": 0.0002,
      "step": 5600
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2402300545642236e-05,
      "loss": 0.0,
      "step": 5602
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2387553458192008e-05,
      "loss": 0.0,
      "step": 5604
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.237280637074178e-05,
      "loss": 0.0,
      "step": 5606
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.235805928329155e-05,
      "loss": 0.0,
      "step": 5608
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2343312195841323e-05,
      "loss": 0.0,
      "step": 5610
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2328565108391093e-05,
      "loss": 0.0,
      "step": 5612
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2313818020940865e-05,
      "loss": 0.0,
      "step": 5614
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2299070933490635e-05,
      "loss": 0.0,
      "step": 5616
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2284323846040407e-05,
      "loss": 0.0,
      "step": 5618
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.226957675859018e-05,
      "loss": 0.0,
      "step": 5620
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.225482967113995e-05,
      "loss": 0.0,
      "step": 5622
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2240082583689722e-05,
      "loss": 0.0032,
      "step": 5624
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2225335496239492e-05,
      "loss": 0.0,
      "step": 5626
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2210588408789264e-05,
      "loss": 0.0,
      "step": 5628
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2195841321339035e-05,
      "loss": 0.0002,
      "step": 5630
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2181094233888808e-05,
      "loss": 0.0002,
      "step": 5632
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2166347146438579e-05,
      "loss": 0.0,
      "step": 5634
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2151600058988351e-05,
      "loss": 0.0001,
      "step": 5636
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2136852971538121e-05,
      "loss": 0.0002,
      "step": 5638
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2122105884087893e-05,
      "loss": 0.0001,
      "step": 5640
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2107358796637665e-05,
      "loss": 0.0,
      "step": 5642
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.2092611709187436e-05,
      "loss": 0.0,
      "step": 5644
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.2077864621737208e-05,
      "loss": 0.0001,
      "step": 5646
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.2063117534286978e-05,
      "loss": 0.0001,
      "step": 5648
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.204837044683675e-05,
      "loss": 0.0001,
      "step": 5650
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.203362335938652e-05,
      "loss": 0.0,
      "step": 5652
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.2018876271936293e-05,
      "loss": 0.0,
      "step": 5654
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.2004129184486065e-05,
      "loss": 0.0,
      "step": 5656
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.1989382097035835e-05,
      "loss": 0.0,
      "step": 5658
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.1974635009585609e-05,
      "loss": 0.0,
      "step": 5660
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.195988792213538e-05,
      "loss": 0.0,
      "step": 5662
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.1945140834685151e-05,
      "loss": 0.0,
      "step": 5664
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.1930393747234922e-05,
      "loss": 0.0,
      "step": 5666
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1915646659784694e-05,
      "loss": 0.0,
      "step": 5668
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1900899572334464e-05,
      "loss": 0.0,
      "step": 5670
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1886152484884236e-05,
      "loss": 0.0,
      "step": 5672
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1871405397434007e-05,
      "loss": 0.0,
      "step": 5674
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1856658309983779e-05,
      "loss": 0.0,
      "step": 5676
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.184191122253355e-05,
      "loss": 0.0,
      "step": 5678
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1827164135083321e-05,
      "loss": 0.0001,
      "step": 5680
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1812417047633093e-05,
      "loss": 0.0,
      "step": 5682
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1797669960182863e-05,
      "loss": 0.0,
      "step": 5684
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1782922872732635e-05,
      "loss": 0.0,
      "step": 5686
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1768175785282408e-05,
      "loss": 0.0,
      "step": 5688
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.175342869783218e-05,
      "loss": 0.0,
      "step": 5690
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.173868161038195e-05,
      "loss": 0.0,
      "step": 5692
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1723934522931722e-05,
      "loss": 0.0,
      "step": 5694
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1709187435481492e-05,
      "loss": 0.0,
      "step": 5696
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1694440348031264e-05,
      "loss": 0.0,
      "step": 5698
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1679693260581037e-05,
      "loss": 0.0,
      "step": 5700
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1664946173130807e-05,
      "loss": 0.0,
      "step": 5702
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1650199085680579e-05,
      "loss": 0.0,
      "step": 5704
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.163545199823035e-05,
      "loss": 0.0,
      "step": 5706
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1620704910780121e-05,
      "loss": 0.0,
      "step": 5708
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1605957823329892e-05,
      "loss": 0.0,
      "step": 5710
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1591210735879664e-05,
      "loss": 0.0,
      "step": 5712
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.1576463648429436e-05,
      "loss": 0.0,
      "step": 5714
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1561716560979208e-05,
      "loss": 0.0,
      "step": 5716
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1546969473528978e-05,
      "loss": 0.0,
      "step": 5718
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.153222238607875e-05,
      "loss": 0.0,
      "step": 5720
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1517475298628522e-05,
      "loss": 0.0,
      "step": 5722
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1502728211178293e-05,
      "loss": 0.0,
      "step": 5724
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1487981123728065e-05,
      "loss": 0.0001,
      "step": 5726
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1473234036277835e-05,
      "loss": 0.0,
      "step": 5728
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1458486948827607e-05,
      "loss": 0.0,
      "step": 5730
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1443739861377378e-05,
      "loss": 0.0,
      "step": 5732
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.142899277392715e-05,
      "loss": 0.0,
      "step": 5734
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1414245686476922e-05,
      "loss": 0.0,
      "step": 5736
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1399498599026692e-05,
      "loss": 0.0,
      "step": 5738
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1384751511576464e-05,
      "loss": 0.0,
      "step": 5740
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1370004424126236e-05,
      "loss": 0.0,
      "step": 5742
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1355257336676008e-05,
      "loss": 0.0,
      "step": 5744
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1340510249225779e-05,
      "loss": 0.0,
      "step": 5746
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.132576316177555e-05,
      "loss": 0.0,
      "step": 5748
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1311016074325321e-05,
      "loss": 0.0,
      "step": 5750
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1296268986875093e-05,
      "loss": 0.0001,
      "step": 5752
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1281521899424864e-05,
      "loss": 0.0001,
      "step": 5754
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1266774811974636e-05,
      "loss": 0.0,
      "step": 5756
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1252027724524408e-05,
      "loss": 0.0,
      "step": 5758
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.1237280637074178e-05,
      "loss": 0.0,
      "step": 5760
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.122253354962395e-05,
      "loss": 0.0,
      "step": 5762
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.120778646217372e-05,
      "loss": 0.0,
      "step": 5764
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1193039374723492e-05,
      "loss": 0.0,
      "step": 5766
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1178292287273263e-05,
      "loss": 0.0,
      "step": 5768
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1163545199823037e-05,
      "loss": 0.0,
      "step": 5770
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1148798112372807e-05,
      "loss": 0.0,
      "step": 5772
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1134051024922579e-05,
      "loss": 0.0,
      "step": 5774
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.111930393747235e-05,
      "loss": 0.0,
      "step": 5776
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1104556850022121e-05,
      "loss": 0.0,
      "step": 5778
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1089809762571894e-05,
      "loss": 0.0001,
      "step": 5780
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1075062675121664e-05,
      "loss": 0.0,
      "step": 5782
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1060315587671436e-05,
      "loss": 0.0001,
      "step": 5784
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1045568500221206e-05,
      "loss": 0.0,
      "step": 5786
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1030821412770978e-05,
      "loss": 0.0001,
      "step": 5788
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.1016074325320749e-05,
      "loss": 0.0,
      "step": 5790
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.100132723787052e-05,
      "loss": 0.0,
      "step": 5792
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0986580150420291e-05,
      "loss": 0.0,
      "step": 5794
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0971833062970063e-05,
      "loss": 0.0001,
      "step": 5796
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0957085975519835e-05,
      "loss": 0.0,
      "step": 5798
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0942338888069607e-05,
      "loss": 0.0,
      "step": 5800
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.092759180061938e-05,
      "loss": 0.0,
      "step": 5802
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.091284471316915e-05,
      "loss": 0.0,
      "step": 5804
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0898097625718922e-05,
      "loss": 0.0,
      "step": 5806
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0883350538268692e-05,
      "loss": 0.0,
      "step": 5808
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0868603450818464e-05,
      "loss": 0.0,
      "step": 5810
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.0853856363368235e-05,
      "loss": 0.0,
      "step": 5812
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0839109275918007e-05,
      "loss": 0.0,
      "step": 5814
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0824362188467777e-05,
      "loss": 0.0,
      "step": 5816
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0809615101017549e-05,
      "loss": 0.0,
      "step": 5818
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0794868013567321e-05,
      "loss": 0.0,
      "step": 5820
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0780120926117092e-05,
      "loss": 0.0,
      "step": 5822
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0765373838666864e-05,
      "loss": 0.0,
      "step": 5824
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0750626751216636e-05,
      "loss": 0.0,
      "step": 5826
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0735879663766408e-05,
      "loss": 0.0,
      "step": 5828
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0721132576316178e-05,
      "loss": 0.0,
      "step": 5830
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.070638548886595e-05,
      "loss": 0.0,
      "step": 5832
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.069163840141572e-05,
      "loss": 0.0,
      "step": 5834
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.0676891313965493e-05,
      "loss": 0.0,
      "step": 5836
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0662144226515265e-05,
      "loss": 0.0001,
      "step": 5838
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0647397139065035e-05,
      "loss": 0.0,
      "step": 5840
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0632650051614807e-05,
      "loss": 0.0,
      "step": 5842
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0617902964164577e-05,
      "loss": 0.0,
      "step": 5844
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.060315587671435e-05,
      "loss": 0.0,
      "step": 5846
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.058840878926412e-05,
      "loss": 0.0,
      "step": 5848
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0573661701813892e-05,
      "loss": 0.0,
      "step": 5850
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0558914614363662e-05,
      "loss": 0.0,
      "step": 5852
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0544167526913436e-05,
      "loss": 0.0,
      "step": 5854
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0529420439463206e-05,
      "loss": 0.0,
      "step": 5856
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0514673352012978e-05,
      "loss": 0.0,
      "step": 5858
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.049992626456275e-05,
      "loss": 0.0001,
      "step": 5860
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0485179177112521e-05,
      "loss": 0.0,
      "step": 5862
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0470432089662293e-05,
      "loss": 0.0,
      "step": 5864
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0455685002212063e-05,
      "loss": 0.0,
      "step": 5866
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0440937914761835e-05,
      "loss": 0.0,
      "step": 5868
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0426190827311606e-05,
      "loss": 0.0,
      "step": 5870
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0411443739861378e-05,
      "loss": 0.0,
      "step": 5872
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0396696652411148e-05,
      "loss": 0.0001,
      "step": 5874
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.038194956496092e-05,
      "loss": 0.0,
      "step": 5876
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0367202477510692e-05,
      "loss": 0.0,
      "step": 5878
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0352455390060463e-05,
      "loss": 0.0,
      "step": 5880
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0337708302610236e-05,
      "loss": 0.0,
      "step": 5882
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0322961215160007e-05,
      "loss": 0.0,
      "step": 5884
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0308214127709779e-05,
      "loss": 0.0,
      "step": 5886
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.029346704025955e-05,
      "loss": 0.0,
      "step": 5888
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0278719952809321e-05,
      "loss": 0.0002,
      "step": 5890
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0263972865359092e-05,
      "loss": 0.0,
      "step": 5892
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0249225777908864e-05,
      "loss": 0.0,
      "step": 5894
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0234478690458634e-05,
      "loss": 0.0,
      "step": 5896
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0219731603008406e-05,
      "loss": 0.0001,
      "step": 5898
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0204984515558178e-05,
      "loss": 0.0,
      "step": 5900
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0190237428107949e-05,
      "loss": 0.0,
      "step": 5902
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.017549034065772e-05,
      "loss": 0.0,
      "step": 5904
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0160743253207491e-05,
      "loss": 0.0,
      "step": 5906
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0145996165757263e-05,
      "loss": 0.0,
      "step": 5908
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0131249078307035e-05,
      "loss": 0.0,
      "step": 5910
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0116501990856807e-05,
      "loss": 0.0001,
      "step": 5912
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0101754903406577e-05,
      "loss": 0.0,
      "step": 5914
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.008700781595635e-05,
      "loss": 0.0,
      "step": 5916
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.007226072850612e-05,
      "loss": 0.0,
      "step": 5918
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0057513641055892e-05,
      "loss": 0.0,
      "step": 5920
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0042766553605664e-05,
      "loss": 0.0001,
      "step": 5922
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0028019466155434e-05,
      "loss": 0.0,
      "step": 5924
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0013272378705206e-05,
      "loss": 0.0,
      "step": 5926
    },
    {
      "epoch": 2.44,
      "learning_rate": 9.998525291254977e-06,
      "loss": 0.0,
      "step": 5928
    },
    {
      "epoch": 2.44,
      "learning_rate": 9.983778203804749e-06,
      "loss": 0.0,
      "step": 5930
    },
    {
      "epoch": 2.44,
      "learning_rate": 9.96903111635452e-06,
      "loss": 0.0,
      "step": 5932
    },
    {
      "epoch": 2.44,
      "learning_rate": 9.954284028904291e-06,
      "loss": 0.0,
      "step": 5934
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.939536941454063e-06,
      "loss": 0.0,
      "step": 5936
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.924789854003835e-06,
      "loss": 0.0,
      "step": 5938
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.910042766553606e-06,
      "loss": 0.0,
      "step": 5940
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.895295679103378e-06,
      "loss": 0.0,
      "step": 5942
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.88054859165315e-06,
      "loss": 0.0,
      "step": 5944
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.86580150420292e-06,
      "loss": 0.0,
      "step": 5946
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.851054416752692e-06,
      "loss": 0.0,
      "step": 5948
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.836307329302463e-06,
      "loss": 0.0,
      "step": 5950
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.821560241852235e-06,
      "loss": 0.0,
      "step": 5952
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.806813154402005e-06,
      "loss": 0.0,
      "step": 5954
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.792066066951777e-06,
      "loss": 0.0,
      "step": 5956
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.77731897950155e-06,
      "loss": 0.0001,
      "step": 5958
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.76257189205132e-06,
      "loss": 0.0,
      "step": 5960
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.747824804601092e-06,
      "loss": 0.0001,
      "step": 5962
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.733077717150864e-06,
      "loss": 0.0,
      "step": 5964
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.718330629700636e-06,
      "loss": 0.0,
      "step": 5966
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.703583542250406e-06,
      "loss": 0.0,
      "step": 5968
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.688836454800178e-06,
      "loss": 0.0,
      "step": 5970
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.674089367349949e-06,
      "loss": 0.0,
      "step": 5972
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.65934227989972e-06,
      "loss": 0.0,
      "step": 5974
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.644595192449491e-06,
      "loss": 0.0,
      "step": 5976
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.629848104999263e-06,
      "loss": 0.0,
      "step": 5978
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.615101017549035e-06,
      "loss": 0.0,
      "step": 5980
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.600353930098806e-06,
      "loss": 0.0,
      "step": 5982
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.585606842648578e-06,
      "loss": 0.0,
      "step": 5984
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.570859755198348e-06,
      "loss": 0.0,
      "step": 5986
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.55611266774812e-06,
      "loss": 0.0,
      "step": 5988
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.54136558029789e-06,
      "loss": 0.0,
      "step": 5990
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.526618492847664e-06,
      "loss": 0.0,
      "step": 5992
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.511871405397434e-06,
      "loss": 0.0,
      "step": 5994
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.497124317947207e-06,
      "loss": 0.0001,
      "step": 5996
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.482377230496977e-06,
      "loss": 0.0,
      "step": 5998
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.467630143046749e-06,
      "loss": 0.0,
      "step": 6000
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.452883055596521e-06,
      "loss": 0.0,
      "step": 6002
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.438135968146291e-06,
      "loss": 0.0,
      "step": 6004
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.423388880696063e-06,
      "loss": 0.0,
      "step": 6006
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.408641793245834e-06,
      "loss": 0.0,
      "step": 6008
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.393894705795606e-06,
      "loss": 0.0,
      "step": 6010
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.379147618345376e-06,
      "loss": 0.0,
      "step": 6012
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.364400530895148e-06,
      "loss": 0.0,
      "step": 6014
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.34965344344492e-06,
      "loss": 0.0,
      "step": 6016
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.33490635599469e-06,
      "loss": 0.0,
      "step": 6018
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.320159268544463e-06,
      "loss": 0.0001,
      "step": 6020
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.305412181094235e-06,
      "loss": 0.0,
      "step": 6022
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.290665093644007e-06,
      "loss": 0.0,
      "step": 6024
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.275918006193777e-06,
      "loss": 0.0,
      "step": 6026
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.26117091874355e-06,
      "loss": 0.0,
      "step": 6028
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.24642383129332e-06,
      "loss": 0.0,
      "step": 6030
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.231676743843092e-06,
      "loss": 0.0,
      "step": 6032
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.216929656392862e-06,
      "loss": 0.0002,
      "step": 6034
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.202182568942634e-06,
      "loss": 0.0,
      "step": 6036
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.187435481492406e-06,
      "loss": 0.0,
      "step": 6038
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.172688394042177e-06,
      "loss": 0.0,
      "step": 6040
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.157941306591949e-06,
      "loss": 0.0,
      "step": 6042
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.143194219141719e-06,
      "loss": 0.0,
      "step": 6044
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.128447131691491e-06,
      "loss": 0.0,
      "step": 6046
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.113700044241263e-06,
      "loss": 0.0,
      "step": 6048
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.098952956791035e-06,
      "loss": 0.0001,
      "step": 6050
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.084205869340806e-06,
      "loss": 0.0,
      "step": 6052
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.069458781890578e-06,
      "loss": 0.0,
      "step": 6054
    },
    {
      "epoch": 2.5,
      "learning_rate": 9.054711694440348e-06,
      "loss": 0.0,
      "step": 6056
    },
    {
      "epoch": 2.5,
      "learning_rate": 9.03996460699012e-06,
      "loss": 0.0,
      "step": 6058
    },
    {
      "epoch": 2.5,
      "learning_rate": 9.025217519539892e-06,
      "loss": 0.0,
      "step": 6060
    },
    {
      "epoch": 2.5,
      "learning_rate": 9.010470432089663e-06,
      "loss": 0.0,
      "step": 6062
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.995723344639435e-06,
      "loss": 0.0,
      "step": 6064
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.980976257189205e-06,
      "loss": 0.0,
      "step": 6066
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.966229169738977e-06,
      "loss": 0.0,
      "step": 6068
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.951482082288747e-06,
      "loss": 0.0,
      "step": 6070
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.93673499483852e-06,
      "loss": 0.0,
      "step": 6072
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.92198790738829e-06,
      "loss": 0.0,
      "step": 6074
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.907240819938064e-06,
      "loss": 0.0,
      "step": 6076
    },
    {
      "epoch": 2.5,
      "learning_rate": 8.892493732487834e-06,
      "loss": 0.0,
      "step": 6078
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.877746645037606e-06,
      "loss": 0.0,
      "step": 6080
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.862999557587378e-06,
      "loss": 0.0,
      "step": 6082
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.848252470137148e-06,
      "loss": 0.0,
      "step": 6084
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.83350538268692e-06,
      "loss": 0.0,
      "step": 6086
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.81875829523669e-06,
      "loss": 0.0,
      "step": 6088
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.804011207786463e-06,
      "loss": 0.0,
      "step": 6090
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.789264120336233e-06,
      "loss": 0.0,
      "step": 6092
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.774517032886005e-06,
      "loss": 0.0,
      "step": 6094
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.759769945435776e-06,
      "loss": 0.0,
      "step": 6096
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.745022857985548e-06,
      "loss": 0.0,
      "step": 6098
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.73027577053532e-06,
      "loss": 0.0,
      "step": 6100
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.71552868308509e-06,
      "loss": 0.0,
      "step": 6102
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.700781595634864e-06,
      "loss": 0.0,
      "step": 6104
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.686034508184634e-06,
      "loss": 0.0,
      "step": 6106
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.671287420734406e-06,
      "loss": 0.0,
      "step": 6108
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.656540333284177e-06,
      "loss": 0.0,
      "step": 6110
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.641793245833949e-06,
      "loss": 0.0,
      "step": 6112
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.627046158383719e-06,
      "loss": 0.0,
      "step": 6114
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.612299070933491e-06,
      "loss": 0.0,
      "step": 6116
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.597551983483262e-06,
      "loss": 0.0,
      "step": 6118
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.582804896033034e-06,
      "loss": 0.0,
      "step": 6120
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.568057808582806e-06,
      "loss": 0.0001,
      "step": 6122
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.553310721132576e-06,
      "loss": 0.0,
      "step": 6124
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.538563633682348e-06,
      "loss": 0.0,
      "step": 6126
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.523816546232118e-06,
      "loss": 0.0,
      "step": 6128
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.50906945878189e-06,
      "loss": 0.0,
      "step": 6130
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.494322371331663e-06,
      "loss": 0.0,
      "step": 6132
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.479575283881435e-06,
      "loss": 0.0,
      "step": 6134
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.464828196431205e-06,
      "loss": 0.0,
      "step": 6136
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.450081108980977e-06,
      "loss": 0.0,
      "step": 6138
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.43533402153075e-06,
      "loss": 0.0,
      "step": 6140
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.42058693408052e-06,
      "loss": 0.0,
      "step": 6142
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.405839846630292e-06,
      "loss": 0.0,
      "step": 6144
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.391092759180062e-06,
      "loss": 0.0002,
      "step": 6146
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.376345671729834e-06,
      "loss": 0.0,
      "step": 6148
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.361598584279604e-06,
      "loss": 0.0,
      "step": 6150
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.346851496829376e-06,
      "loss": 0.0,
      "step": 6152
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.332104409379147e-06,
      "loss": 0.0,
      "step": 6154
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.317357321928919e-06,
      "loss": 0.0,
      "step": 6156
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.302610234478691e-06,
      "loss": 0.0,
      "step": 6158
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.287863147028463e-06,
      "loss": 0.0,
      "step": 6160
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.273116059578235e-06,
      "loss": 0.0,
      "step": 6162
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.258368972128005e-06,
      "loss": 0.0,
      "step": 6164
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.243621884677777e-06,
      "loss": 0.0,
      "step": 6166
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.228874797227548e-06,
      "loss": 0.0,
      "step": 6168
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.21412770977732e-06,
      "loss": 0.0,
      "step": 6170
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.19938062232709e-06,
      "loss": 0.0,
      "step": 6172
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.184633534876862e-06,
      "loss": 0.0,
      "step": 6174
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.169886447426633e-06,
      "loss": 0.0,
      "step": 6176
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.155139359976405e-06,
      "loss": 0.0,
      "step": 6178
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.140392272526177e-06,
      "loss": 0.0,
      "step": 6180
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.125645185075947e-06,
      "loss": 0.0,
      "step": 6182
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.11089809762572e-06,
      "loss": 0.0,
      "step": 6184
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.09615101017549e-06,
      "loss": 0.0,
      "step": 6186
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.081403922725263e-06,
      "loss": 0.0,
      "step": 6188
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.066656835275034e-06,
      "loss": 0.0,
      "step": 6190
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.051909747824806e-06,
      "loss": 0.0,
      "step": 6192
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.037162660374576e-06,
      "loss": 0.0,
      "step": 6194
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.022415572924348e-06,
      "loss": 0.0,
      "step": 6196
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.007668485474119e-06,
      "loss": 0.0,
      "step": 6198
    },
    {
      "epoch": 2.55,
      "learning_rate": 7.99292139802389e-06,
      "loss": 0.0,
      "step": 6200
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.978174310573663e-06,
      "loss": 0.0,
      "step": 6202
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.963427223123433e-06,
      "loss": 0.0,
      "step": 6204
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.948680135673205e-06,
      "loss": 0.0,
      "step": 6206
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.933933048222975e-06,
      "loss": 0.0002,
      "step": 6208
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.919185960772748e-06,
      "loss": 0.0001,
      "step": 6210
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.904438873322518e-06,
      "loss": 0.0,
      "step": 6212
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.88969178587229e-06,
      "loss": 0.0,
      "step": 6214
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.874944698422062e-06,
      "loss": 0.0,
      "step": 6216
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.860197610971834e-06,
      "loss": 0.0,
      "step": 6218
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.845450523521604e-06,
      "loss": 0.0,
      "step": 6220
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.830703436071377e-06,
      "loss": 0.0,
      "step": 6222
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.815956348621149e-06,
      "loss": 0.0,
      "step": 6224
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.801209261170919e-06,
      "loss": 0.0,
      "step": 6226
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.786462173720691e-06,
      "loss": 0.0,
      "step": 6228
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.771715086270461e-06,
      "loss": 0.0,
      "step": 6230
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.756967998820233e-06,
      "loss": 0.0,
      "step": 6232
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.742220911370004e-06,
      "loss": 0.0,
      "step": 6234
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.727473823919776e-06,
      "loss": 0.0,
      "step": 6236
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.712726736469548e-06,
      "loss": 0.0,
      "step": 6238
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.697979649019318e-06,
      "loss": 0.0,
      "step": 6240
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.68323256156909e-06,
      "loss": 0.0,
      "step": 6242
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.668485474118862e-06,
      "loss": 0.0,
      "step": 6244
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.653738386668634e-06,
      "loss": 0.0,
      "step": 6246
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.638991299218405e-06,
      "loss": 0.0,
      "step": 6248
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.624244211768176e-06,
      "loss": 0.0,
      "step": 6250
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.609497124317948e-06,
      "loss": 0.0,
      "step": 6252
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.594750036867719e-06,
      "loss": 0.0,
      "step": 6254
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.5800029494174905e-06,
      "loss": 0.0,
      "step": 6256
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.565255861967262e-06,
      "loss": 0.0,
      "step": 6258
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.550508774517033e-06,
      "loss": 0.0,
      "step": 6260
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.535761687066804e-06,
      "loss": 0.0,
      "step": 6262
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.521014599616575e-06,
      "loss": 0.0,
      "step": 6264
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.506267512166347e-06,
      "loss": 0.0001,
      "step": 6266
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.491520424716119e-06,
      "loss": 0.0,
      "step": 6268
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.476773337265891e-06,
      "loss": 0.0,
      "step": 6270
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.462026249815662e-06,
      "loss": 0.0,
      "step": 6272
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.447279162365434e-06,
      "loss": 0.0,
      "step": 6274
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.432532074915205e-06,
      "loss": 0.0,
      "step": 6276
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.417784987464976e-06,
      "loss": 0.0,
      "step": 6278
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.403037900014748e-06,
      "loss": 0.0,
      "step": 6280
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.388290812564519e-06,
      "loss": 0.0,
      "step": 6282
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.37354372511429e-06,
      "loss": 0.0,
      "step": 6284
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.358796637664061e-06,
      "loss": 0.0,
      "step": 6286
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.3440495502138325e-06,
      "loss": 0.0001,
      "step": 6288
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.3293024627636045e-06,
      "loss": 0.0,
      "step": 6290
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.314555375313376e-06,
      "loss": 0.0,
      "step": 6292
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.299808287863147e-06,
      "loss": 0.0,
      "step": 6294
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.285061200412918e-06,
      "loss": 0.0,
      "step": 6296
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.270314112962691e-06,
      "loss": 0.0,
      "step": 6298
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.255567025512462e-06,
      "loss": 0.0,
      "step": 6300
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.2408199380622335e-06,
      "loss": 0.0,
      "step": 6302
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.226072850612005e-06,
      "loss": 0.0,
      "step": 6304
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.211325763161776e-06,
      "loss": 0.0,
      "step": 6306
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.196578675711547e-06,
      "loss": 0.0,
      "step": 6308
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.181831588261319e-06,
      "loss": 0.0004,
      "step": 6310
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.1670845008110904e-06,
      "loss": 0.0,
      "step": 6312
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.152337413360862e-06,
      "loss": 0.0,
      "step": 6314
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.137590325910633e-06,
      "loss": 0.0,
      "step": 6316
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.122843238460404e-06,
      "loss": 0.0,
      "step": 6318
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.108096151010175e-06,
      "loss": 0.0001,
      "step": 6320
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.0933490635599465e-06,
      "loss": 0.0,
      "step": 6322
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.078601976109718e-06,
      "loss": 0.0,
      "step": 6324
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.063854888659491e-06,
      "loss": 0.0003,
      "step": 6326
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.049107801209262e-06,
      "loss": 0.0,
      "step": 6328
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.034360713759033e-06,
      "loss": 0.0,
      "step": 6330
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.019613626308805e-06,
      "loss": 0.0,
      "step": 6332
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.004866538858576e-06,
      "loss": 0.0,
      "step": 6334
    },
    {
      "epoch": 2.61,
      "learning_rate": 6.9901194514083475e-06,
      "loss": 0.0,
      "step": 6336
    },
    {
      "epoch": 2.61,
      "learning_rate": 6.975372363958119e-06,
      "loss": 0.0,
      "step": 6338
    },
    {
      "epoch": 2.61,
      "learning_rate": 6.96062527650789e-06,
      "loss": 0.0,
      "step": 6340
    },
    {
      "epoch": 2.61,
      "learning_rate": 6.945878189057661e-06,
      "loss": 0.0,
      "step": 6342
    },
    {
      "epoch": 2.61,
      "learning_rate": 6.931131101607432e-06,
      "loss": 0.0,
      "step": 6344
    },
    {
      "epoch": 2.61,
      "learning_rate": 6.916384014157204e-06,
      "loss": 0.0,
      "step": 6346
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.901636926706976e-06,
      "loss": 0.0001,
      "step": 6348
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.886889839256747e-06,
      "loss": 0.0,
      "step": 6350
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.872142751806518e-06,
      "loss": 0.0,
      "step": 6352
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.857395664356291e-06,
      "loss": 0.0,
      "step": 6354
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.842648576906062e-06,
      "loss": 0.0,
      "step": 6356
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.827901489455833e-06,
      "loss": 0.0001,
      "step": 6358
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.813154402005605e-06,
      "loss": 0.0,
      "step": 6360
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.798407314555376e-06,
      "loss": 0.0,
      "step": 6362
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.783660227105147e-06,
      "loss": 0.0,
      "step": 6364
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.768913139654918e-06,
      "loss": 0.0,
      "step": 6366
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.7541660522046895e-06,
      "loss": 0.0,
      "step": 6368
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.7394189647544615e-06,
      "loss": 0.0,
      "step": 6370
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.724671877304233e-06,
      "loss": 0.0,
      "step": 6372
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.709924789854004e-06,
      "loss": 0.0001,
      "step": 6374
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.695177702403775e-06,
      "loss": 0.0,
      "step": 6376
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.680430614953546e-06,
      "loss": 0.0,
      "step": 6378
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.665683527503318e-06,
      "loss": 0.0,
      "step": 6380
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.6509364400530905e-06,
      "loss": 0.0001,
      "step": 6382
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.636189352602862e-06,
      "loss": 0.0,
      "step": 6384
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.621442265152633e-06,
      "loss": 0.0,
      "step": 6386
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.606695177702404e-06,
      "loss": 0.0,
      "step": 6388
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.591948090252175e-06,
      "loss": 0.0,
      "step": 6390
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.5772010028019474e-06,
      "loss": 0.0,
      "step": 6392
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.562453915351719e-06,
      "loss": 0.0,
      "step": 6394
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.54770682790149e-06,
      "loss": 0.0,
      "step": 6396
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.532959740451261e-06,
      "loss": 0.0,
      "step": 6398
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.518212653001032e-06,
      "loss": 0.0,
      "step": 6400
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.5034655655508035e-06,
      "loss": 0.0,
      "step": 6402
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.488718478100575e-06,
      "loss": 0.0,
      "step": 6404
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.473971390650346e-06,
      "loss": 0.0,
      "step": 6406
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.459224303200118e-06,
      "loss": 0.0,
      "step": 6408
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.44447721574989e-06,
      "loss": 0.0,
      "step": 6410
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.429730128299661e-06,
      "loss": 0.0,
      "step": 6412
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.414983040849433e-06,
      "loss": 0.0,
      "step": 6414
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.4002359533992045e-06,
      "loss": 0.0,
      "step": 6416
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.385488865948976e-06,
      "loss": 0.0,
      "step": 6418
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.370741778498747e-06,
      "loss": 0.0,
      "step": 6420
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.355994691048518e-06,
      "loss": 0.0,
      "step": 6422
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.341247603598289e-06,
      "loss": 0.0,
      "step": 6424
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.326500516148061e-06,
      "loss": 0.0,
      "step": 6426
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.311753428697832e-06,
      "loss": 0.0,
      "step": 6428
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.297006341247604e-06,
      "loss": 0.0,
      "step": 6430
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.282259253797375e-06,
      "loss": 0.0,
      "step": 6432
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.267512166347146e-06,
      "loss": 0.0,
      "step": 6434
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.2527650788969175e-06,
      "loss": 0.0,
      "step": 6436
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.2380179914466896e-06,
      "loss": 0.0,
      "step": 6438
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.223270903996461e-06,
      "loss": 0.0,
      "step": 6440
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.208523816546232e-06,
      "loss": 0.0,
      "step": 6442
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.193776729096004e-06,
      "loss": 0.0,
      "step": 6444
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.179029641645775e-06,
      "loss": 0.0,
      "step": 6446
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.1642825541955465e-06,
      "loss": 0.0,
      "step": 6448
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.149535466745318e-06,
      "loss": 0.0,
      "step": 6450
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.13478837929509e-06,
      "loss": 0.0,
      "step": 6452
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.120041291844861e-06,
      "loss": 0.0,
      "step": 6454
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.105294204394632e-06,
      "loss": 0.0,
      "step": 6456
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.090547116944404e-06,
      "loss": 0.0002,
      "step": 6458
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.0758000294941755e-06,
      "loss": 0.0,
      "step": 6460
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.061052942043947e-06,
      "loss": 0.0,
      "step": 6462
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.046305854593718e-06,
      "loss": 0.0,
      "step": 6464
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.031558767143489e-06,
      "loss": 0.0,
      "step": 6466
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.01681167969326e-06,
      "loss": 0.0,
      "step": 6468
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.002064592243032e-06,
      "loss": 0.0,
      "step": 6470
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.9873175047928044e-06,
      "loss": 0.0,
      "step": 6472
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.972570417342576e-06,
      "loss": 0.0,
      "step": 6474
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.957823329892347e-06,
      "loss": 0.0,
      "step": 6476
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.943076242442118e-06,
      "loss": 0.0,
      "step": 6478
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.928329154991889e-06,
      "loss": 0.0,
      "step": 6480
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.9135820675416605e-06,
      "loss": 0.0,
      "step": 6482
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.898834980091432e-06,
      "loss": 0.0,
      "step": 6484
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.884087892641204e-06,
      "loss": 0.0,
      "step": 6486
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.869340805190975e-06,
      "loss": 0.0,
      "step": 6488
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.854593717740746e-06,
      "loss": 0.0,
      "step": 6490
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.839846630290518e-06,
      "loss": 0.0,
      "step": 6492
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.8250995428402895e-06,
      "loss": 0.0,
      "step": 6494
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.810352455390061e-06,
      "loss": 0.0,
      "step": 6496
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.795605367939832e-06,
      "loss": 0.0,
      "step": 6498
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.780858280489604e-06,
      "loss": 0.0001,
      "step": 6500
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.766111193039375e-06,
      "loss": 0.0,
      "step": 6502
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.751364105589146e-06,
      "loss": 0.0,
      "step": 6504
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.736617018138918e-06,
      "loss": 0.0,
      "step": 6506
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.721869930688689e-06,
      "loss": 0.0,
      "step": 6508
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.707122843238461e-06,
      "loss": 0.0,
      "step": 6510
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.692375755788232e-06,
      "loss": 0.0,
      "step": 6512
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.677628668338004e-06,
      "loss": 0.0,
      "step": 6514
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.662881580887775e-06,
      "loss": 0.0,
      "step": 6516
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.6481344934375466e-06,
      "loss": 0.0,
      "step": 6518
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.633387405987318e-06,
      "loss": 0.0,
      "step": 6520
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.618640318537089e-06,
      "loss": 0.0,
      "step": 6522
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.60389323108686e-06,
      "loss": 0.0,
      "step": 6524
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.589146143636631e-06,
      "loss": 0.0,
      "step": 6526
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.5743990561864035e-06,
      "loss": 0.0,
      "step": 6528
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.559651968736175e-06,
      "loss": 0.0,
      "step": 6530
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.544904881285947e-06,
      "loss": 0.0,
      "step": 6532
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.530157793835718e-06,
      "loss": 0.0,
      "step": 6534
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.515410706385489e-06,
      "loss": 0.0001,
      "step": 6536
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.50066361893526e-06,
      "loss": 0.0,
      "step": 6538
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.485916531485032e-06,
      "loss": 0.0,
      "step": 6540
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.471169444034804e-06,
      "loss": 0.0,
      "step": 6542
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.456422356584575e-06,
      "loss": 0.0,
      "step": 6544
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.441675269134346e-06,
      "loss": 0.0,
      "step": 6546
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.426928181684117e-06,
      "loss": 0.0,
      "step": 6548
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.4121810942338885e-06,
      "loss": 0.0,
      "step": 6550
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.397434006783661e-06,
      "loss": 0.0,
      "step": 6552
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.382686919333432e-06,
      "loss": 0.0,
      "step": 6554
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.367939831883204e-06,
      "loss": 0.0,
      "step": 6556
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.353192744432975e-06,
      "loss": 0.0,
      "step": 6558
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.338445656982746e-06,
      "loss": 0.0,
      "step": 6560
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.3236985695325175e-06,
      "loss": 0.0,
      "step": 6562
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.308951482082289e-06,
      "loss": 0.0,
      "step": 6564
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.29420439463206e-06,
      "loss": 0.0001,
      "step": 6566
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.279457307181831e-06,
      "loss": 0.0,
      "step": 6568
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.264710219731603e-06,
      "loss": 0.0,
      "step": 6570
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.249963132281375e-06,
      "loss": 0.0,
      "step": 6572
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.2352160448311465e-06,
      "loss": 0.0,
      "step": 6574
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.220468957380918e-06,
      "loss": 0.0,
      "step": 6576
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.205721869930689e-06,
      "loss": 0.0,
      "step": 6578
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.19097478248046e-06,
      "loss": 0.0,
      "step": 6580
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.176227695030231e-06,
      "loss": 0.0,
      "step": 6582
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.161480607580003e-06,
      "loss": 0.0,
      "step": 6584
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.146733520129775e-06,
      "loss": 0.0,
      "step": 6586
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.131986432679546e-06,
      "loss": 0.0,
      "step": 6588
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.117239345229317e-06,
      "loss": 0.0,
      "step": 6590
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.102492257779089e-06,
      "loss": 0.0,
      "step": 6592
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.08774517032886e-06,
      "loss": 0.0,
      "step": 6594
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.0729980828786315e-06,
      "loss": 0.0,
      "step": 6596
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.0582509954284036e-06,
      "loss": 0.0,
      "step": 6598
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.043503907978175e-06,
      "loss": 0.0,
      "step": 6600
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.028756820527946e-06,
      "loss": 0.0,
      "step": 6602
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.014009733077717e-06,
      "loss": 0.0,
      "step": 6604
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.999262645627488e-06,
      "loss": 0.0,
      "step": 6606
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.98451555817726e-06,
      "loss": 0.0,
      "step": 6608
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.969768470727032e-06,
      "loss": 0.0,
      "step": 6610
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.955021383276803e-06,
      "loss": 0.0,
      "step": 6612
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.940274295826575e-06,
      "loss": 0.0,
      "step": 6614
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.925527208376346e-06,
      "loss": 0.0,
      "step": 6616
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.910780120926117e-06,
      "loss": 0.0,
      "step": 6618
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.896033033475889e-06,
      "loss": 0.0,
      "step": 6620
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.88128594602566e-06,
      "loss": 0.0002,
      "step": 6622
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.866538858575432e-06,
      "loss": 0.0,
      "step": 6624
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.851791771125203e-06,
      "loss": 0.0,
      "step": 6626
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.837044683674974e-06,
      "loss": 0.0,
      "step": 6628
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.8222975962247455e-06,
      "loss": 0.0,
      "step": 6630
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.807550508774518e-06,
      "loss": 0.0,
      "step": 6632
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.792803421324289e-06,
      "loss": 0.0,
      "step": 6634
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.77805633387406e-06,
      "loss": 0.0,
      "step": 6636
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.763309246423832e-06,
      "loss": 0.0,
      "step": 6638
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.748562158973603e-06,
      "loss": 0.0,
      "step": 6640
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.7338150715233745e-06,
      "loss": 0.0,
      "step": 6642
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.719067984073146e-06,
      "loss": 0.0,
      "step": 6644
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.704320896622917e-06,
      "loss": 0.0,
      "step": 6646
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.689573809172688e-06,
      "loss": 0.0,
      "step": 6648
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.67482672172246e-06,
      "loss": 0.0,
      "step": 6650
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.660079634272231e-06,
      "loss": 0.0,
      "step": 6652
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.6453325468220035e-06,
      "loss": 0.0,
      "step": 6654
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.630585459371775e-06,
      "loss": 0.0,
      "step": 6656
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.615838371921546e-06,
      "loss": 0.0,
      "step": 6658
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.601091284471317e-06,
      "loss": 0.0,
      "step": 6660
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.586344197021088e-06,
      "loss": 0.0,
      "step": 6662
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.5715971095708595e-06,
      "loss": 0.0,
      "step": 6664
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.556850022120632e-06,
      "loss": 0.0,
      "step": 6666
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.542102934670403e-06,
      "loss": 0.0,
      "step": 6668
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.527355847220174e-06,
      "loss": 0.0,
      "step": 6670
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.512608759769946e-06,
      "loss": 0.0,
      "step": 6672
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.497861672319717e-06,
      "loss": 0.0,
      "step": 6674
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.4831145848694885e-06,
      "loss": 0.0,
      "step": 6676
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.46836749741926e-06,
      "loss": 0.0,
      "step": 6678
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.453620409969032e-06,
      "loss": 0.0,
      "step": 6680
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.438873322518803e-06,
      "loss": 0.0,
      "step": 6682
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.424126235068574e-06,
      "loss": 0.0,
      "step": 6684
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.409379147618345e-06,
      "loss": 0.0,
      "step": 6686
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.394632060168117e-06,
      "loss": 0.0,
      "step": 6688
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.379884972717888e-06,
      "loss": 0.0,
      "step": 6690
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.36513788526766e-06,
      "loss": 0.0,
      "step": 6692
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.350390797817432e-06,
      "loss": 0.0001,
      "step": 6694
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.335643710367203e-06,
      "loss": 0.0,
      "step": 6696
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.320896622916974e-06,
      "loss": 0.0,
      "step": 6698
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.306149535466746e-06,
      "loss": 0.0,
      "step": 6700
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.291402448016517e-06,
      "loss": 0.0,
      "step": 6702
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.276655360566288e-06,
      "loss": 0.0,
      "step": 6704
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.261908273116059e-06,
      "loss": 0.0,
      "step": 6706
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.247161185665831e-06,
      "loss": 0.0001,
      "step": 6708
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.2324140982156025e-06,
      "loss": 0.0,
      "step": 6710
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.217667010765375e-06,
      "loss": 0.0,
      "step": 6712
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.202919923315146e-06,
      "loss": 0.0,
      "step": 6714
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.188172835864917e-06,
      "loss": 0.0,
      "step": 6716
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.173425748414688e-06,
      "loss": 0.0,
      "step": 6718
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.1586786609644594e-06,
      "loss": 0.0,
      "step": 6720
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.1439315735142315e-06,
      "loss": 0.0,
      "step": 6722
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.129184486064003e-06,
      "loss": 0.0,
      "step": 6724
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.114437398613774e-06,
      "loss": 0.0,
      "step": 6726
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.099690311163545e-06,
      "loss": 0.0,
      "step": 6728
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.084943223713316e-06,
      "loss": 0.0,
      "step": 6730
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.070196136263088e-06,
      "loss": 0.0,
      "step": 6732
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.05544904881286e-06,
      "loss": 0.0,
      "step": 6734
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.040701961362632e-06,
      "loss": 0.0,
      "step": 6736
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.025954873912403e-06,
      "loss": 0.0,
      "step": 6738
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.011207786462174e-06,
      "loss": 0.0,
      "step": 6740
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.996460699011945e-06,
      "loss": 0.0,
      "step": 6742
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.9817136115617165e-06,
      "loss": 0.0,
      "step": 6744
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.966966524111488e-06,
      "loss": 0.0,
      "step": 6746
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.952219436661259e-06,
      "loss": 0.0,
      "step": 6748
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.937472349211031e-06,
      "loss": 0.0,
      "step": 6750
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.922725261760802e-06,
      "loss": 0.0,
      "step": 6752
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.907978174310574e-06,
      "loss": 0.0,
      "step": 6754
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.8932310868603455e-06,
      "loss": 0.0,
      "step": 6756
    },
    {
      "epoch": 2.78,
      "learning_rate": 3.878483999410117e-06,
      "loss": 0.0,
      "step": 6758
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.863736911959888e-06,
      "loss": 0.0,
      "step": 6760
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.848989824509659e-06,
      "loss": 0.0,
      "step": 6762
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.834242737059431e-06,
      "loss": 0.0,
      "step": 6764
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.819495649609202e-06,
      "loss": 0.0,
      "step": 6766
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.804748562158974e-06,
      "loss": 0.0,
      "step": 6768
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.7900014747087453e-06,
      "loss": 0.0,
      "step": 6770
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.7752543872585165e-06,
      "loss": 0.0,
      "step": 6772
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.7605072998082877e-06,
      "loss": 0.0,
      "step": 6774
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.7457602123580593e-06,
      "loss": 0.0,
      "step": 6776
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.731013124907831e-06,
      "loss": 0.0,
      "step": 6778
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.7162660374576026e-06,
      "loss": 0.0,
      "step": 6780
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.701518950007374e-06,
      "loss": 0.0,
      "step": 6782
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.686771862557145e-06,
      "loss": 0.0,
      "step": 6784
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.6720247751069162e-06,
      "loss": 0.0,
      "step": 6786
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.657277687656688e-06,
      "loss": 0.0,
      "step": 6788
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.642530600206459e-06,
      "loss": 0.0,
      "step": 6790
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.627783512756231e-06,
      "loss": 0.0,
      "step": 6792
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.6130364253060024e-06,
      "loss": 0.0,
      "step": 6794
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.5982893378557736e-06,
      "loss": 0.0,
      "step": 6796
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.5835422504055452e-06,
      "loss": 0.0,
      "step": 6798
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.5687951629553164e-06,
      "loss": 0.0001,
      "step": 6800
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.5540480755050876e-06,
      "loss": 0.0,
      "step": 6802
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.539300988054859e-06,
      "loss": 0.0,
      "step": 6804
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.524553900604631e-06,
      "loss": 0.0001,
      "step": 6806
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.5098068131544026e-06,
      "loss": 0.0,
      "step": 6808
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.4950597257041738e-06,
      "loss": 0.0,
      "step": 6810
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.480312638253945e-06,
      "loss": 0.0,
      "step": 6812
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.465565550803716e-06,
      "loss": 0.0,
      "step": 6814
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.450818463353488e-06,
      "loss": 0.0,
      "step": 6816
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.436071375903259e-06,
      "loss": 0.0,
      "step": 6818
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.421324288453031e-06,
      "loss": 0.0,
      "step": 6820
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.4065772010028023e-06,
      "loss": 0.0,
      "step": 6822
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.3918301135525735e-06,
      "loss": 0.0,
      "step": 6824
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.3770830261023447e-06,
      "loss": 0.0,
      "step": 6826
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.3623359386521164e-06,
      "loss": 0.0,
      "step": 6828
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.3475888512018876e-06,
      "loss": 0.0,
      "step": 6830
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.332841763751659e-06,
      "loss": 0.0,
      "step": 6832
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.318094676301431e-06,
      "loss": 0.0,
      "step": 6834
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.303347588851202e-06,
      "loss": 0.0,
      "step": 6836
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.2886005014009737e-06,
      "loss": 0.0,
      "step": 6838
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.273853413950745e-06,
      "loss": 0.0,
      "step": 6840
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.259106326500516e-06,
      "loss": 0.0,
      "step": 6842
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.2443592390502874e-06,
      "loss": 0.0,
      "step": 6844
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.229612151600059e-06,
      "loss": 0.0,
      "step": 6846
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.2148650641498306e-06,
      "loss": 0.016,
      "step": 6848
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.2001179766996023e-06,
      "loss": 0.0,
      "step": 6850
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.1853708892493735e-06,
      "loss": 0.0,
      "step": 6852
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.1706238017991447e-06,
      "loss": 0.0001,
      "step": 6854
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.155876714348916e-06,
      "loss": 0.0,
      "step": 6856
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.1411296268986875e-06,
      "loss": 0.0,
      "step": 6858
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.1263825394484588e-06,
      "loss": 0.0,
      "step": 6860
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.1116354519982304e-06,
      "loss": 0.0,
      "step": 6862
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.096888364548002e-06,
      "loss": 0.0,
      "step": 6864
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.0821412770977732e-06,
      "loss": 0.0,
      "step": 6866
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.067394189647545e-06,
      "loss": 0.0,
      "step": 6868
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.052647102197316e-06,
      "loss": 0.0,
      "step": 6870
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.0379000147470877e-06,
      "loss": 0.0,
      "step": 6872
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.023152927296859e-06,
      "loss": 0.0,
      "step": 6874
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.00840583984663e-06,
      "loss": 0.0,
      "step": 6876
    },
    {
      "epoch": 2.83,
      "learning_rate": 2.9936587523964022e-06,
      "loss": 0.0,
      "step": 6878
    },
    {
      "epoch": 2.83,
      "learning_rate": 2.9789116649461734e-06,
      "loss": 0.0,
      "step": 6880
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.9641645774959446e-06,
      "loss": 0.0,
      "step": 6882
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.949417490045716e-06,
      "loss": 0.0,
      "step": 6884
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.9346704025954875e-06,
      "loss": 0.0,
      "step": 6886
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.919923315145259e-06,
      "loss": 0.0001,
      "step": 6888
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.9051762276950303e-06,
      "loss": 0.0,
      "step": 6890
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.890429140244802e-06,
      "loss": 0.0,
      "step": 6892
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.875682052794573e-06,
      "loss": 0.0001,
      "step": 6894
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.8609349653443444e-06,
      "loss": 0.0,
      "step": 6896
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.846187877894116e-06,
      "loss": 0.0,
      "step": 6898
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.8314407904438877e-06,
      "loss": 0.0,
      "step": 6900
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.816693702993659e-06,
      "loss": 0.0,
      "step": 6902
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.80194661554343e-06,
      "loss": 0.0,
      "step": 6904
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.7871995280932017e-06,
      "loss": 0.0,
      "step": 6906
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.7724524406429734e-06,
      "loss": 0.0,
      "step": 6908
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.7577053531927446e-06,
      "loss": 0.0,
      "step": 6910
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.742958265742516e-06,
      "loss": 0.0,
      "step": 6912
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.7282111782922874e-06,
      "loss": 0.0,
      "step": 6914
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.7134640908420587e-06,
      "loss": 0.0,
      "step": 6916
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.6987170033918303e-06,
      "loss": 0.0,
      "step": 6918
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.683969915941602e-06,
      "loss": 0.0,
      "step": 6920
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.669222828491373e-06,
      "loss": 0.0,
      "step": 6922
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.6544757410411444e-06,
      "loss": 0.0,
      "step": 6924
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.6397286535909156e-06,
      "loss": 0.0,
      "step": 6926
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.6249815661406876e-06,
      "loss": 0.0,
      "step": 6928
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.610234478690459e-06,
      "loss": 0.0,
      "step": 6930
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.59548739124023e-06,
      "loss": 0.0,
      "step": 6932
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.5807403037900017e-06,
      "loss": 0.0,
      "step": 6934
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.565993216339773e-06,
      "loss": 0.0,
      "step": 6936
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.5512461288895445e-06,
      "loss": 0.0,
      "step": 6938
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.5364990414393158e-06,
      "loss": 0.0,
      "step": 6940
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.5217519539890874e-06,
      "loss": 0.0,
      "step": 6942
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.5070048665388586e-06,
      "loss": 0.0,
      "step": 6944
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.49225777908863e-06,
      "loss": 0.0,
      "step": 6946
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.4775106916384015e-06,
      "loss": 0.0,
      "step": 6948
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.462763604188173e-06,
      "loss": 0.0,
      "step": 6950
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.4480165167379443e-06,
      "loss": 0.0,
      "step": 6952
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.433269429287716e-06,
      "loss": 0.0002,
      "step": 6954
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.418522341837487e-06,
      "loss": 0.0,
      "step": 6956
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.403775254387259e-06,
      "loss": 0.0,
      "step": 6958
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.38902816693703e-06,
      "loss": 0.0,
      "step": 6960
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.3742810794868016e-06,
      "loss": 0.0,
      "step": 6962
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.359533992036573e-06,
      "loss": 0.0,
      "step": 6964
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.344786904586344e-06,
      "loss": 0.0,
      "step": 6966
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.3300398171361157e-06,
      "loss": 0.0,
      "step": 6968
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.3152927296858873e-06,
      "loss": 0.0,
      "step": 6970
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.3005456422356586e-06,
      "loss": 0.0,
      "step": 6972
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.2857985547854298e-06,
      "loss": 0.0,
      "step": 6974
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.2710514673352014e-06,
      "loss": 0.0001,
      "step": 6976
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.256304379884973e-06,
      "loss": 0.0,
      "step": 6978
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.2415572924347443e-06,
      "loss": 0.0,
      "step": 6980
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.226810204984516e-06,
      "loss": 0.0,
      "step": 6982
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.212063117534287e-06,
      "loss": 0.0,
      "step": 6984
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.1973160300840583e-06,
      "loss": 0.0,
      "step": 6986
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.18256894263383e-06,
      "loss": 0.0,
      "step": 6988
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.1678218551836016e-06,
      "loss": 0.0,
      "step": 6990
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.153074767733373e-06,
      "loss": 0.0,
      "step": 6992
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.138327680283144e-06,
      "loss": 0.0,
      "step": 6994
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.1235805928329157e-06,
      "loss": 0.0,
      "step": 6996
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.1088335053826873e-06,
      "loss": 0.0,
      "step": 6998
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.0940864179324585e-06,
      "loss": 0.0002,
      "step": 7000
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.0793393304822297e-06,
      "loss": 0.0,
      "step": 7002
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.0645922430320014e-06,
      "loss": 0.0,
      "step": 7004
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.0498451555817726e-06,
      "loss": 0.0,
      "step": 7006
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.035098068131544e-06,
      "loss": 0.0,
      "step": 7008
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.020350980681316e-06,
      "loss": 0.0,
      "step": 7010
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.005603893231087e-06,
      "loss": 0.0,
      "step": 7012
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.9908568057808583e-06,
      "loss": 0.0,
      "step": 7014
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.9761097183306295e-06,
      "loss": 0.0,
      "step": 7016
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.961362630880401e-06,
      "loss": 0.0,
      "step": 7018
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.9466155434301728e-06,
      "loss": 0.0,
      "step": 7020
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.931868455979944e-06,
      "loss": 0.0,
      "step": 7022
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.9171213685297156e-06,
      "loss": 0.0,
      "step": 7024
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.902374281079487e-06,
      "loss": 0.0,
      "step": 7026
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.8876271936292582e-06,
      "loss": 0.0,
      "step": 7028
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.8728801061790297e-06,
      "loss": 0.0001,
      "step": 7030
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.8581330187288013e-06,
      "loss": 0.0,
      "step": 7032
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.8433859312785725e-06,
      "loss": 0.0,
      "step": 7034
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.828638843828344e-06,
      "loss": 0.0,
      "step": 7036
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.8138917563781156e-06,
      "loss": 0.0,
      "step": 7038
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.7991446689278868e-06,
      "loss": 0.0,
      "step": 7040
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.7843975814776582e-06,
      "loss": 0.0,
      "step": 7042
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.7696504940274294e-06,
      "loss": 0.0,
      "step": 7044
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.7549034065772013e-06,
      "loss": 0.0,
      "step": 7046
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.7401563191269725e-06,
      "loss": 0.0,
      "step": 7048
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.732782775401858e-06,
      "loss": 0.0157,
      "step": 7050
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.7180356879516295e-06,
      "loss": 0.0,
      "step": 7052
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.7032886005014012e-06,
      "loss": 0.0,
      "step": 7054
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.6885415130511724e-06,
      "loss": 0.0,
      "step": 7056
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.6737944256009438e-06,
      "loss": 0.0,
      "step": 7058
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.6590473381507154e-06,
      "loss": 0.0,
      "step": 7060
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.6443002507004869e-06,
      "loss": 0.0,
      "step": 7062
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.629553163250258e-06,
      "loss": 0.0,
      "step": 7064
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.6148060758000295e-06,
      "loss": 0.0,
      "step": 7066
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.6000589883498011e-06,
      "loss": 0.0001,
      "step": 7068
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.5853119008995723e-06,
      "loss": 0.0,
      "step": 7070
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.5705648134493438e-06,
      "loss": 0.0,
      "step": 7072
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.5558177259991152e-06,
      "loss": 0.0,
      "step": 7074
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.5410706385488866e-06,
      "loss": 0.0,
      "step": 7076
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.526323551098658e-06,
      "loss": 0.0,
      "step": 7078
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.5115764636484295e-06,
      "loss": 0.0,
      "step": 7080
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4968293761982011e-06,
      "loss": 0.0,
      "step": 7082
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4820822887479723e-06,
      "loss": 0.0,
      "step": 7084
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4673352012977437e-06,
      "loss": 0.0,
      "step": 7086
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4525881138475152e-06,
      "loss": 0.0,
      "step": 7088
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4378410263972866e-06,
      "loss": 0.0001,
      "step": 7090
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.423093938947058e-06,
      "loss": 0.0,
      "step": 7092
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4083468514968294e-06,
      "loss": 0.0,
      "step": 7094
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.3935997640466009e-06,
      "loss": 0.0,
      "step": 7096
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.3788526765963723e-06,
      "loss": 0.0,
      "step": 7098
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.3641055891461437e-06,
      "loss": 0.0,
      "step": 7100
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.3493585016959151e-06,
      "loss": 0.0,
      "step": 7102
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.3346114142456866e-06,
      "loss": 0.0,
      "step": 7104
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.3198643267954578e-06,
      "loss": 0.0,
      "step": 7106
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.3051172393452294e-06,
      "loss": 0.0,
      "step": 7108
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.2903701518950008e-06,
      "loss": 0.0001,
      "step": 7110
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.2756230644447723e-06,
      "loss": 0.0,
      "step": 7112
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.2608759769945437e-06,
      "loss": 0.0,
      "step": 7114
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.246128889544315e-06,
      "loss": 0.0,
      "step": 7116
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.2313818020940865e-06,
      "loss": 0.0,
      "step": 7118
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.216634714643858e-06,
      "loss": 0.0001,
      "step": 7120
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.2018876271936294e-06,
      "loss": 0.0002,
      "step": 7122
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.1871405397434008e-06,
      "loss": 0.0,
      "step": 7124
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.172393452293172e-06,
      "loss": 0.0,
      "step": 7126
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.1576463648429437e-06,
      "loss": 0.0,
      "step": 7128
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.1428992773927149e-06,
      "loss": 0.0,
      "step": 7130
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.1281521899424865e-06,
      "loss": 0.0,
      "step": 7132
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.113405102492258e-06,
      "loss": 0.0,
      "step": 7134
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0986580150420292e-06,
      "loss": 0.0,
      "step": 7136
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0839109275918008e-06,
      "loss": 0.0,
      "step": 7138
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.069163840141572e-06,
      "loss": 0.0009,
      "step": 7140
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0544167526913436e-06,
      "loss": 0.0,
      "step": 7142
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0396696652411149e-06,
      "loss": 0.0,
      "step": 7144
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0249225777908863e-06,
      "loss": 0.0,
      "step": 7146
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.010175490340658e-06,
      "loss": 0.0,
      "step": 7148
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.954284028904291e-07,
      "loss": 0.0,
      "step": 7150
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.806813154402006e-07,
      "loss": 0.0,
      "step": 7152
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.65934227989972e-07,
      "loss": 0.0,
      "step": 7154
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.511871405397435e-07,
      "loss": 0.0,
      "step": 7156
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.364400530895148e-07,
      "loss": 0.0,
      "step": 7158
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.216929656392863e-07,
      "loss": 0.0,
      "step": 7160
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.069458781890578e-07,
      "loss": 0.0,
      "step": 7162
    },
    {
      "epoch": 2.95,
      "learning_rate": 8.921987907388291e-07,
      "loss": 0.0,
      "step": 7164
    },
    {
      "epoch": 2.95,
      "learning_rate": 8.774517032886006e-07,
      "loss": 0.0,
      "step": 7166
    },
    {
      "epoch": 2.95,
      "learning_rate": 8.62704615838372e-07,
      "loss": 0.0,
      "step": 7168
    },
    {
      "epoch": 2.95,
      "learning_rate": 8.479575283881434e-07,
      "loss": 0.0,
      "step": 7170
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.332104409379147e-07,
      "loss": 0.0,
      "step": 7172
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.184633534876862e-07,
      "loss": 0.0,
      "step": 7174
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.037162660374577e-07,
      "loss": 0.0,
      "step": 7176
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.88969178587229e-07,
      "loss": 0.0001,
      "step": 7178
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.742220911370005e-07,
      "loss": 0.0,
      "step": 7180
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.594750036867719e-07,
      "loss": 0.0,
      "step": 7182
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.447279162365434e-07,
      "loss": 0.0,
      "step": 7184
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.299808287863148e-07,
      "loss": 0.0,
      "step": 7186
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.152337413360861e-07,
      "loss": 0.0,
      "step": 7188
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.004866538858575e-07,
      "loss": 0.0,
      "step": 7190
    },
    {
      "epoch": 2.96,
      "learning_rate": 6.85739566435629e-07,
      "loss": 0.0,
      "step": 7192
    },
    {
      "epoch": 2.96,
      "learning_rate": 6.709924789854005e-07,
      "loss": 0.0,
      "step": 7194
    },
    {
      "epoch": 2.96,
      "learning_rate": 6.562453915351719e-07,
      "loss": 0.0001,
      "step": 7196
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.414983040849432e-07,
      "loss": 0.0,
      "step": 7198
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.267512166347147e-07,
      "loss": 0.0,
      "step": 7200
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.120041291844861e-07,
      "loss": 0.0,
      "step": 7202
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.972570417342575e-07,
      "loss": 0.0,
      "step": 7204
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.825099542840289e-07,
      "loss": 0.0,
      "step": 7206
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.677628668338004e-07,
      "loss": 0.0,
      "step": 7208
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.530157793835718e-07,
      "loss": 0.0,
      "step": 7210
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.382686919333432e-07,
      "loss": 0.0,
      "step": 7212
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.235216044831146e-07,
      "loss": 0.0,
      "step": 7214
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.08774517032886e-07,
      "loss": 0.0,
      "step": 7216
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.940274295826574e-07,
      "loss": 0.0,
      "step": 7218
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.792803421324289e-07,
      "loss": 0.0,
      "step": 7220
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.645332546822003e-07,
      "loss": 0.0,
      "step": 7222
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.497861672319717e-07,
      "loss": 0.0,
      "step": 7224
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.350390797817431e-07,
      "loss": 0.0,
      "step": 7226
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.2029199233151455e-07,
      "loss": 0.0,
      "step": 7228
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.055449048812859e-07,
      "loss": 0.0,
      "step": 7230
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.9079781743105734e-07,
      "loss": 0.0,
      "step": 7232
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.7605072998082877e-07,
      "loss": 0.0,
      "step": 7234
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.6130364253060025e-07,
      "loss": 0.0,
      "step": 7236
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.4655655508037167e-07,
      "loss": 0.0,
      "step": 7238
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.3180946763014304e-07,
      "loss": 0.0,
      "step": 7240
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1706238017991447e-07,
      "loss": 0.0,
      "step": 7242
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.023152927296859e-07,
      "loss": 0.0001,
      "step": 7244
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.875682052794573e-07,
      "loss": 0.0,
      "step": 7246
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.7282111782922874e-07,
      "loss": 0.0001,
      "step": 7248
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.5807403037900017e-07,
      "loss": 0.0,
      "step": 7250
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.433269429287716e-07,
      "loss": 0.0,
      "step": 7252
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.28579855478543e-07,
      "loss": 0.0,
      "step": 7254
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.1383276802831444e-07,
      "loss": 0.0,
      "step": 7256
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.9908568057808584e-07,
      "loss": 0.0,
      "step": 7258
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.8433859312785724e-07,
      "loss": 0.0002,
      "step": 7260
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.6959150567762867e-07,
      "loss": 0.0001,
      "step": 7262
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.548444182274001e-07,
      "loss": 0.0,
      "step": 7264
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.4009733077717152e-07,
      "loss": 0.0,
      "step": 7266
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.2535024332694294e-07,
      "loss": 0.0,
      "step": 7268
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.1060315587671434e-07,
      "loss": 0.0,
      "step": 7270
    },
    {
      "epoch": 3.0,
      "learning_rate": 9.585606842648578e-08,
      "loss": 0.0,
      "step": 7272
    },
    {
      "epoch": 3.0,
      "learning_rate": 8.110898097625719e-08,
      "loss": 0.0,
      "step": 7274
    },
    {
      "epoch": 3.0,
      "learning_rate": 6.636189352602861e-08,
      "loss": 0.0,
      "step": 7276
    },
    {
      "epoch": 3.0,
      "learning_rate": 5.161480607580003e-08,
      "loss": 0.0,
      "step": 7278
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.686771862557145e-08,
      "loss": 0.0,
      "step": 7280
    }
  ],
  "max_steps": 7281,
  "num_train_epochs": 3,
  "total_flos": 2.563609032226944e+16,
  "trial_name": null,
  "trial_params": null
}